Setting dead times on Cameca SX50/SX100/SXFive

[lucaSX50]

Edit by John: it should be pointed out that the procedure for setting the deadtime constants in Probe for EPMA (editing the SCALERS.DAT file) is the same for the Cameca SX50/SX51//SX100 and SXFive.

Dear All,
We are trying to optimize the dead times for our detectors to work with high currents (300 nA). It turns out that the calculations suggest dead times like 2.3 ms. If I set a time like that in the cameca software, the value defaults to 2 (this is the dead time we are using for all our detectors). If I set the dead time from the SXLocal using: sacq sp4 dtim 2.3 the value is accepted (I receive no error message). However there's not a display of this value anywhere. The graphic software still shows 2 as dead time. Same thing if I use disp from the SXLocal.

Is this just a display limitation or it's actually the hardware that does not support those values?

Is it actually possible to set dead times with fractions of milliseconds in the SX50?

Also ... is the dead time correction something built into the cameca software/hardware but simply not accessible for modifications (i.e. like turning it on/off or selecting when to apply it. Similar to what is possible with Probe for EPMA)?

I am new at this (as you can imagine) so ANY suggestion/explanation/advice will be greatly appreciated.

Thanks.

Luca

[John Donovan]

Hi Luca,
It is quite confusing.

There is a philosophical difference in the way the Cameca and JEOL deadtime electronics works. In the JEOL the measured deadtime is just that. What one measures. It includes a contribution from the detector gas amplification and the pulse shaping electronics. However, it is NOT a constant. As Paul Carpenter has demonstrated and I have confirmed, the deadtime is affected by a number of factors including the bias voltage and the incident x-ray energy.

Because these effects can contribute to a "variable" deadtime constant, Cameca microprobes incorporate a circuit for masking the variation in the "intrinsic" deadtime to a moderate but known (and constant) deadtime value which is set in the hardware.

The intrinsic deadtime of the Cameca electronics and detectors is somewhat larger than the JEOL and is about 1 to 2 microsecs and one can set it to zero and effectively measure only the "intrinsic" deadtime.

On the SX100 the software will not allow one to set the masking deadtime to zero and so one can not directly measure the intrinsic deadtime. However, I set the the "masking" or "forced" deadtime to 1 microsec and it gave 1.03, 1.38, 1.44, 0.91, and 1.13 microsecs for the 5 spectrometers, using Ca Ka when measured using Paul Carpenter's Excel sheet for calibrating your deadtime. See attached documents. One for measurement and one for calculations.

So on the Sx100 instrument the intrinsic deadtime is probably around 1 microsec but when set the "masking" or "forced" deadtime (DTIM parameter) is set to 3 microseconds and then measured, the ACTUAL measured pulse widths should be used for the software correction.

The beauty of this method is that although the underlying intrinsic deadtime may change due to changes in bias voltage and x-ray energy, the measured pulse width (and hence actual deadtime) never varies.

So on the Cameca there are three deadtimes:

1. The intrinsic or native deadtime which can vary with detector bias and x-ray energy.

2. The "forced" or "masking" deadtime used to mask the variation and is set to an integer value by the hardware (DTIM on the SX50). In Probe for EPMA, one sets the Cameca "forced" or "masking" deadtime in the SCALERS.DAT file on line 35.

3. The actual measured deadtime of the integer "forced" or "masking" DTIM value which used for the deadtime correction in software for each spectrometer and crystal combination on lines 72-77 of the SCALERS.DAT config file.

So in summary:

One should determine the intrinsic deadtime with the SX50 DTIM set to 0 us (or the Sx100 set to 1 us). Then one needs to set the imposed deadtime to some larger value (to have it function as a constant mask as originally intended by Cameca) and then, and only then...

measure the actual deadtime at that imposed integer deadtime using the classical method of increasing beam current, for use in the software correction.

[John Donovan]

The deadtime spreadsheet in the previous post is used for automated acquisition of the data set though the Probe Software supplied Remote Automation software interface. It acquires 5 count cycles of 60 seconds over a range of beam currents.

That intensity data is then pasted into the deadtime calculation documents which are attached here.

[John Donovan]

From the PFE User Reference manual:

Line 35                       (Cameca integer deadtimes)
  0        0        0        0        0        0        0        0         "Cameca integer deadtimes"

This line is used to explicitly specify the integer deadtime constants used to set the Cameca PHA hardware (non-extendible or "enforced" deadtime. Therefore this data is used only by the Cameca SX100/SXFive hardware interface and the values are not accessible from within the program. These integer deadtime values are distinct from the single precision deadtimes specified on line 13 above which are used to perform the actual deadtime correction in the analysis routines.

