Tuning PHA Settings

[Probeman]

Tuning our PHA settings for our WDS spectrometer electronics has always seemed, to me at least, a bit of a "dark art".   

It doesn't help that on JEOL instruments one cannot make fine adjustments to the PHA gain (they are fixed values at multiple of 2, i.e., 16x, 32x, 64x, etc.), so that one must set the PHA gain to a rough value and "tune" the PHA by adjusting the high voltage bias.

While on Cameca instruments, one adjusts on the detector bias to obtain a proportional response from the detector (or even better just leave the bias voltage at nominal values of ~1300v for low pressure flow detectors and ~1850v for high pressure detectors), and "tune" the electronics by making fine adjustments to the PHA gain.

Traditionally we are taught to simply adjust the bias or gain (as the case may be) to produce a PHA peak of around ~4v or so on JEOL instruments and ~2 to 2.5v on Cameca instruments.  The idea being that this should keep most of the PHA peak above the baseline level, but still not encroach on the window level.

But as SEM Geologist has pointed out, we generally should not be using the DIFFERENTIAL PHA mode (where both the baseline and window levels are actively filtering pulses), and instead simply utilize INTEGRAL PHA mode where only the baseline level filtering is active.  As SG correctly points out, the count rates utilized in the dead time correction will not be accurate if we have excluded a significant number of pulses from our dead time corrections, i.e., from higher order reflections being filtered by the window level.

In fact, to obtain a linear response from the spectrometer it is better practice to keep our PHA tunings as "open" as possible, so that we do not lose any pulses when we perform analyses at different count rates, thereby causing pulse height depression effects as Anette demonstrated here:

https://smf.probesoftware.com/index.php?topic=1466.msg11271;topicseen#msg11271

That generally means leaving our baseline values at the lowest possible settings, e.g., ~0.2v for Cameca and ~0.5v for JEOL. And if one absolutely must utilize DIFFERENTIAL mode, then leave the window levels at ~4.5v for Cameca and ~9.5v for JEOL.  The idea being to obtain a linear response of intensity with concentration (and/or beam current) as much as possible, and then have the quantitative spectral interference correction do its job!  Remember, for 1st order spectral interferences, the window level filtering has no effect at all, except for a few stray cosmic rays!   

But again, there is generally no reason to utilize PHA DIFFERENTIAL mode at all, except possibly in rare cases where a high order (<1) spectral interference is present, and no suitable interference standard is available. Remember, the quantitative interference correction requires a standard with a known concentration of the interfering element, and *none* of the interfered element (nor any other interfering elements).  I can only think of the interference situation of Na Ka 2nd order interference on O ka, and that is because it is difficult to find a standard material that contains sodium, but no oxygen.  I've tried natural cryolite for this interference correction, but it seems to contain some oxygen, maybe from surface reactions.

I recently was asked to explain how I approach PHA tuning of the baseline and window levels, and based on recent work with several colleagues in our soon to be published new dead time correction calibration method and dead time correction expression, here is how I would explain things:

Here are some places to start for PHA settings in the EPMA user forum:

https://smf.probesoftware.com/index.php?topic=1466.msg11416#msg11416

https://smf.probesoftware.com/index.php?topic=1466.msg11450#msg11450

https://smf.probesoftware.com/index.php?topic=1466.msg11636#msg11636

In summary, one should almost never use PHA differential mode (so that removes question about the PHA window setting!), and instead use only integral mode. The key point is not to set the center of the PHA peak at a specific position, but only to ensure that the tail (and escape peak if present), are fully *above* the base line level. If the right side of the PHA peak is cut off on the right side of the PHA plot, that will not matter when one is in integral mode. Anette and I have tested this question on both JEOL and Cameca instruments.

The worse thing one can do is set one's PHA settings so that at different concentrations or beam currents the PHA peak shifts downwards (to the left) so that is gets cut off by the baseline at lower count rates.  That is why one should always tune one's PHA setting at the highest expected concentration and at the highest expected beam current for that particular probe run.  That way, at lower concentrations and/or beam currents, the PHA peak will only shift to the right, and in integral mode all the photons will still be counted even though the PHA peak may appear to be cut off on the right side of the plot.

We realize that this is not the way we were taught to tune PHAs but that is what we have learned since then!

[Probeman]

Traditionally we are taught to simply adjust the bias or gain (as the case may be) to produce a PHA peak of around ~4v or so on JEOL instruments and ~2 to 2.5v on Cameca instruments.  The idea being that this should keep most of the PHA peak above the baseline level, but still not encroach on the window level.

But as SEM Geologist has pointed out, we generally should not be using the DIFFERENTIAL PHA mode (where both the baseline and window levels are actively filtering pulses), and instead simply utilize INTEGRAL PHA mode where only the baseline level filtering is active.  As SG correctly points out, the count rates utilized in the dead time correction will not be accurate if we have excluded a significant number of pulses from our dead time corrections, i.e., from higher order reflections being filtered by the window level.

