PHA settings

[sem-geologist]

Again (as expected) the gain shift is even higher simply due to the change in the emission line energy.  I didn't have time to check the LiF crystal, but assume we would observe similar results there.

As For LIF, the PHA shift is the least from all of the crystals, as extreme count rates are very hard to reach even on large crystal (i.e. 200kcps). I see that your PHA was done rather with moderate count ratios. I suggest to look into PHA going to higher and higher beam current. I went so on our SX100 up to 3µA (which is not very possible with beam regulator and auto setting the beam, that needs manual setting of C1 and C2 to some non conventional spots, and 200µm aperture without regulator is suggested for that). You will saturate the count ratio with much lesser values, but PHA will be changing up going past that point.

We know the PHA peak shifts to a higher energy as the concentration (count rate) drops, because the detector is under less "load" at low count rates, and hence the bias voltage can maintain a more consistent high voltage.

Nope. You can't see the bias in PHA, because it is decoupled (in electronic terms; in human terms it means that signal is separated with capacitor, which pass through only voltage raise and drop as a signal). That claim would be true if shape of PHA of high count rates would be identical to that of low count rates, and as those pictures shows it is not. The shift is produced at charge pre amplifier output due to bipolar output used. Heck, that is long known problem in spectrometry and Amptek even have separate chip to detect and correct that (also pileups), but on EPMA WDS'es it looks that part was ignored, probably as it was not designed to run into that kind of count rates supposing that we measure only small fraction of X-ray spectra. Large crystals and high brightness cathodes makes it now possible effortlessly to overuse spectrometer counting electronics to extent which never was designed to do.

To see that shift is generated after charge preamplifier and it is not measured shift of bias decrease You need to see that on oscilloscope while stepping beam currents toward very extremes. Amplitudes stays the same! (of course piled up amplitudes will have double, triple, quadruple heights, but single pulse will have exactly the same relative amplitude). PHA peak shifts to left as the probability for pulse to happen and be registered directly after previous pulse at negative voltage of bipolar output (see the blue mock-up picture above, which shows that mechanism, where 2nd pulse has exactly same relative amplitude, but absolute amplitude measured from 0 will be smaller than 1st peak, and WDS MCA measures absolute, not relative height). With going with high intensity peak, large crystal and extremely high current it is possible to shift main PHA pulse out from visible range below 0.4 V in PHA graph. You still will see some "interesting" pattern of double, triple, quadruple... piled-up pulses. Triple and quadruple (and higher) pulses has so deep negative relaxation tail of bipolar output there You can fit whole normal (not piled up) pulse, and its top will not go above 0V, but its relative amplitude still will be the same.

I also was thinking as You that it is bias decrease which do shifts (Why I would not think like that if books are telling so?), but observed change in shape of PHA made me doubt, and thus I conducted extreme experiments trying to drain the bias. And it looks that You can't drain the bias with intensive X-rays. Used HV generators do extremely good job keeping it in check. This also made me 100% sure that gas proportional counter has no dead-time, or even in ultra extreme cases (ultra high beam current) it is impossible to make that part anyhow to saturate (gas counter + charge preamplifier). That is completely different compared to EDS, where different preamplifier feedback design (particularly that capacitor which needs to be discharged, to reset the voltage level at preamplifiers output) introduces dead-time already at preamplifier part. This is also Why I am sceptical about solid state counters for WDS. With some proper redesign (updating to current knowledge of low-noise design and fast DSP) gas proportional counters could easily do a few million counts per second with near 0% dead-time, and much better energy resolution.

Thanks for the very useful information on the internal electronics.

Thanks, but I made some not correct claims. After looking through so many schematics (not only Cameca) my mind starts mixing up things. Old Cameca WDS board has not two Phillips 68070 cpu's but only one; two chips (which I correctly remember to be two), where ADC signal is sent to, are actually one of early FPGA's. While it would had some pins available for 12 bit ADC, I think limiting factor was rather the limited number of gates in those early FPGA's. Gate number limits how complicate logic can be implemented, and those same FPGAs also need to control the WDS motors, so it is rather not enough FPGA gates and not i/o pins. Also ADC sends to processing all 8 bits, I had to mix that one with some other schematics in my head. Sorry for that 1 bit misinformation (my initial claim that it is only 7bits). I already fixed few boards, and most of schematics sits in my head. But my points about technological limitations and board filled-up and all those leading to limitation on old WDS boards to 8bits is still valid.

[Probeman]

We know the PHA peak shifts to a higher energy as the concentration (count rate) drops, because the detector is under less "load" at low count rates, and hence the bias voltage can maintain a more consistent high voltage.

Nope. You can't see the bias in PHA, because it is decoupled (in electronic terms; in human terms it means that signal is separated with capacitor, which pass through only voltage raise and drop as a signal). That claim would be true if shape of PHA of high count rates would be identical to that of low count rates, and as those pictures shows it is not. The shift is produced at charge pre amplifier output due to bipolar output used. Heck, that is long known problem in spectrometry and Amptek even have separate chip to detect and correct that (also pileups), but on EPMA WDS'es it looks that part was ignored, probably as it was not designed to run into that kind of count rates supposing that we measure only small fraction of X-ray spectra. Large crystals and high brightness cathodes makes it now possible effortlessly to overuse spectrometer counting electronics to extent which never was designed to do.

Maybe. Whatever the reason for PHA peaks shifting to lower voltages at high count rates, it is clear that higher energy x-rays cause PHA peaks to shift higher, but not enough to affect full range WDS spectrometer wavescans, which I believe was the original question.

I agree it is a shame that these WDS electronics have not been updated. I'm sure that faster electronics now exists that could handle these very high count rates.