Understanding Photomultipliers
Understanding Photomultipliers
Understanding Photomultipliers
Photomultipliers
Understanding photomultipliers 2. Principles of operation 3. The photocathode
1 introduction 1 Photomultipliers are extremely sensitive light detectors This section gives information on :
providing a current output proportional to light intensity.
the light sensitive area of the photomultiplier
They are used to measure any process which directly
2 principles of operation 1 or indirectly emits light. Large area light detection, high the effect of the window on light transmission
e) hemispherical
gain and the ability to detect single photons give the photocathode spectral response For diffuse light sources, e.g. arrays of photomultipliers.
3 the photocathode 2 photomultiplier distinct advantages over other light
photocathode sensitivity units
Application: Fundamental scientific research.
detectors.
1 2
3. The photocathode 3. The photocathode
quantum efficiency QE %
Q S20, trialkali or multialkali (NaKSbCs) the For a particular photocathode, S is calculated from
transmission %
MgF2 Q S W borosilicate
20
multialkali photocathode response extends from uv measured values of η(λ) and tabulated values of I(λ) and
50
W
to the near infra-red. It has high light level capability V(λ) by integration.
15 but may require cooling to reduce dark current.
borosilicate
10
3.4. Photocathode sensitivity
1.0
0.8
5
3.3. Photocathode types quantum efficiency: η (λ) or QE% figure 5 Relative luminous efficiency of human eye V( λ ) and radiant
power of a tungsten filament lamp operated at 2856 K.
quantum efficiency QE %
Photocathodes can be manufactured from a variety of 10 ( ), the quantum efficiency at wavelength λ is the
compounds and each type has a characteristic spectral average photoelectric yield per incident photon and
is normally expressed as a percentage. It is the most The luminous sensitivity specification has been adopted
response. The best choice is usually the one with the by all photomultiplier manufacturers. Values of S
maximum response over the wavelength region of fundamental unit concerning the performance of the
photomultiplier. Important practical considerations such range from 20 μA/lm to over 400 μA/lm, depending on
interest. 1
KBr Csl RbTe CsTe the photocathode type. For this test the lamp output is
as resolution, signal/noise ratio and detectivity are all
There are other considerations, such as operation at high related to quantum efficiency. adjusted to 1 millilumen and approximately 80% of the
light levels, and dark current, which are covered in detail photocathode area is illuminated.
elsewhere. (See table 5.3(a) page 10 and figure 23 radiant sensitivity (responsivity): E(λ)
0.1
100 200 300 400
filter measurements (CB, CR, IR)
page 15). Radiant sensitivity is defined as the photocathode
wavelength nm It is clear from (2) that S is derived by integrating
current emitted per watt of incident radiation at
35
wavelength λ and is expressed in mA/W. It is related terms which are wavelength dependent, with a high
figure 4c Solar blind photocathodes deposited on MgF2 are
bialkali sensitive to uv light only. to quantum efficiency in the following way contribution coming from long wavelengths. Luminous
30
λη(λ) sensitivity figures cannot, therefore, be used to compare
quantum efficiency QE %
3 4
3. The photocathode 4. The electron multiplier
corning blue (CB) effective diameter and uniformity The electron multiplier is a very low noise, high gain, type size gain timing linearity magnetic ap(t)
wide band amplifier with the capability of providing an most immunity
.. .. . . . ...
sensitivity for sources emitting in the blue region of the photomultiplier output is examined using a flying spot CF
photomultipliers are discussed in 4.2.
spectrum. This filter is most relevant to photomultipliers scanner to check uniformity as illustrated in figure 7.
. .... .. .
--- SbCs
electron
40
trajectory BeCu
focusing
20 45mm typ. electrodes
table 4.3 Performance differentiators of secondary emission
surfaces
0
300 400 500 600 700 800 900 1000 ---
wavelength nm
first dynode 4.4. Gain
figure 6 Filter transmissions.
Gain in a photomultiplier is derived by current
amplification. Each dynode amplifies the incident
cathode QE% measurement electron current and the overall gain is given by the
figure 8 Photoelectron trajectories between photocathode
product of the individual dynode contributions. With
ET Enterprises has spectral response equipment which and first dynode.
45mm typ. many stages of gain, a small photoelectric signal is
--
--
5 6
4. The electron multiplier 4. The electron multiplier
In general, a single high voltage power supply is The time constant of the circuit is: τ = RC, where: In the extreme case τ→∞,C integrates the charge and
used, with a resistive voltage divider network, to 4.5. Anode sensitivity A/lm the output rises with a time constant τs and remains
R0RL ...(6)
provide suitable inter-electrode voltages. The more R= there:
R0 + RL N e G [exp(-t/τs) - 1]
dynode stages in the photomultiplier, the higher is the The voltage required to attain the specified anode and v0(t) =
C ...(9)
gain at a specific overall applied voltage and the higher sensitivity is recorded on the test ticket. Specified in this
way the output is referred to the cathode sensitivity, S,
C = C0 + CL
the maximum gain attainable. This is illustrated in There are applications where the output is not pulsed
figure 10. It should be noted that gain figures and quoted in μA/lm. The gain can be derived from this and S but it varies unpredictably with time. In film scanning
by using the following relationship: Consider the output voltage response to a light source
curves are obtained with a particular voltage for example, a transition from light to dark gives a step
which decays with a single exponential time constant, τs,
distribution specified by the manufacturer. The gain Anode Sensitivity (A/lm) X 106 change in output, measured in microseconds. Here
G= ...(5
curve isdifferent if the voltage distribution is then: i(t) = N e G exp (-t/τS) ...(7) the consideration is one of choosing the output time
Cathode Sensitivity (μA/lm)
different from that specified. τS constant to match the fastest anticipated transition,
ET Enterprises specifies the voltages required on each N e G R [exp(-t/τS) - exp(-t/τ)]
and: v(t) = τ - τ ...(8) while simultaneously providing some smoothing of
photomultiplier to achieve two fixed anode sensitivities S unwanted statistical fluctuations. For example, if the
(the nominal and the maximum) with inter-dynode where:
20 fastest transition t r has a time constant of 1 μs, and, if
voltage distributions as specified in our photomultiplier N is the number of photoelectrons in the light pulse, (C0+ CL)RL is chosen to be < 3x10 -7, the output circuit
data sheets. It is sometimes useful to construct a gain- e is the electronic charge, will reproduce the transition quite faithfully. For R L=
voltage curve using these two points on log-log paper. G is the photomultiplier gain.