The typical procedure is to set the deadtimes on the Cameca PHA hardware all to zero and then to measure the "intrinsic" deadtime of the system using a range of beam currents from approximately 10 to 200  nA on a pure metal x-ray line such as Si Ka (PET and TAP) or Ti Ka (PET and LIF). A sufficient counting time should be used to obtain .2% precision or better. The best method is to find the "worst case" deadtime for each spectrometer since the deadtimes may vary somewhat as a function of detector bias and x-ray line energy.

Assume that the "intrinsic" deadtimes measured (when all spectrometers are set to "DTIM"=0) are 2.23, 3.14, 3.45, 3.78 and 2.15. The next step would be to set the DTIM deadtime parameter for each spectrometer to a value large enough to completely "mask" this intrinsic" deadtime, that is values of 3, 4, 4, 4 and 3. Now since these integer deadtime are not accurately set by the Cameca hardware, the operator must now re-run the deadtime calibration measurement using these new values and note the actual deadtimes. In this case depending on the instrument, the measured deadtimes will be somewhat larger, say, 3.75, 4.65, 4.12, 4.89 and 3.32.

In the example just described, the "DTIM" values of 3, 4, 4, 4 and 3 should be entered on line 35 for setting the Cameca hardware and the measured values that correspond to them, that is, 3.75, 4.65, 4.12, 4.89 and 3.32 should be entered on lines 72-77 below to use in the software correction.

The integer "DTIM" deadtime values must be between zero and ten. If a value of zero is read, then the program will load the single precision deadtime from line 13 and truncate to integer each value for setting the Cameca PHA hardware to an approximate value.

[Mike Jercinovic]

Luca and John,
Here is the reply I sent to the SX users list earlier today after Les Moore responded...
First, the hardware imposed deadtime is corrected using the same value for the software deadtime correction.  This is, of course, fine in theory as long as, for example, the 2 us called for actually actually ends up 2us.  However, we find this does not hold up, so when you enter 2us, you may actually end up with something more like 3 or 4 us actual deadtime when you measure it doing a classical deadtime test.  On the SX50 with the SXRayN50 software/firmware, these two things are coupled, which is a real problem.  On PeakSight for the SX100 platform, this has been de-coupled so you give it a hardware deadtime, then adjust the software correction to give good linearity.  So, if you are using PfE for quantitative analysis on your SX50 (you have it there at VPI, right Luca?), then you should be able to adjust the software correction independently no matter what the hardware imposed deadtime is set to.

The second issue is with the picoammeter.  No matter how well tuned it may have been, this will go out of spec at some point, and may need tuning frequently.  If the current measurements are not linear, then your deadtime tests can give you a wrong apparent deadtime, or, in many cases, different apparent deadtimes depending on the current regime you measure.  There are five gain loops in the picoammeter for each current range: <0.5nA, 0.5-5nA, 5-50 nA, 50-500nA, and 500-10000nA.  In a perfect world, these are all linearized.  In actuality, the gains and offsets tend to fall out of linearity, so you may see a completely different deadtime slope in the 5-50 range compared to the 50-500 range.  Unfortunately, the 5-50 range resistors do not have trimmers (so you have to change resistors to affect the gain/slope), the others do, so they are all trimmed to match the 5-50 range.  You will really see what is going on if you do your deadtime test, set the deadtime according to the full count/current range, then redo the test and plot intensity (cps/nA) vs. current.  If your deadtime correction is perfect, you should be able to see a nice flat cps/nA plot for a calibration throughout the current range (of course, being sure to do this measurement on a material, like metals, that can take the high current).  How big of a problem is this?  If you calibrate and analyze at the same current, it's obviously not a problem at all.  If you calibrate at, say 20nA, then do the analysis at 300nA to get high sensitivity, then you have a big problem if your current measurement ranges do not correspond well, especially if you intend to run major elements along with trace elements at high current.  This is, of course, what we do in our facility, so we have to deal with this issue.

Mike J.

[Philipp Poeml]

Hi John,

ok, so I did the measurements using the spreadsheet. I used V Ka on PET LiF LiF PET. I set the Cameca SX100 deadtime to 1. Now I have some nice regression curves, with 4 different values (Mean DT All + Last and Regression DT All + Last).

