Testing new multi-channel analyser PHA capability for Cameca SX

[Probeman]

This is an old thread but since people have been discussing PHA recently on the SX50 listserver, and also comparing the MCA (multi channel analyzer) PHA versus the traditional SCA (single channel analyzer) PHA, I decided to run some tests on my instrument which (except for Spectro 1 which has a cracked crystal and therefore lower intensities), should be a good comparison of a (mostly) normally running SX100 instrument.

If you look at the filenames of the images attached below (login to see attachments), you'll note that I ran PHA using MCA and SCA on Ti Ka on all five spectrometers at both 10 nA and 100 nA. I used 0.1 sec per point and 40 points on the (traditional) PHA SCA scans.

As noted previously in other topics, the MCA method produces PHA data much faster but at much lower energy resolution. I guess that's Heisenberg for you!   

It's also interesting to toggle between the 10 nA and 100 nA PHA scans (both MCA and SCA) and one can see the PHA peaks at 100 nA shifting to higher voltages from the lower intensities at 10 nA.

[BenjaminWade]

Hi all
I have finally found some time to start using MCA in PfE, and am struggling at the moment to get my PHA plots from Peaksite to match with those I acquire in PfE. I have an SXFive with dynamic baseline behaviour, and as such have modified my ini file to the following parameters:

PHAMultiChannelMin="0"
PHAMultiChannelMax="5"

When I run a PHA through Peaksite, visually I can see the plot continuing all the way up to 5V, however when I export the PHA plot through Peaksite the voltage column only goes up to a maximum of 4.826V? Looking at the rest of this post I note Gareth's finished at 4.98V where it was assumed the last 0.02V was the last channel missing, but mine is missing ~0.174 V...

What is strange is that a visual inspection of the PHA plot in Peaksite I can even see little inflections in the curve right up to 5V where there must be datapoints...


As a result, when I overlay a Peaksite PHA plot (using data exported from Peaksite which only goes up to 4.826V), to one that I acquire  in PfE which acquires all the way up to 5V (using 0.1s/256 intervals), there is an obvious mismatch as below:


The shapes of the plot are identical, however there is a lateral offset of which I was hoping was going to be 0.174V, but its actually around 0.130V. Also there is a strange increase in channel intensity when I acquire through PfE, and I dont know why this is.

Doing just a visual comparison the PHA plot acquired in Peaksite I see that it is centred around 3.6V and has a Y maxima of ~250. When I run the PHA scan through PfE I can see the PHA plots acquire in Peaksite, and Peaksite displays the last acquired scan from PfE. Visually I can see it is almost identical to that acquired through Peaksite in both peak maxima and peak centre (ie centred at ~3.6V, Y maxima ~250). See image below of the side by side comparison. You could be forgiven to think they are the same plot but they are actually subtly different.


So visually I think I am OK that the MCA PHA scan in PfE is doing the same as what Peaksite is doing, but it would have been nice to confirm that by overlaying some plots on each other!! Does anyone have any experience with what might be going on here with regards to my voltage maxima from export of 4.826V, and why my maxima channel counts are totally different in my PfE PHA scan....I am tearing my hair out.

Cheers

[Probeman]

Hi Ben,
Please don't tear your hair out!   

Here's the thing with the Cameca MCA PHA method: the function call only returns the intensity data, not the voltage values. The voltage values are calculated based on the INI file PHA Min/Max and what those should be is not exactly clear to me.

I'd suggest two things. First, turn on DebugMode in PFE and you will see much of the low level communications during the PHA acquisition. In fact if you turn on VerboseMode you will even see the raw intensities coming back from the instrument in the PFE log window.

But due to the design of the MCA electronics (only 8 bit MCA), I wouldn't attempt to glean much scientific information from these scans. Instead I would turn off the MCA mode and acquire your PHA scan one point at a time as god intended!   

See here to acquire "traditional" PHA scans on your Cameca instrument:



You will obtain quite different (higher resolution) scans and the voltage values will be exactly what they claim to be.

[BenjaminWade]

Hi John
Ha right OK thanks I will continue to acquire my PHA as the creator intended....but now that we are opening pandora's box this takes me to my next problem and why I wanted to start using MCA for PHA in PfE. I just cannot get good traditional PHA scans in PfE no matter what beam current, x-ray line, window, software/ini settings I use, they always come out extremely "jagged" and useless. This has always been the case for me so I have just been using Peaksite for PHA scans, and PfE for Gain and Bias scans.

Below are some examples of "traditional" PHA scans I get in PfE (Count Time of 1/256 intervals).





You can see they are pretty horrible. I don't know if its any way related, but I also get really strange artifacts in my Gain scans in PfE. See some Gain scans below of different elements on different spectrometers (all at same beam current, counttime/interval of 1/256). Although not entirely obvious on SP5/LIF I think there is still weirdness at those points. If I use PET on SP5 I still see the bumps as well.