SbCs 10kΩ, for example, (C0+ CL) needs to be ≤ 30pF. With
15 high gain d1
this time constant, transitions faster than 100 ns will
not be followed.
first dynode gain, δ 1
v (t) (V)
-1.0
0
0 100 200 300 400 500 600 700
anode τ =30ns
Vk-d1 (V)
τ =100ns
-1.5
R0 C0 CL
figure 9 Variation of first dynode gain, δ1, with k-d 1voltage. v(t) RL
τ=∞
108
-2.0
14 stage
ground for HV + 0 10 20 30
t (ns)
12 stage
107
figure 11 The equivalent circuit for a photomultiplier. figure 12 The output v(t) for various time constants,
assuming N = 100, G = 1 x 10 6, τs = 5 ns, C = 10 pF
gain (G)
106
10 stage
The equivalent circuit applies to both dc and pulsed
By inspection of (8), the output voltage is a faithful
applications. In the case of pulsed light sources, the
representation of the input current when τ<< τs and
nature of the output signal depends upon the time
this is referred to as current mode operation.
105 profile of the input. Many applications involve light
Voltage mode operation applies with τ>> τs, in which
sources with an exponential profile. The following
case the pulse height in volts is proportional to the total
analysis is useful to determine the amplitude and the
input charge. For high or variable pulse rates there is a
104 timing profile for such input stimuli.
1000 1500 2000 2500 danger of overlap, in voltage mode operation. It is usual
applied volts (V) practice to use a value of 1/τ less than one quarter of
the event rate to avoid pulse pile up effects. In pulsed
figure 10. Variation of photomultiplier gain G with applied
light applications, the choice of τ depends principally
voltage, illustrating effect of number of dynode stages.
on the characteristic decay time of the light source
and the anticipated event rate. For example, in Na(Tl)
applications, where τs = 230 ns, choosing τ = 1 μs
ensures good integration of the signal while permitting
event rates of up to about 50 kHz, before the onset of
serious pile-up effects.
7 8
5. Limits on performance 5. Limits on performance
Photomultipliers are high gain, wide bandwidth optical currents of 100 μA. Limits on maximum photocurrent
Light
detectors and as with all sensors, limitations apply to the also apply, governed by the resistivity of the photo-
ET Enterprises Input
quality of performance with regard to: IMPORTANT - This test ticket must be sensitive layer. Exceeding these limits results in non-
returned with tube if making a guarantee
claim. linear performance. Side window photomultipliers, with
maximum gain (sensitivity) dark current a metal substrate supporting the photosensitve layer,
Order No. 67670 REF - 921
time or frequency response stability and hysteresis Type: 9900B07 can sustain much higher currents than end window
---
linearity of response noise-in-signal Special selection
types. Photocurrent limits are given in table 5.3(a).
Serial: 15049
tr
5.1. Maximum voltage Test result data 10%
cathode type lk (nA/cm2)
Bgd 389s-1
Side Window 5000
significant, resulting in unstable performance and high Background counts
>0.3pe
dark current with the possibility of breakdown under figure 14 Illustrating photomultiplier time response (rise time, full
width at half maximum, transit time). table 5.3 (a) Maximum cathode currents at 20oC.
extreme conditions. Permanent damage will occur if
the photomultiplier suffers breakdown. The sensitivity
limit is determined by the maximum allowable gain of Photomultipliers with plano-concave windows and linear b) pulsed anode current limits
the multiplier, which depends on the type of multiplier focused or circular focused multipliers give the best time Pulsed anode currents greater than 100 μA can be
figure 13 Sample photomultiplier test ticket. Parameters are
structure. The maximum allowable inter-electrode performance. Other factors affecting timing include: drawn with the proviso that the mean anode current,
measured using standard voltage dividers.
voltages are determined by the spacing of the electrodes averaged over one second, is below 100 μA. The
i) the number of dynodes. Fewer stages give better
in the photomultiplier. Excessive voltage causes photomultiplier output current will be linear with respect
electrical breakdown, and this must be borne in mind in 5.2. Time response timing.
to light input until the onset of space charge effects.
ii) the overall voltage. Higher field strengths
voltage divider network design (Section 8). Similarly, the The current at which non-linearity first appears depends
improve timing. The best timing is attained with
maximum allowable overall voltage is limited to less than The response of a photomultiplier to a delta pulse of on the type of multiplier structure and on the operating
high k-d1 voltage, with operation at the maximum
the sum of the stated maximum inter-dynode voltages. light is governed by the electron trajectories within the conditions. Linearity can be improved with a voltage
recommended gain. The time response varies
Maximum allowable inter-electrode voltages are given tube. Photoelectrons created by the light pulse follow divider distribution which has higher inter dynode
approximately as 1/V1/2.
