WDS Detector Dead Time calibration acquisition now in StartWin application

[John Donovan]

As many know, proper calibration of the WDS detector dead time constants is important for accurate quantitative analysis, particularly at high beam currents, where count rates can often reach thousands of counts per second.

Previously, calibration of the WDS detector dead time constants was tedious, requiring many manual measurements at multiple beam currents. A few years ago, Paul Carpenter and I automated much of the process by utilizing an Excel spreadsheet macro which calls the Remote Automation Server interface to automatically acquire these calibration data from the instrument. Subsequently, this data is then pasted into another Excel spreadsheet where it is fitted and plotted to calculate the actual dead time constants for each WDS detector, which can then be utilized in quantification software such as Probe for EPMA. These Excel macro acquisition and calculation spreadsheets are posted here:

https://probesoftware.com/smf/index.php?topic=1038.msg6890#msg6890

A full discussion on editing the PFE SCALERS.DAT file (which is where these dead time constants are stored), is here:

https://probesoftware.com/smf/index.php?topic=394.0

Recently however, because Microsoft has increasingly "locked down" the execution of Excel macros and ActiveX servers with each new operating system, Paul Carpenter has suggested that we implement a method for acquiring the necessary dead time calibration data in one of the Probe for EPMA applications.

The Remote ActiveX server is still fully functional and works fine under Windows 7 and Windows 10, but for ease of use we decided that providing a "built-in" method for acquisition of the dead time calibration intensities would be useful.  Therefore the latest version of Probe for EPMA (v. 12.5.5) now includes a new version of StartWin which contains a new button for acquisition of the dead time calibration intensities, as seen here:



To perform a dead time calibration on your instrument simply tune up your WDS spectrometers to a peak of high intensity, usually Si Ka on Si metal using PET and TAP crystals, or Ti Ka on Ti metal using LiF and PET crystals. 

On Cameca instruments, one should set the "enforced" electronic dead time values to as low a value as possible (e.g., using the Cameca SX Configuration software or editing the Cameca "integer" dead times in the SCALERS.DAT file), in order to measure the actual "intrinsic" dead times on each detector, and then selecting an integer dead time slightly larger than the "intrinsic" dead time of the detector. And finally acquiring the actual dead time of the specified "enforced" dead times as described in this post:

https://probesoftware.com/smf/index.php?topic=33.msg2153#msg2153

Note that on the SX100, one can set the integer dead time to zero, but on the SXFive one cannot set it to zero, so simply set it to 1 micro-second. The "intrinsic" dead times on Cameca instruments are around 3 micro-seconds so that is OK.

On JEOL instruments, simply acquire the dead times, since there is no "enforced" electronic dead time feature present to deal with.

Then simply launch the new StartWin application, click the DeadTime button and specify a starting and stopping beam current. The software will then automatically acquire the dead time calibration intensities in the correct format for pasting into the Excel dead time calculation spreadsheet written by Paul Carpenter. 

Note that examples of these Excel acquisition macro and calculation spreadsheets are installed to the Probe Software Remote application folder (usually C:\Program Files (x86)\Probe Software\Remote), but remember, these Excel spreadsheets *must* be copied to a "writable" folder such as the Documents folder, in order that they can be updated (all Windows application folders are always read only!). 

To utilize this new dead time calibration acquisition feature in StartWin, please update as usual from the Probe for EPMA Help menu and please let us know what you think!

[John Donovan]

We made a small tweak to the dead time calibration acquisition code in StartWin last night.

Because Paul Carpenter's Excel spreadsheet for calculating the dead time constants is formatted for 5 WDS spectrometers, the cut and paste operation from StartWin to Excel works perfectly for instruments with 5 WDS spectrometers.

But if your instrument has fewer than 5 spectrometers you will need to do some additional copying and pasting to allow the embedded formulas in his Excel spreadsheet to work properly.

So we decided to modify the StartWin dead time calibration acquisition code and automatically generate additional columns for the missing WDS spectrometers, if your instrument has less than 5 WDS spectrometers as seen here:



Question: are we making it just all too easy for everyone?   

[Probeman]

I just wanted to discuss an issue that I believe is largely overlooked in the EPMA community.   And that is the calibration of the dead time constants on our JEOL and Cameca WDS spectrometers. 

This discussion is important for everyone to read, even if you do not have the Probe for EPMA software. Yes, I will be discussing it in terms of the Probe for EPMA software because by using the StartWin utility these dead time calibrations are very easy to do as they are almost entirely automated and only take an hour to perform!

