NOTES:

KEVIN’S WEBSURFER
HANDBOOK X
FOR CRYSTAL RADIO

MY WEB PAGE VISITED

Kevin Smith
2012

i
284
 NOTES:

Printing / Binding Instructions

1. Choose “fit to page” in print menu
2. Print document double sided on letter size paper
3. Cut the entire printed document in half
4. Fold over making sure the page numbering is continuous
5. For the cover: Print just the first page on card stock paper
Cut the cover in half as well
6. Assemble the covers on the document
7. Punch the left side for a binding, spiral or comb as desired

www.lessmiths.com/~kjsmith/crystal/cr0intro.shtml

KJ Smith

ii
283
NOTES:

INTRODUCTION

In March of 2009 while shopping for some sensors on ebay,

quite on a lark I typed "Crystal Radio" as a search term not
really knowing what to expect. My search was not totally
random, I have always been interested in radio
communications and in radio propagation as a geophysical
probe of the ionosphere, (more about that here). Additionally
as a child I had once a small "Spy Pen" crystal radio what
always facinated me, and as an adult I had given a crystal radio
toy kit to my second son who showed much interest in
computers and electrical things. He has since gone on to his
Master's degree in EE and now workd designing computer
chips at Intel in Silicone Valley. So, maybe it was the radio?

Vaguely I thought to find a crystal radio or radio kit of some

sort to play with, something more serious than a cheap plastic
toy. What I found was suprising, so much there. I explored the
listings, odds and ends mostly of course, its ebay after all, and
a few completed radios new and vintage, and one in partictular
which caught my eye. There were several kits by someone
calling himself Xtalman including one he calls Dunwoody
which seemed to exactly fit my idea, a kit, easy enough for a
beginner but serious enough to satisfy a more educated mind. I
decided to pull the trigger, the radio was offered "buy it now",
so I did! So began my journey into a black hole of interest,
somebody really ought to post warnings on all crystal radios..
"WARNING, may cause serious refocusing of lifestyle".

In the time since, I have plunged deeply into the web literature,
joined the Xtal Set Society, and built homebrews of my own in

iii
282
addition to purchasing a few sets, (described separately). In the The Xtal Set Society Dedicated to once again building and
following pages I present each radio with the ideas and experimenting with radio electronics
motavations that move me to own, admire, and use them. It has Great source for plans and books, kits and parts. I joined the
become a facinating journey for me which I wish to share so society as soon as I finished my Mystery Set.
that others may understand and catch the bug for themselves,
so that other radio builders may view and comment, with Dave Schmarder Dave's Homemade Radios, literally 10's of
suggestions perhaps for improvement, and so us newbies may different sets all with interesting features and each a pleasure
heed the WARNING, and STILL take the plunge! to look at. Much technical background and good design ideas
can be found as well. I like this page.

Scott's Crystal Radios Mount Your Own Crystals

Here I learned, and successfully applied, a good technique to
mount my own crystals.

Part 2, Inspiration:, Sites with radio plans and projects, so

many ideas.

Mike Peebles Crystal Radios - Peebles Originals!

Tom's Handmade Radios Nice selection of radios, superb

craftmanship, for sale!.

Birmingham Alabama Crystal Radio Group Nice contest page,

many lovely contest radios

Jim's Crystal Radio Page Crystal radios with ferrite cores, and
a Mystery Set.

Larry J Solomon Solomon Radios, several variations on the

Mystery Set and others..

Rainer Steinfuehr Gollum´s Crystal Receiver World.

So Many more, if I missed your site my sincere apologies.

iv
281
The Stay Tuned Website is probably the site I visit most often CONTENTS:
of all. This excellent effort contains an enormous amount of
information, vintage plans and projects, Contest Radios, 1) Resources for Theory and Design.
Museums, and more. Download Professor Coyle's coil
calculators, you will use them constantly. This is the only RADIO LABORATORY.. p 003
source for MB Sleeper's 101 radio receiving circuits I have Antenna Tuning p 007
found. Coil Q p 019
Q Analytical p 029
The Smith-Kettlewell Technical File ENERGY-FREE Capacitors and ESR p 047
RADIO: THE CRYSTAL RADIO SET REVISITED Diode Protocol p 055
This is also a great introductory page with a great deal of Diode Test p 065
design tips and background. There is no author and its Radio Test p 099
embedded in an eye research center website, oh well.. I highly
recommend this page. A Library of Crystal Hookups and Theory: p 113

Owen Pool WB4LFH Crystal Radio Resources 2) Radios I have built with my own little mitts:
Owen Pool's great site with much fundamental background,
projects, teaching, contest results and so much more. SPARK GAP TRANSMITTER.. p 119
Coil p 122
Steve Mcdonald VE7SL Radio Notebook Crystal Radio DXing Interruptor p 128
Great site for discussions of DXing with crystal radios and Spark Gap p 134
double-tuned design fundamentals. Oscillation p 139
Performance p 143
Mike Tuggle DESIGNING A DX CRYSTAL SET
Mike's discussion of his Lyonodyne-17 and the concepts SLEEPER #1.. p 155
behind its design.
DUNWOODY.. p 159
Dick Kleijer crystal-radio.eu MYSTERY.. p 165
Dick Kleijer's site contains great discussions of theory and the
HOMEBREW p 171
results of numerous experiments with his crystal radios.
DE FOREST.. p 177
BREMER-TULLY.. p 185
GLOBAL DX'TER.. p 189

v
280
FLEMING.. p 195 Acknowledgements:
http://www.lessmiths.com/~kjsmith/crystal/ack.shtml
MINSTREL BOY.. p 203
HAMMARLUND.. p 209 Radio hobby in general and certainly crystal radio is a
mentored hobby with most practitioners able and willing to
TEFLON.. p 219
help one get started and do well in this hobby. In my
researches for information I have been greatly aided by the
3) Accessories on Breadboards
many helpful web resources built by this community. I wish to
offer here my sincere thanks for your efforts. I certainly would
LOADING COIL.. p 229
not have come as far as I have without this help! The following
WAVE TRAP p 233 list is only partial to the number of sites I visited and if your
page is not listed I offer my apologies. This list represents
AUDIO AMPLIFIER p 235
those sites I found most significant and helpful in terms of 1)
Crystal Radio Fundamentals and 2) Set Design Inspiration.
4) Radios I have acquired (read: paid waaay too much for,
with no regrets)
Part 1, Fundamentals:, Sites where I spend most of my time.
STEINITE.. p 243
Lance Borden BORDEN RADIO COMPANY
AMERICAN RADIO STORES.. p 249 The start of my adventure, serious crystal radio kits, a
BBC COMMERATIVE.. p 253 seriously fine gentleman.

5) Other thoughts, speculations, ideas, ravings... Kenneth A. Kuhn Crystal Radio Engineering
Kenneth Kuhn's excellent pages on the fundamentals and
JACQUEMARD.. p 261 design of crystal radios and all their associated parts,
excellently written, accessable to the non-specialist yet precise
Steel Galena, my thoughts:.. p 265 and clear. I spent hours studying these documents.
KGOW1650, a local PEST!!.. p 275
Alan R. Klase CRYSTAL SET DESIGN 102
6) Acknowledgements:.. (links) p 279 Alan Klase's site is chock-a-block full of great work and ideas
for interesting sets. His Design 102 gave me good insights in
8) Notes p 282 the various incarnations a crystal radio may take.

Darryl Boyd Stay Tuned Website: see #58 Crystal Radio

Circuits 101

vi
279
 vii
278
viii
277
More disturbingly, it takes the joy out of the hobby.
Mercifully, my handy dandy wave trap goes a long way to
solve much of the trouble, but not far enough. I need to direct
my efforts in trapping this station now.

CHAPTER I

Resources for Theory and Design

Question: Is this spectrum normal? Is it common for one

station to so dominate the scene? I sometimes wonder if they
arent cheating on their power output there. I know, probably
not. My next project may be finding a way to use this energy
to charge my batteries or light my house, or heck, to power my
radios! The following plot shows that while by day I am
merely in trouble, at night things are near hopeless. The band
scan is a night spectrum. Sad thing is, I don't partictularily care
to listen to sports on radio, (or television for that matter).

276 1
 Some thoughts on KGOW 1560 "The Game".
http://www.lessmiths.com/~kjsmith/crystal/kgow.shtml

Over the months of playing

with my hobby, my radios,
learning, reading, i have
come across a small fly in
the ointment of an
otherwise very enjoyable
hobby. From time to time
while listening to my radios
I would find the airwaves
filled with a single station, KGOW blasting across my entire
tuning range. It was curious and interesting at first, radios
would receive strong clear signals with or without a crystal or
rectifier which seemed to me thoroughly at odds with what I
supposed I was learning about how radios work. Discussions
on Rap-n-Tap provided some, but not deep insights into my
problem. Mostly I was advised to check twice and thrice all
my connections. The most frustratingly bothersome aspect was
that, it affected all my radios. Maybe my antenna and/or
ground was responsible. Even when hooking up my "Sleeper
#1" (shunting a diode across the phones) gave me the problem,
in fact gave me the problem in spades. Turns out I am able to
receive KGOW clearly merely by attaching the antenna and
ground across the phones, no diode, no nada.

Looking somewhat deeper into the mystery, I chose to look

closer, not at my antenna and ground, but at the BC band itself.
Using my Icon R75 and RxPlus software, I made a band scan
of the BC activity. What I found was most troubling for my
location. The plot below shows a screen capture of the scan.
Clearly KGOW stands out 2 to 3 times stronger than any other
local station. It dominates the scene and saturates my post.

2 275
 My radio "laboratory"
http://www.lessmiths.com/~kjsmith/crystal/lab.shtml

Having now acquired several

crystal radios by purchase
(two) or building my own
(four), I begin to wonder
how they stack up against
each other and against other
radios I read about and
dream of on the net. It can be
fairly common to hear radio
performance described in qualitative terms, "high Q", "sharp
tuning", "highly sensitive", etc. Without measured numbers to
back them up, it is extremely difficult to judge the subjective
notes of different radio users. It is inevitable I believe that in
this hobby that one ends up getting together as much test
equipment as possible, (read "affordable") in order to know
more objectively about the radios one builds and/or uses. This
page is the tale of my own effort to equip a small "laboratory"
of sorts for just such purpose.

My first needs were related to my building projects and

included meters for measuring capacitance and inductance. I
have excellent-quality bench meters for other hobby projects
including a Keithley 192 and 181 voltmeter with IEEE 488
computer interfaces. These are voltmeters only though, nothing
else. My first purchase then was a capacitance/inductance
meter. It worked great for capacitance, but the lowest range for
inductance, 20mH unfortunately is useless. So, this meter was
quickly followed by a stand-alone inductance meter ranging
down to 200uH. I use these frequently when making or
measuring coils, and measuring capacitors purchased on ebay,
better to design a radio around a specific componant.

274 3
 with a surface area of 1,600 square feet. The native lead
My next needs follow logically, what about the radios I build?, occurred along a part of the margin of the ore body in a zone
how do they do? For the non-electrical engineer, testing radios about 9 inches thick separated from the wall rock by clay
is a bit of a daunting task and this page is by no means gouge. The ore generally was argentiferous steel galena partly
intended to be a tutorial. I do give my experiences so that or completely altered to cerussite and anglesite with minor
perhaps someone with the same idea may benefit and possibly minerals including wulfenite, litharge (?), quartz and limonite.
avoid the mistakes I have certainly made. I describe my The native lead occurred as threads, sheets to one-half inch
equipment and the protocols used in my testing, but I thick and pods up to 4xIX2.S inches.
STRONGLY refer the reader to the following two pages for
technical guidance and explanations, most of what I do came
directly from here:
Dick Kleijer crystal-radio.eu http://www.crystal-
radio.eu/enqmeting.htm
Gollum´s Crystal Receiver World
http://home.snafu.de/wumpus/gollum/testing.htm article by
C.A. Lauter
I heartily express my appreciation to our European friends for
making available on the web these excellent technical tutorials
for crystal radio testing.

*****

Testing and understanding your set from the Antenna at the

front, the tank in the middle and on to the Diode at the back
and then the Radio complete, the whole shebang!

ANTENNA TUNING UNIT.. Design notes and Models,

Effect of earth resistivity..

COIL Q.. Summary of web research on expected values of Q

for various coil designs.

4 273
sheared, massive, and 'steel' (a dense, extremely fine-grained, DIODE TEST.. Describes protocols and results of testing
almost cryptocrystalline type). Sheared and steel galena many different diodes, diode classes, crystals, and even two
textures were caused by post-ore shearing of coarse-grained different vacuum tube diodes.
galena and it was not uncommon to find the three varieties
side-by-side. These textural variations had important DIODE PROTOCOL.. Describes a useful protocol for
implications for exploration because the more massive and determination of Diode Is, n, and Ro.
free-grained types, which are less friable, were not affected as
much by oxidation. As galena was released by erosion from DIODE CALIBRATION.. A simple set of modelled
the oxidized vein subcrops, it tended to develop a protective characteristics with variations in Is and n.
white rind of the Pb sulfate anglesite, and to collect downhill
from the vein as a dispersion (float) train. RADIO TEST.. Describes protocols and results of radio
performance testing on my sets.
THE GEOLOGY AND MINERAL DEPOSITS OF THE
TOBY-HORSETHIEF CREEK MAP AREA, NORTHERN
PURCELL MOUNTAINS, SOUTHEAST BRITISH Conclusion
COLUMBIA., OPEN FILE 1990-26. By Alasdaire Pope So, for an investment in a few key pieces of good equipment,
mine consist of massive panidiomorphic galena, sphalerite, yes at some non-trivial cost, you can appreciate the hobby so
pyrite, sucrosic cerussite and banded dolomite, galena, much more. I have an oscilloscope as well, but so far have not
sphalerite and pyrite. In hand specimens bleischweif texture explored its value. I prefer the recording and spreadsheet
(steel galena), indicating deformation by simple shear methods as they give a better record of the results and allow
(McClay, 1980) is frequently seen. graphing and easy comparison. Knowing how your sets
perform, being able to test your diodes, knowing your coil
PROCEEDINGS OF THE THIRTY-THIRD ANNUAL inductance are all important aspects of knowing your radios. I
MEETING OF THE N4INERALOGICAL SOCIETY OF certainly recommend this!
AMERICA AT BOSTON, MASSACHUSETTS. Ciao!
NATIVE LEAD, PRESIDIO COUNTY, TEXAS JORRN T.
I_ONSDALAEN D KATHRYNO . DICKSON University of
Texas, Austin, Texas
Approximately 3,000 pounds of ore containing native lead was
recovered from three small tabular ore bodies in the western
part of the shafter District, presidio county, Texas. Two of the
ore bodies were vein-like in limestone with clay gouge walls.
The metallic lead was localized at the centers of the veins. The
third ore body was flat lying disk-shaped about 4 feet thick

272 5
 Steel galena owes its fine-grained texture either to mechanical
deformation of larger crystals or to a beginning transformation
to anglesite. It is supposed to be rich in silver but this is by no
means always true. PbS+4O=PbSO4 (Anglesite)

Geology of Colorado and Western are Deposits By Arthur

Lakes, 1893.
A coarse galena is generally poor in silver, while fine grained
"steel galena" is generally rich in silver, but the reverse may
also be the case.

Waldschmidt, 1921. Hecla East Vein, Coeur d'Alene

The coarsest galena examined showed cleavage faces from 2 to
4 mm, square, but generally it is much finer grained than this.
The fine-grained galena is commonly known as steel galena
because freshly fractured surfaces have a steel-like luster. In
all the specimens examed, this steel galena appeared to be in
an intimate mixture of galena, tetrahedrite, sphalerite, pyrite,
gnague, and other minerals present in the ore.
It appears that the original ore body consisted chiefly of
galena, sphalerite, pyrite, pyrrhotite, tetrahedrite, freibergite,
quartz and siderite. When movement took place along the
O'Neil Fault and possibly along with minor slips, the ore body
was grreatly crushed, during which time the galena acted as a
putty-like mass for the suspension of broken fragments of the
harder minerals. This breaking resulted in the formation of the
typical steel galena.

Great mining camps of Canada 1. The history and geology of

the Keno Hill Silver Camp, Yukon Territory.
Keno Hill mineralization "silver-lead-zinc veins in lastic
metasedimentary terranes". Galena is present mainly as the
common coarse-grained, well-crystallized and friable type, but
also displays several other textures, such as fine-grained,

6 271
 ANTENNA TUNING
Notes on simulations for a Tuggle Front End
http://www.lessmiths.com/~kjsmith/crystal/atu.shtml
By Kevin Smith

Introduction:

When designing my radios I always make extensive use of

Mike Peebles and Dan Petersen's "Professor Coil" spreadsheet.
This tool along with the many tutorials and explanations online
have vastly eased the design work around homebrew coil
winding. We can now model a coil with good accuracy prior to
the actual job and thus better match the coil to the other
components we plan to include. One quickly becomes aware
that most coils for the broadcast band vary around a nominal
220 or so uH. More than this and most variable capacitors will
have too much bottom-end capacitance to tune the top of the
band. Less than this and you need caps with fairly high values
Steel Galena from the Tintic Mining District, a classic silver
when fully meshed, 500+pF or more.
locality in Utah. (The excessively sparkally surface in the
photo is from the flash.) Note also the presence of Anglesite
All the above is pretty straight forward. On the other hand,
alteration, (the clearish anglesite crystals separated from the
when I have been reviewing double-tuned set designs, I often
galena by a dark finer grained almost greenish alteration zone).
note that the coil used for the ATU (antenna tuning unit) will
This is good radio material, but you may need hunt a bit more
have an inductance quite a bit lower than the main tank coil.
for hot spots than the finer-grained stuff above.
Professor Coil doesn’t address this aspect of coil design and I
have found little discussion online concerning ATU design.
Steel Galena Notes off the web:
Mostly this seems to be dealt with in passing or as a digression
when dealing with other subjects. The most useful web sites
USGS Bull 625, 1917
for addressing this aspect of crystal set design include Dick
Galena of the Coeur d'Alene District contains appreciable
Kleijer's excellent work and Ramon Vargas's detailed analysis
silver, is generally not so well crystallized, and has a much less
of the "Tuggle Front End", that’s about it. Neither site
perfect cleavage.
discusses the ramifications of varying different parameters
such as the Earth Resistance, Coil Inductance, etc although
Mineral Deposits, Waldemar Lindgren, 1919
Kleijier's site includes a great calculator page to allow one to

270 7
ask these questions. In my explorations of ATU design
(primarily Tuggle) I have made extensive use of Kleijier'e page
and for this I am deeply in his debt. My hat's off to Dick,
thanks so much.

The following discussion presents the results of many models

cranked through the design calculator. I wished to understand
what factors play main roles and which have minor parts.

For a conventional detector-circuit coil one can pretty much

control all the main factors, really only the inductance and
capacitance of the circuit. The main and only surprise comes in
the form of "stray capacitance" resulting from the spacing of
the coil wires. This can readily be 1) guessed at and 2)
minimized by good coil winding technique. For the antenna
circuit by contrast many of the needed parameters, earth
resistance, antenna capacitance and inductance, and antenna
resistance in the forms of actual wire resistance and radiation
From Leadville Colorado, the specimens below are the best
resistance are, for most of us, unknown and only guessed at.
examples of "steel galena" I have yet come across. The
The following figure illustrates the antenna and ATU with a
samples have a dull weathered surface skin deep hiding a
list of the main "components" that need to be understood
massive, fine-grained lead-silver mining ore of a classic Rocky
and/or modeled.
Mountain type. On a fresh surface the sample has a fine texture
and flat grey color. This is not museum material, and its
The Antenna Equivalent Circuit consists of an AC voltage
difficult (to impossible) to locate on ebay, but its the real
source Va (the
mccoy when it comes to crystal radio. Don't even bother with
signal of interest),
the pretty pretty Missoury galena, it rightly belongs behind a
in series with some
glass display, not in a radio. Chips from this sample, potted in
radiation resistance
woods metal, rectifies strongly almost regardless of where the
Rr, antenna
cat's whisker touches, instant sound.
capacitance Ca,
antenna inductance
La, antenna
resistance Rr, and
finally a ground

8 269
Steel galena is difficult to find for sale due primarily to its lack resistance Rg for the return path to complete the circuit. In
of excellent crystal cleavage. Galena, as is well known, has addition the ATU consists of a coupling capacitance C2, and a
beautiful LARGE crystal structure. Who would buy that ugly- tank with inductance L1 and capacitance C1. The capacitance
looking grey junk? and inductance of the ATU is needed to tune out the reactance
of the various components of the antenna for which we do not
Kevin have data. What to do?

Without an antenna analyzer (expensive) one can only estimate

the many component values based on published antenna
models. Ken Khun's engineering page has a number of nice
models and a good discussion of the antenna parameters we
will be trying to understand and use. For my personal setup my
antenna consists of about 75' of 14awg wire averaging about
10-12' high and with another 25' of lead-in. For such an
antenna Khun models a 30m antenna 3m high which
approximates my situation closely enough. Wire resistance is
negligible as is the radiation resistance which he shows to vary
between some 0.1 to 1.5 ohms. Such an antenna is capacitive
by nature and will have about 220 - 375pF along with some
small 20uH inductance, not too far off a standard "dummy"
antenna. This pretty much leaves earth resistance Rg as the
main unknown parameter.

For Rg one has the option of punting and taking the "Standard"
value of 25 ohms. In my modeling I have found this parameter
to be critical and highly sensitive. I do not recommend
guessing here, it is recommended one go to the internet page of
their state, county, or local government and search for reports
on ground or soil resistivity (or conductivity). As this is an
important agricultural and engineering parameter, it has been
surveyed for most places and should be available with some
effort. Effort well rewarded. Earth resistivity for the Texas
Gulf Coast is a mercifully low 10 - 15 ohm meters, but for
many regions this will not be the case.

268 9
 "Substitution of other atoms for lead in galena is not very
Earth resistivity depends on a number of variables including extensive. Among the elements which do accur in small
the material, moisture, mineral salt, and temperature. A table amounts are: Sb, As, Bi, Ag, Ti, Zn, Cd, Fe, Mn and Cu. In
of typical ranges in ohm meters for some different soils many cases these may be present in impurity minerals
follows: (acanthite Ag2S, sphalerite, chalcopyrite, etc.) rather than in
the galena structure."
Loam 5 - 50 ohm meter
Clay 4 - 100 Why is this interesting for the radio engineer? The following is
Sand/Gravel 50 - 1,000 my own theory, (speculation mostly). Most of you who have
Limestone 5 - 10,000 worked with galena detectors have noted that hot spots
Sandstone 20 - 2,000 generally occur on or adjacent to fractures, lines or
Granite 1,000 - 2,000 discontinuities on the crystal face, seldomly on the smooth
Slates 600 - 5,000 lustrous face itself. Rectification occurs at the junction of two
minerals, or of a mineral with a cat's whisker metal. Many
Moisture up to about 17% dramatically lowers resistivity, explanations have been offered, most having to do with the
mineral salts are needed and pure water is an insulator, and as position and nature of the crystal lattice at the point of contact,
the temperature approaches freezing the resistivity also rises and the perceived presence of "impurities" (atoms other than
dramatically. All this factors into your estimation of Rg. The lead) in the galena crystal matrix. As noted above, "Steel
actual resistivity seen by the circuit depends on the earth Galena" is a often an ore of silver containing not only galena,
resistivity and the type of grounding system you have installed. but abundant admixtures of associated minerals generally too
The more metal in the ground, and deeper, the lower the small to see and thus seldomly described except in detailed
resistance. Know thy earth! A map of USA soil resistivity mining reports as above. In addition, the very small grain size
follows: of common steel galena allows numerous physical and mineral
discontinuities on the small detector surface area, and thus
numerous hot spots. I speculate that many/most of these hot
spots may in fact be at the junctions of galena and other
minerals along with the light contact of the ever-so-sharp cat's
whisker itself. Other galena (lead) deposits such as the
Missouri / Tri-State region are not associated with silver
mineralization (Tri-State ores typically have only about 30ppm
Ag). These show large museum quality crystal faces but have
poor detector properties.

10 267
"The coarsest galena examined showed cleavage faces from 2
to 4 mm, square, but generally it is much finer grained than
this. The fine-grained galena is commonly known as steel
galena because freshly fractured surfaces have a steel-like
luster. In all the specimens examed, this steel galena appeared
to be in an intimate mixture of galena, tetrahedrite, sphalerite,
pyrite, gnague, and other minerals present in the ore."....
"It appears that the original ore body consisted chiefly of
galena, sphalerite, pyrite, pyrrhotite, tetrahedrite, freibergite,
quartz and siderite. When movement took place along the
O'Neil Fault and possibly along with minor slips, the ore body
was greatly crushed, during which time the galena acted as a
putty-like mass for the suspension of broken fragments of the
harder minerals. This breaking resulted in the formation of the
typical steel galena."....

