Author Topic: WDS hardware ideas, improvements  (Read 6753 times)

Brian Joy

  • Professor
  • ****
  • Posts: 243
Re: WDS hardware ideas, improvements
« Reply #15 on: May 01, 2022, 10:53:27 PM »
I am not at all suggesting an SDD detector for WDS. I agree that would be silly. 

Instead I am imagining the use of a simple photon detector at room temperature, perhaps a pin-diode detector. Even just to eliminate the hassle of dealing with P-10 gas bottles would be great!   :)

I am not an electronics expert at all, so I will defer on the technical details, but I wonder if the Bragg crystal gives us all the energy resolution we require. Obviously, we need to filter the thermal noise from the detector, but that shouldn't be difficult I am hoping. 

I wasn't thinking clearly when I wrote “Si wafer” when referring to the JEOL/Bruker solid state detector.  The design was of the SSSD type (solid state scintillation detector), but I no longer remember any details of its construction.  I’d be glad to look into this and then post in a new topic.  Would this be helpful?
« Last Edit: May 01, 2022, 11:42:22 PM by Brian Joy »
Brian Joy
Queen's University
Kingston, Ontario
JEOL JXA-8230

NicholasRitchie

  • Professor
  • ****
  • Posts: 106
    • NIST DTSA-II
Re: WDS hardware ideas, improvements
« Reply #16 on: May 02, 2022, 05:06:16 AM »
I agree with Mike that Cameca has done a considerable amount of innovation (large area crystals also being an important one) over the last 10 to 15 years, but I think Nicholas' point is that with Cameca now (mostly) not selling microprobes, those innovations will essentially fall by the wayside.

Looking forward can we expect JEOL to start innovating more?  I don't know, but let's try and list a least 10 ways JEOL could improve their current microprobe. I'll start:

1. Implement solid state detectors for WDS spectrometers
2. ...
solid state detectors... again. Can't understand that obsession. It would make sense only if diffraction crystals would be focusing to point and not line. As Solid state detectors are round. The G(F)PC has few key advantages for being used in WDS:
    1) The active volume where X-rays can be registered is very narrow cylinder around the biased wire. That allows more simple diffraction crystal manufacturing process and peak is not blurred by large detection area (which large SDD would do). The active volume is used in its fullest. For SDD It would need some "line"-like aperture in front, so only very little fraction of active detector surface would be used.
    2) Cooling. G(F)PC does not need that. SDD working at room temperature and higher would have comparably terrible resolution.
    3) solid state detectors produce unnecessary spectral artifact like tails i.e. with incomplete charge collection. Si Ka escape line kicks in with lower energy lines than Ar Ka escape (or Xe escape). It is in particular important for Pb, Th, U in geochronology where Silica detector would introduce artifacts which are not on GFCP.
What I propose is combining advantage of G(F)PC with advantage of SDD (or more precise learn from advantages of current solid state detector counting electronics). Currently implemented GFPC counting is very strongly affected with pileups and does not allow to filter out 2nd order lines fully. current PHA is passive dumb PHA. That is like selecting regions for mapping on EDS for overlapping peaks (i.e. Zr La and P Ka; SKa and Pb Ma). But that problem (EDS mapping) already was solved a decade ago with introducing of deconvolution method! That is the main missing point for total WDS PHA empowerment. Currently PHA is set like those old time EDS maps at baseline and window size - the region of histogram. But PHA curve could be collected during counting and saved like EDS spectra for later treatment. Currently integral and diff counts are preserved at WDS board during and after counting. With such curve preserved the exact window, or with deconvolution by fitting Gaussian curve the precise part of counts corresponding to selected measured line could be much more precisely quantisized. Also this would allow to have much better peak-pile up correction than what EDS normally have by utilizing deterministic fixed dead time (as currently is). Actually we don't need much shorter deadtime (shorter deadtime means shorter shaping time and it comes at cost of worse energy resolution). SDD does not have much shorter dead time and still can cope much better with higher count rates than WDS. (i.e. 1Mcps)

I am not at all suggesting an SDD detector for WDS. I agree that would be silly. 

Instead I am imagining the use of a simple photon detector at room temperature, perhaps a pin-diode detector. Even just to eliminate the hassle of dealing with P-10 gas bottles would be great!   :)

I am not an electronics expert at all, so I will defer on the technical details, but I wonder if the Bragg crystal gives us all the energy resolution we require. Obviously, we need to filter the thermal noise from the detector, but that shouldn't be difficult I am hoping. 

