Cooling a photomultiplier tube


A photomultiplier tube is a device used to measure very low levels of light. It’s a common tool of particle physics when trying to detect just a few photons. It turns out that running a tube at room temperature will not provide the best results. To improve the accuracy and sensitivity of his equipment [David Prutchi] built this thermoelectric photomultiplier tube cooling rig.

You can’t actually see the tube in this image but it looks similar to a vacuum tube or Nixie tube. The difference being that the components inside the glass dome make up the detector instead of an amplifier or filament display. To make a physical interface with the glass [David] wrapped it in magnetic shielding and finished with a layer of aluminum foil tape. This cylinder was then snugly fit inside of an aluminum heat sync. two Peltier coolers were attached to the outside of the heat sync, using Arctic Silver thermal compound to help transmit heat. A thermocouple was also added to monitor the temperature of this first stage of cooling. All of this fits into an aluminum enclosure which was filled with expanding spray foam before having a trio of fan-cooled heat syncs attached to it.

25 thoughts on “Cooling a photomultiplier tube

  1. That guy always has an excellent writeup (more HaD authors should read and mimic his writing methods).

    Why people still mess with Peltier coolers is beyond me, theoretical Delta T is just that, theoretical and actual results are almost always dismal, and not high enough to provide any true amount of cooling.

    Just put your mux setup in a bath of LN2 like the astro guys do for their CCD detectors. Cheap, self regulating temp, easy to maintain and manage, and a Delta T in the hundreds not tens of degrees.

    1. With LN2 your options are to replenish the bath as it warms, which isn’t practical in many situations, or to recompress the gas which is going to be very inefficient to support such a small volume. Anyway, it sounds like the project was to see if it could be done, and not necessarily to make a practical solution.

      1. Bulk LN2 is pretty inexpensive, so I’m not sure why you say it would be impractical. I’d have to check, but I think my lab pays around ten cents per liter. The home hobbyist is obviously going to pay several times that, which is still very do-able. I can’t imagine a PMT in a vacuum dewar going through more than a liter or two during the night (the heat load of the PMT should be pretty small unless it is broken).

        To put this in perspective, the cryo traps in my largish vacuum system use less than 70 liters per week, running 8-12 hours per day (ie, ~2 liters per hour). (I’m running a molecular beam experiment so there’s a *huge* gas load on the traps)

    2. A lot of astro guys just use a peltier with a heat sink too. The liquid nitrogen idea is while pretty easy and straight forward is wasteful.
      Another option would be to hookup a refridgeration unit from like a window A/C or water cooler or something. Would need to do some pipe soldering and get it re-charged (could use the ole standby DIY’er recharging gas Propane) If he needed a continuous chill colder than what’s possible with the peltiers then this may be a pretty good option.

      1. Keep in mind the astro guys using peltiers are working outdoors at night, so the Delta T, even inefficient Delta T’s get to a lower final temp due to the lower starting (i.e. ambient) temp.

    3. Just because there are more effective alternatives doesn’t make Peltier effect coolers useless. TECs are cheap, reliable and easy to design into a system.
      If one have unlimited resources the best solution is a multi-level system with a cryostat however if one wants to inexpensively cool a device down to (at most) -40 degrees Celsius in a place where energy is easily available a TEC or a TEC cascade is the obvious solution.

      1. Peltiers are stackable if you want really low temperatures. It’s just that the heat load cannot be very high because each peltier in a series must be able to sink all the power from the previous stage while maintaining the delta.

        As a rule of thumb, the CoP of a peltier is 0.1 for the maximum dT and 1 for zero dT which means that at full delta it’s pumping 10% of the nominal power while operating at the nominal power.

        You can achieve high coefficients of power by undervolting a peltier, up to 4-5 times the input power if your dT is small, but even at 40-50 degrees dT it’s still going to improve the pumping efficiency from 0.1 to 0.5

    4. I take those cheap little Peltier coolers, just a bit larger than a 6-pack of 12oz cans, and improve their cooling capability.

      It’s a fairly easy process. Take the heat sinks off and ensure the faces against the Peltier are smooth and flat. The machining on some is very bad, had one that looked like it’d been cut with a chisel and hammer. Next step is a glass bead cabinet to blast all surfaces of the heat sinks except what goes against the Peltier.

      On the ones where the cold side is a metal tub, the problem is the spot where the Peltier touches is rarely flat. I use a hammer and a steel block to tap it flat then use a piece of aluminum sheet (bead blasted on one side) with heat sink compound between it and the thin tub wall. That reinforcement ensures the metal is held tightly to the Peltier.

