Test Equipment Anonymous

This is going to be a rather short thread because the Keithley 610C that I bought for £10 five days ago works! The reason for the delay in testing was that I the 610C's input connector is a UHF socket and I didn't have a BNC to UHF adapter until half an hour ago. With the adapter, I was able to connect the 610C to a Keithley 6221 current source (with triax to BNC adapter at 6221 end) and go down in decade steps from 10mA to 10pA and the 610C gave the right answer each time.

But I spent a great deal of time analysing the circuit in LTspice with a view to replacing the (presumed dead) input MOSFETs. On the surface, the electrometer is a standard op-amp made of discrete components. But true to form, Joe Keithley drew the circuit in a way that makes it difficult to understand. First off, the circuitry around the input MOSFETs is upside down.

There are three zero controls. The coarse zero switch (rear panel) applies the positive rail via the wiper to a string of ten 249R resistors at the bottom of the diagram. Similarly, the medium zero applies the negative rail via the wiper to a string of eight 806R resistors with a 68k at each end, effectively giving a fine range at the middle where the wiper is active. The fine zero (multi-turn pot concentric with medium zero switch) is misleadingly drawn above the second stage of emitter followers, but is actually in parallel with the medium zero, and although it applies the negative rail at the top of the drawing, it would work equally well if its wiper were connected to the positive rail.

The two MOSFETs live on a sub-board, together with four other devices that are probably JFETs wired as picoamp diodes. The picoamp diodes are probably connected across the gate-source and drain-source terminals of each MOSFET as reverse biassed diodes to protect the MOSFET. They are not shown on the diagram. I believe that the MOSFETs are P-channel enhancement devices (the diagram at the Tektronix website wasn't scanned very well). That seems very odd, but it would make (a little) sense to draw their drains at the top of the diagram, feeding the Q103 and Q104 emitter followers. I am not going to tempt fate by poking a meter around that board.

I had assumed that the MOSFETs were toast, so I spent some time fiddling with the circuit to see how it might be persuaded to work with devices that are available. I have a hundred DN2530N2 which are 100mA 300V N-channel depletion MOSFETs. Not terribly similar, but available. More to the point, I've measured their gate leakage current at VDS = 180V and it was 2fA at at 20C, making them suitable. LND150 is a MOSFET that straddles enhancement and depletion mode and might also be a possibility. I will have a little more of a go at simulation and if I get something that looks viable, I'll post it.

The Tektronix diagram shows capacitors of a not-quite-readable value connected across the base-emitter junctions of the high voltage (+/-120V) output stage. They are not in my 610C.

The diagram shows a mains transformer with an electrostatic screen totally enclosing the secondaries and another to the primary. I have a 610C transformer loose and I measured its capacitances. CPS was 1pF, which is a tenth of what one one would expect with a conventional E/S screen, but not as good as a genuine all-enclosing secondary screen which would give about 0.1pF. If the mains transformer has failed, you're going to need to wind a replacement yourself, complete with E/S screens. Off-the-shelf parts will not do.

I'm glad it worked out. Buying something like that sold as spares or repairs, with a reputation for having an unobtainium part fail, is always going to be a gamble.

Further investigation says that its input leakage current is of the order of 10fA, compared to the 3fA of a 6517B. That's almost certainly down to the construction/switching, not its MOSFETs. It zeroes happily in NORMAL mode (where it measures voltage drop across a shunt resistor), but a lot less happily in FAST mode (where the inverting op-amp drives input current through its feedback current conversion resistor). It's really not fair to complain about that in an instrument designed in 1968. Tomorrow, I shall investigate the monitoring output; that's important when using an electrometer...

The 610C now has a BNC input socket. I found that a bulkhead BNC-BNC barrel exactly fitted the UHF connector's large D-shaped hole, so I popped a barrel in the lathe and machined off the outer part of the BNC at one end, exposing the inner PTFE and socket. With a clean tool, I took a little of the PTFE off to expose the socket and give me something to solder to. A metal pencil case had a chassis mount BNC plug fitted, making a screened enclosure for the 610C's resistance measurements. It works just fine and I can now measure resistors >1G.

Marco Reps has done a video about the Keithley 610c.
He replaces that ancient connector with something new.

I haven't needed to tweak the pots inside, but he's absolutely right about screening when measuring high value resistors. I have an earthed tin that plugs onto the front of a standard HPAK meter. With it, I can measure a 100M resistor that is impossible with ordinary leads. I haven't yet photographed the metal pencil case attachment for the 610C.

I have just fitted a BNC on the back of the 610C in place of its strange original connector. This modification was a lot harder than the first as the hole (15.8mm) was far too large for any BNC. Also, it wasn't absolutely clear as to which wire was which, or whether one might float up to 120V under some tests. I machined a pair of shouldered washers out of Perspex and fitted a chassis mount BNC socket having a long thread. Getting it wrong the first time proved that the green wire needs to go BNC body and the clear PTFE-sleeved wire to the central pin. Like that, I could hang an oscilloscope across the meter whilst measuring a 45G resistor. The meter output is 3V FSD, so the 45G resistor produced -1.3507V at the monitoring output (the manual warns of inverted polarity).

More important is how much hum is in that direct voltage. 610C000.PNG is with the lid of the pencil case closed, 610C001.PNG is with the lid open. With the lid closed, we see 11mV_RMS of nasty hum, complete with some spikes. This is to be expected when there's a mains transformer within the electrometer, but it's 0.3% of FSD, so acceptable. Note that the hum is 50Hz, not 100Hz, indicating that it is magnetic leakage from the mains transformer, not power supply ripple.

With the test jig's lid open, we see far more hum. Yes, it probably has zero DC component, but if there's enough hum to clip the electrometer, it won't. I never operate an electrometer without an oscilloscope on the monitoring output to ensure that my measurement is acceptably hum-free.

I've just calibrated the 610C against a Keithley 6221 current source, which is a much newer instrument and made to a rather higher specification. Measurement was by connecting an HPAK 34470A 7 1/2 digit DMM across the 610C's x1 output, removing moving coil meter linearity errors and making the measurement essentially a range switch resistance investigation. The 610C manual warns that the large value resistors used for the lower ranges should be replaced periodically. But they're seriously expensive, so that's not going to happen. I've stuck a label on the 610C detailing the errors so that correction can be applied.

OK, so 4.5% error wasn't acceptable and I found I had some high value resistors. Sadly, I had to use two or three resistors in series to get the right value, so it's obvious someone has been in there. But the results are a lot better. As we used to say at the BBC, "Good enough for sport."

A success story. A few days diversion and a right result.