Signal to Noise

SIGNAL TO NOISE IN MASSLYNX
Calculation of Signal to Noise in the Chromatogram

view and TargetLynx
©2017 Waters Corporation COMPANY CONFIDENTIAL 1

Signal to Noise in Chromatogram view

Introduction
 Signal to Noise ratios are used in various places for different

applications
 Chromatogram Window
– Used for instrument specification testing
 QuanLynx/TargetLynx Window
– Used to determine data quality for QC purposes

Chromatogram Window
 As an example let’s take a look at a file within MassLynx

named PEST03 under DEFAULT.PRO

Ranges of Signal and Noise
 Scan numbers for signal and noise can

be entered by hand or by dragging on the
chromatogram (right mouse)

Ranges of Signal and Noise
 The result shows the calculated S/N value and the regions used for
the calculation

Noise Region
 The noise region has a horizontal line plotted through the mean value
of all the used points in the noise region

Signal Region
 The signal region has a horizontal line plotted through the whole
signal region at half the height of the signal above the noise

Signal Region
 The signal is found by taking the highest point in the signal range and
measuring its height above the mean noise level.

Noise Processing: NO Extra Processing and
Peak to Peak Display
 The calculation of the noise value will

vary depending on the options
specified. Lets take the simplest case
first.
 This will use all the data points in the
noise region
 The noise value will be the difference
between the highest noise point and
the lowest in the range

Noise Processing: NO Extra Processing
 This method of calculating gives the biggest possible value for noise,
and thus the lowest signal to noise
 S/N PtP 34.88

Noise Processing: NO Extra Processing and
Ignore Zeros
 Removing ‘spurious’ outliers from

the noise region is an accepted
strategy.
 Ignore Zeros will not include any
points of zero intensity in the noise
calculations.
– This will not make any difference in
our example, but if there are data
dropouts or the instrument detection
has been thresholded, this can
have a dramatic effect.

Noise Processing: Ignore Scans
 Other outlier removals can be

applied after zero removal (if any)
 These are all very similar

– Ignore worst 5% of scans
– Ignore scans outside +/- 1 SD
– Ignore scans outside +/- 2 SD
 …but subtly different.

Noise Processing: Ignore worst 5 % of Scans
 Removal of 5% of scans: removes exactly 5% of the scans in the

noise range that are furthest from the mean and then recalculates the
mean noise and noise strength without using these points. (S/N PtP
39.07)

Noise Processing: Ignore Scans outside +/-1
SD or +/-2 SD
 Removal of +/- 1 or 2 SD removes points outside of the specified

range and then recalculates the mean noise and noise range.
+/-2SD should
remove
+/-1SD should around 5%
remove of points
around 32% (S/N PtP
of points 36.39)
(S/N Ptp
66.11)

Noise Processing: Ignore Scans outside +/-1
SD or +/-2 SD
 For noise that has a gaussian distribution around the mean noise
value, 5% removal and +/- 2SD should give similar values (39.07 and
36.39 for this example)
 The more skewed the noise, the smaller the S/N value for +/- 2SD
removal will become relative to the 5% outlier removal.

S/N Calculation: Use Underlying Raw Data
 In SCN509 and MassLynx 4.1,

there is the option to NOT use the
raw data points.
 By deselecting this option a
smooth can be manually applied to
the data before calculating signal to
noise

S/N Calculation: Use Underlying Raw Data or
Smoothed Data
 Applying a basic smooth has a significant effect
 Smoothed: 50.06
 Without smooth: 34.88

S/N Calculation: Use Underlying Raw Data or
Smoothed Data
 The noise range has been reduced by the smooth

S/N Display: RMS
 RMS processing works exactly the

same as PtP processing except that
the final noise value is determined
by the Root Mean Square
 The calculation is done in the
normal manner
(x
j 1
j  x )2
N
 Unless ‘Use (n-1) for RMS’ is
selected (preferred)
 (x
j 1
j  x )2
N 1

S/N Display: RMS
 Technically, RMS is the N version, and Standard Deviation is the N-1

version. The prior to be used if the mean is known ‘a priori’.
PtP Noise
RMS Noise
(~1SD)

S/N Display: RMS
 RMS S/N with no extra point removal is approximately twice the S/N
of PtP with removal of outliers +/- 1SD as we would expect.