In this ore, galena is NOT the silver-bearing mineral. Rather, Modeling:

the silver is in the freibergite and to a lesser extent in the
tetrahedrite. Very interesting. Quoting from classic mineralogy All the above discussion is great but... What is important
texts on galena I find the following: really? The following section presents the results of applying
different parameters to get a feel for the sensitivity of it all. To
Dana's Manual of Minerology: (Hurlbut) begin with I need explain my base assumptions. All the models
"Pb 86.6, S 13.4 per cent. Silver is usually present, probably as are based on Kleijer's calculation spreadsheet which requires
admixtures of silver minerals such as argentite or tetrahedrite." inputs for the following parameters:

Elements of Mineralogy: (Mason and Berry) Frequency

"Commonly, galena is very nearly pure PbS. The silver, Coil inductance
arsenic, and antimony reported in chemical analyses are LC circuit Q unloaded
largely due to the inclusions of argentite of tetrahedrite, small * Complex impedance of antenna or
amounts of which are difficult to detect in a black opaque Series Resistance (Rg+Ra+Rr)
mineral." Series Capacitance Ca
Series Inductance La
An Introduction to the Rock Forming Minerals: (Deer, Howie
and Zussman)

266 11
* If the series values are given, the complex impedance is Some thoughts on "Steel Galena" and its use as a crystal
calculated. detector/rectifier for radio waves.
http://www.lessmiths.com/~kjsmith/crystal/sgalena.shtml
1) Frequency is chosen per your interest, I have taken models
at F = 550-1100-1700 kHz. I have recently become
interested in crystal radios
2) Coil inductance one has control over when winding. I have and am especially
chosen base values that resonate with about 400pF tank interested in the "cystal" in
capacitance, with some sensitivities. the radio. This is due to my
background as a geologist /
3) Unloaded Q I do not have. For modeling I have taken mineralogist / petrologist /
Vargas's measures for a 4.5" coil wound with 660/46 Litz wire. oil exploration geologist. I
I assume this will be about as good a coil as one can wind. I have read in several radio
also made a sensitivity for lower Q. Q is a function of websites concerning steel galena describing it as argentiferous
frequency and I have used the appropriate value of Q to match as well as fine grained and useful as an excellent detector with
the modeled frequency. numerous hot spots. While commonly associated with silver
ores, I have never read of galena itself as being much silver
4) Series Resistance is basically what the earth and ground bearing as a mineral. This puzzle sent me to the web to see
system will deliver plus a small 1-2 ohm contribution from Ra what I could find on this "steel" variety.
and Rr. I have modeled four cases 10, 15, 25 and 50 ohms.
The first two cases reflect my needs on the Gulf Coast with Steel Galena does indeed associate most closely with silver ore
low resistivity soil. The 25 ohm case is a "typical" or deposits, especially in the classic Rocky Mountain silver
"standard" grround. Finally, many will have much higher districts like Leadville in Colorado, Coeur d'Alene in Idaho,
resistivity soils and you need to know just how difficult things and Tintic / Park City of Utah. In all cases the ore bodies
can get! themselves are in sheared and altered sedimentary rocks.
Mineralization is due to hydrothermal movements of water
5) Series capacitance I based for my specific antenna on the with movement aided by the sheared, and crushed zones
model of Ken Kuhn (Mathematical Model of Wire Antenna). having fine to extremely fine grain size, hence its name "steel".
The capacitance of a 30m antenna 3m high he calculated to
range from 220pF and low frequencies to 375pF and high To give a sense of the deposits themselves I wish to quote a
frequencies. I input the correct capacitance to match the couple accurate petrological descriptions:
frequency modeled.
Waldschmidt, 1921: Hecla East Vein, Coeur d'Alene
6) Series inductance I just input 20 uH every time.

12 265
 The data and plots for four different scenarios follow:

264 13
 Innards Photograph from an actual set..

The above plots show the calculated capacitance versus

frequency for four different models of Earth Resistivity / Tank
Inductance. In each case I maintain the same log capacitance

14 263
calculations from above: vs frequency scale for easy comparison. The cases modeled
LW tuning with assumed 450pF cap is 2300 - 2400 uH represent increasing earth resistance presented to the ATU
from a very low 10 ohms to a fairly high 50 ohms. Actual
The following photos show the set and the references earths can go up to two orders of magnitude higher. I chose L1
mentionned above. inductances such that the maximum needed capacitance would
approach 500pFs, easily found on many variable caps. With
(some useful translations for my anglophone friends: respect to the coupling cap (C2, red curve), as the earth
PO = Petits Ondes = Short waves [so named prior to the utility resistivity increases the needed capacitance declines. The tank
of the HF bands was known] cap value needs to resonate with the inductor and increases
GO = Grandes Ondes = Long Waves with smaller-value inductance. In searching for a 500pF max
Bobine Addition = Additional Coil) cap value, the choice of inductance needs to decrease with
increasing earth resistance. Parameters other than Rg, L1 and
Jacquemard Schematic and layout.. C1-C2 have minor impact on the models.

Conclusions:

From the plots one can readily see that above 10 ohm earth the
two variable capacitors on the ATU do not track well. As the
earth resistivity increases, the worse the tracking. Most
locations are not blessed with a low resistivity earth and this
needs to be factored into the ATU design. There is little to be
done, changing the design coil inductance will change the
needed capacitance on BOTH C1 and C2, (larger L1 leads to
smaller C1 and C2 for resonance and vice versa). A possible
solution well worth trying is to use a dual-gang capacitor
where one section has a different value than the other, the
above plots give an easy way to decide the max values needed.
Use the smaller gang on the coupling circuit and the larget on
Circuit Analysis/Redraws of above.. the tank. Experience tells us that ganging the capacitors on a
"Tuggle" front end works well, but from the models I have to
imagine one might squeeze a bit more performance by giving
up the convenience of one-dial tuning on the ATU, especially
where your earth has a fairly high resistance.

262 15
The best design concept will be to know first and critically Professeur Jacquemard Poste a Galene
your earth resistance. If this is just a guessed-at parameter then http://www.lessmiths.com/~kjsmith/crystal/jmrd.shtml
I strongly urge you to put down the Litz, set aside the silver-
plated caps, store the holy-grail diodes and go do some A bit of a mystery this one,
research on your local ground. All the time, expense, and effort I am hankering to build a
on the greatest "state-of-the-art" crystal receiver will be wasted breadboard repro of the set
if the ground is not attended to. If the resistivity of your earth if only to play with this
is high in ohm-meters, get more metal in the ground or bizarre (to me) circuit.
consider a counter-poise. The following table from LEM
Instruments shows the relation between earth resistivity in ohm data and information on
meters and actual earth resistance in ohms as a function of the the set is not small, but not
earthing method used. enough:

Models and Data: from the schematic:

L1: form 25mm dia, 68mm long selenoid, 170 turns 4/10 wire
10 ohm / 220 uH Lowest resistivity Earth Case sensitivities calculation- 0.4mm-26awg wire, 220uH coil
L2: ID 25mm, W 8mm. OD, #turns, guage not given
Ohm pF uH uH Qu kHz cpl tank L3: ID 20mm, OD 34mm. W, #turns, guage not given
Rg Ca La L1 f C2 C1
10 220 20 220 475 550 442 228 estimation from photo:
10 300 20 220 400 1100 68 36 L2: OD about 32-35mm, wire guage is small in the 32+ range
10 375 20 220 200 1700 44 -3 L3: W maybe 6-8mm, wire guage is small in the 32+ range

10 220 20 220 700 550 273 255 from book:

10 260 20 220 640 750 114 122 MW = 545 - 1500 Kc
10 300 20 220 585 1100 55 47 LW = 157 - 272 Kc
10 340 20 220 440 1400 41 20
10 375 20 220 300 1700 36 5 from calculations:
MW tuning with 220uH typically 365 - 450 pF
10 220 20 200 700 550 304 287
10 300 20 200 585 1100 58 54 from clues in the two publications
10 375 20 200 300 1700 38 7 1. most Vcaps are 450pF
2. most GO coils 120turns x two coils

16 261
 15 ohm / 180 uH Low resistivity Earth Case sensitivities

Ohm pF uH uH Qu kHz cpl tank

Rg Ca La L1 f C2 C1
15 220 20 180 475 550 337 328
15 300 20 180 400 1100 61 63
15 375 20 180 200 1700 40 10

15 220 20 180 700 550 220 352

15 260 20 180 640 750 100 176
15 300 20 180 585 1100 49 73
15 340 20 180 440 1400 37 37
15 375 20 180 300 1700 32 17

15 220 20 130 700 550 313 511

15 300 20 130 585 1100 59 110
15 375 20 130 300 1700 38 30

15 220 20 205 700 550 195 302

15 300 20 205 585 1100 45 61
15 375 20 205 300 1700 30 13

25 ohm / 165 uH "Standard" resistivity Earth Case

sensitivities
Ohm pF uH uH Qu kHz cpl tank
Rg Ca La L1 f C2 C1
25 220 20 165 475 550 213 396
25 300 20 165 400 1100 48 84
25 375 20 165 200 1700 32 22

25 220 20 150 700 550 164 462

25 300 20 150 585 1100 41 102
25 375 20 150 300 1700 27 31

260 17
25 220 20 165 700 550 151 416
25 260 20 165 640 750 76 213
25 300 20 165 585 1100 39 91
25 340 20 165 440 1400 30 50
25 375 20 165 300 1700 26 27
CHAPTER V
25 220 20 200 700 550 130 335
25 300 20 200 585 1100 35 72
25 375 20 200 300 1700 24 20
Other Thoughts and Speculations
50 ohm / 150 uH High resistivity Earth Case sensitivities

Ohm pF uH uH Qu kHz cpl tank

Rg Ca La L1 f C2 C1
50 220 20 150 475 550 128 476
50 300 20 150 400 1100 34 108
50 375 20 150 200 1700 24 35

50 220 20 150 700 550 96 490

50 260 20 150 640 750 53 256
50 300 20 150 585 1100 28 113
50 340 20 150 440 1400 22 65
50 375 20 150 300 1700 19 39

50 220 20 125 700 550 110 595

50 300 20 125 585 1100 31 139
50 375 20 125 300 1700 21 49

Kevin Smith
09/2011

18 259
 Coil Q:
http://www.lessmiths.com/~kjsmith/crystal/coilq.shtml
Kevin Smith

Introduction:

I have been building and studying crystal radios for some time
now and slowly begin to learn a few things about these
marvelous sets. In this section I begin my exploration of coil,
coil quality and that mysterious dimensionless factor, Q... This
page thus is a bit preliminary as I have not yet begun any
measurements to determine the Q of my coils. Please bear with
me.

My purpose here is to present some facts and data that has

resulted from my explorations of the web. I have often
wondered at what the quality factor of my coils should be and
as often realized that I really do not have any expectations as
to what is possible. Now, having dome some research, I can
with some confidance say that this is solved. Coils as used in
crystal radio broadcast band reception typically employ coils
with Q factors ranging from 100+ (pretty lousy) through the
several 100's (decent) and on up to 1000 or more for those
remarkable Big-Litz wonders ($$$). Knowing the expected (or
actual) Q of your coil is important in so far as it impacts the
choice of diode to be used in the set. On my page of Diode
Calibration I present a summary graphic which indicates how
the diode Rd value relates to the tank parallel resistance Rp.
This Rp in turn is a function of the coil Q. Schezzam! So here
we are.

In your set construction, with a good effort and good

engineering practice one can easily expect to wind a solenoid
coil in the Q = 200 range without much trouble, even with a

258 19
cardboard form, sealed, of course. With various "open" coils,
spiderweb, basket weave, diamond weave, etc. you may expect
to double that.. possibly. The advantage in open coils derives
from first the separation between adjacent wire turns which
reduces self-capacitance in the coil, and secondly from the
obvious lack of a form. All materials used in the coil will have
some amount of dielectric losses associated with them and the
less material used the better. Air core coils are best in this
respect. I am not here to speak about ferrite coils as I have no
experience with them. Frankly, they seem (to me) a bit like
cheating.

The following graph presents data that I have scoured off the
web. My primary sources include Hund and Groot, 1925*,
Wes Hayward, Dave Schmarder, Ken Khun, Mike Tuggle,
Steve Ratzlaff, and Dick Kleijer. 10,000 thanks for those who
post their data on the web! The following plot gives the Q
value as a function of coil Rs (series resistance. In striving for
a high Q coil, in effect one is eliminating losses and lowering
the series resistance as much as possible. There is more to it of
course. The Q formula Qu = 2pifL / Rs tells us that Q is also a
function of the coil inductance L and the frequency f of the
measurement. All measurements chosen for plotting are made Practical Wireless, September 1972 cover.
around 1 Mhz and most of the coils are in the L = 200 - 350
uH range. So, despite the formula, in this plot the coil Rp is the
main driver. From the plot it should be apparent that winding a
coil with Q = 200 or better should be a no-brainer. If your coil
Q is less, you just aren't trying. At the high end, Q's > 1000
seem to be pretty extreme and these coils are expensive. You
better be using big Litz 660/46 and use a basket design
(although the two best coils on the plot were solenoids).

20 257
 On the plot I also post in orange circles my estimate of what
the loaded Q presented to the diode (Antenna + Tank) might
look like. Loaded Q will always be lower than the unloaded Q
by several times. I have estimated that high-end sets (big litz,
silver-plated ceramic insulated caps, best wiring practices)
may lower the Q by about 2 1/2 times (B Tongue's
performance set has Q = 700). At the low end (vintage
components, small solid wire coils, taps) the load may lower Q
by up to 5 times. Ken Khun in his excellent web book states
Popular Electronics, March 1995 that typical sets at 1Mhz have a loaded Q between 20 and 100,
50 typical and this is where most of the data falls. I scaled the
divisor by Qu to produce the above plot but note that this is
merely an estimate. Somewhere some bloke has no doubt

256 21
tuned a big litz coil with cheapo capacitors. Having an idea of
your set's Ql will allow a better selection of the proper diode
for matching.

Circuit Diagram of same.

The above plot is from a series of measurements on two of my
coils kindly made for me by Steve Ratzlaff, AA7U. The blue
diamonds are the measurements on a 660/46 Litz basket beast
and the yellow diamonds are measurements on a modest
18awg tapped solenoid wound of a cardboard form. The data
from the low Q coil form a lovely continuum across the BCB
band while the litz coil data above 1MHz or so seem to be
declining and erratic. Steve measured the coil a great many
times and found good consistant results. The measurements are
excellent but I must say, I am a bit suspicious that the high-Q
coil may be at the limit of the HP Q-meter calibration, or
perhaps 660/46 litz Q tops out at 0.7-0.8 MHz and crashes
above that. (I have not yet found time to persue this). The plot

22 255
plans? Or, was this a kit? As I cannot locate an actual copy of also includes Q vs f curves at constant Rp values. It is clear
the Sept 72 journal, I still can only speculate. I do hope it is a that, while tuning across the BCB spectrum, the Rp of the coil
true homebuilt. changes continuously.

The following photos show the set and the references

mentionned above.

Calculating the coil Rp from the data (Q:C:f), I have

reproduced the above graph of the coil Rp vs frequency. It is
evident that the trend follows a perfectly straight line for the
Portrait of the beauty.. low-Q coil. I would expect a similar straight-line relation for
the litz coil but again things look strange. My impression here
is that of a much steeper Rp vs f relation where the calibration
on the Q-meter breaks down near 1MHz (Rp ~1.5 MOhm) and
the data wanders until a second linear trend (with about the
same slope) is re-established between 1.4 to 1.7MHz.

Now time for a few calculations, just how good can a coil Q
get?

254 23
The following calculation is made to determine the AC BBC 1922 - 1972 Commerative
resistance of the Litz wire used in a coil: http://www.lessmiths.com/~kjsmith/crystal/bbc.shtml
From different Litz wire manufacturing sites one can find
tables and data allowing the calculation of the resistance of A bit of a mystery this
your favorite litz wire. one, I found this radio on
ebay and it was just SO
The formula for the D.C. resistance of any Litz construction is: pretty that I had to have it.
The circuit is a simple
Rdc = Rs (1.0515)^Nb * (1.025)^Nc / Ns variometer of pancake-coil
design. Being a variometer
Where: has its drawbacks such as
Rdc = Resistance in Ohms/1000 ft. limited sensitivity and
Rs = Maximum D.C. resistance of the individual strands (4544 gawd-awful selectivity. If this piece wasn't so darned cute I
for 46awg wire) would let it adorn the shelf indefinately. The set does have one
Nb = Number of Bunching operations (assume = 2) mysterious element for me. Who made it? When? Some
Nc = Number of Cabling operations (assume = 1) research on the web resulted in two interesting discoveries and
Ns = Number of individual strands (assume = 660) an added mystery. On Darryl boyd's excellent web site I found
in the plans section a mention of this very set, or one very near
Rdc = 4544 (1.015)^2 (1.025)^1 / 660 = 7.27 ohms / 1000 ft. to it taken from the March 1995 issue fo Popular Electronics.
The reference was to the September 1972 British journal
The ratio of AC resistance to DC resistance of any Litz Popular Wireless, the text I have shamelessly copied and
construction is: placed below, the critical text highlighted in red. I
subsequently found an image of the cover of the Prac Wireless
Rac/Rdc = S + K (N Di / Do)^2 * G issue with the set nicely featured. So, I have to assume this set
is homemade at or soon after the publication of the plans in
Where: that issue. I assume the set was made in USA, but cannot be
S = Resistance ratio of individual strands when isolated sure at all.
(1.0003 for 46awg wire)
G = Eddy Current basis factor = (Di * sqrt(f) / 10.44)^4 The new mystery? well, if you look closely at the radio I have,
F = Operating Frequency in HZ (assume 1MHz) the drawing in the Pop Electronics article and the photo of the
N = Number of strands in the cable = 660 radio on the Prac Wireless cover, you will see virtually the
Di = Diameter of the individual strands over the copper in EXACT SAME set right down to the hardware and placement.
inches = 0.0016 Only a few minor differences. My mystery then is this, was
this an extremely faithful homebuilt version from detailed

24 253
 Do = Diameter of the finished cable over the strands in inches
= 0.056
K = Constant depending on the number of strands = 2

Rac/Rdc = 1.0003 +2(660*0.0016 / 0.056)^2 * (0.0016*1000 /

10.44)^4 = 1.393

Therefore the AC Resistance of 660/46 litz wire is

approximately:

The A.C. resistance is: 1.39 * 7.27 = 10.13 ohms/1000ft.

A 5inch diameter basket weave coil made from 660/46 Litz

wire having an inductance of 230 uH will typically require
some 40 turns or about 53ft of wire. At 10.13 Ohm/kft, that
Circuit schematic for the set. comes to an AC wire resistance of 0.53 ohms for the coil
alone. Were all the losses represented by series resistance of
the wire, the coil would have, at 1 MHz an unloaded Q = 2700!
Naturally, wire resistance is not the only source of loss in the
tank. There is a capacitor, metallic objects intruding into the
magnetic field of the coil, diaelectric losses, eddy current and
other losses. Its no wonder that the best coils just top out above
Q = 1000 or so.

This is where I leave things. Time to get measuring. Below I

provide my input data, have at it!

Kevin Smith

*Note that in 1925 the importance of coil Rs was understood,

but the factor Q was apparently not used. I have taken (quite
painfully) the L and Rs data from their plots and calculated the
resulting Q. I recommend to those interested to download the
Portrait, set and box. pdf of their paper.

252 25
 Inside I expected to find a typical tapped selenoid, the box is
Radio-Frequency Resistance and Inductance certainly large enough for one. What I found suprised and
Qu = 2pifL / Rs rather shocked me. It still does! The coil is bundle-wound, a
1Mhz mass of 26-30 dcc wire about 2.25 inches diameter. I was
Coil Type Wire Rs L Q unaware that such a design would work. It is impossible to
awg ohm uH check how the tap wires are attached, they seem to just get
solenoid 660/46 0.90 200 1400 wound into the coil with no apparent connection. It appears to
solenoid 660/46 0.91 200 1375 be continuous wires from one switch set wound around and
solenoid 175/46 2.88 265 579 around and then exiting to the second switch set. That sounds
solenoid 12 2.98 230 485 like eight independent coils wound together. Another odd
solenoid 14 3.28 230 440 circuit characteristic is the fact that the tap points between the
solenoid 16 3.52 230 410 two tap switches are wired to each other (see photo and
solenoid 18 4.01 230 360 schematic). Perhaps connecting in such a manner combines
solenoid 50/46 4.89 275 353 them into a single longer coil, hard to speculate. How this coil
solenoid 32/38 6.20 327 331 works is different from anything I have experienced in crystal
solenoid 20 4.66 230 310 radio. Completing the schematic, rectification takes place via
solenoid 16 7.60 319 264 an ordinary cat's whisker and small potted galena crystal,
solenoid 28 8.80 360 257 whew! This is certainly an interesting design if not otherwise
solenoid 22 5.78 230 250 efficient or selective. It is a welcome addition to my small
solenoid 24 8.10 319 247 menagerie of vintage sets.
solenoid 28 10.10 360 224
solenoid 28 10.30 360 220
solenoid 50/46 10.84 378 219
solenoid 24 6.72 230 215
solenoid 28 9.60 327 214
solenoid 28 9.70 319 207
solenoid 26 8.03 230 180
solenoid 28 327

Basket weave 660/46 1.03 186 1134

Spider 660/46 1.38 241 1000++
Basket weave 660/46 1.08 186 1082
Spider 660/46 0.90 150 1000+
spider 660/46 816

26 251
 Spider 165/46 1.94 232 750
Basket weave 175/46 2.56 265 650
spider 660/46 641
Spider 100/45 2.41 238 620
Spider 100/45 2.44 241 620
Spider 100/45 1.73 149 540
spider 175/46 4.94 373 474
Loose Basket 32/38 5.30 317 376
Spider 40/44 4.16 248 375
Spider 40/44 4.16 225 340
Spider 40/44 2.93 154 330
Spider 32/38 7.50 331 277
Loose Basket 24.00 7.60 317 262
Basket weave 22.00 6.82 265 244
Loose Basket 28.00 8.50 317 234
Spider 24.00 9.50 327 216
Basket weave 32/38 10.40 332 201
Spider 28.00 10.60 327 194
Honeycomb 32/38 12.00 355 186
Spider 28.00 12.00 330 173
Basket weave 24.00 13.80 323 147
Basket weave 28.00 16.40 323 124
Honeycomb 24.00 18.50 347 118
Honeycomb 28.00 27.50 347 79

ferrite rod 50/46 2.58 250 610

toroid 22 3.93 240 384
toroid 18 3.20 195 383
toroid 22 8.20 449 344
toroid 175/46 4.82 244 318
toroid 50/46 5.23 248 298
ferrite rod 22 80
ferrite rod 22 16.94 158 58.6

250 27
2Layer Bank 32/38 9.40 327 219 American Radio Stores, Inc.
3Layer Bank 32/38 13.00 336 162 http://www.lessmiths.com/~kjsmith/crystal/ars.shtml
2Layer Bank 24 14.20 323 143
2Layer Bank 28 16.80 323 121 An interesting vintage
4Layer Bank 32/38 19.50 360 116 crystal set recently came
3Layer Bank 24 21.00 333 100 up for bidding on Ebay, an
3Layer Bank 28 24.00 333 87 American Radio Stores,
4Layer Bank 24 29.00 336 73 Inc radio in good working
4Layer Bank 28 32.50 336 65 condition. This set
Double Layer 28 inf 355 0 interested me because,
from the photos provided,
it looked to have a fairly
classic 1920's circuit with tuning accomplished via two
switches to taps on the coil. I presumed that they are units and
10's taps as outlined on the classic Bureau of standards Circuar
#121 set. A modest bid brought the radio to my doorstep a
couple weeks later.

Looking at the radio when it arrived showed me that this was,

even for the time, a fairly inexpensive set lightly built with
cheap but workable parts. As expected, the radio has fair
sensitivity and abysmal selectivity. (Recall, I bought it for the
classic design not its DX potential). Somewhat to my
consternation, the front panel was attached to the box by four
small brads that could not be removed. No chance to view the
inductor in the interior without compromising the vintage
nature of the radio. I agonized over the question of whether to
tear into the box or preserve the radio. After deliberating some
30 - 40 nanoseconds I was scheming for the most efficient and
least destructive procedure to access the interior. My handy
Dremmel neatly sliced the small brads one by one leaving no
marks on the box or panel.

28 249
 An Analytical Approach to the measurement of coil Q
http://www.lessmiths.com/~kjsmith/crystal/coilqm.shtml
Kevin Smith

Background:

Measuring the values of your radio components is an

interesting and fun part of the Crystal Radio Hobby. Not only
does it help you design better radios, but helps to understand
the physics behind the wires and plates. The radio coil, one of
the principal components is generally that part wound by hand.
All other components, diode, variable capacitor are usually
bought ready-made. The coil sits at the heart of the set and yet
it remains devilishly difficult to measure and characterize.
Even cheapo inductance meters, (I mean, my cheapo meter) do
not properly measure the coil inductance at radio frequency.
Coil quality, well that is a whole other matter. Quality, or the
Q-factor as it is often called is a bench-intensive and
measurement-intensive proposition that even the most
dedicated radio fan may shy away from.

If you wish to know your coil Q-factor you have the choice of
tracking down and paying a lot of money for an old HP or
Boonton Q-Meter, or setting up a test bench and making the
required measurements yourself. A number of techniques for
measuring the Q of a coil have been published and excellent
summaries can be found in "Q Factor Measurements on L-C
Circuits" by Jacques Audet, and "Experiments with Coils and
Q-Measurement" a web page by Wes Hayward.

All the techniques discussed involve variously setting up a

parallel or series tank containing the coil to be tested and an
accompanying capacitor to tune it to the needed measurement

248 29
frequency. The tank is generally energized with a signal
generator and measured with an oscilloscope or sensitive
digital voltmeter. Other components often may include
attenuators, SWR analyzers, Spectrum analyzers etc. Sources
of error enter with coupling, loading, uncertainty as to the
actual internal resistance of the source generator, and the need
for several independent measurements which are then
multiplied or divided together, adding additional error. All this
works very well for the engineer with good bench practice,
good equipment ($$), and the patience to perform the
measurements several times to check repeatability. All these
things plus the required anality coefficient the current author
lacks.

In this paper I propose an alternative approach that takes an

analytical look at the oscillating tank waveform and
determines the coil Q from that. This technique dispenses with
A peek, such as it is, at the set condensers and cam tuning
a majority of the equipment involved with traditional methods
assembly.
but does require a digital oscilloscope. In recent years the
affordability of these scopes has greatly increased. If you have
a digital scope, or access to one, or were looking for an excuse
to purchase one, then read on, this technique may be for you.

Some Theory:

The inspiration for this technique is certainly not new. I first

found the following simple circuit, (Figure 1) on page 22 of
Bucher’s 1919 "Wireless Experimenter’s Manual". Having a
good deal of experience capturing and evaluating damped
radio oscillations for a Spark Gap transmitter, I immediately
saw the utility of the circuit for a simple Q determination based
on the damping decrement of the tank.

30 247
 Figure 1. Simple oscillation circuit, Bucher 1919.

