News:

:) If you can't see some "in-line" images in a post, try "refreshing" your browser!

Main Menu

JEOL WDS detector dead times

Started by David Steele, December 15, 2014, 09:20:53 PM

Previous topic - Next topic

David Steele

Just seeking communal advice/confirmation.... hopefully....

I've finally determined the intrinsic dts on the three P10's and 2 sealed Xe's on our 8530F using Paul C's spreadsheet.  I acquired counts using Startwin and also switched off the on-board dt and beam drift corrections before starting the count accumulation process.

a rough tabulation of the dts is:

Channel     Det     mean dt     regr dt     hi count regr dt
1                P10        1.15         1.15           1.29
2                 Xe         1.34         1.33           1.40
3                P10        1.15         1.10           1.13
4                 Xe         1.24         1.25           1.34
5                P10        1.24         1.22           1.26

Questions:
1) are these values consistent with those from other JEOL EPMAs/gas detectors (P10 and/or Xe), especially on recent 2013/2014-vintage JEOL EPMAs?

2) Logic/conservatism suggests that I should use the higher values, and my preference is to use the (longer) dts determined from the high count rate regressions.  Do others have a preference or comment?

3) the default dt values currently in the PfE Scalars.dat file are 1.9us for all 5 gas detectors on our EPMA.  Is this 1.9us default value used in PfE based on past experiences with JEOL EPMA gas detectors, that is, irrespective of the values I've just determined, should I stick with the longer default (1.9us) values and "know" that I probably won't run into detector dt issues almost irrespective of the count rates I'm using?

4) Can anyone (JD?) confirm that the dt values stored in Scalars.dat are used 'as is' (unmodified) in the Normal or Precision dt corrections selected by the user in the Analytical/Analysis Options window, at least on the JEOL EPMA?  I put this question that way because I believe we are unable to impose a 'hardware' dt correction on the JEOL (as can be implemented on a CAMECA).

Many thanks,
David   

Edit by John:  David, you can edit your posts by using the Modify button, no need to create a second post.

Probeman

#1
Quote from: David Steele on December 15, 2014, 09:20:53 PM
1) are these values consistent with those from other JEOL EPMAs/gas detectors (P10 and/or Xe), especially on recent 2013/2014-vintage JEOL EPMAs?

When I see the default DT values on a JEOL instrument, they always use 1.1 usec. The guy who can answer these questions best is Paul Carpenter. I'll ping him...  but in the meantime I'll attach two of his JEOL 8200 spreadsheets below for Si Ka and Ti Ka.

Quote from: David Steele on December 15, 2014, 09:20:53 PM
2) Logic/conservatism suggests that I should use the higher values, and my preference is to use the (longer) dts determined from the high count rate regressions.  Do others have a preference or comment?

I would run both Si Ka and Ti Ka as you can do both on PET crystals and then see what differences there are compared to the regressions differences.

Quote from: David Steele on December 15, 2014, 09:20:53 PM
3) the default dt values currently in the PfE Scalars.dat file are 1.9us for all 5 gas detectors on our EPMA.  Is this 1.9us default value used in PfE based on past experiences with JEOL EPMA gas detectors, that is, irrespective of the values I've just determined, should I stick with the longer default (1.9us) values and "know" that I probably won't run into detector dt issues almost irrespective of the count rates I'm using?

See my response to 2.

Quote from: David Steele on December 15, 2014, 09:20:53 PM
4) Can anyone (JD?) confirm that the dt values stored in Scalars.dat are used 'as is' (unmodified) in the Normal or Precision dt corrections selected by the user in the Analytical/Analysis Options window, at least on the JEOL EPMA?  I put this question that way because I believe we are unable to impose a 'hardware' dt correction on the JEOL (as can be implemented on a CAMECA).

Note that single line DT values on line 13 in the SCALERS.DAT are now ignored assuming that you have entered non-zero values for each spectrometer crystal combination on lines 72 to 77 in the SCALERS.DAT file.