If you don't have the Probe for EPMA software, these calibrations are still fairly easy to do using Paul Carpenter's Excel spreadsheet and ReadMe file, found here as attachments (login to see attachments):

https://probesoftware.com/smf/index.php?topic=1160.msg7924#msg7924

If these dead time constants are not properly calibrated, attempts to extrapolate from a high concentration standard to a low concentration (high intensity to low intensity) will be problematic. This problem only gets worse with increasing beam currents and today this issue is made even worse by the use of "large area" crystals on both JEOL and Cameca instrument, that yield fantastic count rates, but depend heavily on proper calibration of the WDS spectrometer dead time constants.

The dead time calibration constants used by Probe for EPMA are defined in the SCALERS.DAT file in several places so it is good to review this information. Originally (many moons ago), the dead time constants were defined on line 13 of the SCALERS.DAT Probe for EPMA configuration file as seen here:

"1"      "2"      "3"      "4"      "5"     "scaler labels"
 ""       ""       ""       ""       ""      "fixed scaler elements"
 ""       ""       ""       ""       ""      "fixed scaler xrays"
 2        2        2        2        2       "crystal flipping flag"
 81010    81010    81010    81010    81010   "crystal flipping position"
 4        2        2        4        2       "number of crystals"
 "PET"    "LPET"   "LLIF"   "PET"    "LIF"   "crystal types1"
 "TAP"    "LTAP"   "LPET"   "TAP"    "PET"   "crystal types2"
 "PC1"    ""       ""       "PC1"    ""      "crystal types3"
 "PC2"    ""       ""       "PC25"   ""      "crystal types4"
 ""       ""       ""       ""       ""      "crystal types5"
 ""       ""       ""       ""       ""      "crystal types6"
3.8      3.8      3.5      3.5      3.5     "deadtime in microseconds"
150.     150.     140.     150.     140.     "off-peak size, (hilimit - lolimit)/off-peak size"
 80.      80.      70.      80.      70.     "wavescan size, (hilimit - lolimit)/wavescan size"
 100.     100.     90.      100.     90.     "peakscan size, (hilimit - lolimit)/peakscan size"
 100      100      100      100      100     "wavescan steps"
 50       50       50       50       50      "peakscan steps"

That is, one dead time constant for each spectrometer on line 13.

But eventually it was suspected that there might be a small energy dependence of the dead time calibration, so additional lines were added to the SCALERS.DAT file so that different dead times could be specified for each crystal on each spectrometer as seen here in lines 72 to 77:

0.       0.       0.       0.       0.    "default PHA gain5"
 0.       0.       0.       0.       0.    "default PHA gain6"
 1300.    1290.    1890.    1300.    1800. "default detector bias1"
 1300.    1300.    1850.    1300.    1850. "default detector bias2"
 1500.    0.       0.       1480.    0.    "default detector bias3"
 1500.    0.       0.       1460.    0.    "default detector bias4"
 0.       0.       0.       0.       0.    "default detector bias5"
 0.       0.       0.       0.       0.    "default detector bias6"
 1        1        1        1        1     "default PHA inte/diff modes1"
 1        1        1        1        1     "default PHA inte/diff modes2"
 1        0        0        1        0     "default PHA inte/diff modes3"
 1        0        0        1        0     "default PHA inte/diff modes4"
 0        0        0        0        0     "default PHA inte/diff modes5"
 0        0        0        0        0     "default PHA inte/diff modes6"
3.8      3.8      3.5      3.5      3.5   "default detector deadtimes1"
3.8      3.8      3.5      3.5      3.5   "default detector deadtimes2"
3.8      0        0        3.5      0     "default detector deadtimes3"
3.8      0        0        3.5      0     "default detector deadtimes4"
0        0        0        0        0     "default detector deadtimes5"
 0        0        0        0        0     "default detector deadtimes6"
 0        1        1        0        0     "Cameca large area crystal flag1"
 0        1        1        0        0     "Cameca large area crystal flag2"
 0        0        0        0        0     "Cameca large area crystal flag3"


Basically, one can utilize the detector dead time constants defined on line 13 only, or if the crystal specific values on lines 72 to 77 are non-zero, the program will utilize those instead.

In addition it should be mentioned for those that have Cameca instruments, that there is an additional "enforced" *integer* dead time value defined on line 35 in the SCALERS.DAT file, which should be set to 2 or 3 or 4 depending on the age of the instrument (typically 3 for new instruments and 4 for old instruments).