for each tube type in the photomultiplier data sheets. individual paths to the first dynode, depending on voltages on the later stages. Linearity also depends on
iii) the photocathode diameter. Smaller diameters
their point of origin on the photocathode and on their the dynode secondary emitting surfaces as detailed in
A test ticket, an example of which is shown in figure13, have better timing. The best timing is achieved when
emission velocities. It follows that they land on the table 5.3(b). Recommended divider distributions are
is supplied with every photomultiplier. This includes two illuminating the central area only.
first dynode at different points and at different times. given in Section 8.
overall voltage figures. The nominal sensitivity, and the Secondary electrons travel individual paths between
overall volts required to attain this, is the recommended dynodes, causing further time dispersion. structure transit time rise time transit time Linearity measurements are made using two LEDs and
(ns) (ns) jitter (σ) (ns)
operating condition. The maximum overall sensitivity, A photomultiplier output pulse is characterised by: a double pulse generator. Two pulses are applied with
and the overall voltage at which this is achieved, refers VB 40-110 8-15 2.2-5.7
the second delayed with respect to the first (individual
to the maximum gain beyond which the pmt should not rise time t r BG 50-80 12-18 4.2-6.4 outputs). Then the two pulses are applied in coincidence
be operated. These voltages are measured using the CF,SW 20-35 1.5-2.5 0.5-1.0 (summed output). Non-linearity becomes evident when
standard voltage divider distributions quoted in our data full width at half maximum t(fwhm) LF 20-55 1.8-2.7 0.5-1.2 the summed output pulse differs from the expected
sheets. If a voltage distribution other than that used by transit time t t sum of the individual pulses.
ET Enterprises is chosen, the overall voltage required to table 5.2 Timing performance of ET Enterprises photomultipliers.
achieve the maximum overall sensitivity will be different. These parameters are defined in figure 14. Transit time jitter figures refer to single photon excitation.
peak anode current (mA)
dynode at Vd-d = 100 V at Vd-d = 300 V
Bandwidth is not a parameter usually quoted for structure dynode surface dynode surface
photomultipliers, although it can be derived from the 5.3. Linearity and maximum CsSb BeCu CsSb BeCu
LF 30 50 100 150
approximate formula: currents CF 10 20 30 50
1
f (3dB) = Hz ...(10) VB
3tr 2 4 5 20
a) direct current limits BG 0.1 0.2 0.5 1
The variation in transit time from one light pulse to the High, mean, anode currents cause fatigue effects and
next is a critical parameter when using photomultipliers reduced photomultiplier lifetime; the maximum mean
to detect the time occurrence of events. The standard anode current should be less than 100 μA and more table 5.3 (b) peak anode current for which there is 5% departure
deviation, σ, derived from a large sample of such events from linear operation. V d-d is the inter dynode voltage
stable performance is achieved by operating below
is known as the transit time jitter (sometimes the fwhm 10 μA. With correct voltage divider design all multiplier
of the distribution, equal to ~2.35 σ, is quoted as the structures are linear, within one percent, up to anode
transit time jitter).
9 10
5. Limits on performance 5. Limits on performance
The shot noise formula relates T to the inverse of the
5.4. Noise, dark current 10000 bandwidth through the Fourier transform and predicts
and dark count time base = 5ms/div
the noise on a cathode current of I k as:
100 8000 1 1
(i 2k )2 = (2 e Ik f ) 2 ...(12)
For the purpose of analysis it is convenient to categorise
Ia (μA)
G, the S/N ratio is
or
pulsed. Ik Ik
2000
S/N = 1 = 1
...(13)
(i ) 2 2
k (2 e Ik f ) 2
A film scanner is a good example of dc: the intensity of 0
the light reaching the photomultiplier is continuously 0 200 400 600 800 Equation (13) is illustrated in figure 18.
variable over a wide dynamic range and is subject to 1 channel
t
ps ren
The required bandwidth of the associated output circuit N
in e C S
10-8 10
is RM
Am ur
0 5
analyzer (MCA). When the photomultiplier measures a pulsed light -1
and electronics may be 3 MHz in order to follow these 10
No
source of constant intensity, there is variation in pulse height from 10
4
1
-1
fast transitions. The output produced by a NaI(Tl) crystal one pulse to the next.
