zorch,
this is really useful info!
Of course all of the temporary circuit modifications described above are done with Electronics Power = OFF!
So I suppose these soldering work was done with PCB mounted in place? What about measurements, how had You measured bias voltages with PCB being not powered? A hint: You don't need to power whole electronics off, just unplug the cable and wait for some time for capacitors to discharge.
I did replace leaking C1 on the spectrometer board and the bias voltage at that capacitor rose about 20V. This eliminated the instability seen in the peak scans and allowed us to set the bias back to the old command value of 1830V for PET. While the bad cap was in circuit, we had to increase the bias command to 1850V to get a decent count rate. A cracked capacitor would also result in leakage current because exposed plates would arc past the insulating layers, creating a carbon track that would act much like a resistor. This would lead to a sag in bias voltage at that part of the circuit.
Interestingly, We don't see described behavior, maybe it depends from how does the capacitor crack (parallel to layers vs perpendicular to layers). Also If bringing baseline up with gain, properly working system will give same amount of counts as canonical 1830V. I tested that and I can reduce it down by 100V. This does then less gas amplification, and relies on analog gain. Unfortunately with cracked capacitor that is no more possible as gain then catch the noise introduced by cracked capacitor (in our case), and doing that will saturate the counting system with false, noise pulses.
Now, Why I would want to reduce bias, one would ask?
I though that an additional leakage current might be taking place inside the P10 tube which carries the bias voltage to the detector. But pulling one end of R9 out of the circuit made no change in the voltage at C1. This result makes it clear that there is no problematic leakage taking place in that bias wire tube.
Indeed, you thought very well. There should be no perceptible leakage current in not irradiated detector. But in working irradiated mode the chamber will pass current from cathode to the anode which will be proportional to amplification and number of incident X-rays (plus other minor stuff). Actually measuring precise leakage current is near impossible due to cosmic rays (0 to 20 random incidents per second) and depending from the energy Townsend avalanches will have different amplitudes and pulsed current flow through the detector events, which could be missunderstood as leakage. The current will get significant with high pulse rate and high gas amplification (high bias voltage), and on Jeol side that causes significant PHA shift toward lower energies... but on Cameca that effect is less severe. Albeit it can be even more reduced by reducing bias, and increasing gain - the HV bias regulation has more chance to keep the set bias in check even at decently high count rate (i.e. 100kcps).
A very strange 2V rise in the voltage at C1 occurs when I pull one end of C2 (DC blocking signal coupling cap to the AMPTEK A203 IC). This apparent leakage of C2 occurs despite having cleaned the dust off with IPA. Since the AMPTEK part is a charge sensitive amplifier, one would expect some degree of DC voltage offset to occur in this devices output with leakage current from the bias supply. I'll replace C2 soon and let you know what I find. It should be noted that any DC offsets in this part of the circuit do not show up at the PHA or counting system because downstream capacitors C5 and C10 block DC voltages.
In case of Jeol probe it has high voltage cable from spectrometer HV supply somewhere far far away (in this sub- µs lenghting measured events - it is far far away, even if it is just one or two meters away). In case of Cameca probe, HV supply is integrated inside spectrometer as close as possible to where this HV is needed. Additionally to that the HV supply output is monitored and HV supply input is regulated with very fast OPAMP. In case there is any leak current, it is compensated so that output voltage of supply would stay the same proportional to set voltage reference. Thus reducing C2 leakage wont increase the bias current. Every capacitor has leakage current and that is normal and expected. 2V is only ~ 0.1% in this case, will You find better high Voltage capacitor with lower leakage?
Additionally leakage is very well known in X-ray detection preamplifiers, i.e. leakage through FET in EDS detector preamplifiers, and still they work. In case of this leakage, AMPTEK A203 IC will have few drains. input pin is connected with direct and reversed polarity diodes (inside IC), which will lower the leakage down to ~0.6V. The signal further is decoupled with capacitors for feedback and input to Charge sensitive preamplifier part (inside IC). thus that downplayed to 0.6V DC offset is ignored. Finally, differently to EDS preamplifiers, the feedback of CSP has built-in resistor which not only constantly discharges the built-in feedback capacitor, but also downplays further the DC offset or leakage current.