As a professional Geologist, I must admit that most math

involving more than ten fingers and ten toes (for advanced
computations) is not my specialty. As such I will keep this
aspect to an absolute minimum. Still, a bit of algebra may be
useful to explain damping and logarithmic decrement. RF
oscillations in a tank circuit are damped due to losses
The spiderweb coil with vintage cloth-covered wires, taps and (primarily resistive) associated with that circuit. The amount of
wireing. damping thus is related to the quality, or Q of the tank. In old
texts the damping is generally determined by measuring the
"Logarithmic Decrement" or the amplitude of successive
oscillation peaks and taking the log of the ratio.

d = ln(A1/A2) (1)

A more generalized version of this formula taking into account

measurements over many periods can be expressed as:

d = 1/n ln(Ao/An) (2)

Where n is the number of periods analyzed, Ao is the

amplitude of the first peak and An is the amplitude of the peak
n periods away. This equation allows a very simple
determination of the log decrement with high accuracy.
Because damping is due to resistive losses in the circuit, it can

246 31
be related to the circuit components as follows, (Bucher,
1919):

d = pi (R / 2pifL) (3)

The angular frequency is often expressed as w= 2pif so this

can be substituted into equation (3) above to simplify as

d = pi R / w L (4)

We turn now to looking at Q and its relation to the circuit

components. The general expression for Q is:

Q = 2pifL / R (5)

Simplifying to

Q=wL/R (6)

With the term w L in both expressions, we can solve for w L in

each and set them equal to each other: Circuit as best I can make out

QR = piR / d so:

Q = pi / d (7)

Dreadful, wasn’t it? The inspiration here is that with a simple

determination of the logarithmic decrement, deriving the
circuit Q is a trivial exercise even a geologist can manage. No
signal generators, no attenuators, no multiple measurements.
All this should serve to reduce sources of error and give a
faster and easier way to determine the quality of that most
central component of your crystal radio, the coil.

32 245
 Setup:

The following schematic shows the setup, (Figure 2). In this

configuration, I have used a simple 9V battery to stimulate the
coil and have the scope connected to the tank via a coupling
coil. An earlier configuration with a power supply resulted in
too much noise and ripple to obtain a good reading. The tank is
coupled lightly to the scope with a coupling coil. This coupling
coil is simply two turns of hookup wire about 4 inches in
diameter placed an inch or two from the coil to be tested and I
have found no reason to space it further. Between the coupling
coil and oscilloscope I utilize the 1:100 probe although testing
with my 1:10 probe showed little impact. In this way, the coil
under test is clean and well isolated from the measurement
circuit.

Portrait of the Steinite set

Figure 2. Schematic of test circuit.

For the tank I have used a small CT1C150 military spec cap
("APC"-type) with a capacitance range from about 20 to 160

244 33
pF. These are silver or nickel plated (depending on the source A Vintage Steinite
of information) and ceramic (steatite) insulated and should http://www.lessmiths.com/~kjsmith/crystal/steinite.shtml
have a high Q with little impact on the test. At least this is the
general assumption used in most Q-measurement articles I Steinite standard from the
have run across. In actual fact the capacitor Q is not infinite mid-1920's. This set has
and not even that high so resistive losses associated with the lost its decal but not its
capacitor can be expected to lower the measured Q. One needs classic look. It has a good
to keep this in mind and know that this measurement is on the condition enclosed crystal
resonator, not on the coil itself. More on this topic in a later detector which works very
section. The APC capacitance range is limited but sufficient well, binding and phone
for my intended test frequencies of 1.0 to 1.1 MHz, this is a posts and labels in the
test, not a radio. typical Steinite lettering.
Six taps on the interior spider-web coil and a mica and brass
Keying, or pulsing the circuit is somewhat tricky. In order to plate "book-style" condenser complete the tuning of this unit.
obtain a clean wave trace the pulse contact needs to be fast and Inside all the components are original with original soldering.
sharp. Contact "bounce" and arcing, especially at radio It is unfortunate that the condenser is not working as expected.
frequency are difficult to avoid and pulsing the circuit several Peering in between the coil tap wires I can see what appears to
times may be needed prior to finding a clean trace. I initially be 3 or 5 brass plates separated by mica dielectrics. The tuning
used a transmitter key but found the contactor at RF to be dial on top turns a shaft with a small cam that alternately
messy with visible arcing. The best method I have found for compresses the plates together and lets them spring apart. I can
pulsing the circuit is to connect a pointed probe to one side of see only the outside plate returning to its open position when
the circuit and tap it sharply but lightly against the hot side of the cam is open, the others remain compressed so the variation
the battery. After a few tries one gets a "feel" for the contact. in capacitance is minimal. Well, I can enjoy one station. In fact
Pulse the circuit and get a wave, things are this simple. the tuning is very broad so when "running" the radio has two
or more stations competing for my attention. Well, its lovely
Procedure: just the same.

Hang the coil under test on a stand where it is clear of other The following photos show the set and its interior.
components or metal objects. Connect the coil as per the
schematic. Connect the probe and scope to the coupling coil
leads and set the scope for a good scale, (Figure 3). I generally
use 5uS / division horizontal scale which gives me 40 to 50
oscillations at about 1 MHz. For higher Q coils I may go to

34 243
 10uS / division for a longer recording. Set the scope trigger
appropriately to capture the data.

Figure 3. Photo of the test setup.

Tap, or "key" the battery to pulse the tank and review the
waveform on the screen. Setting the scope on its measurement
mode allows reading the wave frequency with each pulse.
Adjust the capacitance after each pulse to bring the tank to the
desired frequency, near 1 MHz in my case. Once at the desired
frequency you are collecting data. When you get a good clean
oscillation free of contact hash and noise, send the data to the
computer for analysis. For my scope set at 5uS/div it returns

242 35
5000 data points to the computer. At 1 MHz that is about 100
data points per cycle, more than enough sampling to avoid any
question of aliasing.

The decrement measurement is based on measuring two peaks.

The data is in digital format so one should eliminate
"estimation error" by actually reading the value off the CHAPTER IV
spreadsheet, noting both time (in uS) and amplitude. Before
taking readings, care must be taken to center the oscillations
around zero. Amplitude measurements assume that the
waveform is symmetrical around zero yet the scope may or,
Radios I have acquired
more likely, may not be set perfectly for this. In the (read: paid waay too much for,
spreadsheet the solution is simply to take the average of all the
readings and subtract that from each. with no regrets)
Type the values for “to” (initial peak time), “tn” (final peak
time), Ao (initial peak amplitude) and An (final peak
amplifude), and the number of periods analyzed “n”.

T = tn - to in uS (8)
f = 1/T in MHz (9)

d = 1/n ln(Ao/An) (10)

Q = pi / d (7)

Next, disconnect the battery and coil and measure carefully the
capacitance, C, that was used in the test. With this you can
calculate the actual coil inductance from the frequency and
capacitance. Why buy an expensive L meter?

L (uH) = 1E6 * ((1/ 2pif)^2 ) / C (pF) (12)

Example One:

36 241
 An example to show off a bit. This test was made with a close-
wound tapped solenoid on a 3" cardboard form. The graphic in
Figure 4 shows some initial contact hash followed by a lovely
damped oscillation. Measurements on this coil showed a test
frequency of 1.055 MHz, L of 215 uH and C = 106 pF. Q was
calculated at 167. A decent coil but not great.

Figure 4. Oscillogram of a coil under test.

n = 45
T = (tn - to) / n = (47.24 - 4.59)/45 = 0.948 uS
f = 1/T = 1/0.948 = 1.055 MHz

Ao = = 11.2 mV
An = = 4.80

d = 1/n ln(Ao/An) = 1/45exp(11.2/4.80) = 0.01883

Q = pi / d = 3.142 / 0.01883 = 167

240 37
To test the validity of this technique we need a bit more theory,
just to be certain this is working. With the data in spreadsheet
form, it becomes possible to model the coil from a theoretical
standpoint. The model can then be plotted with and compared
to the acquired data to test for correlation. Various web pages,
especially university-sponsored lab exercises provide a good
source for practical techniques and their theoretical
underpinnings. Some good sources I utilized include:
"Damped oscillations in RLC circuits" by Barbara Dziurdzia at
AGH University of Science and Technology in Cracow, "RLC
Circuits" a lab note from Rice University, and "Oscillations
and Resonances in LRC Circuits" from Durham University in
the UK. And very many others.

From the above we find the general equation for damped

electrical oscillations as:

qc = qo * e^(-a*t) * cos(2pif * t + phi) (13)

Where
qc is the charge in volts
qo is the initial charge
a is the damping factor (=Rs/2L (series) and = 1/2RpC
(parallel)
R is the coil resistance in ohms
L is the coil inductance in uH
t is the time in uS
f is the frequency in MHz
and phi is the phase angle

The phase angle and qo charge are initial conditions and

adjustments should be minor to fit the calculated oscillation to
actual. L, the inductance comes from the calculation we made
with equation (12), and frequency comes from equation (9).

38 239
It even looks nice from behind, vintage binding posts,
naturally.. This leaves the resistance of the coil (R) as the primary
variable. Model the above equation in the spreadsheet by
adjusting R until the model results match the measured
oscillation from the coil. From the above example we see
Figure 5 with the modeled damped wave overlaid on the test
oscillation.

Schematic of the kit, (AnyKits.com) A005. The kit is

configured with a maximum gain between 45 and 50, too high
for headphones but too low really for listening with the
speaker easily. You will note the location of R1 (in series with
a 10uF capacitor) between pins 1 and 8 on the LM386. If you Figure 5. Oscillogram with modeled damped wave response.
short the resistor leaving only the capacitor, then the gain will
increase to the maximum of 200. If you do such a project, I Here the background data from the test is in grey and
recommend you download the LM386 datasheet from National overlaying it is the dashed model in black. The fit is nearly
Instruments. perfect using a resistance of 0.22 Mohms. Relating resistivity
to Q we have the following equation (5):

Q = Rp / 2pifL

Substituting f = 1.055 MHz, C = 106 pF, and R = 0.237

Mohms the result is Q = 167, exactly the same as with the

238 39
decrement method. An exact match will not always be the
case, but they ought to be fairly close. The match between
theory and the decrement method gives confidence that this
analytical technique is as robust as it is simple.

qc = qo * e^(-t/2RpC) * cos(2pif * t + phi)

qc = 12.2 exp(-t/2*Rp*106)*cos(6.283*1.055*t + 1)

Where t = 0 to 50 uS
R = 0.237 Mohm in this case for best match.

Q = Rp / 2pifL = 237000 / 6.283 * 1.055 * 215 = 167

Example Two:

To be sure the technique works I have also assessed the quality

of a 5inch diameter basket weave coil wound from 660/46 litz
I was initially worried that the audio input impedance may
wire. Such coils are rightly considered "performance coils" for
cause me trouble. I read so much about this, impedance
their low skin-effect and low resistivity.
matching, transformers, etc etc, almost a fetish really. I have
subsequently learned that the LM386 input impedance is about
800 Megohms, the 10k pot being the limiting factor. So, with
considerable relief, I concluded this ought not present any
problems with the crystal radio circuits. I find the amplifier
works very nicely with my radios. The sound is as good as the
Radio Shack unit.

The amplifier certainly adds to the vintage "look" of my setup.

Well, you be the judge, I hope I come close, please check out
the photos...

From the front, wooden base, full display of components, big

adequate volume control..

40 237
crystal radios one builds with passion, creating something
more than a mere radio, but a work of art, a piece of furniture.
Still, the technology itself is primitive dating back to the dawn
of radio. When I look at the mini amplifier, it looks like a toy
yet inside there is the most wonderful technology, solid state
integrated circuitry. We take this advanced technology and
slap it inside a cheap plastic box, an appliance. Naturally, I
don't complain about the price, but for my project I wish for
something above.

The solution struck me as obvious, give the same care and

respect for the amplifier as I do the radio. I also sought to
maintain a "vintage" visual aspect for the unit, breadboard
layout, antique binding posts and a large adequate volume
control knob, (the pitiful volume control wheel on my RS mini
has already come apart once and threatens to break at any
time). I wanted a nice lacquered base and sturdy speaker Figure 6. Second oscillogram with modeled damped wave
enclosure. The enclosure was the main challenge. I finally response.
ended up with a heavy plactic goblet from a local thrift shop
cut down to a tapered cylinder into which the speaker can be The coil tested values as follows:
placed tightly. The rest of the set is modular with plug-in
connections for the audio in, power, speaker, and volume n = 45
control potentiometer. A question, do I add a brass knife T = (tn - to) / n = (47.82 - 4.70)/45 = 0.958 uS
switch for the on-off? I worry I am out of room on the base. f = 1/T = 1/0.958 = 1.044 MHz
Still pondering this.
Ao = = 18.05 mV
An = = 13.65

d = 1/n ln(Ao/An) = 1/45exp(18.05/13.65) = 0.00621

Q = pi / d = 3.142 / 0.00621 = 506

And:

qc = qo * e^(-t/2RpC) * cos(2pif * t + phi)

C = 123 pF

236 41
qc = 18.55 exp(-t / 2*Rp*123)*cos(6.283*1.044*t + 1) A breadboard Audio Amplifier
http://www.lessmiths.com/~kjsmith/crystal/amp.shtml
Where t = 0 to 50 uS
R = 0.627 Mohm in this case for best match. A little project stemming
from my persistant
Q = Rp / 2pifL = 627000 / 6.283 * 1.044 * 189 = 506 difficulty in hearing.
(Losing one's hearing is
The results above for a big litz coil clearly demonstrate its never fun to admit, but
superiority to a more common tapped solenoid. Published data combined with crystal
though has let me to expect a big litz coil Q to be in the 1000+ radio, well, its trouble!) I
range. I have run this test backwards and forwards, checking hope I do not offend any
all possible causes for this unexpected low Q with no "purists" with this project.
difference in the final value. In fact I am quite impressed with Still, I understand that many crystal radio owners keep a small
the repeatability of the technique and the correspondence amplifier too, for "those times" when, well, you know.. So, this
between the results from the decrement and those from amplifier project, its for, well, those times.. (I suspect Radio
modeling the wave form. Shack would have discontinued their nifty Mini Amplifier
years ago had it not been for the crystal radio community.)
To be certain I even tested the coil using the old -3dB method.
As stated previously, I am uncomfortable using a signal My idea was to build my own unit rather than use a store-
generator in addition to the scope, and I have already found bought amplifier. Additionally, I wished to put my own stamp
trouble connecting the 1:100 scope probe directly to the tank, on it, make it unique. I started with a kit for the basic layout
as is necessary in this method. With that said, the results yield and electronics. I wanted to base things on the NI LM386 chip
a bandwidth of 2.3kHz about a nominal frequency of which is a good part and hard to break. I found several
1018.8kHz for a tank Q = 450. I suspect the method (with the distributers of what appears to me as the exact same kit and so
problems noted) lowballs somewhat the final result. In any chose the cheapest $6 dollar kit. NI sells the LM386 for 34
case, this confirms the test to have a moderate Q and nothing cents a pop and would even provide a few "samples" but then I
like the expected values one reads online. would be stuck laying things out and wireing.. Been there,
done that. A printed circuit board is just SO much easier to
Notes on Capacitor Q and its impact on the measurement use.
value.
One of my inspirations for this project also stems from the
This is where we begin to think more seriously about the nature of modern electronics in general, and the Radio Shack
capacitor used to resonate the tank. The assumption of very mini amplifier in partictular. Using my RS mini, I am always
high Q needs to be questioned. I stated in the setup discussion struck by the contrast between electronics new and old. With

42 235
 my opinion that the capacitor Q ought to be high enough as to
not influence the coil Q measurement in any significant
manner. I said that for three reasons: 1) Just about every
published article starts with such an assumption and 2) I had
not done much research on the subject of capacitor Q and so
had no particular expectations other than assumption 1 and
finally 3) I had no measurements or specific knowledge of the
Q for the capacitor used in my setup.

Reason 1 is a punt and should be rejected, reason 2 has been

remedied and I give some analysis on the interesting subject of
capacitor ESR (Equivelant Series Resistance) on a separate
page so that readers may set their own expectations. For the
third reason I have received the kind assistance of Steve
Ratzlaff, AA7U who offered to make actual Q measurements
on the coils and APC capacitors with his HP4342A Q-meter,
results follow:
Wave Trap
Litz basket coil,
550kc 1127Q;
1000kc 1140Q,
1700kc 674Q (194.2 uH on aade.com LC meter)

Solenoid coil,
550kc 205Q,
1000kc 240Q,
1700kc 222Q (218.2 uH on aade.com LC meter)

Using Steve's Litz wound ferrite rod coil (1000kc 1047Q) as

the reference for external caps:

100 pF silver mica (99.2 pF meter):859Q at 1068kc (Q meter

alone and coil 1025Q)

234 43
APC cap with red/black leads (149 pF meter): 937Q at 904kc Wavetrap, a useful gadget
(Q meter alone and coil 1060Q) http://www.lessmiths.com/~kjsmith/crystal/wt.shtml
APC cap with my Litz leads (149 pF meter): 995Q at 904kc (Q
meter alone and coil 1060Q) This little gadget is most
handy for crunching those
Steve's measurements above provide important datapoints with local behemoths that
which to calibrate my technique. Note first of all that the Litz always seem to be
coil is not a modest 506Q but a robust 1140Q at 1MHz. The directing their strongest
reason for my low measurement was my assumption of a high- signals right in your
Q capacitor. Steve's measurements of two of my APC's are 995 direction. My wave trap
and 937, hardly very high. There is an obvious need for a is a bit of a combination
correction. Wes Hayward in his excellent notes on Q of 1) Osterhoudt's QRM
measurements provides the key in the form of the following coil, 2) a design I found online on Owen Pool's excellent web
formula: site, and 3) a bit of my own seeing as how I had some, but not
all of the ingredients needed from both traps. That is, I used
Ql = (Qc * Qres) / (Qc - Qres) (14) what I found in the junk box.

where Ql is the desired coil Q, Qres is the measured Q of the The trap is formed on a stiff cardboard paper core with 1 1/4
tank/resonator, and Qc is the tank capacitor Q. inch outside diameter. On this I drilled holes for winding the
tank coil, 110 turns of #30 awg enameled wire. Both Poole and
Substituting the values Qres = 532 and Qc = 995 the equation Osterhoudt recommend #32, but 30 is what I had. Over this I
simplifies thus: would a coupling coil, 15 turns space wound of #22 awg
Ql = (937 * 506) / (937 - 506) enameled wire. Tuning of the tank is completed with a 410 pF
Ql = 1100. var cap.

The result is distressingly close to Steve's measured value of Tuning is sharp and deep, it takes a bit of getting used to, but
1140Q. This shows the possible level of accuracy in the then, schezzam!!!.. problem station gone! I luv it.
analytical technique used, both reassuringly high and very
"Litz-like".

Taking the first solenoid and substituting the values Qres =

167 and Qc = 937 the equation simplifies thus:
Ql = (937 * 167) / (937 - 167)
Ql = 203.

44 233
 This is slightly low to Steve's measured 240Q. I am of the
opinion that the original tank measurement may have been a
bit on the low side. Still, again the results are within the
ballpark and the technique works if you have a good measure
on the Q of the capacitor used in your tank. Alternatively,
starting with a known coil Q, the technique will work equally
well to determine the Q of an unknown capacitor.

Conclusion:

If you have access to, or own a digital oscilloscope, or have

been hankering to purchase one, then the time may be now.
The analytical method presented above is simple in practice
and robust in theory. I would also say that, by working with
your wave forms, you learn to recognize your data and soon
get a good "feel" for what is good and what is not. The
technique also produces, should you wish, an auditable report
of the wave and analysis. It is hard to argue against the results
with the waveform sitting there staring right at you!

Measuring your coil Q should not be difficult.

A copy of the evaluation spreadsheet in xls format can be

found here CoilQbyDecrement_kjs.xls

Bibliography:

Jacques Audet, VE2AZX., Q Factor Measurements on L-C

Circuits., QEX - January/February 2012, pp 7 - 11.
http://www.arrl.org/files/file/QEX_Next_Issue/Jan-
Feb_2012/QEX_1_12_Audet.pdf

232 45
 of the inductance versus rotation angle and a photo of the
EE. Bucher, 1919, "Wireless Experimenter’s Manual", Variometer mounted in its stand, ready for use.
Wireless Press, New York.

Wes Hayward, W7ZOI, Experiments with Coils and Q-

Measurement, October, 2007 (Updates 01Dec07, 08Dec08.
http://web.archive.org/web/20101226143919/http://w7zoi.net/c
oilq.pdf

Steve Ratzlaff, AA7U kindly made detail measurements on the

Q of the capacitors and coils used in this report.

Dick Kleijer , 2013, Measuring the Q of LC Circuits,

http://www.crystal-radio.eu/enqmeting.htm

Barbara Dziurdzia, 2013, Damped oscillations in RLC circuits,

AGH University of Science and Technology in Cracow,
http://home.agh.edu.pl/~lyson/downloads/Manual_8.pdf

Rice University EE., RLC Circuits,

http://www.owlnet.rice.edu/~phys102/Lab/RLC_circuits.pdf

Durham University EE., Oscillations and Resonances in LRC

Circuits,
http://level1.physics.dur.ac.uk/projects/script/lcr.pdf

46 231
 Crystal Radio Capacitors and ESR
Kevin Smith

In my ongoing effort to understand crystal radios and the

theory that makes them work I have recently turned to setting
my expectations concerning capacitance. The tuning capacitor
is, next the coil itself, the principal component of the set. Few
builders fabricate their own tuning caps so I suppose it is
possible to take them a bit for granted. I have been reading on
the radio forums some threads concerning radio Q and the
subject of the variable cap Q seems to be getting more
attention. So, what is the deal?

I set about my study by gathering what I could find on the web

and at the close of this note I include a short reference list of
useful pages from which I extracted data the ideas. I wish to
extend special thanks to Ben Tongue who, as usual, has
provided the essential measurements and background. The
Perhaps the less-than-optimal condition was a factor, but I was
plot below of capacitor Q versus Equivalent Series Resistance
in luck to win the coil with an only bid. I had no illusions that
(ESR) summarizes my findings. It is a bit busy as there are a
this was a long-shot deal, the chances of the variometer being
number of interesting and important ideas represented in the
functional were poor. My expectations were not deceived
relationships presented.
when it arrived sometime later. Nope, it didnt work. The
question was, could it be restored, and just how did the thing
Any discussion of component quality needs first to
work, how were the connections made anyway? Inspection
acknowledge that the fight to increase the component Q is in
quickly led to three obvious broken connections and an
reality a fight to eliminate (not possible) or reduce as much as
understanding as to how the connections were intended. A few
feasible the resistive losses associated with that component. It
dabs of solder later and under the meter I had a fully-
helps then to look not just at Q, but at how Q relates to the
functionning variometer! All that was left was to give the
component ESR. There are two fundamental equations that
vintage coil a lovely stand to show it off.
describe this relation: 1) Q = cotan theta = 1/D. D is the
dissipation factor and is an attribute of the dielectric material
Measurements of the Inductance show a large variation, from
used in the construction of the capacitor. Equation 2) ESR = D
163 uH to 1036 uH. This should give good duty to electrically
/ 2pi f C. This equation tells us that capacitor losses expected
"lengthen" my woefully short antenna. Shown below is a plot
due to resistance are both angular frequency and capacitance

230 47
dependant. It is important to remember that not all Q's are Vintage Variometer / Loading Coil
created equal. What we want is to keep the ESR to a http://www.lessmiths.com/~kjsmith/crystal/lcoil.shtml
minimum.
My receiving station is
On the plot there are a number of data trends presented. Note pretty typical I expect,
first that both plot axes are logarithmic, it makes a difference decent ground, so so
on how one thinks of the data. Additionally, note that the plot antenna and I have been
is specific (mostly) for data at 1.0 MHz. I make the hankering for some time to
assumption that a cap that is superior at that frequency is likely add a loading coil to my
to be superior across the BCB. station to better match that
antenna with my radios.
Lack of urgency or just
plain laziness perhaps have kept me from working on this
project. Part of my hesitancy has been the idea of utilizing a
variometer as a loading coil, and these are not easy to find.

Variometers are variable induction coils that function by being

connected in series and changing the mutual inductance
between them. Most variometers though have a relatively
small range of inductances making them of limited suitability
for, well, just about anything in crystal radio. They find a
home mostly these days as a control for regeneration in such
circuits. Alternatively they make a nice variable inductance
load on your antenna, the purpose I was seeking. Only, I really
wished to find/make something with quite a lot of wire to get a
good range of inductances going. It was by some degree of
luck I believe that I noticed some time back on ebay a rather
beat-up looking variometer on offer. The variometer, shown
The first things to discuss are the three slanting lines. These below, was a basketball weave design and importantly, was
are computed from a table of Dissipation Factors at 1.0MHz by wound in at least three layers suggesting a lot of inductance.
Dr. Johnson. I have supplied the capacitances and calculated
the resulting Q and ESR values. When one goes looking at
capacitor data sheets or summary tables one is likely to find
data computed at 100 KHz and 1.0 MHz (mercifully) but that

48 229
 the likely capacitor data will be specific to caps in the uF
range, or 10^6 higher that what we need in a crystal radio tank.
Be careful not to get enamored with the high Q values
occasionally quoted, check the capacitance. From the plot it
should be apparent that not all Q's are created equal. Larger-
value caps have lower ESR and so the Q's are often agreeably
high. I have plotted a few specific cap measurements from
various sources.

Two measurements should stand out; these are indicated by the

larger blue diamonds. These are measurements by Bill
Hebbert and presented by Ben Tongue in section B of his
Article 24. I reproduce his Figure three below, with grateful
apologies for the liberties I have taken,

By replotting Ben's figure I can calculate the relation between

Q and frequency
(table below) and so
compute the estimated
Q at any frequency.
On my main plot I
give Ben's Capacitor
A and B for 1.0 MHz
given the 250 uH
inductor. Ben states
that the plot assumes
20pF stray added. Since the capacitance needed to resonate at
1 MHz and 250 uH is 101 pF, I make the assumption his caps
were set to 81 pF. This accounts for why the diamonds are set
slightly off the 100 pF line of the plot. Given his Q
measurements it is easy to compute the ESR for each cap.
This is the world of the crystal radio builder. I include the
ESR at the BCB endpoints for each cap as well to give the full
picture.

228 49
 pF f Q
tank cap MHz Cap A Cap B
375 355 520 19000 4750
190 170 730 14000 3230
114 94 943 14400 1610
101 81 1000 13148 1828
60 40 1300 11600 870 CHAPTER III
35 15 1710 9800 460

for I = 250 uH and 20pF stray in tank

(Q data in italics is computed) Accessories on Breadboards
The quest for high Q is indeed a quest for low ESR as long as
you work in the capacitance range needed. One interesting
piece of advice I have occasionally seen is that, to increase
your tank Q, it is better to use lower-value coils and higher-
value tuning caps. As much of the losses in the tank derive
from losses in the coil this makes excellent sense. I wish to
ask the question what about those higher-value caps? The plot
immediately suggests that moving to higher capacitance drives
down the ESR, this is a good thing. It may or may not
improve the cap Q but I am not sure it matters as long as
resistive losses are minimized. To better set my expectations I
have again turned to Ben's Cap B data.