What is really weird is that those strange "bumps" in CPS all somehow coincide with each other on the X-axis even on scans done on totally different days at totally different Gain voltage ranges on totally different crystals?? How is that even possible? These "bumps" do not occur in any Bias scans, only in Gain scans. I have tried changing point time settings and interval and it makes no difference, they all still occur at exactly the same places...

Here are gain scans on Si Ka at different count times and interval settings:



The only thing I can think of is that its some strange time dependant thing going on from when you initiate a scan, and at these very specific regular time intervals it somehow changes the acquisition time per spot to change the resultant calculated CPS number?? But I dont really see how that would work as it should change the periodicity when I change the count time?

For reference here are the Bias scans:






Also while we are on the topic and looking in pandora's box , I have a few other problems....

1) There is a tiny bug in exporting Gain/Bias scans from the from the display fitting window, in that when you click on the "Export Data" button and you get the window to save the DAT file, the automated file name it pops up with labels it as the opposite of what it is (in that if you are trying to export a Bias scan it labels it as a Gain scan and vice versa);

2) Mostly a comment, but in Bias scans the first data point is almost always some randomly high CPS value which I just ignore. I was wondering if it was a voltage settling thing after changing the Bias a large amount, as I saw the PHAFirstTimeDelay setting in the ini. I increased this from 5 to 10 seconds which didnt make a difference, so I guess that setting is just for PHA scans not Bias scans?

3) I cannot run PHA scans automated, only manually one at a time through the PHA properties window. If I attempt to run PHA scans through automated actions I get the error message below



These settings are corresponding to the first element its trying to do a scan for, but they all seem fine to me I cannot see anything wrong? I have checked my scalers file to make sure nothing is whack in there but it all seems fine.

Cheers

[Probeman]

Well that doesn't look very good at all!   Try running some scans with only 40 points or so.  Also increase the time per point. Be patient!

I'll try running some non MCA PHA scans on my instrument and see what I get, but it probably won't be until next week.
john

[Philipp Poeml]

My traditional PHA scans look as crappy as Ben's. Exactly the same. One thing we tried was to get "purer" P10 gas, but this did not seem to help. I would also be very interested in a solution to this.

[Probeman]

There's several things to consider here.

First of all the MCA method uses several loops of averaging to "smooth" the intensity data. And because it's always integrating, it can acquire more photons per clock time, just as a ROM (continuous) wavescan is more efficient than a step/count wavescan. 

This is why the MCA PHA method is preferred for normal use. My point was simply that if Ben is interested in making exact measurements of his PHA distributions, he might want to try using the "traditional" PHA method. Because then the X axis (voltage) values are exactly known. Again, analogous to step/count wavescans where the spectrometer position (x-axis) is exactly known and not calculated by interpolation as when performing ROM (continuous) wavescans.

Second of all, one needs to have a sufficient integration  time when performing "traditional" PHA scans. The default in PFE in 0.1 seconds for 40 acquisition points. Running the default PHA scan parameters (0.1 sec/40 points), here is a "traditional" PHA scan from my SX100 this morning:



And here is an MCA PHA acquisition using the same conditions:



My point is simply that one gets better energy resolution and more accurate x-axis values (voltage) when using the traditional PHA acquisition method.

I have no idea why anyone would think that this has anything to do with the purity of the detector gas.  Both PHA acquisition methods are using the same detector!

[Probeman]

1) There is a tiny bug in exporting Gain/Bias scans from the from the display fitting window, in that when you click on the "Export Data" button and you get the window to save the DAT file, the automated file name it pops up with labels it as the opposite of what it is (in that if you are trying to export a Bias scan it labels it as a Gain scan and vice versa);

Hi Ben,
I will look into this. I haven't used the Export PHA button in many years, thanks for the heads up.

2) Mostly a comment, but in Bias scans the first data point is almost always some randomly high CPS value which I just ignore. I was wondering if it was a voltage settling thing after changing the Bias a large amount, as I saw the PHAFirstTimeDelay setting in the ini. I increased this from 5 to 10 seconds which didnt make a difference, so I guess that setting is just for PHA scans not Bias scans?

It makes a lot of sense because the bias scan lowers the bias voltage dramatically for the first point and it could very well require some settling time when the bias voltage is lowered by that much.  But the PHAFirstTimeDelay setting is the wrong one to change. This value is only applied the *first time* the bias voltage is set on that spectrometer (per application instance). You'd have more luck by editing the BiasChangeDelay parameter in the [software] section of the Probewin.ini file.