This is so one can specify a different deadtime for each crystal on each spectrometer because as you will find when running tests using both Ti Ka and Si Ka, you will get different values.  Why then do we call deadtime a "constant"!?    ::)

So, to answer your question, yes they are used "unmodified" (if I understand what you mean) for both the traditional and also the 2 factorial high precision DT expressions in probe for EPMA.  And yes, there is no enforced "hardware" DT capability for JEOL instruments as there is for Cameca instruments (the Cameca enforced "hardware" values are specified on line 35 of the  SCALERS.DAT file and must be integer values).

But, these "software" DT values can also be edited in the user's Probe for EPMA data file, if you have better numbers at some point in time.

I'm also attaching Paul's deadtime calibration document. The Deadtime Excel macro that comes with the Remote COM server app in Probe for EPMA allows one to operate their instrument from Excel. This spreadsheet is used in conjunction with the calculation spreadsheets below to perform the calibrations that David did above.

One can also use StartWin as he did for acquiring the deadtime calibration data, but it's more manual than the Remote Excel macro spreadsheets which run for over an hour (typically) to acquire a full set of deadtime calibration data automatically that can then be simply pasted into the calculation spreadsheet!
The only stupid question is the one not asked!

Paul Carpenter

All,

Here is what I have to offer on this topic.

1. One should not use assumed ("factory") settings for deadtime on their microprobe. To do so can result in up to percent level systematic errors resulting from and incorrect deadtime constant, the magnitude of which depends on the difference in count rate between sample and standard.

2. The use of element specific detector bias values on the Jeol in order to have the PHA pulse at 4 volts (and an equivalent gain setting on the Cameca to put the pulse at 2 V) is an attempt to remove the dependence of deadtime on X-ray photon energy. It also provides the same clearance from baseline noise, room to set the baseline, and headroom for coincidence pulses above the 4 V main pulse.

3. Attention should be paid to the existence of escape peaks and appropriate baseline setting used.

4. So, "what are typical deadtime values". I have observed on production Jeol 8x00, 8530 instruments that the 1.1 usec value is sometimes correct but the deadtime can be as high as 1.7 usec or so. I have seen an 8900 with 1.1 usec deadtimes approximately. I have no information from Cameca users on typical values because it is complicated by the 3 usec enforced dt value, but still it would be worthwhile to see values of 3 + X where X is the apparent deadtime value.

5. Not only should you measure the deadtime, but monitoring the long term changes is important. There is a small change in the deadtime for P-10 counters as the Ar:CH4 ratio changes when the tank slowly empties; there is perhaps a greater change during tank evolution than when you change the tank.

6. Probably most important is to observe the linear range of your deadtime plot which demonstrates to you what the dynamic range of the counting system is. I use a count rate range up to at least 200k cps to see where nonlinearity starts and possible paralyzable behavior exists -- you want to know where that is now rather than discovering it as an artifact later. Stage map runs are likely data sets where a large dynamic range is sampled due to the high probe current that you should be using.

7. Pulse processing and alignment issues are problem areas for probe users and instruments. It is really important to have the counting system characterized for WDS, compare to using an EDS system where the calibration was incorrect and the baseline set too high (cutting out low energy photons) and the detector parameters (resolution, energy calibration) changing as a function of count rate. You would raise hell about those problems with the EDS system, so why not do the right thing and characterize your WDS system. I applaud Dave for doing so and sharing the numbers here.

8. Last thing.  I think that a lower dynamic range is used by Jeol when evaluating the deadtime. The 1.1 usec value should be viewed as a placeholder until you update it with an actual measured value. The dt is not measured on any system to my knowledge prior to or during installation, regardless of manufacturer. This is because resolution and WDS crystal performance, leak rate, blah blah, all that stuff is a spec value and a deliverable benchmark for the installation.

Cheers,

Paul "deadtime" Carpenter :)
Paul Carpenter
Washington University St. Louis

Brian Joy

#3
Recently I evaluated JXA-8230 detector deadtimes for a variety of X-ray lines spanning a wide range of energies.  I used the JEOL 13-element standard block, which I polished under methanol immediately prior to conducting the measurements.  I set the detector gain so that I could center the PHA distribution (at count rate 5k-10k /s) at 4 V by applying a bias between 1600 and 1700 V.  As in Paul's spreadsheet, at given beam current I used an average of five measurements collected at different locations to obtain my plots.

