Low Level Gamma Radiation

It's just ion migration through the glass envelope. I don't know if folks have found a good solution to it, but BITD this was a serious limitation.

Cheers

Phil Hobbs

RH Electronics doesn't seem worried:

It should be easy to add a simple inverter and ground the cathode.

Hamamatsu shows several types of cathodes:

What is a photocathode?

A thin photosensitive film formed on the inner side (vacuum) of the light input window. It converts light entering through the light input window into photoelectrons.

- I didn't know that. Near infrared is 0.75 to 1.3 microns, so the file doesn't have to be very thick.

Types of photocathode

Bialkali photocathode for visible light region

Multialkali photocathode with sensitivity extending to near infrared region

Alkali-halide photocathode for UV light detection

GaAs, InGaAs (group III-V compound semiconductors) with sensitivity from UV to near infrared region

Oops - it is already grounded.

And the positive anode should eliminate any ion migration through the glass.

Well, there are both positive and negative ions in the crystal, so reversing the polarity would just change the nature of what migrates.

John

Migration after 10 dynodes of gain is better than migration at the photocathode where every electron literally counts.

Yup. Grounded-cathode is the usual method with scintillators. You couple the pulses out with a capacitor, so it's not that big a deal.

Cheers

Phil Hobbs

The anode isn't vulnerable to corrosion because it's not deposited on the glass, and it's a nice beefy piece of metal, compared with the very thin transparent photocathode.

Cheers

Phil Hobbs

I seriously do not understand this. With a grounded cathode, the signal you're extracting at the anode end is exposed to ripple from the PMT supply, without benefit of a multi-megohm divider chain. There's also the need to use a DC restorer of some sort to figure out where the baseline is. Both of these problems go away with a grounded anode. Seems like a no-brainer.

-- john, KE5FX

Photomultipliers are current sources; the current gain is 'exposed to ripple from the PMT supply', so that isn't a design feature from which to expect any difference at all. If the pulses are short, filtering against the ripple (longer duration than the pulses) might not be difficult. You'll be rejecting dark current either way.

I'd think transformer-coupling would be a natural way to capture pulses while rejecting lower frequency ripple and DC. In my experience with PMTs we used good regulated DC power.

PMTs were mostly used in pulse counting mode for astronomy. Height of the pulse doesn't matter provided that there is one (or not).

Image Photon Counting System, IPCS from Imperial College being the first new generation microchannel plate based imaging device in the 1980's.

My own experience was mostly of ion counting mass spec systems rather than photon counting. Deconstructed PMT in a hard vacuum and ion beams rather than light. We had to do some elaborate ion optics to provide a photon stop to prevent light from the plasma reaching the detector.

Hard vacuum requires stainless steel and no paint so an effective photon stop is harder to make than it sounds.

ISTR that in pulse counting mode it needed dead time correction once the count rate got high and that there was a cute way to run it in analogue mode by dropping the supply voltage and monitoring the current.

In both cases the supply voltages were as stable as we could reasonably make them (more so for the analogue mode). Cross calibrating the analogue to pulse counting modes as a function of mass was quite bad for the detector but essential if the results were to be meaningful.

Right, as far as that goes.

The photocathode is the high-Z end of the photomultiplier, and so most naturally goes furthest from ground. In free-space applications, we run the anode near ground and the photocathode at -1 to -2 kV, because the PC end doesn't draw any cuarrent to speak of.

That makes it easy to get the (largish) anode current out into normal low-voltage circuitry. (*)

The large scintillator crystal is out there where you can touch it, so for safety reasons it's good if it's kept near ground.

It's important for sensitivity reasons that the scintillator crystal be connected to the PMT faceplate through a continuous path with refractive index at least as large as the glass of the PMT envelope.

The reason is that a major fraction of the light from the scintillator is incident on its exit facet from angles beyond the critical angle for CsI -> air.

I suspect that this is where the difficulty lies: If you have diffuse light in a medium of refractive index n, and want to couple the light efficiently into air, it turns out that your maximum efficiency is

Sooooo, to avoid losing signal, you want to avoid coupling light from your scintillator into air.

The most straightforward way to avoid that is to optically-couple the scintillator to the PMT faceplate, either by direct contact or using some sort of index oil or gel. OK so far? It's also good if your detector survives use, though. Putting the scintillator at ground and the photocathode at -1 kV Photocathode corrosion due to halide ions drifting through the glass faceplate has been a real issue for a long time, and the saPower supply ripple is an easily-soluble problem, especially since a scintillation event is very bright in PMT terms--you aren't trying to do photon counting.

I've used cascades of PNP cap multipliers to do some pretty fun things with PMTs, including getting (to me) impressive linearity for analogue RF modulated detection.

Usually it's important for safety reasons that the metal housing for the big ugly scintillator crystal is near ground.

Overall, for slowish counting applications, floating the PMT anode is a win.