But I think a better solution is for Probe Software to add a small delay after setting the first bias voltage in the bias scan acquisition

3) I cannot run PHA scans automated, only manually one at a time through the PHA properties window. If I attempt to run PHA scans through automated actions I get the error message below



These settings are corresponding to the first element its trying to do a scan for, but they all seem fine to me I cannot see anything wrong? I have checked my scalers file to make sure nothing is whack in there but it all seems fine.

I don't think this has to do with the automation per se, but we are looking into it.  I suspect it has to do with using so many points per PHA scan.  That is, when I use the default 0.1 seconds and 40 points I have no trouble running automated PHA scans from the Automate! window.

[Probeman]

Here is a comparison of "traditional" PHA for Fe Ka on my Cameca Sx100:



Again using 0.1 seconds and 40 points. Note that the escape peak is much better separated from the main peak. And here is the MCA PHA for Fe Ka using the same conditions:



Note that the energy resolution is significantly less in the MCA PHA acquisition, making the escape peak less visible. This is actually a *good* thing in several ways, since the escape peak photons are actually Fe Ka x-rays, just with an energy loss equal to the emission energy of the Ar K shell.  So including them in ones PHA range is more accurate (when considering gain shifting from different Fe emission intensities due to composition) and also better statistics.

[Karsten Goemann]

The MCA although returning 256 channels may in fact only have 32 channels as the values in between look like they are interpolated (linear), see earlier posts in thread. If this is true this would limit the spectral resolution a lot. I've used these settings in probewini.ini on our SX100 because of that:
PHACountTime="0.1"              ; default integration time for PHA acquisitions
PHAIntervals=32              ; default PHA intervals

We could never get good gain scans on our SX100 either, with similar step artefacts to what you've found. My thoughts back then were that the electronics may need more time to settle in between gain changes, but I'm not sure if that's the actual explanation. The Cameca philosophy seems to be to allow the PHA peak to move, rather than always forcing it to a 2V maximum or something like that for every x-ray line measured. With the forced extended deadtime there's probably less danger of running into deadtime correction issues and we did not find pulse height depression an issue at reasonable count rates. We had a general rule of thumb to not go above 20kcps. We tended to use pretty much fixed gain & bias values for a given spectro/crystal combination (which were very stable over time) and just changed baseline and window (PHA diff) as required.

[John Donovan]

Hi Karsten,
The Cameca MCA function call does indeed return 255 discrete values, and they may be smoothed before they are returned by the firmware, but I suspect not. As 255 channels would be a normal 8 bit MCA, which makes sense that Cameca would utilize this.

I'm more concerned with the *non* MCA mode for PHA scanning on the Cameca, and why the returned distribution is so "spiky" when more than 80 or so values are measured.  Probeman can confirm that this does indeed occur on his SX100 instrument also.

The concern is that we use the same code to acquire JEOL and Cameca non MCA PHA scans, using the same function call that we use for normal WDS photon counting.  So just as a sanity check we'd really like to see a JEOL PHA scan on a strong peak using a 1 second integration time and say 200 or so intervals. Do these JEOL PHA scans look as "spiky" as Benjamin's SXFive scans?

The good news is that we at Probe Software found a fixed the several minor bugs that Benjamin reported yesterday, specifically 1), the problem running non MCA PHA scans with over 80 points when performing automated peaking from the Automate! window, 2), the naming of the PHA scan export file and 3), we added a delay after setting the first bias voltage for a bias scan.

Right now this first point bias scan delay is using the RealTimeInterval from the Probewin.ini file, which is usually around 400 ms or so, but we could use another value, in fact, thinking about it maybe we should.

Benjamin: please download and try this latest version of PFE and see if you get a smaller intensity for the first point on your bias scans and feel free to test the other bug fixes.

But as to what is causing these "spiky" PHA scans on the Cameca when running in non MCA mode with 80 or more points is still a mystery.

Karsten: if you get a chance please test a PHA scan using 200 points on your 8530.  Anette just sent me a scan from her JEOL instrument and it looks completely fine.  So it appears to be a problem specific to the Cameca instrument when in non MCA mode.

I wonder if Cameca's firmware is messing with the gain or something when the PHA window is being adjusted during non MCA PHA scans?

[BenjaminWade]

Hi all
Many thanks for the replies, I am glad I am not the only one re: PHA scans and weird gain scans...

John I appreciate your comment re: resolution of SCA vs MCA PHA scans which is why I was hoping to get it working OK in PfE. However I still think something strange is going on with my PHA scans and its not a count/integration time issue, or a #of points issue. Below are scans ran with 40 points at 0.1/1/10 seconds and you can see that the strange spikiness occurs in all, just at different amounts of scatter.





With regards to other comments, you have just replied while I am still writing this! Great news with the Automate, I can confirm from testing this morning that it was anything over 80 points that it crapped out on. I must have missed that Bias delay setting in the ini file as well, thanks for pointing that one out.