This Cameca specific issue of "enforced" dead times is discussed here:

https://probesoftware.com/smf/index.php?topic=33.0

This whole dead time issue is of course even more critical for Cameca instruments because they tend to have dead time constants of around 3 to 4 microseconds or so, while JEOL instruments tend to have dead time constants around 1 to 2 microseconds or so.  That is when both are new!

I bring this dead time calibration issue up because it is critically important to *not* rely on the manufacturer's "default" dead time constants, and instead perform these calibrations yourself, especially as the instrument ages and the dead time constants increase over time due to contamination.

Especially when performing consensus k-ratio measurements as the FIGMAS group is attempting to do, with a number of high purity synthetic materials as described here:

https://probesoftware.com/smf/index.php?topic=1415.0

My own opinion is that performing these k-ratio measurements without proper dead time calibrations will yield inaccurate results. Especially at even moderate count rates seen at 10 or 20 nA or higher.  In fact I suspect that most of the reason why many labs "insist" on using "matrix matched" standards is because their dead time constants are not properly calibrated!   

In the next post I will document an example of how dead time calibrations affect accuracy even at relatively low beam currents.

[John Donovan]

Probeman has been discussing the importance of proper dead time calibrations for high accuracy EPMA and I wanted to chime in on a subtle but important aspect to the setting of PHA parameters.  This is in regards to PHA settings for quantitative analysis versus PHA settings for the dead time calibration.  So how are the PHA settings different for these two situations?

Well, in normal quantitative analysis we are usually adjusting our PHA settings when the beam is on our primary standard which normally has a relatively high concentration of the element of interest.  This normally means a relatively high count rate. Which means that when we move to our unknown sample, the count rate will usually be *lower*, hence the PHA peak will tend to shift to the right.

Therefore we normally would like to place our PHA peak a little to the left of the center of our PHA range as seen here (roughly):



On a JEOL instrument that means the peak should be around 3.5 to 4 volts (of a 0 to 10 volt range), and for a Cameca usually around 2 to 2.5 volts (for a 0 to 4.5 volt range), so that there is room when the PHA peak shifts to the right for lower concentrations/intensities.

However when we perform dead time calibrations we are usually start at a low beam current and adjust our beam current up (say from 10 nA to 200 nA) to obtain an increase in count rate and therefore observe the increase in the dead time effect.  So because we will be *increasing* our count rate when performing a dead time calibration, when we initially adjust our PHA settings we want to be sure to adjust our PHA peak so that it is somewhat to the *right* of the center of our PHA range! This is shown here (roughly):



This means that as we increase our beam current (and hence increase the count rate), the PHA peak will shift to the left and we will have room for it to do so.

[Probeman]

Remember, when acquiring dead time calibrations, we can acquire the data for Carpenter's Excel spreadsheet manually (using 10 beam currents from say 10 nA to 200nA with 5 replicates per beam current at 60 seconds each) or use several automated methods including the PFE Startwin application:

https://probesoftware.com/smf/index.php?topic=1160.0

or the PFE Remote Server Deadtime.xlsm macro VBA script from within Excel attached to this post: (login to see attachments):

https://probesoftware.com/smf/index.php?topic=1038.msg6890#msg6890

or the JEOL or Cameca scripts used by your instrument service engineer.

And typically we will utilize the Si Ka line on Si metal using a combination of TAP and PET crystals, or the Ti Ka line on Ti metal using a combination of PET and LiF crystals.

And remember, if we tune up the spectrometers started at an initial beam current of say 10 nA and will then be stepping up the beam current during the dead time calibration acquisition, we will want to tune of PHA distribution placing the PHA peak slightly to the right of the center as seen here:



Of course this is quite a different PHA setting for normal acquisitions when we tune our spectrometer on our (high concentration) primary standard, and then move to unknowns (with typically lower concentrations) with lower intensities as shown here:



Of course the above images show Cameca PHA distributions, but the JEOL PHA distributions will visually look very similar, just with x axis ranges from 0 to 10 instead of 0 to 5.

If anyone does perform a dead time calibration on their instruments, please feel free to share with us here. The question is: do you also observe your dead times slowly increasing as your instrument ages as described here?

https://probesoftware.com/smf/index.php?topic=1442.msg10671#msg10671

[Brian Joy]

And typically we will utilize the Si Ka line on Si metal using a combination of TAP and PET crystals, or the Ti Ka line on Ti metal using a combination of PET and LiF crystals.