The small plot below takes data presented in his Figure 5.

Data presented include Q at various capacitances and
frequencies. I cross-plotted his data for pF and Q for the
appropriate frequencies using both 250 uH and 180 uH
inductances. Reading the actual Q off the tiny plot was an
estimate at best, but it is interesting. This allows one to
estimate his cap Q at any capacitance setting. The table below

50 227
 provides the
estimated Q, and
calculated D, and
ESR for the
capacitances from
Fig 5 (with 20pF
stray added). The
values of Q do not
improve by moving
to higher
capacitances but the
ESR of the cap is slightly (at low tuning freq) and well
improved (at the high tuning freq). This is the goal. It is
worthy to note that the tuning frequencies do not line up with
the original data from Ben's Fig 3 as seen on my main plot.
While Ben specifically noted that his figure 3 included 20pF
strays, in his figure 5 he speaks in absolute capacitances. I did
my best to correct for this but there remains some ambiguity in
the presentation. The point remains though; moving to higher
capacitances in the tank will lower the tank losses due to ESR.

250 inductor (w/20pF stray)

tank pF KHz D Q ESR ohm
385 513 0.0002083 4800 0.177
220 679 0.0003448 2900 0.404
134 870 0.0005714 1750 0.917
101 1000 0.0007638 1309 1.502
34 1726 0.0034973 286 23.046

180 inductor (w/20pF stray)

tank pF KHz D Q ESR ohm
385 605 0.0002381 4200 0.172
220 800 0.0003571 2800 0.355

226 51
134 1025 0.0006250 1600 0.852
48 1712 0.0019341 517 6.425

On the main plot then I have included for Ben's Cap B a

second Q vs. ESR curve to illustrate the estimated benefit of
using a higher tuning capacitance range. The new curve slides
mostly to lower ESR (especially at the high end of the tuning
range) and also to slightly lower Q. The improvements made
appear modest from the modeled data and, for the case of a
high-Q capacitor the possible improvements may be slight
indeed. It is comforting to know that at least there should be
no negative impact. Lower ESR certainly improves tank
overall efficiency. This is a win-win situation where the set
designer wins twice!
A look at the resonance curves above (including an over-
In seeking the "Holy Grail" tank capacitor I would add to the coupled 2-inch spacing case) is very interesting. I include this
list of valuable features. We have ceramic stator supports, as a nice demonstration of resonance at various conditions of
silver plating on the vanes, and good wiper contacts as features coupling. The 2-inch coupling case clearly demonstrates the
desired. I would add to this an extended tuning range if the problem of over-coupling in a radio. Two resonance peaks
capacitor is truly to be considered "Holy". With a standard show up and tuning brings in different stations at the same
230 uH coil the tuning for the BCB ranges from 40 to 380 pF time. Total power is still very high, just not useful. The 8-inch
or so. With a 180 uH coil the capacitance needed ranges from case is pretty close to critical with good power and the two-
50 to 490 pF, (stray capacitance not accounted for). peak problem eliminated. Bandwith is still pretty high at 22khz
(the curves are made with the 4 parallel HP's). Finally when
Two questions immediately come to mind with this analysis: separated to 14 inches the power loss is not too bad and
1) To how low a coil inductance can you go? bandwidth has narrowed to 10 giving a loaded Q of 107 as
2) If you find a few of these super-caps can you mentionned above.
send me one?
It ain't Litz, but PD Close!
kjs 6/2013 (in France) kjs 11/2012

52 225
tuggle tuner on this set is in the 50-80 ohm range giving an
excellent match and maximum power transfer to the tank.

At the back end things get more complicated, a diode is References:

present! For my testing I have included a step to determine,
under test conditions of course, the optimum load resistance AVX Basin Capacitor Formulas.
for maximum power transfer to the phones. Maximum power http://www.avx.com/docs/Catalogs/cbasic.pdf
output means that the load resistance matches the output
resistance of the tank/diode combination. The principal factors Ben Tongue 1999, Sensitivity and selectivity issues in crystal
which influence this include the frequency and tank coil Q radio sets including diode problems; measurements of the Q of
(which don't change from test to test), the diode Rd and variable and fixed capacitors, RF loss in slide switches and
characteristic, and the signal voltage being rectified. loss tangent of various dielectrics.
http://www.bentongue.com/xtalset/24Cmnts/24Cmnts.html
Signal level and diode characteristic are inter-related and
together determine the resistance of rectification. At very low Wikipedia, Types of Capacitors, Ohmic Losses, Dissipation
signal voltages rectification takes place on the "square-law" Factor and Q.
portion of the diode with large swings in resistance with signal http://en.wikipedia.org/wiki/Types_of_capacitor#Ohmic_losse
sine wave. Larger signal levels rectify higher on the s.2C_ESR.2C_dissipation_factor.2C_and_quality_factor
characteristic curve (peak-detection region) with consequant
lower resistances and less variation. The Rl shown on my table Dr. Gary L. Johnson, 2001. Lossy Capacitors
thus is specific for my test. My radio signal is in the 6-8 mV http://www.eece.ksu.edu/~gjohnson/tcchap3.pdf
region placing it in the peak-detection portion of HP5082-2835
diode and possibly borderline square-law / peak detection on
the FO-215. This explains why the optimum Rl is lower that of
the HP but pretty close to the FO-215. It may also help explain
why the FO bottomed out on the 14-inch separation, there was
just no longer much signal for it to work with.

224 53
 Pin = Power into the set in microwatts
Pout = Power out the back end (into the phones) also in
microwatts
Eff = Set efficiency or Sensitivity and is the ratio on Pout to
Pin * 100
BW = -3dB Bandwidth of the set at 1100khz (f res)
QL = Loaded Q of the set = f res / BW
Rx = Input resistance of the set in ohms
Rl = Load resistance on the tank in kohms
Rd = Junction resistance of the diode in the set

The power numbers are fairly straight forward and represent

the power calculated across the dummy antenna resistor and
load resitor. Input to the set is from a signal generator with a
200mVpp signal output. Losses due to the dummy were
accounted for and the input power is that presented to the set
itself. Looking at the Efficiency and QL numbers one sees
nicely the tradoff between sensitivity and selectivity. The
larger the spacing between the sets, the higher the QL, but at a
sacrifice of sensitivity. While the numbers look low, they are
still good for crystal sets and the 107 QL is a superlative
number. By comparison, I have looked at a High-Performance
built by Dave Schmarder (see my Radio Test page) with big
Litz and super capicitors and this set max's out around QL =
120-125. I would opine that any QL above 100 is
"Performance Quality".

Note for a bit the Rx numbers. Impedance matching for

maximum power transfer starts at the antenna and antenna
tuner. The dummy antenna used for the measurements mimics
a fairly typical longwire with an output impedance between 25
and 50 ohms. Of the many radios I have measured, most have
set input resistances between 100 and 250 or so ohms. The

54 223
 Diode Modeling Update
Kevin Smith
http://www.lessmiths.com/~kjsmith/crystal/dmodel.shtml

The following explains my first attempt to compare a

measured diode characteristics with calculated plots based on
the Shockley Equation.

I have been taking serious look into diode modeling and

measurement recently in an attempt to better understand these
solid-state successors to the humble crystal and cat's whisker
detector. There is much good information on diodes to be
found in many excellent web sites of course, but there is
nothing quite like actually making the measurements and
working with them to get a good feel. This page reports some
of my protocols and test setups which I have found to be
useful.

Performance summary: Most probably the single best web article for the measurement
of diode Is and n parameters is given by Ben Tongue in his
Article 16. In this article he describes an interesting circuit for
the measurement as well as protocols for making them. For my
purposes I did not wish to make this a construction project and
felt I might get along with good quality meters and my
already-built "Diode Test Jig". Essentially Ben Tongue's
method consists of making two precision measurements of
As mentionned at the start of the article, this set was designed Voltage and Current through a diode at small signal levels,
as a high-performance rig. Lets take a look at the numbers and essentially about 3 and 6 times Is (sufficiently low that the
assess how well I did. On the spreadsheet above I have voltage drop across the series resistance Ro can be ignored).
indicated the essential measures of the set in four The measurements are then substituted into the Shockley
configurations, two spacing tests (8 inches and 14 inches) with equation Is*{exp[(qe/(n*k*T))*(V-I*Rs)]-1} and solved
4 parallel HP 5082-2835 diodes and two spacing tests with the simultaneously for the two data pairs. Mike Tuggle has
holy grail FO-215 diode. First an explanation of the units: provided a nice excel spreadsheet to do the math. This
spreadsheet, Cal_n_Is.xls forms the basis of my technique and

222 55
I am thus indebted to both Ben Tongue and Mike Tuggle for from 4 to 7 or more khz off peak tuning when I remove my
making this project feasible. hand from the dial. The set is fairly modular and readily
modified so I am keeping my eyes open for better quality caps.
In addition to simply calculating the main parameters Is and n, I am afraid I must learn to live with hand capacitance. Oh well,
I also wished to measure enough points to plot a characteristic we see about the next set!
curve, and to compare the measured curve with one calculated
directly from the Shockley equation. My first spreadsheet then Portrait of set, ummmm..
combined Tongue/Tuggle calculation protocol and a graphical
view of the match between measured and theoretical. I did
make a few methodology modifications which I thought/hoped
would allow added accuracy:

1) Using the full Shockly equation without simplifying for an

assumed 25dC room temperature. While the 25dC assumption
is generally good and probably within the measurement error
(or perhaps not), including the actual measured temperature
eliminates doubts of inaccuracies due to this parameter. With
the power of PC's and spreadsheets there is no reason simplify
the equation.

2) Used Vd = 0.04 and 0.05 V (adequately close to Ben

Tongue's recommended 0.039 and 0.055 V) whenever
possible. Note for some diodes with significantly different
forward voltage drops, Si diodes in partictular I had to use
higher values for Vd.

3) Reporting: I decided that, as there is no unique solution to n

and Is, each is dependant on the valuse Vd and Id, I feel that
reporting both Vd and Id is necessary for repeatability.
Naturally I also include the ambient temperature in the report
as well, and

4) I included a calculation for Ro (or Rg if you are using

Tongue's reference) as that is the goal of the exercise.

56 221
With high-efficiency coils, very loose coupling is permitted A screen shot shown below:
with acceptable power levels, thus allowing very narrow
bandwidths to be achieved. This can only be accomplished
with separate base boards. To further minimize losses I have
used short direct lengths of large 16 AWG stranded copper
hookup wire and importantly, I have insulated all components
and electrical connections from the board with ceramic
standoffs. Finally, for the coils I have utilized a very lovely
sky-blue teflon-insulated, silver plated wire of 18 AWG,
strictly military spec, ($$$). I know, I KNOW.. This isn's a
Litz Blitz set and yes, the performance suffers a little because
of this. Still, I have good performance measures to tell me I am
certainly on the right path. Note also that wire of such size and
insulation is quite thick and to get the desired inductance
required very large coils of some 4 3/4 inch diameter. The
coils dominate the set visually and provide the name.

The tuning capacitors I would allow are most likely the To use the sheet one need only adjust the voltage Vd to that
weakeast link in the chain. I have chosen fairly generic though shown in the first column and record the current Id. If only
not bad-quality caps. The antenna tuner has a dual-gang cap Vd1/Vd2 and Id1/Id2 are measured then the next step is to
with sections 465pF and 375pF respectively, well matched for adjust the value for "n" until the match between the two
my needs. Note, from my studies of antenna tuning (elsewhere equations is exact. I even provided a simple ratio calculation to
on my pages) I have noted that a tuggle tuner does not easily test for a match.
necessarily require equal-size sections on the tuning cap. In While the math is good and results excellent, I quickly found a
fact, I would state that tracking is better if the ground section couple deficiencies in my method,
max's out at a lower capacitance that the tank section. The cap 1) measurements require high precision and its nearly
has aluminum blades and ceramic insulation and is a decent impossible to land EXACTLY at the voltage required. I
capacitor. For the tuning tank I chose a standard 410pF single needed only to get quite close, let the meters stabilize for 2 - 5
gang unit from the Crystal Set Society. These are aluminum minutes, and then record BOTH Vd and Id.
blade caps with phenolic insulation, OK but not superb. Both 2) the work is rather tedious and, for a large number of diodes
tank and atu include 3:1 reduction gears to help spread the it pays dividends to measure only the needed Vd and Id and let
band and allow more careful tuning. I did not allow for hand the plot aside. My second spreadsheet thus dispenses with the
capacitance which I have noted is indeed a problem. When I plot. It is with my second spreadsheet that all my data tables
tune the set to resonance with a signal generator, I find I am and results are posted.

220 57
3) better comparisons between diodes can be made if one My Teflon Warrior
targets specific Currents rather than Voltages, discussion on http://www.lessmiths.com/~kjsmith/crystal/teflon.shtml
this below.
With this current radio I am making my entry into the
A view of my current data reporting spreadsheet showing ALL "Performance Radio" constructors club. I have tried to follow
input parameters as well as determined values of n, Is, and Rd. all the best practices of high-end crystal set design while
Data input fields in blue: (target Id1 close to 1.0 uA and Id2 giving the set my own personal breadboard mark. Two central
close to 0.5 uA). themes guided my design on this set, simplicity and simplicity.
Calculated fields in red and black. From the circuit diagram below you will note that I have
Adjustment field "n" in green: (adjust value of n until the ratio chosen, not surprisingly, a simple double-tuned circuit layout
of the two calculations = exactly 1). with a Tuggle front end and a Tank/Detector circuit with no
circuit decorations whatsoever. No taps. No selectivity
enhancement. No Benny. No Decoration. No nada. This is a
basic radio wherein I have striven to keep losses, including
insertion losses, to an absolute minimum. Should I wish to
(Note: the engineers in the crowd will have noticed in the include selectivity enhancement and/or benny at a later time,
above report than I am carrying a degree of precision not they will go on an outboard platform along with transformer
justified by the level of accuracy in my measurements. The and phone plug. The radio itself is simplicity realized.
final results of n, Is, and Rd should be rounded off to not more
than three significant figures.)
A copy of my spreadsheet can be downloaded here.. As can be seen in
the set photos
While measuring a good number of diodes, both Germanium, below, this design
Schottky, Silicon, and even a few LED's, I noted that Tongue's follows standard
recommended measurements at Vd = 0.039 and 0.055 V practice of placing
resulted in current reports spread out over more than two the open and
orders of magnitude. As the determination of Is and n is not closed circuits of
unique but a function of Id, I felt uneasy by this method. While the set on separate
most Germanium diodes I have measured have a fairly narrow baseboards. This
range of Forward Voltage drops, (Vf), that of Schottky's can is important as it
range up to a tenth of a volt. This is the factor responsible for allows easy and
the huge range of measured Id values, see the following plot of wide adjustment
I/V characteristics for some selected diodes: of the coupling
between the coils.

58 219
 In order to get what I feel is a better comparison between
diodes, I have decided to target not some pre-determined
voltage, but rather target currents of Id = 0.5 and 1.0 uA. On
the above plot on the right (with Log Id vs Vd scale) it will be
clear that I am aiming at the same part of the characteristic
curve regardless of Vf. Hopefully this will allow good
comparison between diodes with rather different Vf, even as
far as including Silicon and LED's in my mix. On va voir....

Shockley equation Id = Is*{exp[(qe/(n*k*T))*(V-I*Rs)]-1}

where:
n = ideality factor
Is = Saturation current in Amps
Id = Diode Current in Amps
Vd = Diode Voltage
k = boltzmann = 1.38E-23 J/K
T = temp K = 300
K = dC + 273.15 Kelvin
qe = electron charge = 1.609E-19 cmb

218 59
 effects I can easily hear when tuning the set. Oh well.. Better
Often simplified to: be forewarned next time! This does not necessarily mean that
Id = Is*(exp(Vd/(0.0256789*n))-1) [at room temperature] double-tuned rigs MUST be built in modular fashion with
separate tuning and detection units, but it certainly suggests it.
Diode Resistance Rd = VT * n / Is With the lack of surface real-estate to lengthen the coil
Where: separation, I made a small modification on the set as seen in
T = 300K the third photo above. There you see the possibility to rotate
VT = k*T/qe = 0.0256789 the outer coil away from vertical thus further lowering the
SO: Rd = k * T / qe * n / Is coupling. I can get a max 40 degree rotation which just gets the
coils to critical coupling. No more ghosts on the band! Still, I
I cannot express enough my indebtedness and thanks to Ben think my next double-tuned rig will indeed be modular.
Tongue and Ken Khun for their great work and web Experience is such a good teacher!
documentation in matters Diode!
kjs 07/2011
A photo of the current "bench" setup. Keithey 195A pico-
ampmeter + Keithley 192 DMM bench meter.

Kevin 09/2011

60 217
 Spreadsheet for calculation of diode n and Is (@1.0uA)
Diode n Is Rd
nA k Ohm
Germanium Diodes
FO-215 ITT 1.11 175 163
FO-215 ITT 1.10 196 144
1N277 black 1.50 2293 17
1N277 black 1.60 2296 18
1N270 bonafide 1.28 915 36
1N270 blue 1.26 857 38
1N270 blue 1.66 2310 19
D18 russia 1.20 194 160
A few words on resonance: When making my measurements I D18 russia 1.27 193 170
noted a curious effect in the resonance curves. They became GAZ 51 Tesla 1.14 140 211
increasingly asymmetrical with wider spacing of the coils and
actually had a double hump at the maximum spacing of 10 cm. GAZ 51 Tesla 1.52 617 64
This is the exact opposite of how I expect resonance to work. I OA 5 Tesla 1.85 3384 14
set to measuring the Vout across a significant portion of the OA 5 Tesla 1.48 1996 19
MW band for each of the coil separations and present a graph D9E russia 1.42 2161 17
below: D9E russia 1.54 2414 16
It should be readily apparent from the above graph that my set 1N34A bonafide 1.82 2153 22
is badly over-coupled even at its widest separation of 10cm. I 1N34A bonafide 1.25 1392 23
suspect critical coupling for the set is with a coil separation in 1N34A green 1.57 1389 29
the 15 - 20 cm range. Yet one more design consideration to 1N34A green 1.30 1185 28
keep in mind when making a new set! The double hump causes 1N34A green 1.32 1554 22
two problems for the set, one: it tunes each station twice and
two: has the possibility to bring in two stations at the same 1N34A 37 orange 1.10 1458 19
time if they have the right separation on the band. Both these 1N34A 37 orange 1.36 1370 26

216 61
 1N34A red 1.34 833 42 Sensitivity on the other hand, well, it sucks. 2 to 11 percent
1N34A red 1.32 993 34 Vdc/Vpp are very low numbers, many of my single tuned sets
range from S = 10 to 50 and my Minstrel Boy getting over 60
D310 russia 1.01 1215 21 percent. What happened? The data show the SEC to be
Diode n Is Rd involved here. Note that with capacitance in series the
nA k Ohm sensitivity is about 10% but when shunted across the tank the
GD 402A russia 1.55 970 41 selectivity improves, but the sensitivity crashes down to 2-3%.
GD 402A russia 1.71 1360 33 I really expected better for this set. This is my second double-
tuned rig, my first was a simple set using solid magnet wire
UK A russia 1.56 1565 26 close-wound on 3" cardboard cores and having 6 twisted taps,
UK A russia 2.00 5766 9 all Q lowering. This sets' coils have no taps and no cores and
UK B red russia 1.64 2049 21 ought to have significantly better Q. The primary difference
UK B red russia 1.85 3777 13 between the two sets is the presence of the benny and SEC on
UK C ora russia 1.92 3462 14 my new set. I am worrying about insertion losses more than
offsetting the improvements in the coils.
UK C ora russia 1.95 3354 15
UK D blk russia 2.00 4901 10 The severe loss of sensitivity when using the SEC on this set
UK E blk russia 2.00 5862 9 also reminds me of my experience with my Fleming set, a
UK F blk russia 1.82 3523 13 modified MRL#2 using SEC. There I also found loss of
UK G blue russia 1.86 3123 15 sensitivity when the SEC was in the circuit. At the time I
concluded that the circuit was inappropriate for testing the
SEC concept, now I am not so sure. This set employs a high-
Average 1.39 1334 54 performance circuit (if not high-performance parts) and is fully
correct for the SEC inclusion. One additional test I intend to
Schottky Diodes make when time allows is to completely remove the SEC and
HP 5082-2835 1.07 12 2206 Benny ornamentation from the set and re-test the sensitivity. I
1SS98 1.08 12 2302 suspect that insertion losses are killing me. Dave Schmarder
1SS98 1.06 13 2188 has the most experience with this design feature, I wonder
1N5711 blue 1.07 6 4541 what his thoughts are? Have you tested a set with and without
the SEC? What have you found? I am suspicious of this feature
1N5711 blue 1.07 6 4816
now.
1N60 1.09 178 158
1N60 1.10 161 176
BAT 46 1.12 141 204

62 215
capacitor which places the capacitance either 1) in series with BAT 46 1.18 172 177
the top of the tank and detector, or 2) shunted parallel across 1N34A ? schottky 1.11 317 90
the tank. I make a full set of measurements with the SEC cap
set fully to each the series and shunt positions. The above 1N34A ? schottky 1.24 482 66
paired measurements are made for each of three coil coupling 1N5819 1.19 750 41
positions. The two coils can be shifted along two rails to bring 1N5819 1.14 807 36
them from essentially touching to a maximum separation of Diode n Is Rd
about 10 cm, or about one coil diameter. This is the max nA k Ohm
separation obtainable on my breadboard as the Hammarlund
caps are quite large and take up much space. To evaluate the 1SS16 NEC 6.24 14288 11
impact of coupling, I measured with the coils separated 1cm, 1SS16 2.27 4101 14
5cm and the max 10cm.

OK, enough with the preliminaries let’s look at the data! Average 1.13 276 1137

Silicon Diodes
1N914 1.95 5 9293
1N914 2.02 7 7618
1N4148 2.00 6 8934
1N4148 2.00 6 9323
The table at left presents the measurements (Res = resonant 1N4007 1.51 1 64311
peak Vout mV : BW = bandwidth and -3dB : Q = under load :
1N4007 1.51 1 77954
S = sensitivity = Vdc/Vpp) vertically and the different cases
horizontally. Bandwidths for the cases all range between 16 - 6A10 1.57 2 19275
18 kHz for SEC in series and 11 - 14 kHz for SEC in shunt. 6A10 1.54 2 20379
These are pretty good numbers for a standard (non-DX 1N4736A Zener 1.15 0 1.91E+09
Contest) rig and correspond to Ql in the 65 - 85 range. Context 1N4736A Zener 1.14 0 2.50E+09
for this comes from Ken Khun's excellent engineering pages
KB 130 russian 1.14 0 5.83E+08
where he notes for crystal sets that loaded Q normally ranges
between 10 (poor) to 50 (typical) to 100 (very good). My set KB 130 russian 1.16 0 4.36E+08
here then is pretty good in this respect and confirms my UK H russian 1.94 8 6263
eardrum assessment above. UK H russian 1.77 4 12548
UK I russian 1.78 0 260253

214 63
 KD 401A russian 1.51 0 232930
KD 401A russian 1.38 0 1539619
D 220 russian 1.26 0 322405
D 220 russian 1.24 0 537707
D 223A russian 1.20 0 10262661
Diode n Is Rd
nA k Ohm
D 223A russian 1.20 0 8565213
UK J russian 1.46 4 9849
UK J russian 1.40 2 20013

Average 1.52 2 2.37E+08

Light Emitting Diodes

Red 2.15 0 1.70E+12
Red 2.07 0 5.52E+12 A few words of the performance of my Hammerlund set are in
Amber 1.56 0 2.09E+17 order. When I first started using the set I noted with
Amber 1.62 0 6.46E+16 satisfaction that, by my eardrum meter, the set seemed to
indeed tune quite sharply. This is certainly my principal goal in
Yellow 1.59 0 8.03E+17
making this double-tuned set. The only fly in the ointment was
Yellow 1.38 0 2.87E+20 that the radio seemed quite insensitive, the stations coming in
Green 2.05 0 4.23E+13 very weakly and needing amplification. The SEC control did
Green 2.07 0 4.34E+13 have an interesting and dramatic effect when used, but it also
Water Green 3.27 0 4.85E+13 hugely lowered an overall low sensitive set. With these
thoughts in mind, I have spent some time making some
Water Green 1.93 0 5.79E+19
preliminary measurements on the sets' resonance, sensitivity
White-Gn 1.44 0 1.05E+20 and selectivity.
White-Gn 3.04 0 6.88E+09
To characterize the set completely, I measure under a number
Average 2.01 0 3.76E+19 of different cases. I wish to test the effect of the selectivity
enhancement circuit (SEC). The SEC uses a 20pF differential

64 213
 Diode Test
Kevin Smith
http://www.lessmiths.com/~kjsmith/crystal/dtest.shtml

So, what is the best diode out there to use? Do crystals always
show poor sensitivity compared to germanium? What about
Fleming's marvelous device, the vacuum diode? I start by
using a test jig, the design shamelessly modified from Dan
Petersen in the Nov 2009 Crystal Set Society Newsletter, (if
you aren't a member, go sign up to be.. NOW):

This is a very simple circuit allowing power input, attaching a

diode for testing, and sampling both the curent and voltage
across the diode under test, (DUT). I made a simple
modification by adding a voltage divider, about 20:1 at the
power input. The power supply can be connected either at the
front across the divider, or after the divider in front of the 100k
current limiting resistor for higher voltage readings. I get the
input voltage from a modest 18V, 2A regulated DC power

212 65
supply. This unit has both volt and amp controls with digital Something was needed here! At left is a photo of a very sweet
readouts, nice to use and dosnt take up much room on my brass shaft converter machined by my clever and capable EE
desk. Here is a case where I recommend you look at getting a elder brother, many beaucoup thanks for your addition to the
new unit. I looked long and hard on ebay, but found few project! Not being an engineer myself I was able to follow
bargins and few units with the same features. Still, the supply very poor practice and just wired the whole lot together,
regulator was never designed to adjust the output to precise crossed my fingers and hooked the sucker up.. Ya Hey! There
small values making it somewhat difficult/tedious to adjust. be sounds here!
Adding a small rheostat to the test jig may aid for making fine
adjustments on the voltage. Another piece of equipment you The set is playing as I type these words and I am most hopeful
will need for this test is an amp meter. I use a nice small digital for a successful performance goal. Although much testing and
meter I found new on ebay for $15 shipping included. It reads measuring yet remains to be done, the set by ear calibration
amps to a lowest range of 200uA, more than enough precision tunes quite sharply. The coils have 4.2 inch OD and 3.2 inch
for the task. For that price get two. The DVM is, of course, my ID and can slide upwards of 4-5 inches separation for looser
trusty Keithley workhorse. Easy protocol as follows: coupling. The selectivity control has a dramatic effect on
1) Set up the test with the diode and meters clipped to the jig. sensitivity and, I am sure, on selectivity as well.. to be
2) Adjust the power supply until the reading on the DVM measured soon so stay tuned.
meter is where you plan to measure. Adjust the voltage (DVM
readout) in steps from 0 to 1.0 Volt in 0.1 V increments and The photos below present the set in its final functionning form.
record the current (Amps) in a spreadsheet.
3) Repeat at each tenth volt recording while the spreadsheet
makes a sweet graph.