10
mean cathode current (A)
10-10 10
3
2
has a characteristic time of 230 ns, and between events,
-1
figure 15 dc detection: illustrating the output signal for three
10
10
the photomultiplier output is ideally zero. This is an
3
2
-1
different levels of illumination; measured at g = 10 6.The rms noise noise
10
10-12 10
example of pulsed operation. contribution can be predicted by using the chart in figure 18. 1
In common with the measurement of all physical
10
5
processes, repeated measurement does not always
-1
0
The purpose of this section is to examine how statistical
10
10-14
6
10
give exactly the same value. In the photomultiplier the
-1
fluctuations and photomultiplier dark current contribute -1
10
time base = 0.2μs/div
7
variation in output signal is related to the statistical
-1
voltage = 5.0mV/div 10
to non-ideal performance in the two modes of operation
10
-2
10-16
fluctuation in the number of photoelectrons created
8
-1
above. An understanding of this will ultimately lead the 10
10
-3
user to the achievement of best performance through (photoelectron noise). There is an additional contribution
10-18
both selection of a tube with the right parameters and from the statistical variation in the number of secondary 108 106 104 102 10-0 10-2
pulse amplitude
operating the chosen device optimally. electrons created per incident electron on each dynode bandwidth f (Hz)
11 12
5. Limits on performance 5. Limits on performance
Note (11) is unaffected (this in fact is one of the major This holds true until the onset of feedback at high
attributes of photon counting) The equations become: tube type σ(g)/g
2
g2/g a /
δ1
1
(δ1 /(δ1 -1))2 5.6. Dark count spectrum gain where i q starts to increase more steeply. For
1 dc measurements, this suggests that for each tube
S/N = (MT) 2 ...(14) 6097 0.76 1.58 1.26 10 1.05
The dark count spectrum depicted in figure 21 shows a there is a window of optimum performance for signal/
9814 0.52 1.27 1.13 16 1.03 peak corresponding to the emmision of single electrons background. In the photomultiplier characterised by
while (12) becomes
from the cathode. Pulse height has been scaled in figure 22, this is at a gain of 2x107. To find the optimum
Ik 1 9924 0.57 1.32 1.15 10 1.05
S/N = 1 ...(15) photoelectrons equivalent, using the peak as reference. operating point it is sufficient to plot the ratio of signal
(2 e Ik f ) 2 a
table 5.4 Calculated noise factors by applying (17) to the curves of The contribution iq to dark current, from pulses in the (derived from a steady light source) to dark current, as a
where, for a multiplier which obeys Poisson statistics figure 19. The entries for δ1, in the penultimate column, are typical
spectrum of figure 21 is given by integration of the function of applied volts.
d1 gain values for tubes of these types.
[ ] δ 1
The figures for a in the last column are calculated from (16). curve where At low gain ID is dominated by leakage. For optimum
a= 2
...(16) ∞
δ 1 iq = ∫ n(q) q dq 0
...(18) performance in low light level applications, operate the
photomultiplier at high gain, but below the point a which
This is a theoretical expression assuming Poisson The interpretation and practical use of these noise factors where n(q) is the number of pulses/sec with charge q.
- dark current increases more rapidly than gain.
statistics apply at each dynode of mean gain δ. However, is straightforward. Referring to the 9814 photomultiplier, If I’ represents the contribution from leakage currents
in practice, the secondary emission process is poorly illustrated in Table 5.4, the S/N predicted by (11) and (13), flowing into the anode, due to applied biassing voltages
described by the simple statistical model assumed in which allows only for photocathode statistics, is further on the dynodes, then: 10- 7
(16). The measured output pulse height distribution degraded by 1/1.13 or 0.89 by the action of the multiplier.
A more detailed account of statistics and noise is given in ID = I + iq
/
...(19)
for a photomultiplier excited by single photons (and
10- 8 lD
hence single electrons at d1) does not conform to the the ET Enterprises Technical Reprint Series. where ID is the measured dark current. For a detailed
gain g
600 B&G 5.5. Dark current and dark count 10- 1 100
(solid lines) and 30 metres underground (broken lines) to illustrate fluctuations inherent in the signals shown in figures 16
2
[ ] [ ]
1 the contribution from cosmic radiation. Note the absence of the and 17 are due to statistical fluctuation in the number
var(g) 2 g2
1
2
a = 1+
/
= ...(17) cosmic ray peak at about 100 photoelectrons equivalent, when
g
2 2
1 of photoelectrons in each pulse the fluctuation in gain
g underground.
0 1 2 3 4 5 6 for each pulse, and a contribution from the dark count
time (min) Figure 22 illustrates a number of points of practical spectrum. It is important to understand the sources
significance. The dark current is dominated by the of background which contribute to figure 21. Such
figure 20 Dark current from a 9829B pmt with a thin quartz
leakage component at low gain. I’ varies linearly with knowledge helps explain why measured pulse height
window (at 2 x 10 7 gain). distributions deviate from ideal.
applied voltage while i q varies linearly with gain.
13 14
5. Limits on performance 5. Limits on performance
and photocathode. In a typical event, the initial big pulse from 5-2500keV. Note that there is a spectrum of
5.7. Sources of background
20
remains.
table 5.9 Background levels in photomultiplier windows.
equivalent cathode dark rate (/cm2/s)
Region C: 2<C<15 photoelectrons equivalent. Counts Decays per minute refer to a 50mm diameter window of weight 30g. 1
in region C are made up of afterpulses and from natural
radioactivity in the window.
5.10. Photomultiplier stability
There is a finite probability that a signal or background
pulse will be followed by a satellite or afterpulse. These The overall sensitivity of the photomultiplier is known 0.1
1 10 100 1000
are caused by ionisation of residual gases within the to vary: days
photomultiplier by the energetic electrons from the
initial pulse. If the ion reaches the photocathode it is with operating time while maintaining constant
figure 25 Illustrating long term stability (1 year) for SbCs
likely to cause a multielectron pulse; if intercepted illumination, applied voltage and ambient dynodes as a function of mean anode current, under
by a dynode then a contribution to region A or B conditions. conditions of constant applied voltage and illumination.
may result. Afterpulses are particularly undesirable on switching the output current, after removing and hysteresis
in correlation studies or applications where temporal re-applying either the light input or overall voltage. When a light source of constant intensity is interrupted
information is involved. The 9863B and 9130B ranges
temperature (oC) with change in ambient temperature and external and then re-applied, the photomultiplier may not
of photomultipliers provide performance with very low
electromagnetic fields. These environmental effects immediately recover its previous anode current level.
afterpulse rates. figure 23 The dark counts from a photomultiplier as a function
are discussed in section 6. The effect shown in figure 26 is attributed to charging
of temperature (above a threshold of 0.2 photoelectrons).