Do you find that the deadtime correction is consistent for different X-ray lines on a given spectrometer?  For instance, do you get the same result for Si Ka and Ti Ka using PET on a given spectrometer?  What about other X-ray lines?

[Probeman]

It's a good question. Here are my tests from 2010 on Si and Ti for spectrometers 1 to 5:

Si:   3.73   4.23   3.23   3.66   2.65   (usec)
Ti:   3.57   3.74   3.80   3.51   3.59   (usec)

Except for spectro 3 and 5 the Ti dead times seem a little shorter.  These are the 2 atm detectors on my system which is interesting...

The 1 sigma on the others (1 atm) are around 0.07 to 0.18 usec, so some of the differences are above 3 sigma.

[Brian Joy]

So then how do you decide which value to use, for instance on your spectrometer 5?  Couldn't your choice introduce significant error in deadtime-corrected count rates for some X-ray lines?

[Probeman]

The computer decides which value to use based on the spectrometer and crystal from the values in the SCALERS.DAT file.

And one can always edit the values in the software post acquisition.

The biggest problem I'm seeing are people using the default factory dead time constants for the life of their instrument!   

[sem-geologist]

 

Such a difference between Ti and Si dead times on the same spectrometers... Now this makes sense why it is called PHA differential mode. (which I don't use, and have no such modern problems then using integral mode). What is Your solution for measuring something in between Si and Ti? Linear interpolation of dead-time?

Jokes aside, I Think dead-time and PHA needs very detailed explanation and a complete full attention of everyone doing EPMA right now. I wan't to contribute and show step by step how this works on Cameca probes. I can't show and pinpoint stuff on the schematics legally, as that is confidential... but I can absolutely legally do a "tear-down", look for traces (where it is possible) on PCB's and make a pictures, look for component datasheets and discuss how most probably those work. I can plug the oscilloscope (We have unfortunately only 2 channel oscilloscope 50MHz) at some points which further explains why we are seeing what we are seeing. The most important thing to grasp for every one  right now is that this is not linear as there is few competing issues and it is not possible to correctly correct this with linear regression (what those mentioned spreadsheets tries to do), the proportion of missing counts scales exponentially to applied beam current. For short fraction of that curve linear fit can look ok, but this example with Si and Ti shows exactly how it complicates things (Ti is for high intensity region, Si is for low intensity region; 10 to 200nA it will cover different portions of exponential curve).

Except for spectro 3 and 5 the Ti dead times seem a little shorter.  These are the 2 atm detectors on my system which is interesting...

Gas Pressure and Ar escape peak response I think have most to do with this, Si Ka alpha also has 2nd and 3rd order bremstrahlung which is cut out with beloved differential mode. Ti Ka at differential mode has less stuff from being cut out thus it gets better, but at high pressure it also is much more intense than on low pressure, thus these observed discrepancies.
And that brings me to this question: Does Probesoftware accounts time for rejected peakspulses? Rejection happens in the digital domain, but spectrometer analog domain is blocked the same as for counted peakpulses and should be taken into account.

Getting back to contribution question: Should I continue on some of the posts, or should I open blank new (YADPPFCP-Yet Another Deadtime PHA Post For Cameca Probes)?
What is preferable form? I think it should be mixed, some static images and text is ok, other things is much easier to communicate with short videos, albeit I don't posses high quality gear, but it should be enough to make a point.

[Probeman]

I finally found an abstract of Paul Carpenter's presentations from the TC 2016 on the importance of the dead time (and PHA and spectrometer alignments) for high accuracy work. See attached pdf (login to see attachments).

My only comment on the abstract would be to the figure showing PHA settings where he has left room on the low energy (left) side of the PHA peak, but not much on the high energy side. I would prefer to leave room on both sides, though it depends of course on ones PHA tuning procedure.

If one tunes their PHA (as most do) on a high concentration of the element such as a primary standard, one should leave room for pulse height depression on the high energy (right) side of the PHA peak, as the PHA peak will tend to shift right towards higher energies at lower concentrations of the secondary standard and unknowns.

Of course if one tunes on the secondary standard or unknown, then leaving room on the low energy (left) side of the PHA peak only, will be sufficient.

However, I prefer to leave room on both sides of the PHA peak, and utilize a wide open differential value (~9 volts for JEOL, ~4 volts for Cameca), or use integral (not differential) mode only. 

