The Photomultiplier

Using a 931-A Photomultiplier

Another project that I believed impractical for this course has proved possible: using a photomultiplier. The RCA 931-A nine-stage photomultiplier was an inexpensive but good detector of feeble light, costing about $11 in the 50's. Photomultipliers are still available, but are quite expensive because they must be very good for use in scintillation detectors or astronomy. Such tubes are well beyond our budget, and require elaborate mountings. When the 931-A came down in price recently to little more than $20, it suddenly became practical to consider experimenting with one. It's an ideal photomultipler for the experimenter.

The only unusual requirement is for a 1000 V power supply that can furnish a few mA. We have already made such a supply, for the oscilloscope, using an expensive and heavy power transformer. It seems that high-voltage, low-current transformers are not available these days, so we must cast about for an alternative. The availability of good silicon rectifier diodes and relatively cheap electrolytic capacitors suggests the use of a voltage multiplier, in particular a sextupler, which will supply 1000 V with an input of 120 V. This circuit has the great advantage that the capacitors and diodes are in series across the high voltage, so that high voltage ratings are not necessary. The sextupler was constructed on a 3-1/8 x 4-5/16 solderable perfboard (the solderable feature is not essential). It is fed by the 120 V secondary of a 1:1 transformer that isolates it from the grounded AC supply. I put the pilot light on the secondary, rather than the primary, but it makes no difference. If you use a 3-prong plug, and connect the positive end of the doubler, the chassis ground, the AC ground and the ground wire together, you will be safe enough, though the 1000 V supply will be hot with respect to ground. If your AC supply is wired improperly (not unknown with do-it-yourself electricians) you will simply blow a fuse when it is plugged in. If you use a transformer, the DC will not be connected to ground at any point, a desirable feature in this case. The transformer I used cost only $4.50, and is ideal for this application, but the supplier (All Electronics) is apparently running out of them.

It is necessary to say that if you are not scared by the prospect of a 1000 V supply with big capacitors, you should not attempt this project. The supply is positively lethal unless you avoid the slightest possibility of contact with it. This means that everything should be inside a metal box, and the wiring should be neat. I used a 6-1/4 x 7-1/4 x 2-3/4 two-piece box with an aluminum bottom and a black crackle finish steel top. The box should be grounded to one side of the secondary of the transformer, not to AC ground. There will be no particular problem from having it float, and if you connect it to anything, it will be happy with the local ground. Hands must be kept away when tests are being made with the box open, one hand definitely in the pocket. Of course, the resistor string on the photomultipler socket that serves as a bleeder is absolutely necessary for safety.

The 931-A should not be subjected to more than 1250 V overall, or 250 V between the ninth dynode and the collector, and the current should not exceed 1 mA. The recommended supply is 1000 V, divided equally between collector, dynodes and cathode. Our supply gives a little over 1000 V when energized with 120 V, so it is ideal. While testing, I bring it up with a Variac. No signs of distress have yet been noticed. Your DMM is probably good up to 1000 V, so you can measure the voltage easily. I keep my hands about a foot away and touch nothing when power is applied. Once you put the lid on, there is no danger. With these voltages, the gain is a remarkable 24 mA/μW, or 24 A/lm at the wavelength of maximum sensitivity, 400 nm. The photocathode has an S4 response, which extends from 300 nm to 700 nm, peaking at 400 nm. The photocathode is 5/16" wide by 15/16" high, centered 1-15/16" above the seating plane. It faces the key of the socket pin, and is located behind the collector grid, deep in the tube. The socket is on the top of the power supply box. I made a shield out of a length of paper towel roll with a cardboard top, cut out and taped down all around. It was sprayed matte black inside and out, and a small square hole was cut with an X-acto knife at the height of the middle of the cathode.

There are 9 dynodes at which secondary electron emission occurs. Between 4 and 5 secondary electrons appear for each electron striking a dynode, for an overall gain of 800,000 ( = 4.53⁹). The dark current (after the tube is allowed to quiet down) is equivalent to an input of 2.5 x 10^-9 lumens (60 nA), and the noise is equivalent to 9.5 x 10^-13 lumens. 1 mA photocurrent corresponds to an input of 42 μlm, or an illumination of 21 millifoot-candles! Obviously, we do not expose the 931-A to bright lights.

The 931-A uses an 11-pin small-shell submagnal base, which looks like an octal base but has more pins. A similar base is used for some 3PDT relays, so a relay socket can be used. This is a different, smaller 11-pin base than that used by the 2BP1 cathode-ray tube, and is equally difficult to find. The resistive voltage divider of 100k resistors is built up on the socket. I did not use the 0.1 μF capacitors shown in the circuit diagram, since they are not needed for DC tests, and the tube will work well without them anyway. There are only 3 connections from the multiplier to the power supply: the cathode (pin 11), collector (pin 10), and dynode 9 (pin 9). The hole for the socket was cut with a Greenlee chassis punch (it's the same diameter as for an octal socket) which made the job easy. Otherwise, cutting the hole in steel is a very annoying job, and I'd advise an all-aluminum box if you don't have a chassis punch.

Construction and testing proved to be a one-day job. Connect a DMM across the 100k output resistor to measure the DC photocurrent (I used binding posts for V_o and ground). Note that you are close to ground here: all the high voltage is at the other end of the tube. The dark current was practically zero, but I easily got 7 or 8 volts output with a flashlight pointed in the general direction, when the voltage across the photomultiplier was 600 or 700 V. You will find that the tube is extremely sensitive to light, unlike the usual phototube or phototransistor. Don't turn the power on when the tube is unshielded! It probably won't ruin the tube, since the current will be limited, but won't do it any good.

High-voltage power supplies have been mentioned here, and on the Cathode-Ray Tube page. A third alternative is shown at the right. This circuit can be put together very quickly if you happen to have an electromagnetic door buzzer available (this is one of the good things that are becoming unavailable). The transformer, a small 6.3 V, 300 mA device, is not insulated for high-voltage service, but seemed to survive well enough in this circuit. A high-voltage 0.47 μF capacitor was used as a filter, which discharges rapidly enough through a DMM that a bleeder is not necessary. For a practical circuit, the capacitance should perhaps be larger, and a 1M bleeder should be added. Remember that ordinary resistors can only take 250 V. Watch the input voltage carefully to make sure the output voltage does not rise too high; it is near the limit of the 1N4007 diode. Use two diodes in series if necessary. With a larger filter capacitor, the circuit becomes more dangerous, of course, and care should be taken with it. This circuit can give an appreciation for the size of inductive spikes. If the load is really low, such as a Geiger-Muller tube, the buzzer could be replaced by a pushbutton that is pressed enough times to build up a voltage. The diode can be replaced by a spark gap, since only voltages in the correct direction are large enough to break down the gap. The big pulses come when the primary circuit is opened and the magnetic field collapses, not when it is closed. A Ruhmkorff Coil operated in exactly this way. Of course, the buzzer is not meant for continuous operation.

Composed by J. B. Calvert
Created 23 March 2002
Last revised 27 April 2002