Traces of naturally occurring 40K, 238U and 232Th are effects within the photomultiplier which temporarily alter
stability the photomultiplier gain.
present in all window material, although ET Enterprises
selects special glasses to minimise these contributions.
5.9. Low background glass The action of drawing anode current alters the secondary
emission coefficient of the dynodes, in particular the
The decay of these isotopes produces light by Cerenkov
ET Enterprises manufactures photomultipliers with low later stages where current densities are highest. The overshoot
radiation and electrons by direct interaction with the
and ultra low levels of naturally occurring radioisotopes. slow variation of anode current with operating time is 10
photocathode. More detailed information on low
usually termed ‘drift’ and its magnitude depends on
-la(mA)
Their use is recommended in low background undershoot
background considerations is given in section 5.9.
the dynode surface (figure 24) and on the mean anode
Region D: D>15 photoelectrons equivalent. These very
scintillation counting, with both organic and inorganic
scintillators. In these applications the interaction of the current (figure 25). ~
~ ~
~
large pulses are a direct consequence of the passage of isotope decay products with the scintillator itself results The change in multiplier gain is not necessarily
cosmic rays (mainly muons and electrons) through the in large amplitude signals. permanent; if the photomultiplier is switched off it will
photomultiplier window. At sea-level this rate is about slowly recover its initial performance. Where stability is
level
Absolute levels of radionuclides, quoted in (ppm) or
light
15 per minute for a 51 mm diameter photomultiplier. of prime importance, the recommendation is to maintain on
Relativistic particles produce Cerenkov light emission in (ppb), are given in table 5.9. These figures refer to levels
the anode current below 1 μA. off
the windows. Since emission is biassed towards the uv, of activity in the window only. Although other parts of
0 1 2
the quartz window variants give the largest pulse height. the photomultiplier contribute, the window is the major time (ms)
In each event, between 15 and 200 photoelectrons are source.Total decays per minute given in the penultimate
produced depending on the window material, thickness column, refer to the total number of γ rays of all energies figure 26 (a) Sample curves illustrating hysteresis effects after
switching light input ON/OFF/ON. Photomultiplier type 9106B
operated at 10 4 gain. The overshoot is of the order of 0.5%.
15 16
5. Limits on performance 6. Environmental effects
dynodes dynodes
1 5 by (14). A low energy tail is always present in 55Fe application.
0.1 spectra. This stems from contributions originating in
0.01 4 regions C and D in the background spectra. 6.1. Temperature
on 2 was selected for low counts. With an unselected Photomultiplier dark current and dark count are
off tube, the noise can encroach to 1-2 keV and affect the critically dependent on temperature. This is illustrated
0 2 4
1 resolution. in figure 23 where for most cathode types there is
time (ms) a doubling in dark current every 5°C rise above room
0 0 (a) dia. 44mm x 44mm
10 101 102 103 2000 temperature. Dark current varies considerably even for
Res 6.9%
figure 26 (b) Sample curve illustrating memory effects after count rate (kHz)
photomultipliers of the same type. ET Enterprises can
switching light input OFF/ON/OFF. Photmulitplier type 9781B
Max provide photomultipliers selected for low dark current or
figure 27 The change in gain with increasing count rate is known
The cause of rate effect is not fully understood but what corresponds to the capture of the entire energy of the photoelectrons/keV is given in the second row. 0
1.0
of the window. Figure 34 illustrates the precautions
to be taken when a NaI(Tl) crystal is coupled to a
6.5. Shock and vibration
0.5
6.3. Electric fields photomultiplier. It is not acceptable practice to ground
Photomultipliers will withstand the demands of portable
the can of the crystal. Instead it must be maintained
0.0 Photomultiplier stability and lifetime are strongly instrumentation and everyday handling. There are,
at cathode potential, and in addition, insulated from
1.5 influenced by electric fields. It is important to distinguish however, levels of shock and vibration which will cause
box & grid 9924B the surroundings.
between external and internal electric fields. Internal mechanical failure or impair electrical performance.
-HV
1.0
electrical fields are generated in the glass envelope -------------------------------
10MΩ
Excessive shock can give rise to mechanical failure
0.5 and immediate surroundings of the photomultiplier by Nal(TI) should any support or electrical connection fracture.
relative output
photomultiplier
0.0 the application of high voltage. These fields are often crystal
Excessive vibration gives rise to mechanical failure of
O/P
responsible for erratic and unstable photomultiplier ------------------------------- anode the envelope or of the internal structure. In addition
1.5 linear focused 9102B behaviour. However their effect can be minimised by loose material may be produced which can accumulate
insulating epoxy or electrostatic/ voltage
1.0 application of appropriate shielding techniques. sleeve silicon grease magnetic shield divider and cause internal short circuits.
interface
0.5
Gain stability and dark counts are strongly affected by figure 34 Stable performance with negative high voltage is Shock and vibration can manifest itself at the anode
0.0 field gradients in the vicinity of the photocathode. This is achieved by eliminating potential gradients in the vicinity of the as microphony – caused by changes in electrode
1.5
illustrated in figure 33 where dark counts are recorded photocathode. The electrostatic shielding and the can of the crystal capacitance from movement of electrodes. In addition,
venetian blind 9956B are both maintained at cathode potential by this arrangement.