What do I do with the values? How do I proceed from here?

Cheers
Ph

[John Donovan]

ok, so I did the measurements using the spreadsheet. I used V Ka on PET LiF LiF PET. I set the Cameca SX100 deadtime to 1. Now I have some nice regression curves, with 4 different values (Mean DT All + Last and Regression DT All + Last).

What do I do with the values? How do I proceed from here?

Hi Philipp,
The procedure for editing the SCALERS.DAT file is described in this post:

http://probesoftware.com/smf/index.php?topic=394.msg2131#msg2131

and also in the Configuration Files section of the Probe for EPMA Help Reference Manual accessed from the Help menu and also in the posts above.

[Philipp Poeml]

Hi John,

Thanks for the reply. However, my question remains. I read all these documents, but I still have these questions. I know where and what scalers.dat is, but what to put into it? What value should I take and where exactly should I put it? I find this confusing. Do I choose that value for all? Do I choose the value for last? Fitted or measured? What would I set in PeakSight? What do I put as the Cameca hardware dead time?

It is not clear to me what to do with these values. Maybe too stupid...

Thanks!
Philipp

[Probeman]

Thanks for the reply. However, my question remains. I read all these documents, but I still have these questions. I know where and what scalers.dat is, but what to put into it? What value should I take and where exactly should I put it? I find this confusing. Do I choose that value for all? Do I choose the value for last? Fitted or measured? What would I set in PeakSight? What do I put as the Cameca hardware dead time?

The values that you should have in the SCALERS.DAT file on line 35 for the Cameca integer "enforced" deadtimes are whatever the hardware deadtime values were when you made your deadtime calibration measurements.  The point being that the integer hardware values force the deadtime to a rough "enforced" value but then for the software correction in PFE you want to use an actual measured value.

Which deadtime value from the regressions should one use?  They should be quite similar but if not that may indicate a problem with the detector and then it gets complicated. Did you read Paul's documentation on his Excel dead time spreadsheet?

I guess I'll quote from my reference manual in case the post I linked to above wasn't clear enough:

Line 35      (Cameca integer deadtimes)
  0        0        0        0        0        0        0        0         "Cameca integer deadtimes"

This line is used to explicitly specify the integer deadtime constants used to set the Cameca PHA hardware (non-extendible or "enforced" deadtime. Therefore this data is used only by the Cameca SX100/SXFive hardware interface and the values are not accessible from within the program. These integer deadtime values are distinct from the single precision deadtimes specified on line 72-77 below which are used to perform the actual deadtime correction in the analysis routines.

The typical procedure is to set the deadtimes on the Cameca PHA hardware all to zero and then to measure the "intrinsic" deadtime of the system using a range of beam currents from approximately 10 to 200  nA on a pure metal x-ray line such as Si Ka (PET and TAP) or Ti Ka (PET and LIF). A sufficient counting time should be used to obtain .2% precision or better. The best method is to find the "worst case" deadtime for each spectrometer since the deadtimes may vary somewhat as a function of detector bias and x-ray line energy.

Assume that the "intrinsic" deadtimes measured (when all spectrometers are set to "DTIM"=0) are 2.23, 3.14, 3.45, 3.78 and 2.15. The next step would be to set the DTIM deadtime parameter for each spectrometer to a value large enough to completely "mask" this intrinsic" deadtime, that is values of 3, 4, 4, 4 and 3. Now since these integer deadtime are not accurately set by the Cameca hardware, the operator must now re-run the deadtime calibration measurement using these new values and note the actual deadtimes. In this case depending on the instrument, the measured deadtimes will be somewhat larger, say, 3.75, 4.65, 4.12, 4.89 and 3.32.

In the example just described, the "DTIM" values of 3, 4, 4, 4 and 3 should be entered on line 35 for setting the Cameca hardware and the measured values that correspond to them, that is, 3.75, 4.65, 4.12, 4.89 and 3.32 should be entered on line 13 above to use in the software correction.

The integer "DTIM" deadtime values must be between zero and ten. If a value of zero is read, then the program will load the single precision deadtime from line 13 and truncate to integer each value for setting the Cameca PHA hardware to an approximate value.