Test Results
First, I offer a quick look at my diode modeling setup. I am
currently upping the ante on resolution in order to determine
diode parameters Is and n, hoping for success soon! Setup and
protocol..

Over a period of a few weeks I ran I-V characterization tests

on a good number and variety of diodes, crystals and my two
tubes to see how very thing stacks up. The following presents
the resulting curves, diode photos, and some puzzling
questions/conclusions I churned up in the process. I start off
with the realization that when one orders diodes, it is by no

66 211
tried this before on a version of an MRL-2 but it was a means certain what one will get. This seems especially true for
disappointment, not the right set for the test. As this set was to the supposedly ubiquitous 1N34A. Unless you see the part
employ a high-performance circuit, an SEC would be number actually on the diode, you probably need to test it to
appropriate. The set itself is only moderate performance know what it really is. So, lets take a look!
though, the caps having fairly high-loss brass plates and my
coils wound with solid wire. This set also ploughs new ground
for me in my first basketweave coils. Basketweave coils recuce
stray capacitance and increase coil Q by eliminating the need
for lossy coil formers. As these were my first, I merely used
solid wire I had in the house. As they are wound of 20awg
solid wire, they are quite springy and I had to hot-glue the
bejeezes out of them to get them to stay put. I do one day
expect to build a truly hi-performance set, but this is just a
stepping stone to test components and design. As such, I tossed
a benny into the circuit for good measure.

The design calls for non-tapped coils to increase coil Q. While

the coil Q is not great, it is certainly considerably better than
any of my earlier cardboard tube or even PVC selonoids and it
ought to have a decent tank parallel resistance. Matching
(hopefully) to this I used a germanium FO-215 diode from
Mike Peebles. The combination in these diodes of low n (~1.1)
and Is (190) gives a diode with moderately high junction
resistance on the order of 150+ kOhm which should work well
with the solid-wire coils in the tank.

The selectivity enhancement

circuit called for a cool 20pF
differential cap which I
found at Surplus Sales of
Nebraska. The cap itself is a
cute little item but the shaft
was a short 1/4in long and
only 0.19in in diameter.

210 67
Germanium Diodes Hammarlund, Double-Tuned
http://www.lessmiths.com/~kjsmith/crystal/ham.shtml

This fine radio features three beautiful and high-quality

Hammarlund square-law 510pF variable capacitors. I have
been hankering for some time to build a radio with more than
just OK to mediocre performance, which translates to a
double-tuned set with a properly tuned front end. The
capacitors mentioned before were, at first, a poor choice for
this project as I needed for the antenna tuner a two-ganged
variable cap and the Hammarlund units were only single.
Additionally, I only had one! With what I can only call
considerable luck, two additional identical capacitors were
listed over the last several months on ebay giving me three
total. The caps are high quality and very heavy construction
and I figured I ought to be able to dismantle them enough to
give a common shaft to two thus ganging them as needed. This
turned out to be
the case and the
project was a go.

Additionally, with
each radio I build,
I like to test and
learn new circuit
concepts and
ideas. This radio
was to be made
with a farily
simple double-
tuned circuit, a
tuggle-tuned front end and variable coupling. I also wished to
test and play with the "Hobbydyne" or selectivity enhancement
ideas presented by Dave Schmarder on his excellent site. I had

68 209
 1N34A series

208 69
1N270-277 series When I did get around to hooking the set up to the signal
generator I found to my relief that indeed the set has good (not
great) selectivity and really excellent sensitivity. Bandwidth
around a nominal 1100KHz signal to -3dB is 22KHz giving
this set a loaded Q of 51. This is right in the middle range of
set loaded Q to be expected for a "typical" crystal set given by
Kenneth Khun in his notes on Resonant Circuits. Importantly
as well, this set does a much better job at rejecting KGOW
than my other sets. I attribute this to the main coils being
toroids. KGOW blasts down my direction with a very powerful
ground wave and I suspect that the other set coils, mostly open
selenoids and / or spiderweb, are picking up the signal and
flooding my set across the tuning range. Toroid coils confine
the magnetic fields within the torus and thus couple poorly if at
all to surrounding fields, including KGOW. Thank you
Magnavox for making such lovely coils!!

70 207
View from above. Here we see a set of Germanium diodes and photos of the
diodes under test. Immediately you should notice two sets of
curves, what I call the "1N34A" set and the "1N270-277" set.
In this analysis, the second set appears clearly superior in
sensitivity and deservedly so. In the photos it is clear that one
bunch of diodes I purchased as 1N34A were in fact better
1N277, the black band and thin gold contact wire as well as
common curve unite these. "True" 1N34A's have a more
robust contact wire. I also purchased a 1N270 and received
pretty orange diodes with the number "37" stamped on it, a
fairly robust contact wire and a curve that most closely
resembles a 1N34A, go figure. Mercifully, I DID manage to
get some bona-fide 1N34A's with the part number on the
diode, as well as a single bona-fide ITT 1N270 diode with the
part number. These are my base for comarison.

In addition to the standard Germanium Diodes tested above, I

also have latched my hands on a few "other" germanium
diodes for the fun of testing. These include two vintage
Performance summary for the set is
Russian types, a D9E (in the 270 class) and a D18 (transitional
quite pleasing. When I first started
between 270 and 1N34A). Additionally, I got a few FUZZ
listening to the radio I immediately
diodes popular, I suspect, with the guitar gadget crowd. These
got the sinking feeling that the
are larger packages in metal cases so I cannot see how the
selectivity would be no better then
internal contact is made other that they are stated to be gold
many of my earlier sets, a
bonded point contact type. Clearly these are highly sensitive
disappointing prospect especially
diodes in my 270 class, the OA5 looking best of the lot.
as this set was meant to be a higher
Interestingly, I recently purchased from good Mr. Peebles a
performer. Part of the problem what just the sets' general
few of his "Holy Grail" ITT FO-215 diodes. My measured
quirky and difficult three-dial tuning method. I soon got used
characteristic for one of these is a dead laydown on the gold-
to fiddling with the dials and found I could usually "tune out"
bonded OA5. From the photos, these are in different packaging
most any nearby or interfering station from the one of interest,
with the FO-215 in a traditional glass casing. What makes the
it just took some effort.
FO-215 so great is the fact that its resistance Rd is an
interesting 150k ohm or so which matches very well with

206 71
typical moderate Q tank curcuits. Interestingly, I find that an Unfortunately the primaries do not have taps. 180 uH is a bit
old Russian D18 diode has very similar characteristics to the low for MW work and would require a variable cap with a
FO-215 and so much also be placed into the "Holy Grail" class fairly high top end. I chose three identical 2-gang 270 pF caps
of germanium diode. Take your pick! tied in parallel to give three 20 - 540 pF tuning. The link
circuit consists of two serial 35 pF secondaries for a total of 70
Spreadsheet for calculation of Germanium diode n and Is pF, not enough to tune with so I would need to add further
(modified from Mike Tuggle's spreadsheet) inductance. Here is another new feature for me. I made for this
set my first honeycomb-style coil, green, of course. I gave the
link coil 110 uH for a total link inductance of 180 uH, just as
in the primaries. All tune with the three variable caps.
Additionally I added a series-parallel knife switch for the
antenna circuit capacitance. For my setup the series position
will work best, but this will add utility later on. (One green
feature I tested and subsequently abandoned was a green LED
diode, but even with bias I found it to be hopelessly in-
sensitive).

Averages: n = 1.38 Is = 1315 nA Rd = 52 kOhm

The above spreadsheet is based on measurements of the diodes

shown above. For the determination of Is and n, I chose at
random two examples from my collection of various diodes,
(all germanium in this case) and measured via a modified
version of the methodology outlined by Ben Tongue and Mike Portrait of set, Wearin O'the Green..

72 205
sharpen the tuning overall. This method forms a triple-tuned Tuggle. For this work I have chosen to set Id2 about 0.5uA and
set with its consequent quirkiness and occasionaly annoying Id1 about 1.0uA and then read the needed voltage. This is the
tuning. Still, I find this part of the fun, I am not looking for inverse of the normal method but the justification is that for
tuning efficiency, just sharpness. any diode regardless of forward voltage drop the measurement
is made at the same part of the LOG I vs V characteristic. This
The lovely green color and large size of the coils would be a allows comparison between all diodes. I have included the
prominant design feature and this immediately suggested an actual room temperature in the calculations in order to
Irish theme for the set. I settled on naming it "The Minstrel eliminate this variable as a source of doubt or error. All the
Boy" from a patriotic Irish song written by Thomas Moore, measured parameters as well as the determined results are
reproduced at the bottom of this page. Of course the name also presented in order to facilitate repeatability.
reflects the idea that I expect to listen to music on the set, it
will be a minstrel for me. This decision then set in motion the
need for other green features, green hookup wire and making a
link coil from green magnet wire. Finally a 1N34A diode I
bought a couple years back from Dave Schmarder with the
green band to cap things off.

The toroid coils have identical features including a primary

with 180 uH and a secondary coupling coil with 35 uH.

204 73
Schottky Diodes The Minstrel Boy, my triple-tuned triple-threat!
http://www.lessmiths.com/~kjsmith/crystal/minstrel.shtml

I am always on the
lookout to find interesting
and different circuits with
which to construct my
radio projects. About a
year or more ago I found
on ebay an interesting deal
for a set of three vintage
toroid Magnavox (Model
D) air-core coils. They are wound from a beautiful lime-green
cloth-covered wire and I just had to make an effort. I assumed,
correctly as it turns out but not obvious from the description,
that they would come with internal coupling coils. I also
assumed they would have a tap somewhere on the main coil
but in this I was to be disappointed. I was also disappointed to
discover that in one of the three coils the secondary connection
was broken. Oh well, my project needed only two coils and I
knew these would make a fine radio. I already had a link-
coupled design in mind.

Toroid coils are interesting in that they confine the magnetic

field within the center axis of the toroid. Because of this, stray
coupling seldomely causes a problem, but then again, coupling
in general is not easy. So, I had the idea these would be my
ticket to exploring designs with capacitive or link couplings
between circuits. Eventually I began scouring the net and my
books/references for circuits employing link coupling. This
coupling method seems to come in two main varieties, tuned
(such as found in the Crystal Set Societys' "My Marconi" set),
and untuned (as found say in MRL's #42 set). Tuning the link
is not needed, but it was something I wanted in an effort to

74 203
virtually no difference between the Narrow and Broad
performance. Oh well... At least I understand now why
Osterhoudt removed the choke from the MRL#2 set.

Q for this single-tuned radio is comparable to other radios I

have built. This is a circuit demonstration radio presenting
Fleming's valve. Nor have I finished with the SEC concept.
The circuit utilized here is not ideal for the SEC, ground is
taken off the top of the coil, its an odd, if interesting,
configuration. I will give the SEC another shot with some
other radio. Overall, I am very pleased with the set and enjoy
listening to the local Spanish music format station with the
warm glow of the tube keeping me company. Oh that glow....

202 75
Schottky diodes are very sensitive diodes that work excellently
in crystal radios. Their construction and theory are different
and I confess to not fully understanding these components.
Still, from the characteristic curves, they are excellent! Again,
note that you dont always get what you bargain for. Here I
found what was supposed to be 1N34A's to be some sort of
Schottky of unknown pedigree. The 1N5819, while having the
lowest forward voltage drop, has a rather very high Junction
Capacitance and will not perform well. Posts on the
RaidoBoard Crystal Radio Forum however highly recommend
the 1N5711 for crystal sets although the characteristic curve
dosnt look that fabulous. Many web pages out there on Shottky
diodes, I recommend you do your homework. Of all the
schottky's, I note that Ben Tongue recommends most highly
the HP5083-2835. The high resistance Rd makes them useful
for DX sets with very very high Q tanks. Even so the diodes Looking at the performance in the Broad setting we find the
need to be paralleled with up to 4 or 5 diodes to match SEC has a definite impact. As before, the valve has higher
sensitivity than the diode, but here we see a classic tradeoff
correctly the Rd with the tank Rp. I have found these
somewhat hard to find and expensive, especially when one between sensitivity and selectivity. Turning the SEC to its
requires using several in parallel. I recently measured a few maximum selective position (lowest capacitance) helps, (Q
1SS98 diodes and I discover them to have characteristics increases from 9 to 20 with the valve) but at the price of lower
extermely similar to the HP and I feel they deseve more sensitivity, (Decrease from 32% to 26%). More disturbingly,
attention. Sadly, on ebay they seem to be just as difficult to the overall sensitivity in the broad setting, which ought to be
find and expensive. No free ride! higher than in the narrow setting, (that is, greater than 40%) is
quite the opposite, (less than 30%). Additionally, setting the
Spreadsheet below for calculation of Schottky diode n and Is SEC to its most selective position merely returns the radio to a
selectivity already found in the narrow setting (narrow Q = 18,
(modified from Mike Tuggle's spreadsheet)
broad Q = 20) while lowering the sensitivity. So, I consider the
SEC in this circuit a bit of a failure. My suspicion is that the
insertion losses associated with the SEC are damaging and no
real selectivity boost is given that was not already found in the
narrow setting. To address this, I removed the choke and
minimized the capacitance of the SEC, effectively removing it
from the radio. Re-testing the radio now showed a return to
good sensitivity and moderate Q. Actually, with this test I see

76 201
(Russian D9E, a 1N270 equivalent) and the tube/valve
(Russian 6C19P) in place.

Averages: n = 1.48 Is = 1275 nA Rd = 1030 kOhm

Looking first at the radio in the Narrow switch setting. The
table at left indicates the peak output DC signal in mV at
The above spreadsheet is based on measurements of the diodes
resonance, the measured bandwidth at -3dB above and below
shown above. For the determination of Is and n, I chose at
resonance, calculated radio Q and Sensitivity (Sensitivity here
random two examples from my collection of various diodes,
is simply 100 * DCout / ACin). The table shows that the SEC
(all schottky in this case) and measured via a modified version
circuit itself provides no change in performance. This is
of the methodology outlined by Ben Tongue and Mike Tuggle.
expected as noted above, the SEC doesn’t impact the circuit in
For this work I have chosen to set Id2 about 0.5uA and Id1
the narrow setting. I ran the valve at 6.3 V on the heater
about 1.0uA and then read the needed voltage. This is the
filament and am surprised to see that the radio sensitivity is
inverse of the normal method but the justification is that for
higher with the valve than with a diode. Dissappointingly, the
any diode regardless of forward voltage drop the measurement
valve had a poor effect on the radio Q, going from Q = 25 with
is made at the same part of the LOG I vs V characteristic. This
the diode to Q = 18 with the valve. I have to imagine the valve
allows comparison between all diodes. I have included the
has a rather different impedance / capacitance characteristic
actual room temperature in the calculations in order to
from the diode. Still, the good sensitivity is pleasing to see.
eliminate this variable as a source of doubt or error. All the
measured parameters as well as the determined results are
presented in order to facilitate repeatability.

200 77
Silicon Diodes

The glow.... Oh the glow!

PERFORMANCE:

So, is is a keeper? Does it perform well? Is Fleming's valve a

revolution? The following tests answer all these questions and
them some. We start by noting from the circuit diagram above
that this is essentially two radios depending on whether it is
switched in the Narrow or the Broad setting. Importantly, in
the narrow position the SEC option is essentially out of the
circuit. This will be seen in the measurements as well. A note
here, all measurements were made at a nominal 1100KHz with
a signal generator and the output of the generator was a sine
wave with 2.0V p-p. This is a strong signal and may impact the
results, especially concerning sensitivity, but it makes for
easier measurements and ready comparison between sets. For
this set I made identical measurements with both the diode

78 199
The actual set with its controls, taps, tuning and SEC shown
below:

Silicon diodes have good characteristics, but an unacceptably

high forward voltage drop making them a very poor choice for
crystal radio unless used with bias. The 1N4736A is a Zenner
diode.

Spreadsheet for calculation of Silicon diode n and Is (modified

from Mike Tuggle's spreadsheet)

198 79
 Additionally I purchased a small variety of tubes for testing.
The results of this study are discussed in full under the Radio
Lab : Diode Test section of my page. Of all the tubes tested, I
finally chose the Russian 6C19P as the tube having the most
sensitive characteristic and a handsome appearnce. The tube is
a power-hungry beast though, (1.1 Amp current draw at its
rated 6.3 V), and this radio requires a power supply, but its
worth it! The following graph plots the characteristics of seven
tubes I tested for the project.

While the project radio was to feature the Fleming valve, I

Averages: n = 1.52 Is = 2 nA Rd = 2.4E8 kOhm naturally wish to have an interesting circuit as well. As always,
it would be a breadboard set and single tuned for ease of use. I
The above spreadsheet is based on measurements of the diodes also wished to utilize and test Dave Schmarder's selectivity
shown above. For the determination of Is and n, I chose at enhancement circuit (SEC) in the set. This last item to
random two examples from my collection of various diodes, compensate for the typical poor selectivity in single-tuned
(all silicon in this case) and measured with a modified circuits. One circuit that caught my eye from an early time was
methodology to that outlined by Ben Tongue and Mike Modern Radio Labs' MRL#2 set. To this set I would add the
Tuggle. In this case, with the radically different forward SEC modification. In such configuration I notice that it bears a
voltage drop of these diodes from Germanium or Schottky remarkably close resemblance to Set #6 in W.J. Mays' Boys
diodes, I have kept the Id values constant (about Id2=0.5uA Book of Crystal Sets (BBCS). The MRL#8 set also shows
and Id1=1.0uA) and varied Vd. I have included the actual design affinities and includes the choke missing in MRL#2, a
room temperature in the calculations in order to eliminate this fact noted by Mr. Osterhoudt himself. In the end my circuit
variable as a source of doubt or error. All the measured design, shown below, must be considered a close, (but not
parameters as well as the determined results are presented in exact), member of a family of radio circuits consisting of
order to facilitate repeatability. MRL#2, MRL#8, and BBCS#6.

80 197
Most avenues of research proved futile until he recalled his Light Emitting Diodes
prior work on the Edison Effect. He literally retrieved the
original bulbs he had experimented on previously and set up a
test to apply the effect for radio wave rectification. His
experiments were certainly successful. This idea proved a
novel and unanticipated use of a prior-known effect amounting
to true discovery and thus was awarded British (24,850) and
American (803,684) patents in 1905, these successfully
defended against De Forest’s subsequent infringements.
Pondering this interesting history, I began wondering about the
Fleming diode and thinking, in crystal radio, this is not
amplification, so why not? Also, one must consider the glow,
oh the glow… I just had to build this.

Finding a correct tube for the project is a bit of a story in itself.

I have never worked with tubes and, while it appeared natural
that diode tubes ought to exist as well as triode/pentode/etc
(they do), I had no idea what tube was what nor how to know
what would work for radio rectification. Many diode tubes I
found are for
power supply work
and may not be
appropriate for
radio detection at
all, (actually they
work fine).
Jumping into the
search both feet, I
waded through the
datasheets of
many/most diode
tubes to get to
know them well
enough to choose.

196 81
 Powered by Fleming's amazing valve
http://www.lessmiths.com/~kjsmith/crystal/flm.shtml

Crystal radio circuits

come in a seemingly
endless variety of forms
and possibilities. That and
the historical context of
their importance at the
dawn of the radio-age
makes them a fascinating
hobby. I enjoy making
these radios, learning and exploring new and interesting circuit
ideas, incorporating vintage components into my sets, and
finding historical connections between the radios and the
fathers of the technology.

My latest project involves a

passive receiver rectifying the
signals with a vacuum diode.
Professor John Ambrose
Fleming had consulted for
Edison Electric in Britain and
had experimented extensively
with Edison two-electrode
bulbs in the 1880’s, presenting
As long as I am measuring various and sundry diodes, I figure his results on the “Edison
I ought to include that most ubiquitous of modern diode, the Effect” to the Royal Society of
LED. Found everywhere, these diodes are rapidly becoming London in 1889, and the Royal
the low-energy light source of choice for many lighting Institution in early 1890, and
applications. I have read occasionally of someone asking finally his definitive paper to
whether these ought to be useful for radio applications as well. the Physical Society of London in 1896. Later after the turn of
To this question the answer is generally a resounding "No!". the century he began consulting for Marconi seeking better
The turn-on voltage is waaay above any reasonable value mechanisms for radiotelegraphic detection of radio signals.

82 195
 expected to be delivered by an antenna to a crystal set. Still,
this simple answer avoids the actual question, what in fact does
the characteristic curve of a LED really look like? Where is the
turn-on voltage with respect to the published junction voltage,
(assuming you can find that).

To provide just such a look I visited my local electronics store

and bought a small handful of LED's, most with the junction
voltage listed and took them home to measure. Typical LED
junction voltages seem to range from about 1.8v to 2.1v or
more. The turn-on voltages look closer to 1.6v-1.7v. Anyone
used to working with Carborundum crystals or silicon diodes
will be used to biasing their rectifier to get good sensitivity.
These LED's once on have a very sharp rise and with a proper
bias should work quite well as detector diodes. As a bonus you
will get a sweet glow as well. OK, not as cool as the glow of a
vacuum diode, but certainly more sensitive!

Spreadsheet for calculation of Light Emitting Diode n and Is

(modified from Mike Tuggle's spreadsheet)

Averages: n = 2.01 Is = Tr nA Rd = 3.8E19 kOhm

194 83
The above spreadsheet is based on measurements of the diodes What I found then, was a farily sensitive but, to be truthful,
shown above. For the determination of Is and n, I chose at woefully selective set tuning from 6 to about 17-18 MHz. Not
random two examples from my collection of various diodes, bad for a single-tuned rig.
(all led's in this case) and measured with a modified
methodology to that outlined by Ben Tongue and Mike Final thoughts. Upon returning to my home base, I quickly
Tuggle. In this case, with the radically different forward hooked the set to my longwire to make first impressions of
voltage drop of these diodes from Germanium or Schottky actual reception. In midday I was able to pick up a spanish
diodes, I have kept the Id values constant (about Id2=0.5uA language format broadcast and two english language religous
and Id1=1.0uA) and varied Vd. I have included the actual format stations and a brief encounter with WWV. At night the
room temperature in the calculations in order to eliminate this bands are flooded with competing stations, success! Note that
variable as a source of doubt or error. All the measured all this with my handy dandy Radio Shack mini-amplifier.
parameters as well as the determined results are presented in When I connected a crystal earphone and used my poor
order to facilitate repeatability. hearing, there was nothing as expected. One pity is the
ongoing presence, even on this set, of KGOW 1.65MHz
Why know the values of Is, n, and Rd of your diodes? forever audible in the background. I consider the set a success
as I was not expecting selectivity in a single-tuned rig. Still,
I have been attempting to get a handle on the main diode with the bands dominated by religous and spanish broadcasts, I
parameters Is, n, and Rd and how they impact the operation of cant say it differs much from the local Houston AM fare. Oh
a crystal radio. My method of choice is to use a graphical well...
approach, plotting charts where variations between plotted
parameters become apparent and easy to visualize. I have
developed an interesting chart from the measurements above
that compares many diodes with respect to Is, n, and Rd, and to
understand how this fits into a matching situation with a fuzzy
indication of typical tank impedance limits.

Ultimately one seeks a diode with an Rd that matches with and

conjugate to the impedance it sees from the tank (+antenna). In
the following discussions I will start with the cross-plot of Is to
Rd and then proceed to a look at how received RF power plays
a part in the selection and why low Is and n are desired.

84 193
The plot above charts the received signal strength (p-p mV)
versus input frequency for the four possible switch
configurations (Detector on tap 1 or 2, Antenna on selective The plot above shows the relationship between diode Is and
[top of coil] or broad [on tap t1]. These four conditions were Rd. Rd is calculated via the equation Rd = VT * n / Is where
tested with the tuning cap fully meshed (low frequency) and VT is the thermal voltage = k * T / q = 0.0257V at room
fully open (high frequency) to give an idea of the actual tuning temperature and Is and n are from the measured diodes. k is
range as compared to the calculated range. All measurements Boltzmann's constant = 1.38E-23 J/K and q is the electron
were made with no load on the output, and the antenna trim charge = 1.609E-19 coulombs. On such a plot I can show lines
cap in the full meshed position. Other conditions not shown of constant n, and plot the values for individual diodes for
were also tested. which I have spent considerable effort to determine the
parameters Is, n and Rd (where Is is the diode saturation
The results indicate that placing the antenna low on the coil, on current, n is the ideality factor, and Rd is the diode resistance).
the T1 position gives the best sensitivity and selectivity in On the plot it should be evident how changes in n or Is impact
tuning. The detector, a 1N277 germanium diode placed high a diodes' Rd.
on the coil at T2 also gives dramatic increase in sensitivity.

192 85
Additionally I have plotted RF signal Rs as orange circles from
published coil unloaded Q (see my section on Coil Q). To
convert from Qu to Ql I have made the assumption that loaded
Q is about 1/5 unloaded Q for low-end sets while in
performance sets it may approach 1/2.5 the unloaded Q. This is
only an estimate then, but starting from actual data. To make
the conversion to Rs I simply plugged the data into the
standard equation R = 2pi*f*L*Q using a scaling factor to
divide down Qu to Ql from high Q (1400/2.5) to low Q
(100/5). This provides a nice visual display of the expected
range for Rs in many crystal sets. I plot the Rs against the
value of Ql used. Note that these data are for the case at about
1MHz and inductances between 200 and 300uH, as published.