I am sure you could come up with something pretty interesting if you thought about it.   8)

As Rick Wuhrer has shown, even a very poor resolution SDD offers benefits as a counter/detector for WDS

1) Even poor energy resolution could eliminate all higher order peaks thus simplifying the measurement process
2) Lower the background by eliminating off-energy X-rays that scatter off imperfections in the crystal improving detection limits

There is no reason that SDDs need to be round.  In fact, many aren't.  Custom SDDs could be designed to enhance throughput at reduced resolution with better matched shape.
"Do what you can, with what you have, where you are"
  - Teddy Roosevelt

Probeman

  • Emeritus
  • *****
  • Posts: 2384
  • Never sleeps...
    • John Donovan
Re: WDS hardware ideas, improvements
« Reply #17 on: May 02, 2022, 09:34:33 AM »
As Rick Wuhrer has shown, even a very poor resolution SDD offers benefits as a counter/detector for WDS

1) Even poor energy resolution could eliminate all higher order peaks thus simplifying the measurement process
2) Lower the background by eliminating off-energy X-rays that scatter off imperfections in the crystal improving detection limits

There is no reason that SDDs need to be round.  In fact, many aren't.  Custom SDDs could be designed to enhance throughput at reduced resolution with better matched shape.

I agree the SDD energy resolution could be useful to eliminate higher order reflections, but this has to be considered against the major hassle of needing to cool the SDD detector.

My practice is to always keep the PHA settings wide open to avoid non-linear response from the gas detector from gain and energy shifting. So except for unusual situations (e.g., 2nd order Na Ka on Oxygen K) we just use the baseline PHA filter to avoid thermal noise and use the quant interference correction in PFE to correct for these spectral interferences.

My intuition is that the Bragg crystal has more than enough energy resolution without the need for an SDD. A simple uncooled pin diode photon detector should be more than enough energy resolution is my guess and it should have 1 Mcps response time. But it's just a guess.

As for "off-energy X-rays that scatter off imperfections in the crystal"  I believe you are discussing polygonization artifacts from crystal manufacturing that gives us our long tails in the WDS peak shape.  Those x-rays are mostly the same energy as the observed emission line, but just dispersed over a wider range of sin theta.  I don't think an SDD energy resolution will be sufficient to filter those out.

I'm going to ask John Donovan to put these solid state detector posts into WDS detector topic because it is an interesting topic.
« Last Edit: May 02, 2022, 10:02:12 AM by Probeman »
The only stupid question is the one not asked!

NicholasRitchie

  • Professor
  • ****
  • Posts: 106
    • NIST DTSA-II
Re: WDS hardware ideas, improvements
« Reply #18 on: May 02, 2022, 01:48:20 PM »
As for "off-energy X-rays that scatter off imperfections in the crystal"  I believe you are discussing polygonization artifacts from crystal manufacturing that gives us our long tails in the WDS peak shape.  Those x-rays are mostly the same energy as the observed emission line, but just dispersed over a wider range of sin theta.  I don't think an SDD energy resolution will be sufficient to filter those out.

I think that Rick was suggesting a different effect.  Polygonization artifacts lead to X-rays from the peak being measured scattering slightly off axis and being measured in the background near the peak.  As you observe, these could not be distinguished by an SDD.   I think the benefit of the SDD is in eliminating a background produced by the broad range of characteristic and continuum x-rays that enter the detector and scatter off imperfections in the crystal leading to a very broad almost uniform background over all energies.  What he suggested is that the background has two contributions - 1) true continuum; 2) scatter from imperfections like micro-cracks and dust.  He reported that the SDD could eliminate the second.
"Do what you can, with what you have, where you are"
  - Teddy Roosevelt

Probeman

  • Emeritus
  • *****
  • Posts: 2384
  • Never sleeps...
    • John Donovan
Re: WDS hardware ideas, improvements
« Reply #19 on: May 02, 2022, 04:28:26 PM »
As for "off-energy X-rays that scatter off imperfections in the crystal"  I believe you are discussing polygonization artifacts from crystal manufacturing that gives us our long tails in the WDS peak shape.  Those x-rays are mostly the same energy as the observed emission line, but just dispersed over a wider range of sin theta.  I don't think an SDD energy resolution will be sufficient to filter those out.