Extended Format SCALERS.DAT Lines 42-83
Lines 42-65 (default baseline, window, gain and bias and deadtime PHA settings for each crystal)
  0.       0.       0.       0.        1.       1.       .3      .45       "default PHA baseline voltages1"
  0.       0.       0.       0.        1.       1.       1.       1.       "default PHA baseline voltages2"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA baseline voltages3"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA baseline voltages4"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA baseline voltages5"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA baseline voltages6"
  0.       0.       0.       0.        9.       9.       9.7      8.       "default PHA window voltages1"
  0.       0.       0.       0.        9.       9.       9.       9.       "default PHA window voltages2"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA window voltages3"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA window voltages4"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA window voltages5"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA window voltages6"
  0.       0.       0.       0.        32.      16.      32.      64.      "default PHA gain1"
  0.       0.       0.       0.        32.      32.      64.      64.      "default PHA gain2"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA gain3"
  0        0.       0.       0.        0.       0.       0.       0.       "default PHA gain4"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA gain5"
  0.       0.       0.       0.        0.       0.       0.       0.       "default PHA gain6"
  0.       0.       0.       0.        1674.    1764.    1750.    1650.    "default detector bias1"
  0.       0.       0.       0.        1750.    1800.    1800.    1700.    "default detector bias2"
  0.       0.       0.       0.         0.       0.      0.       0.       "default detector bias3"
  0.       0.       0.       0.         0.       0.      0.       0.       "default detector bias4"
  0.       0.       0.       0.         0.       0.      0.       0.       "default detector bias5"
  0.       0.       0.       0.         0.       0.      0.       0.       "default detector bias6"
These lines are used to set the individual default PHA settings for the baseline, window, gain and bias on the different spectrometer and crystal basis. These crystal based parameter values replace the older default PHA settings in lines 24-27. The older values are still read and used as defaults for backward compatibility.

You only need to enter values for as many crystals as you have for each spectrometer (up to six crystals per spectrometer). The range of allowable values is the same as the PHA parameters for lines 24-27.

Lines 66-77 (default Inte/Diff mode and Deadtime PHA settings for each crystal)
  0        0        0        0          0        0       0        0        "default PHA inte/diff modes1"
  0        0        0        0          0        0       0        0        "default PHA inte/diff modes2"
  0        0        0        0          0        0       0        0        "default PHA inte/diff modes3"
  0        0        0        0          0        0       0        0        "default PHA inte/diff modes4"
  0        0        0        0          0        0       0        0        "default PHA inte/diff modes5"
  0        0        0        0          0        0       0        0        "default PHA inte/diff modes6"
  0        0        0        0          0        0       0        0        "default detector deadtimes1"
  0        0        0        0          0        0       0        0        "default detector deadtimes2"
  0        0        0        0          0        0       0        0        "default detector deadtimes3"
  0        0        0        0          0        0       0        0        "default detector deadtimes4"
  0        0        0        0          0        0       0        0        "default detector deadtimes5"
  0        0        0        0          0        0       0        0        "default detector deadtimes6"
These lines are used to set the individual default PHA settings for the inte/diff mode and deadtime on the different spectrometer and crystal basis. These crystal based deadtime parameter values replace the older default settings in line 13.

[Philipp Poeml]

Hi John,

thanks for the explanations. It is a bit clearer now. But there is something going wrong. Ok, as described: I used V Ka on PET LiF LiF PET. I set the Cameca SX100 deadtime to 1.

I got some dead time values out of the spreadsheet: 4.06  2.78  2.29  3.06 (all four values are pretty much the same for any one spectrometer, that seems right).

I then continued (as described in your text) and set the Cameca dead time to 5  3  3  4 and measured everything again. And now something is weird, because I get for
SP1 fit all 7.87 and fit last 6.78
SP2 fit all 6.97 and fit last 1.56
SP3 fit all 6.33 and fit last 1.64
SP4 fit all 6.07 and fit last 4.82

There is something weird going on here, right? What am I doing wrong?

[Probeman]

thanks for the explanations. It is a bit clearer now. But there is something going wrong. Ok, as described: I used V Ka on PET LiF LiF PET. I set the Cameca SX100 deadtime to 1.

I got some dead time values out of the spreadsheet: 4.06  2.78  2.29  3.06 (all four values are pretty much the same for any one spectrometer, that seems right).