Matching the diode to the tank is a matter of finding a diode

with both sensitive qualities (low n) and an Rd close to the Rs
presented to the diode (as per Ben Tongue's suggestions for
Peak and Square Law detection). Those diodes with Is's in the
100 - 250 nA (100-280 kohm for n=1.1) or so range have the
possibility to match while connected to the top of the tank coil
without the use of Q-lowering taps, while untapped matching Initial testing on the set was dismal. NO RECEPTION. There
to high-end high-performance big-Litz baskets will require followed many hours of double checking all wireing, solder
diodes with Is's in the 35 - 60 nA (500-800 kohm for n=1.1) joints, fidelity to the schematic, etc. Everything checked out,
range. Most "typical" germanium diodes have high Is values no good. Time to seek professional help, which happily
(>500, >60kohm) and will require a tap. Diodes such as FO- corresponded to my annual visit home. There my radio
215, BAT 46, 1N60 and GAZ51 can be matched to many tanks engineer brother took the radio and gave it his close attention,
without taps. For tanks with high quality Litz coils one will be puzzled over it as had I, and concluded a defective part
using HP 5082-2835 or 1SS98 diodes, generally 3 to 4 in somewhere. This he soon found in the tuning capacitor which
parallel to lower the Rd to the desired range. (Curiously, I have had had a connection slip off its set point. This easily
not found any diodes with Rd values in the 300 to 4500 kohm corrected, the radio seemed ready for testing. Unfortunately he
range.) Note that both resistance and impedance are frequency- did not have an antenna handy (he works in uhf+ research
specific so a match at one frequency will not remain matched stuff) so all I could do was hook the set to a signal generator
across the broadcast band. and test its performance, summarized below.

86 191
at 3 turns and 8 turns from the ground (center) of the coil. I
ended with 9 turns on the final set with testing giving me a 7.6
mH coil. The final schematic presented below.

The above is an interesting plot taken from the text discussion

in "Crystal Set Analysis" by Berthold Bosch. It presents
received signal strength across the tuned circuit for various
scenarios from threshold audibility to local blowtorch. This
plot should be considered a single example specific to his
location and antenna/ground system. In the text he describes
his antenna an an inverted L 43m long (140') and 10m high
(32'), an excellent antenna most of us do not have the real
estate to erect, but offsetting this is a poor ground with Rg =
210ohms. Given the offsetting conditions I would imagine this
is a good generic example of what received signal voltage and
power levels presented across the LC circuit to the diode are
likely to be. Bosch cited 40nV as the threshold of audible
detectability (impedance matched conditions with 16kohm RF
impedance / 4kohm DC phones) and for this plot I pushed it
back to 20nV as it gave a superior regression fit.

190 87
 Global DX'ter
http://www.lessmiths.com/~kjsmith/crystal/sw2.shtml

Finally, a shortwave set

comes off the design
bureau and out into the
world. This is a type of set
I have been planning and
pondering on for some
time. With many crystal
radio design possibilites
this will certainly not be
the only SW set I plan, but it is my first effort, my chance to
"do it" on shortwave. For a first effort I settled on a fairly
simple design, a single-tuned rig with some degree of control,
but not overly much. I took as my base a design by Rainer
Steinfuehr on his Crystal Reciever World site. This design
features a high Q spider coil with two taps, and adjustable
capacitance on both the antenna and tuning circuits. It made a
rather lovely small set indeed.

The specifics of the design were entirely my own and were

based on the frequency range I wished to target and the
specifications of the materials I had available for the project. In
partictular, I had available small good-quality 10 - 115 pF
variable caps for which my clever elder brother had lathed
This plot at left takes some actual diode I/V measurements lovely 1/4 inch shaft extenders for use with standard tuning
rather than the models presented below and plots them along knobs. My target spectrum was about 5 to 18 MHZ as that is
with the above RF voltage across the LC circuit versus RF where the dominant international BC action takes place. With
power in mW. Here one sees that pushing for very very low Vf luck that would also allow receiving WWV at 10 and 15MHz
(via high Is / low n combinations) pushes the limit of as well. Digging into the design spreadsheets, it seemed an
detectable signal power from the antenna. At such minute inductor with 7 to 9 mH would best deliver the goods. I made a
power levels I imagine the reverse leakage current becomes spider form from a circlar cut of 1/8 inch Delrin plastic and
significant and probably offsets the low operating point gains wound 11 turns of 18awg magnet wire, carefully soldering taps
sought.

88 189
 Crystal Galena and Pyrite Diodes
Performance summary of my
Bremer-Tully Set is so so with a
set Q (loaded) of 16, -3db
bandwidth of 68 kHz and a
sensitivity calculation only 13%,
not so great. I rather expected as
much as this is a single-tuned set
of very simple circuit design.
Compensating slightly is the fact that it tunes with a single dial
which is certainly more user-friendly. It does require a wave
trap to cut out powerhouse KGOW blazing away and flooding
the entire BC band on this set if left untrapped. Most of my
radios have higher sensitivity so this set definitely benefits
from my audio amplifier.

Bank winding in six banks: 4 wires / side, 10 windings per

bank, times 6 = 60 turns. Coil form 3 inches in diameter, space
wound 0.8 inches wide calculates to 376 uH, pretty close to my
measured 377 uH. Measurements: Primary B-P = 20 uH. The above curves demonstrate the wide variation in properties
Secondary F-G = 377 uH, G-T = 210 uH, T-F = 65 uH. I and qualities that can be found in natural crystals of galena or
know, 65 + 210 not equal 377, I measured it several times to pyrite, the two most common and best quality natural stones.
be sure! In each crystal test I first poked around the crystal for some
time to 1) determine a typical sensitivity for the crystal in
question and 2) to locate the best hot spot with which to test.
This turns out to be a non-trivial exercise on the diode test
setup. In a crystal radio one need only listen for the loudest
spot. With the test setup one needs test both the forward and
reverse current in order to determine if the whisker is on a hot
spot or not. Very tedious work! (In retrospect, if I were starting
over making a test jig, I would definately add a DPDT switch
to readily change between forward and reverse current
measurements. I'd probably toss in a rheostat for fine
adjustments as well).

188 89
For many crystals there are limited number of possible hot
spots, but these may be hot indeed. For most of my "Steel
Galena" samples (Tintic Utah, or Leadville Colorado) there are
numerous hot spots under virtually every place I touch the
probe, but in general the sensitivity is good to so so. These
crystals are very kind to work with in terms of finding spots
and avoiding frustration. Mirror galena on the other hand may
have quality hot spots, but any hot spots at all are rare and
frustratingly difficult to locate. Here my Philmore detector
crystal shines with an almost ideal "Galena" response. To
chase down this rabbit I purchased some lovely mirror galena
from Sweetwater Missouri. I broke off a few appropriate-size
chunks to pot in woods metal and test. At first I was very
excited with the high currents I was seeing at moderate
voltages. Figured I had struck gold. When these crystals failed
miserably to rectify anything in my radios, I re-measured
things in both forward and reverse directions. These crystals
obey Ohm's law and act like typical resistors, not suitable for
Circuit schematic for the set.
radio work at all.

For my pyrite crystals the work has been especially tedious

and frustrating. With one of the crystals one "hot spot"
alternated, entirely on its own, between hot and bad while I
was making the measurements. I would start over and over,
sometimes getting interesting readings then suddenly it would
drop to low values and I'd start over, back and forth. I present
this data as best as I have measured, and I don't intend to go
back! You see at least one of the crystals, my "China 1", (from
a lead/silver mine in Hunan) gave a sweet classic-looking
curve. More to the point, "ideal" curves for natural minerals,
are difficult to come by. Most crystals you use will be less than
ideal. The good news is that, while listening to your set,
poking about for a good spot is far easier than what I have

90 187
tune down to 500 kHz. The closest variable cap I have found is gone through to produce these curves. Your ear will take care
a cute little 2-gang 10-270pF unit. This is as low as I have of you!
found at the low end for a standard-configuration variable cap.
The goal then is to try and keep the circuit capacitance as low
as I can, which I translate into keeping the circuit as simple as
possible.
My circuit is further constrained by the coil itself. It is
essentially an "oatmeal box" type setup with an aperiodic small
inductance primary and large inductance secondary. Tuning
then must be single on the detector side only. My only option
is whether I connect the crystal/diode to the tap post or to the
top of the coil. I went with the tap, so shoot me! One
complication I maintained, as I do on all my sets, is to keep a
switch between using a crystal and cat's whisker or using a
diode. In this case I chose a Schottky 1N5819, the most
sensitive diode I have yet tested, (diode test). I have yet to test
just how high on the BC band I can tune, but KGOW at 1560
kHz sure gives me trouble.

In the photo I indicate groupings based on an easy measure of

performance. I note the current in milliamps for each set where
the plate voltage is set at 0.5V. The greater the current the
better are your chances to get a sensitive crystal, assuming
Ohm's Law is not followed! Crystals in the "dead zone" on the
left will have their woods metal re-melted for new detector
crystals and the bad ones tossed, its tough love for crystals. I
find that easily half the potted crystals I make are tossed this
way, and only a few can be considered superlative.

186 91
Vacuum Diodes A Bremer- Tully Transformer
http://www.lessmiths.com/~kjsmith/crystal/brt.shtml
When I decided early on to construct a radio based on the
vacuum diode (see my Fleming Radio section), I had to find a I recently won on eBay and
suitable tube for the project. That began with an extensive lovely NOS Bremer Tully
search through the datasheets checking vacuum diode air core transformer its
properties and reviewing the characteristic curves for a large inspection tag still
number of diode tubes. In the above figure I plot the faithfully wired to the base,
characteristics as best I can on a commpn plot for comparison. (and so it will remain on
I was hoping to find a good candidate with sensitive the final radio). It is
characteristics and low energy consumption. As you will find, absolutly gorgeous and I
such a beast did not exist. Following this research began a was the only bidder when
period ot purchase and testing. it sold, for such a great price! Honestly, I was not totally sure
what I was getting when I bid other than it looked like a coil
I have tested seven different tube types including 6.3V rectifier and, from the ebay photo, maybe would work for shortwaves.
diodes that take a lot of power to run and are not really suitable Calling it a transformer was confusing although I understand
for battery use, a 9v dual diode tube (20D1), a pentode/diode all coupled coils are transformers. Perhaps that is why I was
tube (1S5) that is designed to run on a battery at 1.5V and the only bidder, well, so much the better! When it arrived I got
50mA, and a couple miscellaneous but cute tubes. In testing a good look, very interesting winding technique. The tuning
the 1S5, I found the filament never glowed incandescant at coil is wound with somthing like #30 dcc or dsc wire. The
1.5V and, on measuring, was barely sensitive to anything. I winding is in 6 bank-wound groups of about 10 windings each.
cannot imagine this tube would make much of a diode for Inside is a small primary antenna coil. I promptly got out my
crystal radio use. I tested different manufacture 1S5 tubes from handy-dandy inductance meter and measured 377 uH on the
two different suppliers. No dice. Finally, I tried to push the tuning and 20 uH on the primary. In addition, there is a tap on
tube to operate at higher-than-specified voltages. At 4.3V the the tuning coil one "stack" in. 377 uH is a bit on the high side
I-V curve was still very flat, but shifted slightly higher, for BC work and I'm not sure I'll ever get the top end of the
crossing the Amp axis at about 0.4 mA, (see graph). At 5.5V band but still, the transformer works beautifully.
one of my tubes gave up the ghost and I didnt do more. I pretty
much rule the 1S5 out for my crystal radio work. Looks like i'll My calculations for a setup using a 377uH coil suggest that a
be needing the power supply when I get around to minimum total capacitance of 23 pF will allow tuning up to
building/operating that Fleming Radio. 1700kHz. This is a tall order, most variable capacitors I have
found have a minimum value in the 20 - 25pF range. Add
A question remains as to what in fact is different about these another 15 - 25pF of stray circuit capacitance and I am way
tubes to give such different characteristics. While the I/V over the line. The good news is that only 270pF is needed to

92 185
 characteristic of most vacuum diodes pretty much follow the
Langmuir-Child Law, the steepness of the I/V characteristic is
largely due to the geometry of the tube elements and the
volume of electron space-charge between them. JB Calvert's
Theory of Vacuum Tubes informs one that this is the property
called Perveance in a vacuum diode. Measuring this requires
plotting I raised to the 2/3 power against V and taking the
slope of the best-fit line through the data. This slope raised to
the 3/2 power is the perveance. The following chart illustrates
graphically the relation bewtween the tube characteristic and
perveance. For the 5726 tube I plot the characteristic in dark
blue for the tube heated with the design 6.3 volts. In light blue
I have plotted the measurements and raised them to the 2/3
power. One sees immediately that the new curve is linear. The
slope of the curve raised to the 3/2 power yields the perveance
factor for the tube.
(Note in addition that when running the tube at a more
sensitive point with 4 volts only on the heater the perveance is
slightly less than spec.).

184 93
 however, is NOT the inventor of the oscillating and
regenerative feedback circuit, which forms the basis of all
radio amplification and truly opened radio to a mass mode of
communication. This honor goes to E.H. Armstrong, the man
who dug into the tube and worked out correctly for the first
time its basic theory and operation. The two men fought a
bitter 19-year litigation battle over this honor that in the end
was awarded to De Forest by the Supreme Court in what
amounts to a technicality.

From what I have read of De Forest’s writings, patents, and

exchanges with Armstrong in the Proceedings of the Society of
Radio Engineers, it is pretty clear that De Forest never really
understood how his tubes worked. So, he was never in a
position to develop the tubes to their full potential. Armstrong
for his part seems to have treated De Forest with academic
contempt. Armstrong went on to invent the super-heterodyne
(1918) and super-regenerative (1922) amplifier circuits and
created almost single-handedly FM radio (1933) as we know it
today. His life ended in tragedy, representing the closing
chapter in the age of great individual inventors, finally crushed
by the greed of corporate America. Virtually all advances
With my measurements calculating the perveance is practical following his were the product of team efforts in company or
and quickly done, results as follows: government research laboratories.

Tube Perveance R2 If therefore you have misgivings as to my honoring De Forest

6C19P 2.54 0.997 with a radio, please understand I do so not for his Supreme
5726 2.19 0.998 Court victory but to honor the inventor of the triode tube, a
6DN3 5.40 0.997 debt which even Armstrong would always acknowledge.
20D1 2.02 0.985 Should you still feel uncertain, then just enjoy the sweet irony
5C12P 0.32 0.999 in honoring the father of the triode tube with a crystal radio.
1S5 0.063 0.999
2D1S 0.155 0.975
6G2 0.050 0.999

94 183
De Forest circuit diagram (aka Sleeper #18). The perveance numbers above range from 0.02 to over 2
mA/V^3/2 (with >2 being good for small-signal detection) and
Performance clearly shows my choice of the 6C19P to be an excellent one. I
summary of am surprised to see a huge 5.4 for the 6DN3 diode, a color
my DeForest television damping diode. It takes a Novar 9-pin socket and is
is really just a large tube so doing more with this tube may have to wait a
abysmal, it bit.
has a low set
QL of 7 Calvert's 2001 published data
reflecting a huge bandwidth of 165 KHz but my sensitivity
calculation is 37% for the condition of tight coupling. This set Tube Perveance
clearly trades Q and bandwidth for sensitivity. The set is 6AL5 2.42
technically double tuned with variable capacitance in the 6H6 0.50
detector circuit and variable inductance via taps on the antenna 7Y4 0.58
circuit. The inductors are loosely coupled via rotating the inner 2X2A 0.017
inductor with respect to the outer. What I discover in fact is 6V3-A 2.3
that rotating the inductor changes the tuning and so it "feels" 6AX4GT 1.42
more like a variometer rather than a variocoupler. Rotating the 6AV6 0.085
rotor 30 degrees to loosen the coupling does not narrow the 1A3 0.075
bandwidth, in fact it increases to 395 KHz. I have much to test
still on this set, it is a disappointment as I expected a much The high-perveance diode tube types I tested show an
better QL. Measurements were made with the detector circuit interesting property in that their characteristic curves do not
farily optimized for a test frequency of about 1100 kHz, but I pass through the origin of the graph. These curves are wholly
wonder if I can better optimize the tap settings for the antenna acceptable as crystal radio detectors although they will require
circuit. More to come. power to operate. What I have noticed with such tubes is that
when I turn them off after using them with a radio, on cooling
What’s in a name? they go through a period of much increased sensitivity (very
loud) before fading to nothing. It is as though running them at
The name De Forest in the history the full 6.3V lowers their full potential as crystal radio
of radio is one few engineers can rectifiers. The reason is in the high perveance of the tubes, the
hear without a fairly strong anode is already proximate to the space charge before any
opinion. Dr. Lee De Forest is the plate voltage is applied. Effectively, while the plate (anode)
acknowledged inventor of the "Zero" voltage is measured with respect to ground, the plate
triode radio tube. De Forest itself is still positive with respect to the cathodic space charge.

182 95
Small currents will continue to flow even with a negative bias
to the plate, (Contact potential).

In order to explore this idea, I ran a series of tests with the

tubes running at lower operating voltages (effectively
diminishing the space charge thus lowering the perveance) as
shown:

View from behind with my cat's whisker crystal detector,

SPDT knife and 1N34A diode.

Here you see the impact of lower operating voltages. With the
tube operating between 3 and 4 volts the characteristic curves
begin to pass through the origin as in regular solid-state diodes
and crystals. It is precisely in this voltage range that I found
the radio sensitivity, as measured by loudness to my ear, to be
most pronounced. It appears that the diode characteristic needs
to pass through the origin of the I-V graph for highest
sensitivity. This takes me to the 20D1 tube which operates at
9v but only 200mA. This will still take a power supply to use,
but the good thing is that the characteristic curve passes

96 181
 through the origin. Running at its design parameters the tube
looks a lot like the 6C19P at 4v.

View from above highlighting the tap wiring, SO much work!

180 97
Photo of the lab with my diode test jig, amp meter, and diode on each switch. The wood knobs have a nice easy action, the
vacuum tubes under test. switch passing readily between contact buttons.

The radio works quite well, in its tightly-coupled setting it is

quite sensitive with a strong signal. Its selectivity on the other
hand leaves much to be desired. This can be expected from the
circuit layout with only the secondary circuit capacitively
tuned and the detector tied at the top. I was hoping that by
varying the coupling I could achieve some increase in
selectivity, but it de-tunes the station with an additional
unacceptable loss of sensitivity. Take this outside the urban
environment with such crowded airwaves and the set ought to
perform admirably. Signals come in strong with the steel
galena nearly as well as with the diode. I like this set.

Portrait of set, ummmm..

98 179
 RADIO TEST
As most of the components for this set were ready made, the Kevin Smith
construction planning felt a bit much like assembling a kit. I http://www.lessmiths.com/~kjsmith/crystal/rtest.shtml
had originally intended to use modern tap-selector switches as
well. In order to make the construction challenging, and more
importantly, to maintain the vintage look and feel of the set, I OK, now that you have built a radio, or several, how do they
quickly abandoned the modern switches and chose to fabricate perform? Without a bit of objective testing you will never
instead classical wiper switches in the early style. Of course, I really know. I have often read descriptions of radio
also include my own style cat’s whisker and SPDT knife performance such as "tunes very sharply" or "separates two
switching to a more modern diode. The diode is my only nod closely spaced stations" and/or etc. These are qualitative
to post war technology. In the crystal cup I have some lovely descriptions that do not really give you much information. To
Utah steel galena potted in wood’s metal, finely crystalline, hot know your radio, it takes some measuring. I am making the
as hell all over its surface. assumption here that, as a homebrew radio builder, you will
find tinkering and measuring your sets just as interesting as
As one can clearly see, my obvious preference is for the their design and construction.
breadboard style of early radio. Much of the fascination and
beauty of these sets lies in the component hardware. To For testing, I have pretty much lifted the excellent procedure
enclose them behind panels or put them in boxes seems such a outlined by Charles Lauter with a few modifications to suit my
waste. The base is a box with about ¼ inch of space inside to own philosophy. I will give my own protocol but first let’s
hide the bolts and wires. I did need a small panel to place the look at the needed equipment and test setup. Testing a crystal
wiper switches on. This is made of 1/8 inch black Delrin set requires two pieces of homemade equipment, a dummy
plastic resin, a very hard and nice to work with material. In antenna in front of the set which allows measurement points,
fact though, it is a bit too shiny, too “plasticy” for my taste. and a measurement load at the back end. Other pieces of
Well, that is what I had. equipment include a signal generator, an rms voltmeter, a
resistance meter, and a good digital voltmeter. Circuit
Fabricating the wiper switches, buttons, and leads to the schematics for the homebrew components are shown below.
inductor was an odyssey in its own right, amply satisfying my
desire for more challenge. Attaching the leads to the primary,
soldering without damaging the windings, and the added touch
of shrink-wrap “mittens” was a labor of love. I do hope it is
justified by the final presentation. The rest of the construction
and wiring was a piece of cake by comparison. The wiper
switch knobs are wood with a felt buffer against the panel.
Behind the panel compression springs keep a steady pressure

178 99
 De Forest, Vintage beauty
http://www.lessmiths.com/~kjsmith/crystal/deforest.shtml

My latest effort, I call this

my “DeForest” after the
principal component of the
radio. Over the last few
months of visiting ebay and
finding essential radio parts
and browsing, I have
stumbled across an
For the tests I originally found an old analog signal generator occasional vintage piece
on ebay which I sadly found to be wholly inadequate. The unit that I just loved and needed to bid aggressively on. My intent
was by no means cheap but the dial was difficult to read with is not to just stash these in a drawer somewhere for the
any precision and the "play" in the dial made readings pleasure of owning, but to use them in projects. This project is
hopelessly inaccurate. Finally, it gave no indication of the built around a lovely De Forest Radio Telephone and
attenuation or voltage output levels, rms or pp, of the signal. I Telegraph Company Vario-Coupler. The vario-coupler, wound
have since found a nice, reasonably-priced new digital unit, in beautiful apple-green silk covered wire 24 gauge or so, has a
(max's out at 2Mhz so strictly for AM band work) that gives tapped primary stator measured at 560 uH and an untapped
good readings, gives me the output voltage level in pp, and can internal rotor of 285 uH. Other vintage components in this set
even produce different wave forms. I recommend if you are include a Windham square-law tuning variable capacitor, 20 to
seeking to equip your lab, go this way from the start! 550 pF, and an Atwater-Kent 3 ¼ inch tuning knob. The
second 3 inch knob on the vario-coupler is a vintage Radion
To calculate the signal input power you will need a good rms Dial made by the American Hard Rubber Co.
voltmeter. This is a non-trivial item and, after my experience
with the vintage signal generator I was leery of getting a used I built the radio according to Sleeper hook-up #18. It is a fairly
meter. In any case, an expensive dedicated single-use piece of straightforward hookup but with sophisticated features
equipment seemed extravagant. For this measurement I have including a double-tuned design, a tapped primary tuned via
found entirely acceptable results using an oscilloscope which, single and multiple taps, and variable coupling. The secondary
of course, is useful for so much more and belongs in your lab coil is tuned with a vintage variable capacitor. As the
regardless! secondary coil is the rotor inside the primary, there is no
provision for tapping and the detector circuit is connected to
It is on the back end of the measurement where I diverge from the top of the coil. This set could use some selectivity
Lauter's protocol significantly. He recommends placing in all enhancement circuitry I suppose, but oh well...

100 177
 radios to be tested a 1N34A diode and then use a standard 2k
ohm load resistance to make the output measurements. This
may give a standard and comparable result, but it does not
insure a good impedance match and does not reflect the
intended usage and setup of many crystal sets, especially sets
intended for high-performance. Here I am interested to learn
just what is the optimal load resistance for my set under the
test conditions. Note the caveat "under the test conditions".
This may not reflect actual usage unless you take precautions.
You will want to set the generator to an input voltage similar
to what you expect the radio to receive in actual operation.
You will probably want to signal level work the diode in its
peak-detection region but not so high as to saturate the diode,
(In making a loaded Q measurement you want to make sure
that the diode impedance stays constant; otherwise, the change
in diode impedance (as the tank attenuates the signal before the
diode) will change the load on the tank and the bandwidth
measurement won't really reflect the Q of the tank). Finally,
for the output voltage you need a good voltmeter with very
high input impedance that will not load down the set under
test. This really means a good DVM. I use a used Keithley 192
bench meter with a 2M ohm input impedance, and as a backup
a used Keithley 180 bench electrometer. Overkill perhaps, but
for just in case...

Getting back to the load resistance then, my protocol is to find

the optimum resistance to deliver the maximum power (not
voltage!) to the load. This is a bit of a tedious process but will
amply rewarded. Having tuned and peaked the set to the
measurement frequency, I then take a series of
resistance/voltage readings to find the maximum power output.
In my circuit this is accomplished with the variable 1M pot.
The resistance meter and voltmeter cannot be connected to the
load at the same time making this a tiresome road to follow.

176 101
Once I find the optimum load resistance, I set the pot to that
and proceed to the main testing for sensitivity and
bandwidth/Q measurements.

Test Protocol as follows:

1) Connect the set to be tested between the dummy antenna

and measurement load as per the above diagram.
2) Connect the signal generator to the dummy antenna and set
the frequency and voltage output. I typically test in the
center of the BCB at 1100khz and use a 200mVpp output
sine wave.
3) Set the load resistance to somewhere around 50k ohms. The
exact value is not critical as it will be adjusted later.
4) Connect the DVM and tune the set to peak the output
voltage. Care here is amply rewarded, especially with
double-tuned sets.
5) Once peaked it is necessary to adjust the generator
My homebrew in its Galenatron
frequency to re-peak the output voltage as hand-
configuration shows fair Q at 53
capacitance or other factors may have prevented perfect
and poor-fair sensitivity of 15%.
tuning. You now have your set properly tuned for
There is a 1cm separation between
maximum output. Presumably this also means the best
the antenna coil and detector coil
possible impedance match between the set and dummy
which may help. In many radios of
antenna. This is the resonant frequency (f res).
this design I see the coils separated
6) Record the frequency and output voltage.
by 2-3mm only. As I wound the
7) Now find the optimum load resistance. This is done by
coils on a single form, there is no experimenting here, just
making a series of paired measurements of load resistance
wondering. The variocoupler here is only a trap and was not
and output voltage. Power in uW = mVout * Rl khoms.
used in this measurement. With the return to a single-tuned set
8) Set the load resistor pot to the value where power output is
you need expect performance downgrade, but greater ease of
highest, this is the optimum load for the test. Record this.
use for general listening. I feel this configuration is a good
9) With the signal generator re-peak the output voltage and re-
compromise for easy listening without the hassel fo
record the resonant frequency (f res) and output voltage
manipulating two (or more) tuning dials. It is clearly superior
(Vout mV).
to the Dunwoody and De Forest single-tuned sets.
10) Multiply mVout by 0.707 to find the -3dB level.

