Earlier this year I bought an Electrometer 139 from Princeton Applied Research in unknown condition. Of course it was broken, so I had a repair project to do. To make a long story short: I have it working. Hurray!

First of all: what is an electrometer? An electrometer is sort of a multimeter but with an insanely high input impedance, in he order of 10¹²Ω (1TΩ). It can measure voltage up to 199.9V, small currents, high resistance and (surprise) charge in coulomb (A×s).

Apart from a broken diode in one of the supply sections, there was only one HUGE problem: the input stage was faulty. It consisted of some sort of obscure double transistor in a single, six pin TO-5 metal can, labeled MD5 0177 117SC. No information on the net, no schematic, pure unobtainium. Thankfully, the members of the Vintage Test Equipment Facebook group pointed me to another, well documented, electrometer, the Keithley 616, which gave me some insight on the innards of such a piece of equipment. Unfortunately, soon it became clear to me that the input stage of the Keithley was not an exact copy of my contraption. After a few shots in the dark to replace the double transistor by two random new transistors, which yielded zero results, I decided to reverse engineer the whole device. That kept me busy during some long summer nights, weekends, weeks and months. 

I realized pretty soon that the double transistor was the input stage of an op-amp, totally built out of discrete components on a separate board (10 transistors, TIS97 and others). I took the board out of the device (which was a terrible chore, because one of the screws was totally unreachable and I had to drill a hole in a pcb to get a screwdriver to it) and tried to complement it with a pair of jfets, which didn't work. Then I tried with n-channel mosfets, which didn't work. Then I tried bipolar transistors, which worked sort of, but could of course never realize a high input impedance. Finally I tried with p-channel mosfets, which worked! Two humble, matched, BS250's did the job! Great, next step: I planned to  replace the BS250's with two p-channel mosfets in a single package, because that was also done in the original situation (probably better for thermal reasons). I had a dozen of APM4953 double p-mosfets. Found out that most of these were not working, or just fake mainland crap. There were only two good ones, but neither did the job. Don't know why. Fact is that the RDSon of the APM4953 is 0.053Ω while the BS250's is 3.5Ω. Normally smaller is better, but in this case not, apparently. So I decided to put the two BS250's in a piece of heat shrink and solder them in the device. 

Then there was another problem. The input connector is a triaxial connector, which superficially looks like a BNC connector, but it has not one, but two concentric shields. In order to minimize leakage current, the electrometer outputs the measured voltage back on the middle shield and thereby virtually stops any leakage current. I took the chassis part to my local electronics shop. The man working there could not give me a mating plug right away, but said he had "boxes full of stuff" behind the store and promised me he would search through them during quiet hours. After two weeks I went back to the shop and he had found half of the plug. After another two weeks I went back again and he had found the other half. Hail to Radio Twenthe in The Hague! He did not really know how much to ask for it, so he asked me what I was ready to pay and we agreed on that.

With a triaxial connector you also need a triaxial cable and these are truly hard to find! Yes, you can buy 1 km of said cable for a small fortune, but that is not what I intended to spend. What makes it even harder is that Triax is a brand of ordinary coax cable and accessories. And that many cables that are advertized as "triaxial", do have double shields, but those shields are not isolated from each other, so no use to me. I gave up more or less, but then on a Sunday I was searching through some boxes in the attic for a piece of cable for something else and I found two pieces of "Draka Coaxial SDH-Switch board cable" from my previous employer, which turned out to have two isolated shields! So there I went, trying to get the cable in the connector and trying to get all three conductors properly connected. But alas! After many many attempts, I couldn't get it to work. Either a short between two conductors or no connection at all. I had to give up. After a good night's sleep I replaced the triaxial connector on the electrometer with the red banana jack that I stole from its back side. The red banana jack was in turn replaced by a grey one of slightly different size that I stole from an audio generator of which all but one bananas had already broken off.

Now it works, at least in Shunt mode. There is also a Feedback mode, which still doesn't seem to work (Keithley has Normal and Fast, which is the same, I think). Happy with it working at least partly! I realize that with the BS250, the electrometer is never going to meet its original specs by far (max leakage current according to he datasheet 20nA. That's at VGS=15V, while in this application VGS is only 1.7V. In the 100V-setting, that would result in 10⁹Ω). I could have ordered an ALD1117 matched p-mosfet pair from Digikey with leakage resistance 10¹⁴Ω, but haven't come to that yet. I have published the schematics that I reverse-engineered on radiomuseum.org for the benefit of the community.

Charles