Author Topic: Why is the PHA minimum 500 mV and why is the minimum Voltage linked to the gain?  (Read 3261 times)

Philipp Poeml

  • Professor
  • ****
  • Posts: 217
Hi,

I was always wondering about the questions in the subject.
Is there a reason for a minimum of 500 mV in the PHA?
And why is it linked to the gain? From gain 800 or so the minimum Voltage also slowly incerases.
This seems to be a firmware thing, because it applies to PeakSight and PfE. If, for example I am at 1500 gain, and I type 500 mV in PeakSight for the lower limit, the setting is being set to 800 instead (not 800, I am not in front of the machine, I do not know exactly right now). Same for PfE. If I set the PHA at gain 1500 to 0.5 V I can see in PeakSight that instead it is set to 0.8 V. If I press the "read" button in PfE after that, it also does read the 0.8 V.

Thanks
Philipp
« Last Edit: September 23, 2019, 12:42:10 PM by Philipp Poeml »

Probeman

  • Emeritus
  • *****
  • Posts: 2384
  • Never sleeps...
    • John Donovan
I was always wondering about the questions in the subject.
Is there a reason for a minimum of 500 mV in the PHA?
And why is it linked to the gain? From gain 800 or so the minimum Voltage also slowly incerases.
This seems to be a firmware thing, because it applies to PeakSight and PfE. If, for example I am at 1500 gain, and I type 500 mV in PeakSight for the lower limit, the setting is being set to 800 instead (not 800, I am not in front of the machine, I do not know exactly right now). Same for PfE. If I set the PHA at gain 1500 to 0.5 V I can see in PeakSight that instead it is set to 0.8 V. If I press the "read" button in PfE after that, it also does read the 0.8 V.

This is just too weird!  I'd never noticed this before! I guess because you have to click the PHA button in PeakSight to see the baseline and window values and we always set our PHAs from Probe for EPMA.

It definitely appears to be some firmware issue because there's nothing in PFE that forces this value to around 0.5 volts. It will warn you if you specify a baseline less than 0.2 volts, but you'd get a warning at least.
The only stupid question is the one not asked!

Philipp Poeml

  • Professor
  • ****
  • Posts: 217
I assume this is the same for all SX100s. Is that so? Did you try on yours?

Probeman

  • Emeritus
  • *****
  • Posts: 2384
  • Never sleeps...
    • John Donovan
I assume this is the same for all SX100s. Is that so? Did you try on yours?

Yes, I did try it on our SX100. On sp2 the baseline won't go below 0.531 volts.  I had never noticed this before. Too bad, getting Cameca to change this firmware behavior will require a lifetime of effort.  I wonder what their reasoning was/is?
The only stupid question is the one not asked!

Philipp Poeml

  • Professor
  • ****
  • Posts: 217
That's exactly my question. What is the reason?
On our machine the lower limit is 500 mV and it will increase with increasing gain.
Anybody knows?

Karsten Goemann

  • Global Moderator
  • Professor
  • *****
  • Posts: 218
Remember this old thread:
https://probesoftware.com/smf/index.php?topic=217.msg1048#msg1048

The lower baseline limit depends on the gain. It's a linear relationship.

Cameca introduced this for the newer SX100s and it was carried over to the SX5s. Our vintage 2003 SX100 at UTAS, serial #846 (with the "old" boards in the config) didn't have it.

The baseline is to cut off noise, isn't it? Maybe with increasing gain that noise peak gets broader?

Probeman

  • Emeritus
  • *****
  • Posts: 2384
  • Never sleeps...
    • John Donovan
That makes perfect sense, I had forgotten about this.  Increasing the gain does of course broaden the peaks (gain = multiplication), but I'd still rather have control over it myself.

How else can I show students the noise peak?   :)
The only stupid question is the one not asked!

Philipp Poeml

  • Professor
  • ****
  • Posts: 217
Yes, ok for the noise peak. So they are automatically increasing the baseline to get rid of a wider noise peak. Interestingly such a noise peak never shows up in the PHA scans. I guess that means the method works.
Unfortunately like this it is never possible to get the complete Ar escape peak.

Karsten Goemann

  • Global Moderator
  • Professor
  • *****
  • Posts: 218
So even if you manually increase the gain (instead of using PHA Auto) you can't prevent the Ar escape peak from being cut off at the baseline because the baseline limit keeps going up? Have you tried adjusting the bias instead? Do you have enough separation between escape and main PHA peak to completely cut it off? Depends a lot on the x-ray line I guess.

In that sense the older SX100s had an advantage because they had a fixed lower baseline limit of 0.56V no matter which gain was used.

sem-geologist

  • Professor
  • ****
  • Posts: 162
I know this topic was inactive for some time (albeit PHA topic as whole is all the time hot).

I had an urge to reply to some comments like this:
I had never noticed this before. Too bad, getting Cameca to change this firmware behavior will require a lifetime of effort.  I wonder what their reasoning was/is?

and this:
That makes perfect sense, I had forgotten about this.  Increasing the gain does of course broaden the peaks (gain = multiplication), but I'd still rather have control over it myself.

How else can I show students the noise peak?   :)

Thankfully Cameca is not JEOL and legitimately removes the noise with hardware. More over with newer hardware the noise is removed adaptivelly - which is an improvement. That is the huge feature -- not a drawback, compared to JEOL, which leaves the background noise to PHA.

The technical reason:
Everything in PHA graph (and everything not, which is out from the right side of PHA scale (more than 5V or 5.5V)) introduce dead-time, as for background noise level pulse to be recognized as a pulse and give some digital amplitude value (so that it would be plotted on PHA graph, or that PHA filtering would work in real-time during acquisition) the electronics for deterministic set time is blocked from other incoming pulses. Not filtering out the noise from signal, and passing the raw noisy signal into PHA would produce floating (depending from electronic noise, which depends from many variables) large dead-time even at very faint X-ray intensities.

Filtering with hardware (zener diode, offset on pulsehold-trigger chip or DC shifting values with voltage-divider) makes electron background noise to make no impact on dead time.

Actually one can see the electronic noise if background noise deteriorates (increases in amplitude) for some reasons. I had learned this recently in the hard way as had noticed such noise started to show up in PHA at higher gain (much higher than normally set by auto) on one of spectrometer. The noise generated if passed to PHA was creating order of magnitude more counts than real X-ray counts thus very largely impacting the dead time. After comparing signals with oscilloscope I found that signal from that spectrometer is much more noisier, while ground (0V) had the same characteristics on all spectrometers (dominated with noise originating from crappy old Phillips 70A VME +5V PSU, but it does not matter as it is later filtered out with differential OPAMP). After opening spectrometer electronics cover (where preamplifier is mounted) I had noticed that one of HV bypass (3kV, 2nF) capacitors is cracked  :o). Nonetheless, if using "normal/auto" PHA conditions, that higher noise is completely hardware-cut-out and does not influence the measurements at all anyhow. In case of JEOL - any hardware component degeneration would dynamically increase the noise and dynamically change the dead-time of counter. So Cameca way is clever engineered and superior compared to lazy approach on Jeol probes.

BTW: to see the Argon esc peak better (for educational use i.e.; Normally it is poor idea overpush the bias, and lower bias than auto is always better idea) the Bias can be increased a bit temporary, which increase the peak/bkgd ratio in proportional counter, and separates the Ar esc peak from noise.
« Last Edit: June 09, 2022, 08:55:08 AM by John Donovan »