I then continued (as described in your text) and set the Cameca dead time to 5  3  3  4 and measured everything again. And now something is weird, because I get for
SP1 fit all 7.87 and fit last 6.78
SP2 fit all 6.97 and fit last 1.56
SP1 fit all 6.33 and fit last 1.64
SP1 fit all 6.07 and fit last 4.82

There is something weird going on here, right? What am I doing wrong?

You are doing nothing wrong. Your detectors might be a little dirty is all.

Set the integer deadtimes to 5  3  3  4 and the software values to the deadtimes you measured.

[Philipp Poeml]

How can one explain the huge difference in SP2 and SP3 fit all and fit last? Isn't that monster big?

[Probeman]

How can one explain the huge difference in SP2 and SP3 fit all and fit last? Isn't that monster big?

I would ask Paul Carpenter.  After all, his name is "Paul *deadtime* Carpenter" and it is his Excel macro!   

[Probeman]

Hi Philipp,
In the meantime it is worth mentioning that one cannot set the hardware "enforced" deadtimes on the SX100/SXFive to zero (it was possible on the older SX50/51), so setting them all to 1 usec to determine your "base" deadtime is the correct thing to do- as you did.

Attached below are calculations on my Sx100 instrument with the hardware DT set to 1 for both Ti Ka and Si ka.  Note that they results for Si Ka Lower energy), are quite a bit longer than Ti Ka. Hence Paul's comment that these deadtimes are *not* constant and depend on bias and x-ray energy.

But you will note that my Sx100 calculated deadtimes at 1,1,1,1,1 usec are much lower than yours, hence my comment about "dirty detectors".

Just FYI, if I got the values you got ( 4.06  2.78  2.29  3.06 usec) at 1,1,1,1 usec hardware DTs, then I would probably set the hardware DTs to 4,3,2,3 rather than 5,3,3,4 because the enforced deadtimes actual values are usually higher anyway, and you don't want to have a larger DT than you have to.

All this discussion relates to a suggestion I made to Curt Scheppman at Cameca that since they utilized the SIMS vacuum electronics for the new SXFive, they should likewise utilize the SIMS counting electronics which has much faster (shorter) intrinsic deadtimes than the existing SXFive counting electronics.

[Paul Carpenter]

All,

You have to provide a graph of the deadtime plot to get meaningful comments from us. The deadtime Excel sheet calculates a deadtime for each measurement point (x = cps, y=cps/nA), and regression deadtime fit using the Excel least squares function (which is done for two sections of the data set) and also displays an average of the discrete values. For a system that is linear and fast, you should get very similar values for the discrete data and the fitted data.

For systems that exhibit paralyzable behavior, the plot (which has a negative slope) has increasing negative slope as the count rate is increased, with a possible asymptotic relationship to some high count rate; that is, at some high count rate no further output of pulses is obtained with an increase in input count rate. You sure don't want to use this range in any measurements and you want to know about it, hence the reason for doing this exercise and using a large count rate range to detect that behavior.

Most plots have sigmoidal behavior, slightly concave up over one range and slightly concave down over the other range.

It is really important to have the target be fully conductive over the large probe current range being used. This is why the spreadsheet uses both probe and absorbed current, and calculates the ratio abs/probe which you should plot vs. probe current; this should be a plot with no real change in the ratio abs/probe and if so it indicates either charging or a problem with the measurement of probe current.

I recommend that these measurements be made with integral PHA mode because you will surely see some level of gain shift and pulse coincidence so you need to have an essentially wide open PHA setting to allow for this. I recommend doing PHA scans at progressive count rates to demonstrate you have the correct gain/bias and baseline settings. I expect that as the count rate increases, eventually the x-ray pulse moves to lower voltage and merges with the baseline noise because the system loses pulse energy resolution and cannot discriminate between the two.

So again, I ask that in addition to posting numbers, you really need to show us the plot so we can see the behavior. I have not really seen any Cameca deadtime plots so why don't you SX-50 and SX-100 folks show us what the plot looks like for these systems and the effect of including the enforced deadtime value.

You guys are using *ultra pure P-10* for your detectors, right?  If you don't (and I observed this at a lab) you will almost certainly see problems that may well be due to contamination of the wire. That is completely avoidable by using the cleanest grade of P-10, and I have never seen problems on the 4 microprobes I have operated over the years. I had a discussion with Colin MacRae and they replace their Jeol sealed Xe counters every 2 years or so, but I have seen very good long term performance of them. Maybe people are leaving the detectors running at high voltage ~1850V for extended periods of time?

Cheers,

Paul