102 175
general listener radio and far far from the performance 11) Adjust the frequency above and below f res until the output
achievable in high-quality DX designs. But it is teaching me as level falls to the level calculated in step10 above. These
intended. frequencies are f high and f low. Record these.
12) The -3dB bandwidth = f high - f low. Loaded Q (QL) = f
Inspired by Dan Petersen's Galenatron, I wished to play with a res / BW
vario-coupler design. This hookup uses a single form with an 13) Take the generator output peak-peak voltage recorded in
antenna coupling coil of 10 turns and a tapped secondary coil step 2 above and convert it to rms voltage as follows:
with 62 turns. The first 2-gang vcap is unused. The single-gang mVrms = mVpp / 2.829. This is Lauter’s RF voltage E1
variable cap tunes the rotor. So far the best position seems to mV.
be keeping the rotor in-line with the main coil. More learning 14) Measure the RF voltage across the 10 ohm resistor on the
to come. dummy antenna. This is Lauter’s RF voltage E2 mV.
15) Input current to the dummy antenna = Lin = E2/10 (mA)
16) Input power to the dummy antenna = Pin = E1 * Lin (uW)
17) Power loss in the dummy antenna = Pda = Lin^2 * 25 (the
dummy antenna resistance)
18) Power delivered to set = Px = Pin - Pda uW
19) Set input resistance = Rx ohms = Ex / Lin where Ex = E1 -
(Lin * 25)
20) Set output power (uW) = Pout = Vout^2 / Rl in ohms
21) Set % efficiency (sensitivity) = 100 * Pout / Pin

Having built a fair number of sets, and with varying quality, all
the above protocol was developed to test my sets. What
follows now is a discussion of how my sets stack up against
each other and some notes on what one may expect in a crystal
radio performance-wise. I give a summary table below of the
essential data gathered on each of my sets.

Table 1

174 103
The above table of measurements were made at a nominal
frequency of 1100khz and input voltage of 0.2Vpp as indicated
on a digital signal generator following the protocol outlined. I
give the radio name under test should you wish to refer to my
pages for more on each set. The columns as follows:

Vin = 0.2V pp (0.71mV rms) Here I show the performance summary for my set with three
Pin = input power in uW into the set (after the dummy coil separations, 11cm, 13cm and 18cm. With these charts you
antenna) see the value and trade-offs in loose coupling between the
Pout= output power across the load resistor in uW antenna and detector circuits. With the coils at 11cm
% Eff = 100 * Pin/Pout separation they are slightly over-coupled but still with
BW = bandwidth in khz at -3db relatively good Q, 68, and fair sensitivity at 22%. Separating
Ql = calculated loaded Q of the set the coils to 13cm and then to 18cm loosens the coupling
Rx = the set input resistance presented to the antenna allowing narrower bandwidths and higher Q (getting to 125,
Rl = measured optimum load resistance for maximum power woohoo!), but at the sacrifice of much sensivitity, down to
transfer 15%, still reasonable. This is an excellent performance, I must
Rd = Diode junction resistance Rd used in the set be doing something right somewhere. Of course this is still a

104 173
 From the data above some interesting observations can be
made.

Figure 1.

In figure 1 above I show the relation of the input power versus

the set resistance. The quality of the fit reflects that fact that
this is largely a calculated value, but it does illustrate the
importance of having a low input impedance to match that of
the antenna. This is not often considered in discussions on
impedance matching but is critically important. The antenna
has an impedance in the 25 - 50 ohm range, not easy to get this
in a set without quality parts and construction. My sets
typically have an input impedance in the low 100’s ohms and
only two of my sets really get down to the 50-60 ohm range
giving a good match to the antenna. Both of these are double-
tuned sets with a tuggle front end.

172 105
If you are serious about getting DX and doing a lot of heavy My second homemade, A double-tuned open test bed
lifting, you will need a good antenna tuner as part of your set http://www.lessmiths.com/~kjsmith/crystal/homebrew.shtml
design.
Having learned my lesson
Figure 2. with the Mystery set, I
have set myself to make a
double-tuned rig with
open wireing for testing
multiple configurations.
The concept is to seek the
maximum amount of
flexibility in hookup and
coil design. This will mean interchangeable coils with or
without taps, and wireing with allegator clips. I have chosen to
hard place in the set three capacitors, 2 double gang 20 - 460
pF and one 20 - 410 single unit. The detector, as always, will
switch between a diode and my own crystal stand. The
modular design will allow testing coils mainly although vcaps
will also easily be tested with jumpers. The number of leads
and allegator clips, not all used with any one configuration will
give this a bit messy look, but this is not a beauty contest but
for learning. As always I include the option of using either a
At the back end of the set things get a bit messier and are not germanium diode or actual crystal.
east to show with a simple two-component plot. Here I have
chosen to highlight the importance of the proper load The photos and circuits below show two hookups currently
resistance on the set loaded Q. In my set testing I have tried to completed.
pay close attention to the importance of the load on the tank
and diode. Early on I considered construction of an output Double tuned circuit with tapped coils and an optional QRM
transformer unit that would work with different radio’s as well trap (unused in photo). This configuration alone will keep me
as with different phones. Such a design can be found on Ben busy quite a while! I am learning what "sharp" tuning is. The
Tongue’s excellent and highly technical web site. The plans antenna coil is tunes series-parallel but the first, (left) 2-gang
were well beyond what I was able/willing to take on so I vcap while I only use one gang of the second vcap. For lower
turned my attention to discovering just what load resistance inductance coils I can use the second gang in parallel.
exactly do my sets require for maximum power transfer. After
all, if I found only a small range typical values, I might design

106 171
 a generic transformer that would work well with many or all
minus the complicated switching.

In my testing protocol
therefore, I included the
important step of
determining the optimum
load resistance for the set
under test conditions. I
certainly part company
with the school of thought
that says to test all sets
exactly the same way, including same diode (1N34A typically)
and same test load. Sets are designed to work best when
impedance matched and at the back end that generally means a
load matched to the tank/diode combination that delivers the
maxumum power. At right above I show a small plot where I
varied the load resistance over a wide range and calculated the
output power of the set. (I give the plot with both a linear and
a more appropriate log scale). This should demonstrate the
importance of optimizing the load resistance.

Optimum load resistance depends on both the tank Q and that

portion of the diode detecting the signal. That in turn depends
on the signal level. Very weak signals will be rectified in the
square-law region of the diode characteristic and require
relatively high (and variable) load resistances. Normal to
strong signals will be rectified primarily along the peak-
detection portion of the diode characteristic and require lower
load resistance with less variation. Not shown in the above
spreadsheet, I also ran measurements with a 2.0Vpp input.
With this stronger signal, rectification was taking place farther
out on the diode characteristic. As expected, the measured

170 107
optimal load resistances for this case was similar to or lower
than that for the 0.2Vpp input case.

With all that said, the plot in figure 2 is messy but the
impression is that of increasing Q with increasing Rl. I view
the plot as having a number of small groupings of
measurements with similar conditions, each showing higher
Q’s associated with higher load resistance.

Figure 3.

Figure 3 is most obvious with the clear benefit of good set

efficiency on output power. An efficient set will better
approach the maximum power transfer desired. High Q plays a
part here. Still, there is always the tradeoff between selectivity
(high Q) and sensitivity. An efficient set delivers more power
to the load, but to go after Q you will ultimately be sacrificing
sensitivity.

108 169
 Figure 4.

Performance summary for the

radio. My Mystery set, despite
initial mis-calculations on the
coil winding specs, turns out to
be a fair performer. the Here is the bottom line, Q verses Sensitivity, efficiency in this
bandwidth around a nominal case. On this plot it is readily evident that I have built my fair
1100Khz is 43Khz giving a set Q share of loser radios, but also a good number of sets which
of 26. This result is superior to push boundary of what can be achieved in terms of
both my DeForest and my kit Dunwoody, not bad for a first compromise between Q and sensitivity. Looking at the plot it
try! Sensitivity comes in at 19%, second after the DeForest. is fairly easy to imagine a sloping line or zone running from 0
So, I consider this a successful entry into the world of scratch- Q at 60% efficiency down to 120-130 Q at 0% efficiency.
built crystal radios. Beyond this one will probably never go with tank and antenna
tuner alone. It will be time to start adding traps to your bag of
tricks. Recall that the above is for the 1100khs frequency and
that Q is a function of frequency.

In order to explore this final conclusion, I have taken a look at

Dave Schmarder’s excellent bandwidth analysis of his #50
contest set. This set is a superb work of craftsmanship and

168 109
skill, experience and usage of the best ($$$) materials More study, more puzzling over the problem, more
available. I take this to be just about the most of what can calculations, more frustration, more more study.. Did I say
wrung out of the sky with a crystal radio. Dave’s protocol is that? Short of complete disassembly, which I was loath to do, I
rather different from my own and he did not make a specific - was unsure where to go with it. Reading all I could, I came
3dB bandwidth measurement so what I do from here on takes across a set with a modification with a second cap tuning the
many liberties with his work, my sincere apologies in advance! antenna, this I could test! My antenna, by the way, is a 100ft
14g stranded insulated copper wire with an additional 30ft or
Figure 5. so of lead-in. Tuning the antenna seemed promising and when
placing a cap the set suddenly found its voice! With a working
radio, I had the dilemma of what to do with the case. No
second cap would fit. The large vintage capacitor could easily
be replaced by two small modern caps though and testing
showed that in reality they tuned very effectively with a dual-
gang capacitor which is what I settled for in the end. The set
still dosnt tun below about 700 Khz or so. Posing the question
on the Xtal Set Societie's excellent discussion forum Rap'n Tap
brought elightenment from Golfguru who wisely asked the
length of my winding, pointing out that the inductance of the
coil should be calculated from the total length of the windings,
INCLUDING the bifilar primary. Running back to Professor
Coyle with a 3.27 inch length gave me 112 uH and a lower
tuning range of 660 kH or so. So, "Mystery" solved...

The following is a final schematic of the set and photos of the

set as it exists. I listen to it most evenings as I work.

110 167
cap, all displayed on a breadboard box with wireing hidden
inside. In figure 5 above I have attempted to cast Dave’s
measurements into something similar to what I am doing,
The risk here was inexperience and it certainly came home! which is looking at the -3db BW and Q at 1100khz. I finally
My 18 AWG wire turns per inch were rather different from produced the above spreadsheet where I simply smoothed the
values assumed in the "Professor Coyle" spreadsheet, both data for the 1000khz case and estimated the bandwidth at -
because the sheet assumes bare wire and because my ability to 3dB. This makes a LOT of assumptions of course. Mostly I
close wind is not that great! My first attempt was for assume that the +-10khz measurements are well above the
experience, but not for getting a useful coil. Rethinking and resonance skirt so my smooth curve is not unreasonable.
adjusting allowed success, almost, on the second attempt and I
got 46 out of a planned 48 turns on the main coil. Assembly The #50 set then appears to have, at 1000khz, a -3db
was difficult but in the end I had a lovely set that wouldn't bandwidth of around 8-11khz depending on diode and/or
bring in anything! Well, almost nothing, some weak interfering spacing between the tank and antenna unit giving loaded Q's of
stations at the cap's full-open position. My first photo is of this 91-125. These numbers fall in the range of what my own data
incarnation. lead me to expect. Ken Kuhn's page on Resonant Circuit
suggests the "typical" QL at 1Mhz ranges between 20 and 100
and Dave’s #50 set is certainly beating the best of that.

In the final figure 6 I have taken my estimated results above

and looked at how Q and peak voltage (assuming constant
conditions other than the spacing between the antenna and
detector units so this ought to reflect power as well). It is quite
apparent that the set is rapidly losing output power while Q
improvements are flattening out as the spacing between tank
and antenna tuner grows to 18 inches. It would appear that a -
3db loaded Q of 125 is near the limit of what can be expected
in a crystal set.

166 111
Figure 6. My first homemade, A Mystery Set
http://www.lessmiths.com/~kjsmith/crystal/mystery.shtml

Having made the

Dunwoody and experienced
the thrill of listening to the
local broadcasts on a radio
of my own, unpowered save
for what the stations
themselves provide, I was
thoroughly hooked on
crystal radio. I hit the web
for all I could find, which is considerable. My many thanks to
all of you who have provided such useful and informative
pages. I categorize my effort along two axes, 1) finding and
reading sites with background theory and advice on crystal
radio, and 2) sites of crystal radio builders to know what is
possible and find inspiration. I was determined to produce a set
of my own from scratch. Of all the possibilities, and there are
SO MANY, I soon settled on the Mystery Set as an interesting
and challenging goal. I have the impression that many, perhaps
most of the experienced builders who make their presence felt
on the web have built this, or modifications of this set. A
worthy place to start, worthy company.

My intent, after studying so many variations of this set was to

settle on the original set as published in the 1932 Brisbane
Sunday Mail. If I can do well, then only can I consider
modifications. My plan was to take the basic circuit, but
otherwise use my own forms and wire, calculating appropriate
inductor dimentions and winding specific for my needs. I
sought a square coil with large dimentions, 8cm mailing tube
wound with 18 AWG magnet wire, switchable detector
between a diode and homebrew, and a lovely vintage 500pF

112 165
 My library of Crystal Radio Hookups and theory...
http://www.lessmiths.com/~kjsmith/crystal/catalog.shtml

This section of my page started out as a place to put my

compilation of crystal radio circuits. That compilation became
a catalog featuring circuit diagrams hand-drawn by myself
from various sources (as indicated on each page, see catalog
reference). When I first entered the crystal radio hobby, I was
amazed and more than a bit intimidated by the sheer variety of
circuit possibilities. I began to make my own circuit drawings
in order to get a handle on this in addition to finding
interesting circuits to construct myself.

To organize the circuits from fairly easy to rather complicated,

I put together a simple classification based on the number of
Inductors and Tuning Capacitors in each circuit and a few of
their main distinguishing features. Better classifications may
exist, but this works well enough for my purpose. A legend
follows so anyone may break the code. The catalog provides
only minimal information about each circuit. Although an
experienced builder may make the calculations to build a radio
from the data provided, I mostly intend this for information
only. I highly recommend if you wish build a set to consult the
original reference.

Later as my experience and familiarity with crystal radio has

grown, and after having built and tested a certain number of
sets, more detailed and technical information has become
necessary. Scouring the web has turfed up an amazing number
of extremely well written and helpful web resources. While the
number of excellent web sites is large, there have been a
certain few that I found myself returning to time and again.
From these I would print out large sections and read carefully,

164 113
scribbling notes on the side, generating more questions and
learning, learning, always learning.

Eventually I began to assemble what were, to me, the most

useful and informative pages into a handy book format. I
decided to print the darned thing out to keep as a sort of
Crystal Radio Handbook, something to take with me on the go,
or just to read on the couch. I am NOT a "kindle" type.. The
project was no trivial thing. I knew what I wanted, but casting
everything into a common (more or less) format, organizing
things, formatting figures, etc, it was a lot of work. For my
handbooks, a glance at the table of contents shows a book
organized to lead the reader along the path I followed. The
authors are many and I wish to thank 10,000 times all the
people who have contributed so much to this hobby on the
web. My apologies if I did not choose your pages for the book,
the choices of what to include and what to leave out was Performance summary of my
purely my own. The final Handbooks, there are three now, I Dunwoody is so so with a set Q
feel represent university-level material from advanced (loaded) of 18 but my sensitivity
introductory to university senior / graduate presentations on calculation is 17%, not so great.
Crystal Radio. Published handbooks on crystal radio are This may be all that can be wrung
traditionally directed primarily to the "boy scientist" out of a single-coil radio, who
adolescent, introductory in nature. I hope this these handbooks knows? Most of my radio's have
fill a gap rather than pander. higher sensitivity so this set
definitely benefits from my audio amplifier.
Finally, to help as reference, I eventually formatted two
particularly interesting web pages as complete volumes in
themselves. Ben Tongue's pages are essential reading for the
serious hobbiest, even if over my head, (read.. waaaaaaay over
my head). At nearly 400 pages, I have broken it into two
volumes. Kenneth Khun's web book is just great, technical, but
not overly, well explained as one would expect from a
professor. Thanks for your efforts and for all those others
whose pages I missed.

114 163
 I intend these documents to be printed on double-side letter
paper, folded over or cut in half and then bound. The foldbook
paging system in the document only makes sense when this is
done.

Crystal Radio Handbook: V1 Catalog of Crystal Radio Hook-

Ups (ready for printing and binding..129pp)
pdf format, 15.7 meg size

Crystal Radio Handbook: V2 Crystal Radio Theory A

Handbook of useful technical articles, (Bookfold format for
printing, 262pp).
pdf format, 2.21 meg size

Crystal Radio Handbook: V3 Solid-State and Vacuum Diodes

A Handbook of useful technical articles, (Bookfold format for
printing, 268pp).
pdf format, 4.47 meg size
Circuit schematic for the set.
Crystal Radio Handbook: V4 Antenna / Ground and Antenna
Tuning A Handbook of useful technical articles, (Bookfold
format for printing, 164pp).
pdf format, 2.13 meg size

Crystal Radio Handbook: V5 Inductance, Inductors, and

Measuring Q Technical look at Coils, (Bookfold format for
printing, 272pp).
pdf format, 5.58 meg size

Crystal Radio Handbook: V6 Antennas and Propagation A

walk down memory lane, historical and personal, (Bookfold
format for printing, 234pp).
pdf format, 9.49 meg size

162 115
Crystal Radio Handbook: V7 Vacuum Diodes Dedicated The radio now has a NOS Philmore crystal detector stand. In
entirely to vacuum diodes, (Bookfold format for printing, this photo I am testing a Pyrite crystal sitting on a crumpled
210pp). aluminum foil bed, successfully detecting 1590 Khz locally
pdf format, 11.00 meg size here in Houston. In the background is my trusty Icon R75
giving me the straight poop on what station I am listening to. I
Crystal Radio Handbook: V8 Resonance Resonance and was never really happy with the Philmore crystal holder
Coupled Circuits, (Bookfold format for printing, 252pp). though, it had to be improved.
pdf format, 17.70 meg size

Kenneth Khun's Web Book on Crystal Radio Engineering

(ready for printing and binding.., 113pp)
pdf format, 1.11 meg size

Ben Tongue's Web Book on Crystal Radio Systems Volume I

(ready for printing and binding.., 185pp)
pdf format, 1.37 meg size

Ben Tongue's Web Book on Crystal Radio Systems Volume II

(ready for printing and binding.., 208pp)
pdf format, 1.59 meg size

Ramon Vargas Patron's Web Book Contributions to Crystal

Radio (ready for printing and binding.., 111pp)
pdf format, 1.65 meg size

This Web Page.. (because I'm so vain.., 238pp)

Current configuration on the Dunwoody now sporting a good
pdf format, 6.72 meg size
brass crystal cup with set screw, also I have lifted the coil off
the wood base with 1/4" standoffs. I am not sure how much of
a difference that made, but seems good practice. This photo of
the set taken while prototyping a homebrew detector stand of
my own. I figured if I was to become a radio builder, I would
need master the detector. I am a geologost so no radio is a
crystal radio without a crystal, I mean, an actual crystal!

116 161
me back to my roots as a minerologist as well. In my fun I
have subsequently learned to pot my own crystal detectors in
woods metal, and tested a number of minerals as detectors,
though in no systematic way. I did immediately notice that
Galena and Pyrite, the most frequently cited minerals as
detectors, are metallic sulfides so I have acquired additional
samples of Stibnite, Molybdnite, Argentite, and Acanthite,
only successful with the Molybinite. Seems a high degree of
symmetry is needed as well. CHAPTER II
The following photos are of the radio as originally constructed
and a couple morphs as I learned to love this great hobby.
Radios I have built with my own
My Dunwoody in its original configuration as per the kit little mitts…
instructions. Clearly seen is my first-ever attempt at winding a
coil, first attempt at taps in said coil, and a fine radio just the
same.

160 117
 Dunwoody, a kit from Borden Radio right here in Houston
http://www.lessmiths.com/~kjsmith/crystal/dunwoody.shtml

My first real intro to this

fascinating world of
crystal radio came with a
set I purchased on Ebay
from Borden Radio
Company, Lance Borden.
Lance is an engineer at
NASA and radio amateur
with a business selling
crystal radio's and parts. When I first ordered my kit he was
very helpful and encouraged me quite a lot. I appreciate the
well-designed radio as much as his help in my introduction.
The radio itself is a single-tuned set with dual 365 caps tuning
on the antenna and detector sides of the single coil. The coil
features numerous taps, a concept I was quite unfamiliar with
and curious about when I first considered the set for purchase.
I figured I had much to learn and the set seems a good and
serious way to begin.

The Dunwoody comes with a 1N34A diode detector, two 365

variable capacitors, fahnestock clips, a sturdy cardboard coil
form and magnet wire to make the inductor. The included
instructions are clear and easy to follow. One item not included
or part of the radio is a crystal detector, but the instructions
give a hint by stating that the spacing for the germanium
detector are the same as for a Philmore detector. That sent me
to the web and introduced me to the crystal in crystal radio. I
thoroughly had the bug by this time as I searched for, and
eventually acquired a Philmore detector and crystal. In fact the
galena crystal is remarkably hot and often performs every bit,
in my opinion, as well as my germanium diode. This has led

118 159
 A Spark-Gap Transmitter
http://www.lessmiths.com/~kjsmith/crystal/spark.shtml

Why a spark-gap
transmitter? And why here
on a crystal radio page?
Well, the second question
I'll answer first as it is
easiest. The transmitter is
a vintage technology
device that is no longer
current in the modern
world and so fits thematically with the purpose of my crystal
radio page. The first Question is more interesting, why build
it? In late 2009 my elder brother, founder and owner of
Northwest Research Engineering, LLC., in Seattle, challenged
me to construct a spark-gap transmitter. He enjoyed seeing my
vintage-style crystal sets and we enjoyed discussing the early
Hooking parallel configuration to my signal generator can give
history of radio together. So, it seemed a logical step to
an idea of the performance of the radio across the spectrum. I
progress in this direction. Additionally, as his work is in
measured the mV output in 100 KHz increments from 500 to
cutting edge ionospheric research, he has little opportunity to
2300 KHz. Across the broadcast band the set has faitly flat
study such retro technologies up close and personal.
responce with a slight tendancy to resonance about 1500 KHz
and then increasing sensitivity from 1900 KHz on up. Keep in
DISCLAIMER:
mind that the set is hooked to a dummy antenna rather than my
I am not a licensed radio-amateur and in any case a
actual long wire. Still this gives an idea of the overall untuned
spark-gap transmitter is not legal to use in the United
nature of the circuit.
States. As such, this project was taken on as a technology-
demonstration exercise only and is not intended for actual
When I hook in a series configuraiton I am unable to get a
use. Such use is not permitted. Nor is this page intended to
measurement, only 0V DC output. I am working on why this
promote the construction and use of such transmitters by
may be the case. Drives me nuts.
others. I DO give detail on the construction and theory of
the transmitter in order to better understand and appreciate
the technologies used at the dawn of the modern wireless
era.

158 119
The concept behind the project is to build an operating spark-
gap transmitter using technology as close as possible to that
used at the turn of the 20th century. Technology that would be
recognizable to Marconi or Fleming. This meant no chips, no
transistors, no diodes, (OK, a galena and cat's whisker is
technically a diode, but you know what I mean...) no solid-
state. It does not mean using vintage components as these are
for the most part unobtainable and/or prohibitively expensive.
With each component new or homebrew the question is: would
Fleming recognize this? Sometimes I needed to stretch things
somewhat, especially concerning the relays, but still I believe I
have remained true to this original goal.

The transmitter was conceived as two modules, a power

module and an oscillation module. Size constraints forced me
to break the oscillation module further into two breadboards,
one for the variable HV capacitance and the other for the
oscillation transformer. The power module is the most
complex piece having on it the induction coil, interrupter,
capacitor, and the spark gap itself. To keep this page length
manageable and allow enough space to give a good description
of each component, its theory and function, I have broken the
descriptions into separate links below.

INDUCTION COIL.. Sleeper #1

COIL INTERRUPTOR..
The hookup in its parallel and series configuragions. In either
SPARK GAP.. case I really cannot hear any signal with just a headphone or
crystal earplug. With amplification I can get good signal
OSCILLATION CIRCUITS.. strength with the parallel hookup giving slightly more
sensitivity than the series hookup.
PERFORMANCE TESTING..

120 157
 The following circuit diagram and photo shows the complete
transmitter and how the modules integrate together:

As shown, the radio passes most loudly a local powerhouse

spanish format station. It is quite sensitive as one can imagine.
Listening closely and you can certainly hear the other stations
on the broadcast band competing for attention. My question
now, do I make a breadboard extravaganza of this set? On va
voir...

156 121
Induction Coil Sleeper #1
http://www.lessmiths.com/~kjsmith/crystal/sleeper.shtml
Marconi's first successful
radio employed an OK, this is a bit of a cheap
induction coil as its main shot for a radio, but.. it IS a
energy source. Other radio!
methods were quickly
adopted that were more It is always interesting to
efficient and more get down to bare essentials,
powerful, but the reliable to do the absolute
induction coil remained minimum and still get a
common throughout most of the early period of radio. Fleming working instrument. It is
describes a typical "10-inch" coil used for radio work in his not for nothing that MB Sleeper included a circuit diagram for
day. By 10 inches, he refers to the length of spark obtainable this hookup, it is a functionning radio. This circuit is
with the coil. These coils are large and today mostly essentially a detector of radio broadcast energy. It has no
unobtainable except as specialty reproductions at outlandish tuneing functionality. Resonance is extremely broad although
prices. Fortunately the venerable induction coil remains in weakly peaking as determined by the impedence and
manufacture as the common (increasingly uncommon capacitance of the antenna itself. In part my idea is to see if it
unfortunately) canister ignition coil found in most cars built is possible to determine this peak region.
through the 1970's. In Fleming's terms such a coil might be I present the set photo with a 1N5819 Schottky diode but it
called a "1-inch" coil, woefully inadequate for transmitting works well with virtually any diode, only the sensitivity
messages to China perhaps, but ideal for this project. Standard changes. I even tried a Zenner diode but had to attach the
coils generate up to about 30,000 volts and "sport" coils can Radio Shack mini-amp and crank it up. Still, it worked ok. So,
put up to 45,000 volts. more to come..