Cheers

Phil Hobbs

(*) That approach is also a good match to the (IME badly overrated) Cockroft_Walton scheme, where you run each dynode off a tap of a C-W multiplier. The C-W approach naturally gives you vaguely equally-spaced taps, and the high-Z end of the C_W matches the high-Z end of the PMT. It's superficially neat, but has serious problems with linearity and noise (especially at lower gain), not to mention the super ugly supply ripple effects.

You can use black chrome, though, no?

Yup. The microchannel plate is basically a massively-parallel version of the Channeltron electron multiplier.

I was looking at Photonis's MCP page the other day, and they were all smug about nobody using curved or chevron-shaped channels anymore, just their original version.

The reason for the curved channels, of course, was to force ions to hit the walls instead of being accelerated directly at the photocathode, damaging it and causing very very bright ion events.

Cheers

Phil Hobbs

Right, as far as that goes.

The photocathode is the high-Z end of the photomultiplier, and so most naturally goes furthest from ground. In free-space applications, we run the anode near ground and the photocathode at -1 to -2 kV, because the PC end doesn't draw any cuarrent to speak of.

That makes it easy to get the (largish) anode current out into normal low-voltage circuitry. (*)

The large scintillator crystal is out there where you can touch it, so for safety reasons it's good if it's kept near ground.

It's important for sensitivity reasons that the scintillator crystal be connected to the PMT faceplate through a continuous path with refractive index at least as large as the glass of the PMT envelope.

The reason is that a major fraction of the light from the scintillator is incident on its exit facet from angles beyond the critical angle for CsI -> air.

I suspect that this is where the difficulty lies: If you have diffuse light in a medium of refractive index n, and want to couple the light efficiently into air, it turns out that your maximum efficiency is

Sooooo, to avoid losing signal, you want to avoid coupling light from your scintillator into air.

The most straightforward way to avoid that is to optically-couple the scintillator to the PMT faceplate, either by direct contact or using some sort of index oil or gel. OK so far?

It's also good if your detector survives use, though. Putting the scintillator at ground and the photocathode at -1 kV

|doesn't work well, it turns out.

Photocathode corrosion due to halide ions drifting through the glass faceplate has been a real issue for a long time, and | I don't know whether improved faceplate materials have helped that much.

Power supply ripple is an easily-soluble problem, especially since a scintillation event is very bright in PMT terms--you aren't trying to do photon counting.

I've used cascades of PNP cap multipliers to do some pretty fun things with PMTs, including getting (to me) impressive linearity for analogue RF modulated detection.

Usually it's important for safety reasons that the metal housing for the big ugly scintillator crystal is near ground.

Overall, for slowish counting applications, floating the PMT anode is a win.

Cheers

Phil Hobbs

(*) That approach is also a good match to the (IME badly overrated) Cockroft_Walton scheme, where you run each dynode off a tap of a C-W multiplier. The C-W approach naturally gives you vaguely equally-spaced taps, and the high-Z end of the C_W matches the high-Z end of the PMT. It's superficially neat, but has serious problems with linearity and noise (especially at lower gain), not to mention the super ugly supply ripple effects.

We had the opposite problem: detecting the light from a plasma but blocking the ions that would darken our window. I wanted to do a thing with angled slats to bounce the light forward but that the ions would impact, but the semi company stuck with a plate with tiny holes, to let a little light and a few ions through.

Bad idea, but then they stole our detector circuit so I hope their windows all darken.

There is the idea of an active capacitor for an RC lowpass filter, nanely a modest-sized cap whose low end is intelligently driven by an amp, so it looks like a much bigger cap on the top end.

I should add, of course, that having a huge E field inside the glass is bound to reduce the photocathode quantum yield despite the shielding effect of the PC itself.

Cheers

Phil Hobbs

Yup. Back in the long ago (1984ish), I built a power supply filter that got rid of 10 Vpp of 120 Hz ripple on a 2 kV supply for some piezo stacks. It worked by lifting the cold end, sensing the hot end via a

With an LF356 op amp, I got 100 dB of ripple rejection, which came in pretty handy. I could probably have done considerably better with that feedforward trick of Woodward's, where you use another op amp to measure the error voltage of the main one, and add in its output.

Another approach like that is the Kanner Kap, which wasn't first invented by Kanner, but he now owns it on account of the cute name. That's the trick where you use an RC lowpass, and drive the cold end of the cap with some super beefy amplifier to make the top stay still.

Cheers

Phil Hobbs

Yep, I think it's clear enough that allowing the faceplate to act as an HV dielectric is a Bad Idea. But it also seems easier to avoid doing that by tying the exterior frame to the cathode as the manufacturer has done here, than to swap the polarity and deal with the HV at the other end where the signal is extracted.

As far as the need to insulate the crystal is concerned, you'll need to do that anyway to avoid exposing the user to thallium, I'd imagine. :-P

The question of whether it makes a difference in terms of ripple is worth thinking through a little further. I see whit3rd's point, but the effect of stray C seems worth considering before concluding that the polarity doesn't make a difference.

-- john, KE5FX