I think that Rick was suggesting a different effect.  Polygonization artifacts lead to X-rays from the peak being measured scattering slightly off axis and being measured in the background near the peak.  As you observe, these could not be distinguished by an SDD.   I think the benefit of the SDD is in eliminating a background produced by the broad range of characteristic and continuum x-rays that enter the detector and scatter off imperfections in the crystal leading to a very broad almost uniform background over all energies.  What he suggested is that the background has two contributions - 1) true continuum; 2) scatter from imperfections like micro-cracks and dust.  He reported that the SDD could eliminate the second.

I see. Interesting. So have any attempts been made to see what the magnitude of such a "pseudo continuum contribution" might actually be? 

I will admit that we do see occasional WDS artificats that have an unknown cause, though these are usually distinct peaks, though sometimes with various different FWHMs:

https://probesoftware.com/smf/index.php?topic=1374.0

https://probesoftware.com/smf/index.php?topic=992.0
« Last Edit: May 02, 2022, 04:30:13 PM by Probeman »
The only stupid question is the one not asked!

Probeman

  • Emeritus
  • *****
  • Posts: 2384
  • Never sleeps...
    • John Donovan
Re: WDS hardware ideas, improvements
« Reply #20 on: May 02, 2022, 05:52:00 PM »
I am not at all suggesting an SDD detector for WDS. I agree that would be silly. 

Instead I am imagining the use of a simple photon detector at room temperature, perhaps a pin-diode detector. Even just to eliminate the hassle of dealing with P-10 gas bottles would be great!   :)

I am not an electronics expert at all, so I will defer on the technical details, but I wonder if the Bragg crystal gives us all the energy resolution we require. Obviously, we need to filter the thermal noise from the detector, but that shouldn't be difficult I am hoping. 

I wasn't thinking clearly when I wrote “Si wafer” when referring to the JEOL/Bruker solid state detector.  The design was of the SSSD type (solid state scintillation detector), but I no longer remember any details of its construction.  I’d be glad to look into this and then post in a new topic.  Would this be helpful?

I'm sure it would be most helpful. I for one would be very interested in not only what has been attempted in the past, but also what new "state of the art" detector technologies are available today for single photon detection.

John Donovan moved these posts to a related thread so we can focus on WDS hardware improvements here.
The only stupid question is the one not asked!

Mike Matthews

  • Global Moderator
  • Professor
  • *****
  • Posts: 135
Re: WDS hardware ideas, improvements
« Reply #21 on: May 03, 2022, 09:46:56 AM »
It’s not just the ‘hassle’ of P10 solidarity state WDS removes, it’s the vast majority of the instrument instability that disappears as well. By far the main reason for having to recalibrate on our standards is because of efficiency changes in the gas counters (e.g. because of changes in atmospheric pressure). There’s also not necessarily a requirement to cool an SDD, even at room T its energy resolution is still far superior to a gas counter, allowing for complete separation of high order overlaps and other signal artefacts.

Probeman

  • Emeritus
  • *****
  • Posts: 2384
  • Never sleeps...
    • John Donovan
Re: WDS hardware ideas, improvements
« Reply #22 on: May 03, 2022, 06:56:31 PM »
Mark Rivers (Argonne National Lab) offers these comments regarding photon detection:

Quote
I think the answer to the question depends a lot on what energy X-rays you are trying to detect.   There are many room-temperature solid state X-ray detectors with single-photon sensitivity and count rates > 1 M/cps.  The Dectris Pilatus and Eiger detectors have millions of pixels made of either Si or CdTe with these characteristics.  These work well for photon energies above ~5 keV.  For low energy x-rays the signal/noise limits single-photon capabilities because they don't have any electron multiplication.  They also are not offered in single-pixel packaging, and the pixels are very small.

One technology that might be useful for your application is avalanche photodiodes (APDs).  These have solid-state electron gain, are single-photon sensitive, and are very fast.  They are used for hard x-ray detection at our synchrotron (APS at Argonne).  I see lots of scholarly articles about soft x-ray detection with APDs, but I am not sure what is available commercially.

https://www.militaryaerospace.com/sensors/article/14200251/avalanche-photodiodes-electrooptical-xray-detection

https://www.slac.stanford.edu/econf/C0604032/talks/snic_baron.pdf
The only stupid question is the one not asked!

sem-geologist

  • Professor
  • ****
  • Posts: 162
Re: WDS hardware ideas, improvements
« Reply #23 on: May 05, 2022, 08:11:28 AM »
Yeah I know there is this problem with PHA shifts with increasing counting rate. I had identified the exact cause of that problem (Again I need to open that thread with review about detectors and counting pipelines), and thought up some not straight forward or counter-intuitive solution, which works more-less up to 80 kcps (in some cases maybe even up to 100 kcps) for high pressure detectors. Combined with rightly used dead time formula at last we could do something with this haunting pulse pile-up's to some degree and have a near linear response of count rate increase to increase of beam current (which now is nonlinear and the non linearity severely increases going above 10 kcps).