as a function of envelope potential. In this investigation, where ac coupling is used, the coupling capacitor
1.0 a cylindrical screen covering the envelope, but leaving itself may be microphonic. The use of charge sensitive
Important
0.5 the window free, was varied in potential using a Any conductor in contact with the window or amplifiers minimises these effects. Mu-metal shields
separate power supply. The effect on performance envelope of the photomultiplier must be connected are recommended to eliminate the effects of changing
0.0
-20 -10 0 10 20 is dramatic and represents typical behaviour when to cathode potential. When the photomultiplier magnetic fields. When operated at negative high voltage,
photomultipliers are operated under these conditions. is operated at cathode negative with respect to external insulators or potting materials in contact with
applied field (tesla) x 10-4
ground, a 10 M Ω safety resistor between -HV and the envelope may breakdown electrically or produce
the shield is advised. Only very good insulators, light. The method of mounting a photomultiplier is
figure 30 Relative output as a function of external magnetic field. critical: encapsulation can be designed to damp external
e.g. PTFE, should be brought into direct contact
(for photomultipliers operated at nominal gain).
with the photomultiplier window. All too often, shock and vibration.
Note: Earth’s Field is shown in darker colour.
19 20
6. Environmental effects 7. Choice of photomultiplier
Flexible potting materials applied over the full length At first sight this appears to be a formidable task
of the photomultiplier are generally recommended for 6.6. Exposure to daylight because the number of variants offered by ET
7.3. Electron multiplier structure
radial support. Axial load is recommended to prevent Enterprises. However, the choice is rapidly reduced
movement of the window with respect to the housing. The photomultiplier is an extremely sensitive light to within a few, or just one device, when external The choice is often restricted by the diameter of the
detector and should not be operated under daylight factors and performance characteristics of particular photomultiplier; the widest variety of multipliers is
Ruggedized glass and quartz photomultipliers can available in 52 mm diameter tubes.
or room lighting levels. Even when non-operational, photomultipliers are considered. External factors such
be specially manufactured to meet severe operating
exposure to daylight or normal lighting levels causes as the area over which the light extends and physical
conditions experienced for example in military, oil well The linear focused structure is recommended in pulsed
photocathode excitation, especially under uv illumination. restraints on overall dimensions usually play an
logging and space launch applications. ET Enterprises light applications for fastest time response, for best
Although this does not damage the tube, it causes an important role in selection.
will design, qualify, manufacture and test devices linearity and for highest available multiplier gain.
increased dark current (count) level. Initially this may
to specific project requirements. After qualification Performance measurements requirements such as:
be several orders of magnitude higher than the final The circular focused structure is the most compact
it is normal practice to test all project devices at an
dark current value. The effect is particularly pronounced spectral response design with good magnetic immunity, good timing, good
acceptance level, which is normally half of the random
in photomultipliers with quartz windows. Although linearity but is limited in the number of stages available.
vibration qualification level. gain
the photomultiplier can be used immediately after
ET Enterprises has extensive environmental test exposure without any effect on overall sensitivity, the dark current The venetian blind structure is an excellent general
facilities including sinusoidal and random vibration, dark current will not settle to its ultimate value for many purpose type with a wide choice in the number of
linearity stages.
shock, thermal and vacuum test. Representative shock hours. Figure 36 illustrates the dark current decay of
and vibration tests for ruggedized tubes are shown in a 9829QB. For these reasons it is recommended that speed of response
The box and grid structure is most commonly used in 30
table 6.5 while figure 35 illustrates power spectral tubes are not exposed to daylight or fluorescent room
invariably reduce the choice to within a few types. mm diameter photomultipliers and is suited to a variety
density curves for random vibration. The frequency lights but are loaded into instruments under subdued
of applications from high light levels down to the photon
range, acceleration values and test duration are agreed lighting conditions.
counting region.
with each customer.
7.1. End window or side window
table 6.5 Shock and vibration limits for ruggedized glass and metal
ceramic photomultipiers. These limits apply for all orientations, 1000 The overall performance offered by the two types 7.4. The number of dynodes
with the photomultiplier non operational. The frequency profile usually favours the selection of the end window form. In
shown in figure 35 applies to random vibration. It is important to choose a photomultiplier with the
end window tubes there is a wide choice of active areas,
dark current (nA)
parameter test glass/quartz metal ceramic 100 photocathode types and window geometry. Gathering correct number of stages. When too many stages
types types
the maximum light flux on to the photocathode from are chosen to provide the required gain, the electrical
^g 30 250 remote, diffuse and directly coupled light sources is performance is degraded. This is a consequence of
shock duration 11 ms 2 ms
shocks/axis 3 30 10 best with an end window. The side window types are low inter-electrode voltages, causing poor linearity and
(1/2 sine)
recommended when space is limited or when light timing. If too few stages are chosen, there may be
sinusoidal frequency 10-32 Hz 10-40 Hz levels are very high. insufficient gain available within the constraints of the
vibration amplitude (p-p) 10 mm 10 mm
frequency 30-2000 Hz 40-2000 Hz 1 maximum allowed operating voltage. Operation beyond
^g 0.01 0.1 1 10 100
20 30 the maximum overall sensitivity will result in excessive
random
sweep rate
maximum
2 octave/min
0.6 g2/Hz
2 octaves/min
2 g2/Hz
hours
7.2. Spectral response dark current, breakdown and short lifetime.
vibration composite 20g rms 36g rms
duration/axis 2 mins 10 mins figure 36 Dark current decay of 9829Q after exposure to daylight For low light intensities, down to photon counting levels,
The most suitable photocathode type is the one that has
for 1 hour (non-operating). 10 to 14 stages will be required. For intermediate light
the maximum quantum efficiency over the wavelength
levels, 8 - 12 stages are usually sufficient. High light
100 range emitted by the light source. At low light levels
levels require between 6 and 10 stages of gain. An aid
6.7. Helium the signal-to-background ratio may be more important.
to selecting the right number of stages is presented in
At high light levels consideration should be given to the
10
table 7.4.