      Another modification is replacing the noisy brush type motor and inefficient radial fan with a brushless DC computer fan to push more air over the hot side. The motor is often poked into a hole in the insulation, which leaves a spot for heat transfer through the tub. Since the replacement fan is mounted to the cover (which has the vents carefully enlarged for more airflow) that hole gets filled with a styrofoam plug.

      Assembled with just the right amount of heat sink compound and properly sealed around the edges (moisture from condensation damages Peltier elements) one of these little coolers can keep a carton of ice cream frozen.

      The same sort of performance improvements can be applied to the bigger coolers with a finned aluminum sink on the cold side.

      I also do the flattening and bead blasting on heat sinks for CPUs, motherboard chipsets, video cards etc. The higher end CPU coolers don’t need their bases flattened but bead blasting off the colored anodizing increases the surface area slightly. Dull surfaces radiate better than shiny surfaces.

      Most any extruded aluminum heat sink which hasn’t had its base machined is not going to be flat. One can either use a milling machine and a really slow feed rate or use a flat file and keep after it until all the lumps and dips are smoothed out.

      What would be an interesting test it to measure the performance of various CPU coolers on the same board, out of the box VS glass bead blasted, to see if it makes any difference in heat transfer.

      1. You should write up your ice cream cooler method – I’m sure many people besides just me would be interested in making one of those.

        //Mmmmm ice cream…..//

    5. The CCD for for my telescope uses peltiers and it gets down to -30c. At that temp there is virtually no noise for any normal exposure. You can cascade the peltiers to get higher delta-t.

    6. Wow LN2 for astrophotography? You really have no clue do you? The only people who use LN2 for astrophotography are NASA, or should I say were NASA since they use Liquid helium to cool space telescopes now.

      LN2 is a bad idea not only because of the ability to work with it but because of the lack of a stable control method. This isn’t a problem out in space where temperature is constant so it works quite well there. But EVERY astrocamera from your cheap little QHY8s to your ludicrously expensive 16bit behemoths use Peltiers to cool the sensor. This all for one simple reason, control. (Well I lie slightly the cheap cameras don’t give you control and run the peltier flat out). The cameras need a way to set a pre-defined temperature to bring the CCD noise to a repeatable characteristic level. This allows you to do dark and bias frame subtraction without having to waste an hour of precious imaging time actually doing dark frames at the correct temperature.

      Say what you want but peltiers not only have their place but they are damn bloody useful. The only problem is people who use them thinking they provide *extra* cooling when their heatsinks aren’t even able to get their setup close to ambient temperatures. Most properly designed TECs have no problem bringing the critical equipment some 30-40deg below ambient.

      1. Yes, I’m just a clueless old fop.

        Luckily you’re here to set us all straight. You should be a consultant, and tell all these people your fine theories about LN2 cooling before they waste all their time and money selling/using LN2 for CCD Astronomy work.

        As to control, LN2 has PERFECT control, last I checked it has a consistent boiling point pretty much everywhere.

        1. “As to control, LN2 has PERFECT control, last I checked it has a consistent boiling point pretty much everywhere.”

          A liquid’s boiling point is determined by it’s vapor pressure and atmospheric pressure. Atmospheric pressure will not only vary per day but with elevation thus the boiling point is not something you can assume to be constant.

          Granted the resulting change in boiling point for liquid nitrogen will only vary maybe a degree Celsius (-195 C vs -196 C for a hasty search of record barometric pressures) which is probably good enough for most uses, I just don’t others to assume boiling point is constant as you nearly implied.

  2. Just for clarification, lowering the temperature is primarily done to lower the “dark count” of the PMT. This is essentially the number of events detected when, in fact, the tube is in a perfectly dark (in whatever part of the spectrum it operates) environment.

    When you are attempting to count a number of events exactly (i.e. photon counting), the number of false events is critical, and lowering the temperature dramatically helps. I do not know how this impacts energy resolution, but many applications don’t require this to be as accurate as count rate.

    The article mentions this, but the description here is slightly vague…

    This is commonly used to do geological formation evaluation; gamma radiation that interacts with a scintillator crystal emits light, which is detected by a PMT, and counted to determine the composition (hydrocarbon concentration primarily) of the surrounding formation.

  3. i might be wrong but isn’t a PMT actually also a kind of “amplifier” one photon creates a electron cascade wish results in an measurable voltage . one photon=>many electrons.

  4. Hehe The peltier won!! Good counter write up guys!. But yes you need to make a good work to get it cool!. If you do it right you will have a bad time :P So relax and play with water cooling and tecĀ“s:)

  5. I love these guys! The father/daughter team tackles everything from basic optics to straight up quantum entanglement in a very easy to follow/replicate process! DIY Quantum mechanics it is.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s