In fact, to obtain a linear response from the spectrometer it is better practice to keep our PHA tunings as "open" as possible, so that we do not lose any pulses when we perform analyses at different count rates, thereby causing pulse height depression effects as Anette demonstrated here:

https://smf.probesoftware.com/index.php?topic=1466.msg11271;topicseen#msg11271

That generally means leaving our baseline values at the lowest possible settings, e.g., ~0.2v for Cameca and ~0.5v for JEOL. And if one absolutely must utilize DIFFERENTIAL mode, then leave the window levels at ~4.5v for Cameca and ~9.5v for JEOL.  The idea being to obtain a linear response of intensity with concentration (and/or beam current) as much as possible, and then have the quantitative spectral interference correction do its job!  Remember, for 1st order spectral interferences, the window level filtering has no effect at all, except for a few stray cosmic rays! 

But again, there is generally no reason to utilize PHA DIFFERENTIAL mode at all, except possibly in rare cases where a high order (<1) spectral interference is present, and no suitable interference standard is available.

I was recently re-reading a paper by M. Fialin from1999, and he points out something I had not previously considered (or at least possibly forgotten!). And that is that at low sin theta positions, "specular reflection" can dramatically increase the intensity of the background.



Therefore the use of differential mode could be helpful in reducing this "specular reflection" at low sin thetas. This would be important for improving trace element sensitivity, since we know that the peak to background ratio is critical for trace element sensitivity.

[sem-geologist]

Interesting. Could a copy be shared of that paper? (or at least a bit more details picked and shared from the paper). Looking to the year of publication I only can guess the demonstrated wavescans are made with SX50. Why it could be important? I know that on SX100 and SXFive (which electronics I am most familiar with), different to Jeol probes, the diodes are placed in series to incoming signal, which pass to PHA only positive, and only pulses with physical amplitude above ~0.6V relative to electric GND plane (the value depends from temperature, type of diode and etc.). From low energy side there is then not much to block with differential PHA as it is blocked physically already with diodes. As far I had tested PHA diff mode on SX100 and SXFive additionally to reducing background I had also seen significant (if not more pronounced) reduction of peaks on most of Xtals (be it Johhan, or Johhansson type). Reducing background 4 fold, and reducing peak 4 fold gives similar Pk/Bkgd thus no real gain with unnecessary complications of PHA introduced.

[Probeman]

Yes, it was an SX50. But I suspect from the discussion that the "specular reflection" x-rays are a much higher energy and not of similar energy to Cr Ka.  Hence the claimed benefit for using differential mode for trace element analysis...

Here's the Fialin paper (attached).

It would be easy to test trace element sensitivity for Cr Ka in Si using integral vs. differential mode on an SX100/SXFive (or JEOL for that matter).

[Probeman]

A colleague of ours (who wishes to remain anonymous), was intrigued by the Fialin full range PHA scans in the above post:

https://smf.probesoftware.com/index.php?topic=1526.msg13242#msg13242

and decided to reproduce them on his own Cameca instrument.

Here they are:











These scans were performed on boron nitride and on the TAP spectrometers one can see a small Al peak due to the polishing compound. Also note the Argon absorption edge visible on the PET spectrometers.

But compared to the TAP scan that Fialin published, our colleague's LIF scan shows little to no difference, the PET scans a slight difference, and only the TAP scans show even a minor difference in the Integral vs. Differential mode scans. 

These scan were performed at 25 keV and 50 nA. But Fialin was running at 30 keV and 500 nA! I can understand the 5 keV increase could contribute to the greater difference in the Fialin data, but the higher beam current should only improve the statistics as far as I can see.

It would be interesting to see some integral vs. differential scans from a few JEOL instruments...

[sem-geologist]

A colleague of ours (who wishes to remain anonymous), was intrigued by the Fialin full range PHA scans in the above post:

Interesting, albeit the effect is smaller from what I was getting for UMa, but I was using narrowed window. What kind of differential was used? Wide window? window set to particular X-ray line? or Diff Auto (the automatic PHA condition shift depending from spectrometer position)?

Also it makes clear we have these specular reflection on 20k-30k cameca sin theta positions. Remember that weird behaviour as we looked for D-H limit with WDS scans? Where this intensities with slope apeared over energy of beam of 10kV? My PHA investigation in that setup showed some nonsense (no high energies) and Narrow window with PHA diff could not cut that artifact out. Specular reflection at 30k to 20k makes sense as it is wideband photon noise.

[Probeman]

A colleague of ours (who wishes to remain anonymous), was intrigued by the Fialin full range PHA scans in the above post:

Interesting, albeit the effect is smaller from what I was getting for UMa, but I was using narrowed window. What kind of differential was used? Wide window? window set to particular X-ray line? or Diff Auto (the automatic PHA condition shift depending from spectrometer position)?

Also it makes clear we have these specular reflection on 20k-30k cameca sin theta positions. Remember that weird behaviour as we looked for D-H limit with WDS scans? Where this intensities with slope apeared over energy of beam of 10kV? My PHA investigation in that setup showed some nonsense (no high energies) and Narrow window with PHA diff could not cut that artifact out. Specular reflection at 30k to 20k makes sense as it is wideband photon noise.

Our anonymous colleague responds:

I kept the window wide and looked at the PHA distributions on each crystal using the following lines to make sure the pulses remained above the baseline and below the window for what should at least be close to the full range of energies each crystal measures. I used the following lines to test this:

    (L)TAP: Si Ka, F Ka
    (L)PET: Cr Ka, Si Ka
    LLiF: Ge Ka, Sc Ka