maximum cathode current allowed. This is discussed in
power spectral density (g2/Hz)
db
/o
oc
ct
b/
7.5
6d
21 22
8. Voltage dividers 8. Voltage dividers
A series of voltages is applied to the photomultiplier to: A resistive divider poses a power dissipation problem,
7.5. SbCs or BeCu dynode surface accelerate and focus photoelectrons on to d 1; accelerate 8.2. Pulsed applications if properly designed in accordance with (22), for high
and focus the secondary electrons between successive count rates. Practical dividers, ideal for high count rates,
At any applied voltage SbCs dynodes have higher gain dynodes and to collect the secondaries from the last Some photomultipliers are capable of handling peak either incorporate zener diodes or transistors in the final
than BeCu ones but this is not always an advantage. dynode at the anode. It is common practice to derive the currents of the order of 200 mA. To achieve this level of stages. These provide the required level of performance
Since the time response depends on applied voltage, voltages from a single supply using a resistive divider performance requires a divider that differs from that in without having to draw excessive divider current.
BeCu is recommended for time-critical pulsed light network. figure 37 in two important respects:
applications. In addition, it has better pulsed current
i) the potentials applied to the last few stages must be
linearity. Voltage divider design is critical to achieving optimum
increased to overcome space charge effects;
8.3. Active dividers
performance from a photomultiplier. An outline of
SbCs requires lower operating voltages. It also has design considerations is given in this section but for
better hysteresis and rate effect performance due to ii) capacitors are added to supply the charge pulse. Active dividers (figure 39) consist of a series of FETs
a detailed account, the reader is referred to ‘Voltage
its low surface resistivity, enabling it to settle quickly With NaI(Tl) applications it is common to use positive interposed between the dynodes and a high impedance
Divider Design’ RP069 6. All photomultipliers will operate
between changing light levels. Consequently it is the high voltage with the cathode at ground potential. resistive divider. The FETs hold the dynodes at fixed
with the linear divider, shown in figure 37, although
best choice in optically chopped light applications such The coupling capacitor, C, shown in figure 38, potentials. This type of divider ensures constant gain
not necessarily with ideal performance. For example, if
as colour film scanning. Further details are given in isolates the dc potential at the anode from the up to a mean anode current of 100uA, the maximum
linearity and timing are of prime importance, it will be
sections 4.3 and 5. measuring electronics. In high energy physics permitted in most pmt types. Power consumption is
necessary to modify the basic configuration of figure
applications it is usual to couple the anode directly typically 100 mW and they are particularly suited to
37. Voltage dividers actually used by ET Enterprises for
to external electronics to avoid base line shift. portable instrumentation.
testing purposes are listed in the photomultiplier
7.6. Electrical performance data sheets.
k a
As explained in section 5, various limitations apply in Voltage divider networks are available to suit most
photomultiplier performance which may further influence photomultiplier tube types and applications. Ask for
the choice of photocathode, multiplier structure and publication ‘Voltage Divider Networks’ RP085 7.
number of dynodes. Reference should be made to this 3R R R
---
R 1.25R 1.5R 2R 3R R2
section before making the final choice. C
8.1. DC applications
Having chosen the most suitable photomultiplier, it is C1 C2 C3 R1
1nF 1nF 10nF
o/p
recommended that a prototype of the optical detection Direct current applications require negative high voltage +HV
system is built. It is important to confirm the mean applied at the cathode, leaving the anode at ground
o/p
cathode current, the mean and peak anode current and potential for ease of interfacing. The choice of R in
the range of operating voltages. ‘Order of magnitude figure 37 is governed by the maximum anode current
calculations’ are sufficient for this purpose. It is often Ia (max) the application demands, subject to Ia (max) figure 38 A tapered divider suitable for pulse applications. In this figure 39 Active, uniform divider providing high performance and
found that the light levels are either much higher or example, positive high voltage is used. R 2 is the load resistor, low power consumption over a wide dynamic range of operation.
always <100 μA. R should be chosen to satisfy the 10KΩ -1M Ω. R1=10MΩ safety resistor.
much lower than expected and re-selection of the most relationship Ia (max)<0.01IHV, where I HV is the divider
suitable photomultiplier may be necessary. current. This will ensure that gain linearity is preserved
As with the linear voltage divider of figure 37, R is
Manufacturers are continually upgrading equipment up to anode currents of I a (max). Once Ia (max) exceeds
chosen to satisfy
performance, which may place additional demands on one tenth IHV, serious deviation from linearity will -
become apparent. IHV > 100 Ia ...(22)
the photomultiplier. Engineers making changes which
affect the operating voltages or the currents flowing in Where Ī a is the mean anode current
the photomultiplier should always re-confirm that the k a
The degree of taper on the last few stages of the
photomultiplier remains within its performance limits.
divider depends on the tube type and on the level of
When re-design takes place, it is advisable to contact
performance required.