Summary of S/N by RMS and PtP
 Showing the influence of outlier removal

– All on Pest03
– Signal scans: 400:515; Noise scans 14:120
– Using raw data
– Using n-1 for RMS calculations
PtP RMS
Nothing 34.88 129.63
5% 39.07 140.03
1 SD 66.11 228.67
2 SD 36.36 131.91

Signal to noise in QuanLynx

S/N in QuanLynx
 In the QuanLynx program a different set of S/N calculations are performed.

 These are coded differently and so will have slightly different rounding errors
even if the options are selected that should give the same result
 The ‘use n-1’ option is effectively always selected

S/N in QuanLynx method editor
 The parameters used for S/N values are set up in the method editor

S/N in QuanLynx method editor: RMS or PtP
 PtP or RMS noise can be used

S/N in QuanLynx method editor: Noise
calculation factor
 The noise will be multiplied by this factor.

 The calculation of S/N will be Signal / 3 x (Standard Deviation of noise), for
RMS.
 The factor should be 1 in the case of PtP method.

S/N in QuanLynx method editor:
Determination of noise region
 Noise start and end RT can be selected
– This is equivalent to the noise range selection in the chomatogram window
 If both left at zero, automatic S/N detection will be applied (see later)

S/N in QuanLynx method editor: Measure of
peak signal
 The method of detecting the peak height can be selected

Measure the peak signal from baseline
 Peak Baseline
– Recommended way to do it

Measure the peak signal from Average Noise
Level
 Average Noise Level

– The way it is done in the Chromatogram view

Measure the peak signal from zero intensity
 Zero Intensity
– The absolute height of the peak

Obtaining the same S/N in MassLynx and
QuanLynx
 Under MassLynx, Chromatogram view:
– Use the same smoothing parameters

in both MassLynx and QuanLynx.
– Select ‘NO Extra Processing’
– Use the same PtP or RMS value as in
the QuanLynx setup.
– Deselect ‘Use Underlying Raw Data’
– Select ‘Use (n-1) for RMS’.

QuanLynx
 Under QuanLynx method editor:
– Set the same ‘Signal-to-noise method’ as in the chromatogram view.

– Set the ‘noise calculation factor’ to 1 (PtP) or 3 (RMS).
– Specify the same RT for the noise as used in the chromatogram view.
– Select ‘Avg. Noise Level’.

QuanLynx
45.97
46.05
 The numbers are only

different due to
rounding errors in the
different pieces of code

Using Automated Noise Processing in
QuanLynx
 Under the method editor

– If we change the noise range to 0.00, 0.00, then the automated noise processing
will be applied.
– You will get a significant bigger number for S/N
Now S/N is
287.63

QuanLynx
 This is because it is using the following way of calculating noise

– Take the whole RT of the trace
– Using a second derivative measure, discard any ranges that could be described as
a peak
– For each range left that is designated as noise, take the RMS or PtP deviation not
from the average height, but from a line drawn through the noise.
 This results in a technically correct derivation of noise, but
– The regions used to calculate the noise for a given trace are unknown
– Peak-like noise will be discarded

QuanLynx
 See the tiny red lines for an example of noise ranges that might be
chosen.
– You can see why the S/N will be higher this way

Summary
 S/N can be obtained in two ways:

– S/N (PtP) = peak height / peak to peak amplitude of the noise
– S/N (RMS) = peak height / RMS of the noise
 S/N can be made manually in MassLynx from the Chromatogram
view.
 S/N can be made automatically in QuanLynx.
 S/N in MassLynx and QuanLynx will produce the same values
provided the appropriate parameters are used.