We could never get good gain scans on our SX100 either, with similar step artefacts to what you've found. My thoughts back then were that the electronics may need more time to settle in between gain changes, but I'm not sure if that's the actual explanation. The Cameca philosophy seems to be to allow the PHA peak to move, rather than always forcing it to a 2V maximum or something like that for every x-ray line measured. With the forced extended deadtime there's probably less danger of running into deadtime correction issues and we did not find pulse height depression an issue at reasonable count rates. We had a general rule of thumb to not go above 20kcps. We tended to use pretty much fixed gain & bias values for a given spectro/crystal combination (which were very stable over time) and just changed baseline and window (PHA diff) as required.

I am glad its not only me re: the weird steps in the gain scan. I will email Cameca and see what they say. I am not sure its a gain settling issue at least on my system, as 1) the bumps occur so regularly in every scan you do at exactly the same Gain values for a given x-ray line; and 2) I did a laborious manual gain scan in peaksite yesterday waiting 10 seconds after changing the gain and got exactly the same bumps in returned CPS at the same places. So something very weird is going on in the electronics.

Cheers

[Probeman]

Hi Ben,
On your non MCA PHA scans above, it seems that they look more "steppy" than "spiky", at least in the 10 second interval scan.  I wonder what could be going on... Anette just sent me this screensnap showing a 256 point PHA scan on her JEOL 8530 which looks perfectly good:



She says she's never seen anything weird in her PHA scans.  But this is a little hard to compare because the JEOL instruments cannot do MCA PHA scans, and the Cameca PeakSight software cannot do non MCA PHA scans (that I am aware of).

On the gain scan issue, I did not run any of those today. Can you post an example of the weirdness in your PHA gain scans? 

One possibility with bias scans is that there might be a 10 volt resolution limitation in setting the high voltage.  There might be something similar with gain scans. I need to look through the code and see if there are any notes in there.

[Probeman]

One possibility with bias scans is that there might be a 10 volt resolution limitation in setting the high voltage.  There might be something similar with gain scans. I need to look through the code and see if there are any notes in there.

In other words, try some more scans (PHA, gain and bias) in PFE using a long count time, and see if you can observe on the Cameca PeakSight software what the actual increments are being displayed, for the x-axis values. Maybe while in debug mode in PFE and observing the calculated values sent to the instrument.

I suspect we might see something interesting.

[BenjaminWade]

Hi John
Yes perhaps my terminology wasn't quite right, it is perhaps more steppy than spikey.

My gain scan issue is what I previously posted regarding the "bumps"

..You can see they are pretty horrible. I don't know if its any way related, but I also get really strange artifacts in my Gain scans in PfE. See some Gain scans below of different elements on different spectrometers (all at same beam current, counttime/interval of 1/256). Although not entirely obvious on SP5/LIF I think there is still weirdness at those points. If I use PET on SP5 I still see the bumps as well.



What is really weird is that those strange "bumps" in CPS all somehow coincide with each other on the X-axis even on scans done on totally different days at totally different Gain voltage ranges on totally different crystals?? How is that even possible? These "bumps" do not occur in any Bias scans, only in Gain scans. I have tried changing point time settings and interval and it makes no difference, they all still occur at exactly the same places...

Here are gain scans on Si Ka at different count times and interval settings:



The only thing I can think of is that its some strange time dependant thing going on from when you initiate a scan, and at these very specific regular time intervals it somehow changes the acquisition time per spot to change the resultant calculated CPS number?? But I dont really see how that would work as it should change the periodicity when I change the count time?

For reference here are the Bias scans:

My Bias scans are fine and I have no issues there, its just the strange plateau bumps in the Gain scans. From Karsten's comments it appears that perhaps he encountered similar issues as well. As per my quoted post above changing the count time had no effect on removing the artifacts in the gain scan.

I have no idea if my Gain problem is related to the steppy/spiky nature of the PHA scans. Would be interested to see if Philip also has strange Gain scans given he might be encountering the same problem I am having with the PHA scans.

What I did yesterday was manually run a Gain scan in the Peaksite software by manually setting the baseline and window, then starting at 1200V incrementing the Gain in 10V increments, waiting a few seconds then recording the resulting CPS. The resulting data is below and as you can see it looks identical to what I get running a Gain scan in Probe for EPMA.



So I think its some weirdness direct from the SXFive electronics. I have emailed Cameca and am awaiting a reply.

With regards to the PHA scan problem, I will run some more PHA scans and watch to see what is happening to the increments in Peaksite and debug mode. If my PHA steppy problem is related to the Gain problem though I suspect I wont see a problem given I got the same result manually running the Gain scan in Peaksite

Cheers