The Analog Realm

The next step in getting the experiment built out is to get an amplifier for the photomultiplier tube working. As a mostly-digital engineer, I always feel a little bit of angst when I have to venture into the analog realm. High-speed analog design, although it follows mathematical principles, is still something of a black art, and its competent practitioners are worth their weight in gold. Starting salaries of $100-150k per year are typical.

I am not one of those wizards. Most of my analog knowledge was picked up through hard experience as an IC designer. I’ve had a few minor triumphs, such as when my team’s physical implementation of an ARM high-speed memory interface had less skew and better signal integrity (bigger “eyes”) than ARM’s reference design test chip. I know how to remove the physical obstacles from circuits achieving their full potential. But that’s not the same as being able to design those circuits in the first place.

So of course, the prudent thing to do would be to ask experts, read data sheets, watch YouTube videos, and generally try to educate myself on how to solve this particular task. I did all that. The problem is that I kept getting different answers. Every source said something slightly (or vastly!) different.

This is frustrating. We’ve been building amplifiers for PMTs for nearly a century now (the PMT was invented in 1930 and became widespread by a decade later). You’d think that one or a few architectures would have become standardized by now, and that everyone would just look them up in a textbook and build that. But instead, every engineer I talked to sketched a different circuit architecture and implied that the problem was so simple that I should be able to just design it myself.

Some things I know:

• The output of the PMT is a current pulse, not a voltage pulse.
• Therefore, the first stage of the amplifier needs to be a “transimpedance amplifier” that takes a current pulse in and transforms it to a voltage pulse out. This is the hard part.
• There probably should be a second stage that amplifies the voltage pulse further. This is a fairly routine amplifier and there are lots of sources explaining how to design it, so it’s easier.
• Both amplifiers need to be “high-speed”. The core op-amp has a property called “gain-bandwidth product”, which is a frequency. For example, let’s say the GBW of the op amp is 10 MHz. A unity-gain amplifier (no amplification, just reproducing or buffering the input signal) built from that could handle frequencies up to 10 MHz (gain 1 x bandwidth 10 MHz = GBW 10 MHz). But if you want it to amplify the signal by a factor of 10, then it can only handle up to 1 MHz. Amplify by 100, and the max frequency drops to 0.1 MHz (= 100 kHz). Doing 10 nS high, 10 nS low square waves gives a period of 20 nS and a frequency of 50 MHz. Only the very fastest amplifier chips can handle that kind of speed. And I probably need to keep the gain fairly low, like maybe 2x or 5x per stage.
• There might need to be a “pulse shaper” to make sure that the final pulse is digital and at least 10 nS long. This is a separate problem and can be dealt with later. There might also need to be a “level shifter” to ensure that the final pulse being fed into the TDC7201 timing chip is 0-3.3 V. I already have chips for this and it’s easy.

I have been nibbling at the edges of this problem for a couple of years now, without tackling it head-on. But it’s time now. Ideally, I need to have this nailed within a month or two, so I have time to write it up for the next round of experiment proposals for PSI.

One approach would be to just buy a piece of instrumentation that was designed to deal with this. New ones are quite pricey (thousands of dollars), but used ones can be found on Ebay and elsewhere. These may be standalone, or rack modules for the CAMAC (a 1972 standard) or older NIM (a 1968 standard) racks. Unfortunately, these all have issues with power. Standalone instruments usually need to be plugged into a wall socket, as do the racks for CAMAC and NIM modules. The modules themselves often require multiple voltages. For example, the NIM standard includes ±300V, ±24V, ±12V, and ±6V, and any module is allowed to use any or all of those. If a module uses half of those, then it takes at least 4 different batteries to power it.

So, I will attempt to design my own, using no more than 2 batteries. I ordered a couple of OPA3S2859RTWEVM development kits from Texas Instruments, and they arrived today. So tomorrow, I set sail on the analog ocean and try to steer to the dry land of something that works.