Signal to Noise

Uploaded by

Copyright:

Available Formats

Signal to Noise

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Signal to Noise

Uploaded by

Copyright:

Available Formats

SIGNAL TO NOISE IN MASSLYNX

Calculation of Signal to Noise in the Chromatogram

©2017 Waters Corporation COMPANY CONFIDENTIAL 1

©2017 Waters Corporation COMPANY CONFIDENTIAL 2

 Signal to Noise ratios are used in various places for different

©2017 Waters Corporation COMPANY CONFIDENTIAL 3

 As an example let’s take a look at a file within MassLynx

©2017 Waters Corporation COMPANY CONFIDENTIAL 4

 Scan numbers for signal and noise can

©2017 Waters Corporation COMPANY CONFIDENTIAL 5

©2017 Waters Corporation COMPANY CONFIDENTIAL 6

©2017 Waters Corporation COMPANY CONFIDENTIAL 7

©2017 Waters Corporation COMPANY CONFIDENTIAL 8

©2017 Waters Corporation COMPANY CONFIDENTIAL 9

 The calculation of the noise value will

©2017 Waters Corporation COMPANY CONFIDENTIAL 10

©2017 Waters Corporation COMPANY CONFIDENTIAL 11

 Removing ‘spurious’ outliers from

©2017 Waters Corporation COMPANY CONFIDENTIAL 12

 Other outlier removals can be

 These are all very similar

 …but subtly different.

©2017 Waters Corporation COMPANY CONFIDENTIAL 13

 Removal of 5% of scans: removes exactly 5% of the scans in the

©2017 Waters Corporation COMPANY CONFIDENTIAL 14

 Removal of +/- 1 or 2 SD removes points outside of the specified

©2017 Waters Corporation COMPANY CONFIDENTIAL 15

©2017 Waters Corporation COMPANY CONFIDENTIAL 16

 In SCN509 and MassLynx 4.1,

©2017 Waters Corporation COMPANY CONFIDENTIAL 17

 Applying a basic smooth has a significant effect

©2017 Waters Corporation COMPANY CONFIDENTIAL 18

 The noise range has been reduced by the smooth

©2017 Waters Corporation COMPANY CONFIDENTIAL 19

 RMS processing works exactly the

©2017 Waters Corporation COMPANY CONFIDENTIAL 20

 Technically, RMS is the N version, and Standard Deviation is the N-1

©2017 Waters Corporation COMPANY CONFIDENTIAL 21

©2017 Waters Corporation COMPANY CONFIDENTIAL 22

 Showing the influence of outlier removal

Nothing 34.88 129.63

©2017 Waters Corporation COMPANY CONFIDENTIAL 23

©2017 Waters Corporation COMPANY CONFIDENTIAL 24

 In the QuanLynx program a different set of S/N calculations are performed.

©2017 Waters Corporation COMPANY CONFIDENTIAL 25

©2017 Waters Corporation COMPANY CONFIDENTIAL 26

 PtP or RMS noise can be used

©2017 Waters Corporation COMPANY CONFIDENTIAL 27

 The noise will be multiplied by this factor.

©2017 Waters Corporation COMPANY CONFIDENTIAL 28

©2017 Waters Corporation COMPANY CONFIDENTIAL 29

 The method of detecting the peak height can be selected

©2017 Waters Corporation COMPANY CONFIDENTIAL 30

©2017 Waters Corporation COMPANY CONFIDENTIAL 31

 Average Noise Level

©2017 Waters Corporation COMPANY CONFIDENTIAL 32

©2017 Waters Corporation COMPANY CONFIDENTIAL 33

 Under MassLynx, Chromatogram view:

– Use the same smoothing parameters

©2017 Waters Corporation COMPANY CONFIDENTIAL 34