The only reason for "tight" PHA settings was in the past to avoid high order spectral interferences, However, now that we have proper spectral interference corrections in software (e.g., Probe for EPMA and Peak Sight), it is more accurate to allow *all* the interferences in, and correct for them quantitatively in software, rather than try to "filter" out interferences using PHA, which only works (partially) for higher order interferences, and not at all for 1st order spectral interferences.

By using "wide open" PHA settings, we avoid creating a non-linear response of concentration vs. intensity in our detection systems which can greatly affect accuracy!

[Brian Joy]

Hi John,

But you still haven't answered a really basic question.  In 2010, you determined deadtime values of 2.65 and 3.59 μs, respectively, for Si Ka and Ti Ka on your channel 5.  So which value did you end up using?  Did you average them?

I maintain that it's effectively impossible to set the deadtime accurately.  Because of this, I tend to do quantitative work at relatively low count rates.  At 5 kcps (on the standard), the deadtime correction on a JEOL spectrometer should be no more than 1%.  Alternatively, If I need to work at substantially higher count rates (say 20 or 30 kcps), then I make sure that the count rate on the standard is similar to that on the unknown.

Brian

[John Donovan]

But you still haven't answered a really basic question.  In 2010, you determined deadtime values of 2.65 and 3.59 μs, respectively, for Si Ka and Ti Ka on your channel 5.  So which value did you end up using?  Did you average them?

I maintain that it's effectively impossible to set the deadtime accurately.  Because of this, I tend to do quantitative work at relatively low count rates.  At 5 kcps (on the standard), the deadtime correction on a JEOL spectrometer should be no more than 1%.  Alternatively, If I need to work at substantially higher count rates (say 20 or 30 kcps), then I make sure that the count rate on the standard is similar to that on the unknown.

Sorry I missed your question. 

To answer, I would simply say that because of the hypothesis from Paul Carpenter (from some years ago) that X-ray energy affects the detector dead time constant, we introduced (also) some years ago, a separate dead time constant for each crystal (a proxy for energy range) in the Probewin.ini file.

So therefore one can specify a separate default dead time constant for each crystal on each spectrometer.

There could certainly be some intrinsic uncertainty in the dead time determinations (as for all determinations in EPMA), but given that many analysts are prone to acquire data at 10K or 20K cps or more, having the best dead time constants we can ascertain, is probably not a bad course of action.

[Brian Joy]

But you still haven't answered a really basic question.  In 2010, you determined deadtime values of 2.65 and 3.59 μs, respectively, for Si Ka and Ti Ka on your channel 5.  So which value did you end up using?  Did you average them?

I maintain that it's effectively impossible to set the deadtime accurately.  Because of this, I tend to do quantitative work at relatively low count rates.  At 5 kcps (on the standard), the deadtime correction on a JEOL spectrometer should be no more than 1%.  Alternatively, If I need to work at substantially higher count rates (say 20 or 30 kcps), then I make sure that the count rate on the standard is similar to that on the unknown.

Sorry I missed your question. 

To answer, I would simply say that because of the hypothesis from Paul Carpenter (from some years ago) that X-ray energy affects the detector dead time constant, we introduced (also) some years ago, a separate dead time constant for each crystal (a proxy for energy range) in the Probewin.ini file.

So therefore one can specify a separate default dead time constant for each crystal on each spectrometer.

There could certainly be some intrinsic uncertainty in the dead time determinations (as for all determinations in EPMA), but given that many analysts are prone to acquire data at 10K or 20K cps or more, having the best dead time constants we can ascertain, is probably not a bad course of action.

I’d encourage everyone to measure deadtimes for multiple X-ray lines per spectrometer and not to rely on just Si Ka and Ti Ka.  It’s a time-consuming but worthwhile exercise.  I see variation in deadtime from one X-ray line to another, but I can’t see any obvious pattern to it -- it is not a simple function of X-ray energy.  The subject was discussed some years ago here:  https://probesoftware.com/smf/index.php?topic=394.0.

[Probeman]

I completely agree that the more measurements the better. However, we're talking about labs that have *never* measured their dead time "constants"!   If we can get them to calibrate their dead times even once, I'd consider that progress!   

The effect of different energy x-rays is not something that I've looked into but I believe Paul Carpenter has, but please feel free to share your own investigations. 

Maybe he will chime in on this? Or someone else that has looked into this?