When doing quantitative analyses, generally I use count rates no greater than ~10k /s on both standards and unknowns (the only exceptions being if standard and unknown give similar count rates); I do this in part to avoid deadtime corrections much in excess of 1%.  For this reason, the values that I've tabulated below characterize the roughly linear correction region at relatively low count rates.  In some cases, I can see onset of non-linearity above about 20k /s, while in other cases linearity appears to persist to count rates as high as around 90k /s.  In some cases (e.g. Ti Ka on PETL), I have a hard time seeing linear behavior anywhere.

These results really make me wish that JEOL enforced deadtimes electronically.


Detector deadtime values in microseconds.
Brian Joy
Queen's University
Kingston, Ontario
JEOL JXA-8230

Ben Buse

Hi,

I'm getting confused.

On the jeol software each spectrometer has a deadtime the default of which is 1.1us

Now in PFE there is also a deadtime correction the default of which is 1.1 us.

The PFE seems to be an additional deadtime correction on top of the 1.1 us is that right?

I tried it with Jeol deadtime off and pfe deadtime off
Then with Jeol deadtime on and pfe deadtime on
Then with Jeol and PFE deadtime on
And that is what it seems to suggest I just wanted to check

Thanks

Ben

Probeman

#5
Quote from: Ben Buse on October 05, 2016, 07:35:21 AM
On the jeol software each spectrometer has a deadtime the default of which is 1.1us

Now in PFE there is also a deadtime correction the default of which is 1.1 us.

The PFE seems to be an additional deadtime correction on top of the 1.1 us is that right?

Please remember to search for an existing topic before creating a new one! 

Hi Ben,
Not sure that I follow you.

There is just the software deadtime correction applied in either the JEOL software or in the Probe for EPMA software, during post acquisition processing.  There is no additional deadtime correction for JEOL instruments.

On Cameca instruments there is an additional firmware deadtime that is applied to "force" the actual deadtime to a specific deadtime at the time of acquisition. This "enforced" deadtime is to deal with the fact that the actual (measured) dead time is not an exact constant in practice, but varies with x-ray energy and detector bias. It's a good idea that only Cameca implements.

See Paul's discussion here on JEOL dead times:

http://smf.probesoftware.com/index.php?topic=394.msg2133#msg2133
The only stupid question is the one not asked!

Ben Buse

That makes more sense. It must have just been analytical error! Turning PFE on and off did make a much bigger difference!

Thanks

Ben

Karsten Goemann

I've recently measured dead times on our new 8530F Plus using TiKa and PET crystals (each of our 5 WDS has one). PHAs were adjusted to have the peaks at around 4V at 50nA beam current. Baselines were 0.7V which includes the Ar escape peak on the P10 channels. PHA depression pushes the peaks down to 3V on the P10 GPCH detectors for the highest count rates (>200kcps). The effect is slightly less pronounced for the Xe XPCH detectors.

As usual the values depend a little on the fitting range (see example plots for Sp3 and Sp5 attached). As we normally do not operate at excessive count rates I excluded measurements >125kcps for the fit for the H detectors, and >50kcps for the standard gas flow detector. This is purely based on experimenting with fitting ranges and looking at the plots, and excluding those where I felt the behaviour started to look significantly non-linear. I wonder if that is at least partly due to PHA depression.

These are the results:

Sp1: 1.28 us (FCS type, P10/GPC, TAP/PETJ/LDE2/LDEB
Sp2: 1.23 us (L type, Xe/XPCH, PETL/LIFL)
Sp3: 1.39 us (L type, P10/GPCH, PETL/LDE1L)
Sp4: 1.38 us (L type, P10/GPCH, TAPL/PETL)
Sp5: 1.26 us (L type, Xe/XPCH, PETL/LIFL)

Using the full fit range increases the measured dead times to around 1.4-1.5 us.