Below text and examples can look a bit cryptic, unless reader is familiar with charge sensitive preamplifiers and basic circuit components (inductors, capacitors, resistors). Below I show two PHA response graphs (overlays of PHA curves) to increasing of count rate. PHA curves in both sets (plots) were acquired at the same sample with near same current steps (setting C2 coil) to increase the count rate (beam current is not given as it is irrelevant - counter responds to incoming x-rays, not beam current). Count rate is given approximate in raw counts per second. Sample is a NiO standard. The measured line is Ni K alpha, the spectrometer is with LLIF and high pressure P10 gas counter.

description of above image: Auto PHA sets low gain, and high bias (supposing that gas amplification produce less noise than semiconductor amplification). That at higher count rate causes slower charging step transition (longer delta t between cascades) at charge sensitive preamplifier's feedback capacitor (working closer to fully charged state) and signal further processes with CR (OPAMP-based) differentiators translate that to lower amplitude, than it would be at lower count rate.

description of above image: Decreasing the bias to level where we still get same amount of counts (in integral mode and maximum gain) allows to charge the feedback capacitor of charge sensitive amplifier less (keeps the capacitor closer to discharged state), thus average load voltage is lower and capacitor voltage cascade transitions are more rapid. As the result CR (OPAMP) differentiators recognize consistently similar cascade transition at capacitor - as from low count rate to medium count rate the mentioned capacitor stays below half of charge voltage where charging slope is very similar in that range (from half charged to near discharged). Thus only high count rate starts to shift the PHA as it starts to enter the higher charge voltage of feedback capacitor.

Why the upper limit is 80-100kcps? There other mechanics kicks in, which is not possible to fix without reducing the shaping time.
There are two competing processes for PHA shift - one is relevant at low-to-medium-high count rates (the average DC component of charge sensitive preampllifier's feedback capacitor) and that can be reduced by reducing bias as shown at above pictures, which makes cascade voltage smaller and thus keeps the capacitor in its 0-50% loaded range. The other process comes at increased density of pulses (in time domain) where chance of registered pulse to start from negative voltage (at after-pulse of previous preceding pulse, as it is two differentiations it produce bipolar pulse) increases tremendously. The only way to fix that is to use faster pulse shaping amplifier (+ following baseline restorer would not hurt), however that would not anyhow fix the problem of PHA shifts made by capacitor average charge level loading above 50%.

The demonstrated way of reducing bias is a way with no hardware modifications. There is some questionable possibility to "fix" that with simple hardware modification - AMPTEK's A203 C.S.P and S.A datasheet describes that by adding additional capacitor in parallel (pins are exposed) to feedback capacitor, the sensitivity of C.S.P can be reduced. I don't know... I need to think about that... It looks for me that feedback resistor value is more deciding at what level capacitors are loaded depending from count rate, and increased capacity would not solve this problem and have just a negative effect of lost P/B (at that step in pipeline) ratio without any advantage at all.
Interestingly, EDS signal pipeline do away with this problem by having no feedback resistor and dropping capacitors charge with resetting circuit kicking-in after voltage gets over some threshold (at the cost of introducing small dead time at C.S.P.; small - input counts reaching 1Mcps produces about 5%). Interestingly XFlash Nano (tailored Bruker detectors for SX100 and SXFive) has very clever designed floating upper reset threshold, where reset happens reaching ~30% of max voltage at low count rates and is raised gradually above 50% (probably even more, had not tested more counts than 1 input Mcps ) with increased count rate. As side effect peak positions shifts a bit toward lower energy at higher count rates (the same effect as observed on PHA shifting on SX WDS). Our SEM equipped Bruker EDS (XFlash 6|30, newer technology than XFlash Nano) feedback capacitors has two reset thresholds (low, and high) - which moves symmetrically away from middle charge point of feedback capacitor and thus shift of peak at different count rates is less noticeable. However that comes at clearly worse resolution than what we see on XFlash Nano on SX'es at small-medium count rates.

There are other wider implications of investigative experiments I came to that above presented conclusions and I will happily share it further (I really should start separate thread dedicated for pipelines of detector-amplification-counting).
« Last Edit: May 06, 2022, 05:52:01 AM by sem-geologist »