When I first began researching this project I was surprised by

the difficulty in finding useful technical specifications for
ignition coils of various manufacture, or even of knowing what
constituted a "good" versus a "bad" spec when such specs were
found. What little information is generally published includes
the maximum voltage naturally, often the secondary resistance,
sometimes the primary resistance and the turns ratio, (ratio of
primary to secondary windings). At first look there doesn’t
seem to be much correlation between voltage and turns ratio or

122 155
 even resistances. It took a lot of research to learn the needed
theory behind Kettering ignition coils and much of what is
found on the web is inadequate, oversimplified, poorly
founded theoretically, or downright wrong. I will try here to
explain in moderate detail how the Kettering Ignition System
works, the theory, and therefore how the various components
(windings, primary, secondary, capacitor, resistances, etc)
work together. With the right theoretical underpinnings, the
specs do make a lot of sense.

Induction coils operate by storing large amounts of energy in

the form of strong magnetic fields. As currents pass through
the windings their electric fields generate magnetic fields, and
fields in one coil induce fields in the other, primary and
secondary. As long as a current is maintained in one winding
its magnetic field is sustained and a corresponding field is
sustained in the second winding. Conversely, moving the coil
through a magnetic field or removing the field entirely will
generate an electric field, (measured as voltage) in the coil.
The ratio of turns between the windings determines the ratio of
voltages between the coils. Although the input voltage to the
primary coil is small, only 12V for a car battery, it has a large
current and a great deal of magnetic energy will be stored in
the primary coil. By shutting off the primary voltage source the
magnetic field of the primary collapses. As this magnetic field
collapses it cuts the secondary coil generating a very high
voltage output. The rate of field collapse (dB/dt) determines
the strength of the voltage output of the secondary. This high
voltage output spikes only when the primary current is shut off
causing the field collapse. Coils therefore require an interrupter
to repeatedly switch the current to the primary on and off. The
trick is to select materials with needed properties to achieve
the results wanted. The Kettering Ignition system use in this
project is described next.

154 123
 Does it work? What does it sound like on an AM radio? I
recorded the above plot off a small AM radio tuned to the top
of the broadcast band. Plenty of signal to go around, I
understand why these transmitters are no longer allowed to
operate! I give you a couple CQ's and an SSS in
commemoration of Marconi's first transatlantic radio reception
on December 12, 1901. Fleming designed the transmitter
The Kettering ignition system consists of a 12v battery (the power station at Poldhu for the transmission.
source of emf), an induction coil, a capacitor, and a switch of
some sort, (points, relay, vibrator etc..) in a car its points. For CQCQ Yes... it really IS a transmitter..
the coil the primary and secondary are wound in series and SSS
connected together inside the canister. The coil has inductance
and impedance/resistance. As the system has LRC, (L = 73
Inductance, R = Resistance, C = Capacitance) the circuit has a
resonant frequency and this frequency determines the unit time
period "dt" in the equation V = L di/dt. Even without the
capacitor there remains some distributed capacitance in the
coil and so opening the points will never instantaneously
collapse the magnetic field, but it will be extremely fast just
the same. The field is maintained by the current from the
battery to ground, opening the points removes the ground and
stops the current.

As the circuit has a more or less fixed L and R from whatever

coil you have available, you can only vary C by choosing the
capacitor carefully. For cars, this typically means a capacitance

124 153
 in the 0.1 to 0.2uF range. This capacitance gives a resonant
frequency such that the dt will deliver a 30kv jolt to the plugs
from the secondary of the coil within the period allowed by the
engine timing.

A smaller dt (shorter unit time) delivers a higher secondary V

(do the math) but standard coils are not designed for this and
the internal insulation will probably not handle the tension.
Additionally, the primary also delivers a voltage back to the
switch/points/relay. A smaller dt will also result in a higher V
off the primary. Of more importance, the smaller dt means that
the back emf hits the points early before they have had much
time to open. Either the higher V and/or the early arrival will
cause arcing at the points. Arcing at the points creates a low
resistivity path to ground and leaks the power away from the
coil and so the secondary will also not deliver much V to the
gap.
If I can get my arms around the frequency and learn to control Conversely, a dt too long will slow the rate of charging and
this, I see this as an excellent gap for my set. But, would prevent the coil from being charged sufficiently in the time
Professor Fleming recognize this? I must fall back to my available (dwell time). The field collapse will be slower
vacuum relay justification and restate that Fleming, as the lowering the voltage out of the secondary, V = L di/dt. A
inventor of the vacuum tube, was no stranger to placing smaller capacitance means higher frequency/smaller dt and
elements inside a vacuum. He would certainly recognize this. vice versa.
********* As dt is determined by the resonance of the LRC circuit, the
variable you have to work with is generally the choice of
Transmits: capacitor value. Modern high performance coils also factor in.
they have lower R (resistance) and lower L (inductance). lower
L hurts because the energy stored in the coil is as follows:
Energy = 1/2 LI^2. Reducing L lowers the stored energy.
lower current (I) also lowers stored energy. Lower R increases
the amount of current flow per unit time as well as increasing
the frequency of the resonant circuit. Larger di and smaller dt

152 125
means a much larger V. High performance coils are designed oxidized. This results in an uneven and "messy" spark. More
to deliver 45kv instead of 30kv and must have the insulation importantly, I suspect that there is a LOT of arcing taking
appropriate to this. The "performance" of these coils comes place at the gap as well. Arcing is a continuing current that
from two favorable factors, the high voltage assures sparking effectively shorts the closed circuit to ground. This would
in high compression and the small dt allows sparking at high explain why the audio-frequency coil oscillations are present
RPM's in 8 cylinder engines. That is, it’s the small dt that even when the sparking is present. The vacuum gap prevents
allows sparking at very fast cycle times. They still have oxidation and produces a clean spark with every pulse. The
roughly 100:1 turns ratio. electrodes are also made of either zinc or tungston or some
other refractory metal. The lesson to be learned, make your
Photo of the induction coil / ignition coil mounted horizontally electrodes of something more durable than brass and get some
on the power breadboard. Other components seen include the steel wool to FREQUENTLY clean the tips!
interrupter relays, the capacitor used to determine the coil
resonant frequency, and the active spark gap. The thing really
works!

The following are a few easy specs on some popular ignition

coils, only a small sampling:

126 151
circuits have different wavelengths and so will set in motion a PerTronix 45011: V max = 45000 V, P res = 0.6 ohm, S res =
variety of unwanted harmonics. Presumably, with an actual 8.6 ohm, P ind = 5.5 mH
antenna close to the design specs the transmitter will perform Mallory 29217: V max = 58000 V, P res = 1.4 ohm, S res =
flawlessly! 10.0 ohm, P ind = 6.6 mH
MSD 8202: V max = 45000 V, P res = 0.7 ohm, S res = 5.2
********* ohm, P ind = 8.0 mH
Taylor 718203: V max = 45000 V, P res = 0.7 ohm, S res = 4.7
The above analyses are presented with the transmitter and gap ohm, P ind = 8.0 mH
working in their "normal" operation mode. You will recall in
my section on the spark gap that I also have tested a UHF For my project I have used the PerTronix coil above although I
vacuum tube spark gap. The following plots give some early imagine most will do nicely. The Kettering discussion,
interesting results needing presentation here. The RF appropriately placed here in the Induction Coil write-up, is
oscillogram is quite typical of what this spark produces. You needed understanding in the next section on the Coil
will immediately notice that instead of an RF pulse consisting Interrupter selection.
of several distinct phases as before, this gap produces very
clean almost pure HF tones with varying amplitude in the
27MHz range. These two differences, unphased tone and high
frequency need explaining. For the difference in frequency I
am really at a theoretical loss, all the formula I have tell me
that the frequency results from inductance and capacitance,
two parameters that did not change. I can only speculate that,
as the spark resistance ought to be significantly lower in the
vacuum gap, this may be the cause. More testing and more
reading needed!
As for the tone itself, again I need to speculate and hypothize.
If you look again at the initial audio-frequency oscillogram,
you note the obvious induction coil audio oscillations as well
as the initial spark. Normally these oscillations should not be
seen unless the electrodes are separated far enough to prevent
sparking. Initially in my testing this was the case, but with time
the coil oscillations became a normal part of my responce.
Additionally, with time I have had to adjust the electrodes ever
closer to get consistant sparking, thus the frustration. I
hypothize that the electrodes themselves are getting quite

150 127
Coil Interrupter
Open circuit:
In order to function Secondary coil = 17.4 uH (L1)
properly an induction coil Distributed inductance = 0 uH (Lo) (no distributed inductance)
requires an interrupter Distributed capacitance = 25 pF (Co) (Scope probe in X10
circuit. The high-voltage position)
output of the secondary Wavelength = 59.6 sqrt((L1 + La)*Co)
winding results from the Wavelength = 39m = 7.63 MHz
magnetic field collapse of
the primary when its The two following plots give spectral analyses of the
current source is oscillations in the closed and open circuits.
interrupted. A single interruption will cause a single HV spike
and thus a single spark. Repeated sparks require repeated
charging and closings of the primary winding. For my project I
was seeking a pulse-rate some 60Hz or better to give an
audible tone at the receiver. Too rapid though and the
induction coil will not have enough time to charge between
cycles. Putting things into automotive terms, a 4-cylendar
engine idling at 750 rpm must cycle the coil every 40mS, or
25Hz, no problem. For an 8-cylendar engine running at 2500
rpm you are down to 6 mS cycling (170Hz) and many coils are
unable to work well at such rates. High performance coils are
intended work well even at such high rates. For the transmitter
I feel a cycle frequency between 60 and 100Hz would be ideal.

It would be lovely if cycle timing were the only constraint,

general purpose relays typically run around 60Hz. Indeed this
was my initial design, woe is me. Another concern the
interrupter needs to deal with is the presence of a back emf This analysis gives a pretty close prediction to the values
from the induction coil primary winding. With a 100:1 turns actually found on the testing and gives confidance that things
ratio, a coil that sends a 45kV pulse out the secondary will also are behaving as per theory. Other frequencies noted are
send back to the interrupter a 450V pulse. Small general harmonics and/or interference between the open and closed
purpose relay contactors remain rather close together even at circuits. The original design was set so that both open and
their widest separation and so are prone to arcing with modest closed circuits resonated together at 120m. Under test the two

128 149
 voltages. When I first hooked up my relay I got modest
Right about now the observant reader must have noted that the sparking from the gap and contact arcing that was quickly
frequencies measured, 2.11 MHz on the closed circuit, and burning up the relay. It was my belief that the capacitor was
7.66 MHz on the open circuit, (and variations in-between) do there to slow the process and allow time for the relay contacts
not compare well with the design frequency desired. This to separate before the back emf hit. I tried capacitors of
certainly needs to be looked at. Recall my original design was increasingly high capacitance, they did not cure the problem. It
for 2.5 MHz on both circuits with the following parameters: was time to hit books and work the test bench.

Closed circuit: What follows was a

Primary coil = 2 uH better understanding
Capacitance = 2000 pF of relays in general,
Wavelength = 1884 * sqrt(L*C) = 119m = 2.52 MHz and SPDT relays
wired in a buzzer
Open circuit: configuration in
Secondary coil = 20 uH (L1) particular. An SPDT
Distributed inductance = 20 uH (Lo) La=Lo/3 = 6.65 uH relay has a common contactor set between two posts. The
Distributed capacitance = 150 pF (Co) contactor has a spring which keeps it in contact with one of the
Wavelength = 59.6 sqrt((L1 + La)*Co) = 121m = 2.48 MHz posts, the "Normally Closed" (NC) position. The other post
Wavelength = 1884 * sqrt(L*C) = 121m = 2.49 MHz thus represents the "Normally Open" (NO) position. The relay
also contains a small coil that, when energized will push the
Under testing conditions I have made some significant changes contactor towards the NO position and so breaking the NC
to the above. Specifically, the lumped inductance and contact. In a buzzer configuration, breaking the NC contact
capacitance are changed to P = 6.2uH (closed circuit), S = also de-energizes the relay coil so there will be no more push
17.4uH (open circuit), and the capacitance = 1070pF (closed towards the NO post. The contactor will use what little
circuit). Also, the design assumed a "generic" antenna with momentum it has to oppose the spring and move to the NO
distributed values La = 6.65uH and Co = 154pF. Under test position. In reality this momentum is not sufficient and so no
there is no antenna and the scope probe gives about 25pF contact with the NO post is ever made. This is why one always
capacitance and no inductance. Substituting the test values uses the NC post with the SPDT buzzer relay. A further
provides the following analysis: negative consequence of this is that the contactor never
separates far from the NC post. The close proximity allows
Closed circuit: breakdown of the dielectric potential of the air between the
Primary coil = 6.2 uH posts and arcing to occur.
Capacitance = 1070 pF
Wavelength = 1884 * sqrt(L*C) = 153m = 1.96 MHz

148 129
Solutions to this arcing problem come in two flavors, the frequency. High decrements, (in the 1920's a spark-gap
dielectric and the contactor separation. One solution used in transmitter was required by law to keep a decrement of 0.2 or
high power relays is to enclose the contactors in a high- below) indicate a poorly engineered / poorly operated station.
dielectric oil to suppress arcing. This is very effective but the Calculations for the wave decrement as follows:
fluid is difficult to work with, expensive, and it slows the
contact frequency greatly. Recall that I am seeking at least a Decrement = 1/n ln (Xo/Xn) where Xo is the amplitude of one
60Hz contact frequency. Another way is to place the wave and Xn is the amplitude of another wave n periods
contactors in a vacuum. While not as good a dielectric as oil, it following.
is significantly better than air and has the advantages of being
easy to work with and maintains high frequency operation. Damping Ratio = 1/sqrt(1+(2pi/D)^2) where pi = 3.14 and D is
Several companies manufacture vacuum relays and, while the decrement from above.
expensive, this is the route I chose. Note: there is of course a
third solution to abandon relays and go solid state. This is fine A particularly nice example of an isolated fully-quenched
if all you desire is cool sparks, but it is not in my directive. oscillation follows. I assume in this event that enough energy
remained present in the closed circuit to allow an occasional
Even vacuum relays when wired as a buzzer have the contact isolated single spark to jump the gap. Such a spark produces a
separation problem. If the contactors remain very close then nice ring in the open circuit without the complexity associated
arcing can still cause trouble. The method to force the with the main sparking from the induction coil. This plot
contactor to move to the fully open as well as fully closed shows the open circuit ringing nicely at 7.66 MHz frequency
position, allowing maximum separation, is to wire two relays with an acceptable 0.12 decrement.
in series. The resulting contact frequency (duty cycle) will be
half that of a single relay. Happily the vacuum relays operate
at a frequency of about 200-240Hz. That provides a 100-
120Hz operation in a two-relay configuration, very sweet.
Additionally, the two-relay output consists of a nice square-
wave signal bringing joy to the professional radio engineer. I
know this because the circuit configuration was suggested to
me by my clever elder radio engineer brother. Fei chang gan
xie! Of course, this solution doubles the expense of an already
expensive relay. I have tried the configuration with general
purpose relays and find the resulting 30Hz operation
unacceptably slow.

130 147
 One final and somewhat late note concerning the vacuum
relays. These relays, (Gigavac GH-3 or Kilovac HC-3) are
load-switching devices designed to operate to 18 amps. The
Pertronix coil on the other hand runs nicely at a cool 20 amps
(12 V and 6 ohm, Ohm's Law here). The duty cycle is plenty
slow enough to allow a longer charging time so all those amps
shouldn't really be needed. I have added a 0.25 ohm ballast to
the circuit to lower the current to about 14 ohms. Those cute
little relays are very expensive and I have already damanged
(temporarily) one of them by pushing 20 amps through. With
the ballast I DO find the spark power, measured in terms of
brightness, noise level, and max length is diminished
somewhat, such a pity.

So, would Fleming recognize this? It is where I push the

Finally, when the energy loss finally quenches the sparking at
boundaries as I doubt that vacuum relays existed in his time.
the gap, the open, secondary circuit then can freely oscillate,
Still, relay action was well understood in Fleming's day. As for
here with a frequency of 10.76 MHz and lasting some 3.3 uS.
enclosing them in a vacuum? Well, I defend myself by stating
This last phase is the desired transmission frequency and the
that Fleming himself invented the vacuum diode. Enclosing
open circuit normally oscillates into the antenna. Decrement on
electrical devices into vacuum bulbs was not a new idea even
this last phase is an excellent 0.04. Early spark gap transmitters
then. So, while the vacuum relay may not have existed, relays
struggled to reduce or eliminate the initial two phases of the
themselves did and Fleming would have been comfortable
event with quenched gaps.
with the vacuum technology. I think...
Why is decrement important? Most RF damping in a spark-gap
transmitter results from interference between oscillations in the
open and closed circuits of the oscillation transformer. When
these waves interfere two separate wave energy peaks are
radiated from the antenna causing loss of signal strength at the
desired frequency and adjacent channel interference. A good
spark-gap transmitter should quench the closed circuit as
quickly as possible after the energy is transferred to the open
antenna circuit. Once quenched the open antenna can resonate
at its natural frequency. Lacking quenching as with my
transmitter, the oscillation transformer needs to be carefully
tuned to resonance and loosely coupled to radiate at a single

146 131
Circuit diagram of the power module indicating the wiring Zooming in to the radio frequency world the above plot shows
concept for the two-relay solution. And finally, a photo of the the radio frequency wave in some considerable detail. This
set with the two relays in position and ready to go. plot indicates that the oscillations given off from a spark gap
are anything but pretty. The complexity results from the
ringing going through several phases with the closed and open
circuits, and induction coil each having their imprint. The
initial (precursor) phase of the event, lasting here about 0.27
uS is the high-amplitude energy (upwards of 520+ V) of the
spark as the induction coil pours its energy directly into the
circuit at the gap. This is followed by the main phase, a
complex 68 V amplitude ringing lasting some 2.8 uS. The
ringing consists of two combined waveforms, one, a high
amplitude (68 V) 2.11 MHz wave with a second lower
amplitude (37 V) high frequency 27 MHz wave superimposed.
This phase results while the closed (primary) circuit still
retains sufficient energy to cause sparks at the gap with each
main oscillation. The higher frequency wave comes from the
open circuit (secondary) inductively coupling to the closed. A
detail look at the main phase seen below:

132 145
This first plot above shows the response at audio frequency. In
the 100 mS plotted we see 11 pulses at 120 Hz frequency. This
represents the pulse frequency of the dual relay system, two
relays at a nominal 240 Hz each. Each pulse starts with a high
amplitude spike associated with the spark discharge and
followed by two to three damped oscillations with a frequency
of 533 Hz. This oscillation is associated with the natural
frequency of the induction coil itself. Damping of waves is a
natural oscillation feature and is refered to as the wave
Decrement. Decrement is the natural log (Log to the base e) of
the amplitudes of any two successive peaks in the wave train.
It results from the loss of energy in the circuit during the
oscillation period. Energy is lost due to resistive heating,
radiation of waved into space, and interference between
oscillations in the primary and secondary oscillation circuits,
especially where they are closely coupled. Where resonance is
high the interference losses are minimized.

144 133
Spark Gap Performance Testing the Transmitter

The heart of the system is If you build it, will it

the probably the simplest work? And how well? This
component, the spark gap page attempts in a
itself. After all the preliminary manner to
required background answer these questions and
theory for the induction provide a first look at the
coil and relay setup, this performance through the
component was a walk in eyes of an oscilloscope. It
the park. Still I need note will take much testing and
that my gap is not intended for serious duty and so does not observation/head scratching before I ferret out all the
require extra features such a cooling fins and/or etc. Below is phenomena present in the set. This is one of the objectives of
an early mock-up of the gap, a simple set of 1/2" brass balls the project, to study and learn about spark-gap technology. The
tapped for a 1/4" threaded brass support rod. The entire plots below are a first look, some will raise more questions
assembly set off the base with lovely tapered ceramic than answers, but that’s the fun!
standoffs.
All of the scope plots presented here are made with a single
hookup to the scope. The scope probe was attached to the set
antenna output of the oscillation secondary coil. Many other
observation points remain to be made and observed and I will
update this page accordingly when needed. I present
observations of the three primary oscillation frequencies
expected from the set: A) the pulse frequency from the relay,
B) the resonant frequency of the induction coil + power
capacitor, and C) radio frequency oscillations of the transmitter
itself. So little work so far, so many questions remain!

Early use of the spark gap, especially with a weak induction

coil, showed up room for improvement. When sparking the

134 143
 brass balls build up an oxidation layer fairly quickly. This
layer increases the dielectric between them and, especially
when the voltage is marginal, (my first coil was old) the sparks
had trouble to jump the gap. I needed to clean the balls
frequently with steel wool. The 1/2" balls spread out the field
flux over a small area. One improvement was to switch from a
ball-gap to a gap between sharply pointed tips. The photos
below show both configurations. The points concentrate the
field and creates a higher potential. This makes it easier for the
sparks to form. Switching between configurations is as easy as
screwing on or off the balls which cover the pointed tips.

Adjustment of the gap length is easily accomplished by turning

the end knobs of the threaded supports. Small top thumb-
screws lock the posts at the desired position. CAUTION: Do
NOT touch any metal components while the spark gap is in
operation. Handle by the insulated knobs only.
The capacitance module simply consists of high-tension
modern capacitors switchable into or out of the circuit in
parallel. The original intention, still in play, is to use Leyden
Jars, but that is another matter. The four capacitors have values
of 500, 1000, 1000, and 3000 pF giving plenty of combinations
to choose from. Changing inductance is a matter of clipping to
the coil at the right place. Finally, coupling the coils is a matter
of sliding the secondary along rails into or out of the primary
winding. The following photos give portraits of the two
oscillation circuit modules.

When one walks into a room, nothing quite says "Transmitter

present!" like a big oscillation transformer!

142 135
 The oscillation transformer design thus becomes a matter to
choose coil lengths and diameters to give a range of values
inclusive of Lsecondary = 20uH and Lprimary = 2uH. To
make the transformer loosely coupled the secondary needs to
slide within the primary. I chose the following parameters:
Primary: Diameter = 7", Length = 2.0", Turns = 4, uH = 3.81
max, 1/8" copper tubing
Secondary: Diameter = 5", Length = 6.0", Turns = 22, uH =
27.4 max, 1/16" copper tubing

In the photos above the sparks are being produced without any
ballast resistance and are just over 2cm long. With the ballast I
subsequently added to protect the vacuum relays, the spark
power is lowered from 20 amps down to about 14 amps. The
sparks obtained are smaller, about 1 to 1.5 cm. According to
Fleming an induction coil capable of producing sparks up to 10
inches (25cm) in air will give a spark of 6-7 mm when
connected to a load. I find with the tank circuit connected to
the gap I need adjust the spark points very carefully and close,
about half a mm or less to obtain consistant sparks. This can be
a frustrating exercise. Air diaelectrics also have the problem of
arcing when the air between the electrodes begins to ionize. A
solution to this is to enclose the spark gap in a vacuum. It just
so happens that occasional inexpensive vacuum spark gaps are
listed on ebay. These tubes are intended for radar (UHF)
applications and are rated in the few hundred volts range. As

136 141
to imagine an amateur inverted L antenna with about 20 uH my gap is tiny now, I felt this may be a good solution, or,
distributed inductance and 150 pF distributed capacitance. As alternatively I might have some fun blowing up a cool vacuum
antennas are magical mystical things that nobody (meaning gap. The following photo shows that in fact the vacuum gap
myself) really understands this appears entirely reasonable. In performs handsomely as a gap for my transmitter. The gap
any case, the transmitter is tunable to bring it to proper length is mercifully fixed at about a half mm and the low
resonance (hopefully) when attached to an actual antenna. diaelectric allows good sparking with flat (rather than pointed)
These then are my fundamental design parameters: spark surfaces about 2mm in diameter. I certainly will
experiment more with this kind of gap.
Design frequency = 120m
Antenna Co = 150 pF
Antenna Lo = 20 uH (20000cms)
La = Lo/3 = 6.67 uH (6667cms)

L1 = Loading, or secondary coil = ?

From Bucher:
Wavelength = 59.6 sqrt((L1+La)*Co)
Thats'a some glow...!
Assuming a design frequency of 120m and solving the above
equation for L1, I find: **********
L1 = 20 uH (20000cms)
20uH then is the value we wish for the oscillation transformer Other Sparks:
secondary winding.
Of note, the gap itself need not remain strictly a set of pointed,
For the primary winding we find in Bucher the following flat, and/or round surfaces. Once you have a working High
useful equations: Voltage power supply, there are other gaps/toys waiting to be
L = lambda^2 / 3550000 * C, and/or tried.
C = lambda^2 / 3550000 * L The photo below is a quick cobbled together Jacob's Ladder
just to see: Will it work? The copper wire is not perfectly
Therefore one starts with lambda = 120m and then either smooth nor can one adjust the spacing between the two
assume a capacitance or an inductance. For this project I found verticals. Still, I can tell this will be a fun side trip with my HV
a good match with 2000 pF and 2 uH. To be safe both power supply. Stay tuned for more.
inductance and capacitance can be varied around these two
design numbers.

140 137
 Oscillation Circuits: HV Capacitance and Transformer

The oscillation circuit in

the transmitter performs
the twin functions of
converting the spark's
energy to trains of radio
frequency oscillations and
transferring them from the
closed tank to the open
antenna for transmission.
Like any tank circuit the main components consist of
capacitance and inductance, both variable in order to tune the
circuit to the needed frequency. The closed circuit loosely
couples to the open circuit to achieve the best resonance and
lowest decrement. The design aspects of this section is where
the radio theory begins and induction/relay theory ends, we are
talking radio here now.

The design for the transmitter starts with a notional target

transmission frequency and the design equations from two
primary sources: Fleming, 1910, The Principals of Electric
Wave Telegraphy and Telephony., and Bucher, 1920, The
Wireless Experimenters Manual. Additionally it requires some
estimate, guesstimate really, of the antenna distributed
capacitance and inductance.

As the project is demonstration only, I have selected a

frequency at the low end of the HF range just above the MW
broadcast bands. The designs in the 1920's typically aimed at
Schazzam!! about 200m or 1.5 MHz. To keep my design close I chose
120m, or about 2.5MHz. It is a fairly quiet part of the
spectrum. Keep in mind this is a notional project not intended
for actual broadcast. I used charts and data presented in Bucher

138 139