Philippe Pinard

I should have posted it earlier. Here are some dead time values I collected back in Aachen on the JEOL JXA8530F. See my thesis for more details.

The spectrometer configurations was:
#1: P10 with LDE1, LIF, LDEB, TAP
#2: P10 with LDE2, PETJ, TAP
#3: Xe with LIF, PETJ
#4: Xe with LIFH, PETH
#5: P10 with LDE2H, LDE5H

To measure the dead time, the output count rate was varied from 1000 to 80000 cps. The accelerating voltage was 15 kV. The acquisition time was also modified to collect at least 100 000 counts per beam current. As Brian, I observed differences in the dead time depending on the X-ray line used. If I remember correctly, it was quite reproducible. Here are all the results. I added Brian's results in the last column for comparison. Karsten which X-ray lines did you use?


ElementLineSpectrometerCrystalDead timeDead time errorBrian Joy
AgLa2PETJ1.580.05
AgLa3PETJ1.50.07
AgLa4PETH1.490.06
AlKa1TAP1.680.06
AlKa2TAP1.750.05
BKa1LDEB1.80.12
BKa1LDEB1.80.13
CKa2LDE21.980.09
CKa2LDE22.010.07
CKa5LDE2H1.910.04
CKa5LDE2H1.90.03
CrKa1LIF2.310.07
CrKa1LIF2.010.1
CrKa1LIF1.90.13
CrKa1LIF2.160.1
CrKa3LIF1.380.06
CrKa3LIF1.490.06
CrKa3LIF1.270.05
CrKa3LIF1.430.05
CrKa4LIFH1.590.06
CrKa4LIFH1.590.04
CrKa4LIFH1.590.09
CrKa4LIFH1.690.9
CuLa1TAP1.420.050.3/0.6
CuLa2TAP1.580.040.3/0.6
FeKa1LIF2.240.09
FeKa1LIF2.30.06
FeKa3LIF1.140.060.5
FeKa3LIF1.180.030.5
FeKa4LIFH1.470.051.4
FeKa4LIFH1.480.061.4
GaKa3LIF1.450.07
GaKa3LIF1.390.05
GaKa4LIFH1.450.09
GaKa4LIFH1.360.07
NKa5LDE5H2.210.23
NKa5LDE5H1.730.1
NiKa1LIF2.070.09
NiKa3LIF1.160.05
NiKa4LIFH1.40.08
NiLa1TAP1.470.08
NiLa2TAP1.510.07
OKa1LDE11.10.41
OKa1LDE11.360.3
SiKa1TAP1.730.081.15/1.25
SiKa2PETJ1.410.05
SiKa2TAP1.730.051.15/1.25
SiKa3PETJ1.220.050.4
SiKa4PETH1.450.040.95
SnLa2PETJ1.660.05
SnLa3PETJ1.40.05
SnLa4PETH1.270.07
TiKa3PETJ1.340.051.1
TiKa4LIFH1.670.061.4
TiKa4LIFH1.530.091.4
ZnKa1LIF1.910.11
ZnKa3LIF1.420.05
ZnKa4LIFH1.510.07
ZnLa1TAP1.410.06
ZnLa2TAP1.710.03
ZrLa2PETJ1.560.08
ZrLa3PETJ1.360.06
ZrLa4PETH1.280.09

Karsten Goemann

Thanks Philippe, interesting results - there doesn't appear to be a clear trend (e.g. change with x-ray keV).

So far I've only done TiKa using PET crystals.

Did you do PHA scans at the different count rates for your measurements? Did you adjust the bias to have the peak at 4V or something like that for all different x-rays?
I'm wondering if I should adjust the PHA peak to a higher voltage because of the pulse height depression.

Ben Buse

#10
Hi Philippe,
That looks very interesting, I've just dug out some old data I did.
I'm struggling to remember what i did - but scales are the different spectrometers.
I only included data >900 cps otherwise error was great.
Still considerable error, but you seen on spec 3 (scaler 3), change in gain from 32 to 8 (TAP to PET?) and increase in deadtime

Ben