ET Enterprises to keep up to date with the latest
products. R R R
---
R R R R R R1 The size of decoupling capacitors can be calculated
from the basic relationship for a capacitor: Q = CV. It
is sufficient to demand that the inter dynode voltage
-HV o/p
remains constant to ~0.1%, that is ΔV/V ≤0.001, while
peak charge ΔQ is drawn. If C is chosen to satisfy
Q/V
C> ...(23)
V/V
figure 37 A linear voltage divider for dc applications. R may be
the divider will cope with short bursts of pulses of
in the range 10 KΩ -10 M Ω. R1=10 M Ω is a safety resistor, which
prevents the anode floating to dn potential when the external load variable magnitude. It is important to appreciate that
is removed. R1 is often omitted because it can cause preamplifier decoupling capacitors only cover short bursts of pulses.
offsets. A sustained high count rate leads to a high mean anode
current I a and there is the possibility that the relationship
(22) will be violated.
23 24
9. Photon counting 10. Further information
There are three modes of photomultiplier operation, References available on request
100
analogue (current mode detection) 106 1) Gating of Photomultipiers (RP 061)
(a)
75
pulse counting and encoding
relative counts
signal plateau
region
2) Monte Carlo Simulation of Photomultiplier
photon counting 50 Resoulution (RP 080)
105
signal / background
Information on both the intensity and the time signature ratio 3) An Investigation of Photomultiplier Background
25 (c)
of a light source can be obtained by using the photon (RP 075)
counting method. 104
0 4) Integral mu-metal* Shields (RP084)
counts (s-1)
characteristic of a photomultiplier is utilised - the ability known as the single electron response (SER)
(b)
to detect single photons over a large photocathode area 7) Voltage Divider Networks (RP085)
Photon counting is that mode of operation where each 102 10
In practice, photon counting is done using a circuit with
We have a worldwide network of representatives.
detected photon is separately time resolved at the anode a single fixed threshold as opposed to a multichannel
Please contact us for details of your local
of the photomultiplier. This is illustrated in figure 40 analyser. An amplifier-discriminator combines a fast
representatives.
where the intensity of the incident light is sufficiently amplifier and a fixed threshold discriminator with an 101 (d)
1
low that there is no overlapping in the sequence of overall sensitivity of the order of one millivolt. Deciding
detected photons. Because the multiplier gain process upon the optimum operating point of a photon detection afterpulses
is statistical in nature, events which all start as single system depends on the nature of the application and 100 0.1
electrons at the cathode produce a range of output pulse to some degree on personal judgement. It is standard
heights. The narrower the spread in pulse heights, the practice to record signal and background curves as suggested
better suited is the photomultiplier to photon counting. shown in figure 42. Noting that the output pulses from 10-1
operating point
0
0.5 1.0 1.5 2.0 2.5
the photomultiplier have a range of size, characterised
A photon counting photomultiplier is one with a well
by the SER, the shape of the signal curve is explained overall voltage (kV)
defined single electron peak, shown in figure 41. This
as follows. Few counts are recorded at low applied HV
is a pulse height distribution, measured at fixed gain,
because the gain of the photomultiplier is insufficient figure 42 Finding the optimum operating voltage for photon
using a source of single photons and a multichannel
to produce a significant number of pulses that exceed counting. The steeply rising afterpulse curve suggests a preferred
analyser. A characteristic of a single electron response, operating point to the left of the Signal/Background plateau region.
the threshold (region ). As the HV, and hence gain,
SER, spectrum is that increasing or decreasing the light
is increased to 1.2kV, about half the output pulses
intensity changes the area under the curve, but not the
exceed the threshold and all pulses to the right of current mode detection
position of the peak.
the peak in the SER contribute to the measured Every discriminator circuit has an intrinsic dead time.
count rate (region ). As the gain is further increased This refers to the period the circuit requires to deal with
time base = 0.2μs/div most of the distribution in the SER is counted and a each signal. Should a second pulse appear during this
voltage = 5.0mV/div
plateau is reached (region ). Further increase in HV period then it is lost to the count rate. In theory it is
results in a slight increase in counts until the onset of possible to allow for the affects of overlapping pulses by
photomultiplier breakdown (region ). Note that the correcting the measured rate, n. If T is the dead time for
plateau characteristic is in effect an integration of the an ideal, non-paralisable, amplifier-discriminator then the
pulse amplitude
SER moving from right to left across figure 41. Also true rate N is
note that gain follows a power law with respect to HV
and and the gain span across the plateau characteristic is N = n / (1 - nT) ...(24)
typically two decades in magnitude. All photomultipliers Discriminators are not ideal and the correction becomes
produce unwanted afterpulses. Curve (d) shows that the unreliable if the dead time exceeds 50%. For a dead
afterpulse rate is a strong function of HV. time of 100 ns, for example, the correction will exceed
Good experimental technique suggest that operating at a 10% at count rates in excess of 1 MHz. At such high
point on the plateau that maximises signal/background is count rates, the statistical advantages associated with
sound practice. This ratio, shown in figure 42, indicates photon counting become less important and it may be
a wide window of acceptable performance. Taking desirable to resort to current mode detection.
account of the benefits of operating on the flattest part
figure 40 Photomultiplier output produced by a source of single
photons.
of the signal curve (providing stable performance against
gain changes) and the steeply increasing afterpulse rate,
leads to the suggested operating point indicated.
*mu-metal
*mu-metal isis aa registered
registered trademark
trademark of
ofMagnetic
MagneticShield
ShieldCorporation.
Corporation.
25 26
Contact
ET Enterprises Limited
Riverside Way
Uxbridge, UB8 2YF, UK
Tel: +44 (0)1895 200880
Fax: +44 (0)1895 270873
Email: [email protected]
www.et-enterprises.com
ref: upmt/11
27