Setting dead times on Cameca SX50/SX100/SXFive

[sem-geologist]

While comparing the model to real observed values on the machine, I had found some strange behavior in peaksight (but as PfS is using some dll, from cameca for communication and data transfer from machine it can be affected too) depending from set software dead time.

The dead time and count capping is nicely illustrated with comparison of WDS scans of same substance at different probe currents. It is clear that background is exactly same (As it is low count rate) when comparing with cps/nA, but high current peaks are capped.

From the beginning of SXFiveFE installation in 2014 I was always in impression that our spectrometer 5 has worse performance as amplitude in WDS scans always were lower. With low count rates the difference is not big, but with high current it is like half. While doing 660nA on Cr2O3 I spotted that such discrepancy is obviously not right as xray rate meter showed nearly identical intensity bars while at peaks.
When setting interpretor for do count for 1 second, the counts would result in the very close values. So Quanti is also not affected, but WDS scans are. I tried to acquire mapping of Cr Ka at 660, and the intensity for 5 spectrometer is reduced. As I did few WDS scan runs at different set dead times, I saw that there is no discrepancy when dead time is set to 1µs, The WDS amplitudes of peaks are near identical and intensity mapping is also near identical for both 2 and 5 spectrometers.
This is really enigmatic behavior, had you seen this on PfS? The results probably is differently acquired from machine when doing image and WDS scan (needs to be pushed to software constantly with WDS position/stage position movement), than when using hardware counters for measuring counts in few seconds for quantification.
I see that mappings should be done at 1 µs DT.

[John Donovan]

This is really enigmatic behavior, had you seen this on PfS? The results probably is differently acquired from machine when doing image and WDS scan (needs to be pushed to software constantly with WDS position/stage position movement), than when using hardware counters for measuring counts in few seconds for quantification.

Hi SG,
I haven't looked at deadtime corrections in wavescans explicitly, but in PFE we apply the deadtime (and beam drift) corrections to all samples: standards, unknowns and wavescans (and of course X-ray maps). The method we use is described in the first page or so of posts in this topic.

For such high beam currents it seems to me that you should be utilizing the higher accuracy deadtime equation described here:



The difference between the "normal" dead time correction expression (#1) and the "high precision expression" (#2) is quite significant at high count rates with Cameca GFPC detectors.

[sem-geologist]

Thank You for the formula.
I put it into the simulation and find that it is not satisfactory, the problem is that it merges two separate processes into one. Counting piled-up peaks as single peak is not depending from the dead time, but from count rate alone. I am coming up with some improved formula. Unfortunately, I find it works rather only on SXFive (or SX100 with new WDS board), as unfortunately I found out that SX100 (old electronics) counting electronics have paralysable behavior (flats out at about 250kcps (raw, not-dead-time-corrected)). I am trying to understand the process to make the equation universal.

[John Donovan]

The deadtime correction expression is an infinite factorial of which only the first factorial term is generally utilized.  This is because deadtime is a statistical event and the single factorial term is generally good enough at relatively moderate count rates.  The expression #2 just utilizes a 2nd factorial term in this infinite series for better handling of extremely high count rates.  It is not intended to deal with other electronic non-linearities.