Solartron 7081 - Reference Circuit

[Ismsanmar]

Just adding a little more information to the connector debate, I have recently bought a couple of original cables on eBay, which are optional accessories, not the one that is suplied with the multimeter. Specifically, the 3 Terminal Input Lead 3193 and the 5 Terminal Input Lead 3183.

And yes, the insulator between pins is PTFE, as in the socket, which makes them a special order by Solartron and truly a unobtanium connector. Here I add a couple of photos of the original next to the copy I bought at Aliexpress. You can see that the pins are a little bit shorter in the Aliexpres connector.

[ap]

Question: Is the cable shield of these connected to the connector case?

[Ismsanmar]

Question: Is the cable shield of these connected to the connector case?

Question: Are you going to use the information to check if you are building correctly the cables that you sell on your shop? 

No, it's ok the question. Of course that it is not connected. First, the connector case is directly connected to ground earth. Second, on the user manual it is clearly indicated that the shield is connected to the guard pin.

I'm not going to take it apart beyond unscrewing the connector, so I cannot check if the cable has a shied on the main harness, or on every individual cable, or on both. And it's shrinktubed till the solder cups. One thing that I've noticed is that the pins are "oxidized" not only where there was contact with the socket side, but on the entire pin, to the solder cup, indicating that the brass has difused on the gold plating. That means another difference with the currently availabe lineup of connectors: they are direct gold plated over the brass, without nickel barrier.

Now the real question: the connector that you use on the cable that you sell it's the custom original one, or the one that Fischer offers to the public with PEEK and brass pin with gold over nickel plating?

Ps.: There is an errata on your shop page. There are two lines that say: "The cable is available as kit (-S-K or -B-K) and as finished cable (-S or -K)." That last "-K" should be "-B".

[ch_scr]

Thank you for pointing out this circuit. To get a better grip on it, I ended up simulating the core circuit (without all the "bells & whistles");
it seems to be quite the elegant improvement of the "bootstrap" zener ref circuit, yielding not only a driven high and low output but a "4 wire" one (Force/sense) as well!
The ability to neatly integrate the "rail splitter" for single supply is icing on the cake...

[Kleinstein]

The last circuit is not doing force / sense in a good way. There is way too much current for the sense lines and the shift of the center is from the force side and not the drive side. The current though the sense lines is reasonable constant but would should still copensate the current by addition corresponding current sources.

[ch_scr]

The "active" vs "passive" version has better stability of the bias current over temperature, and the output headroom / power dissipation stays "centered" with the "active" one as well.
There is also the question if the transistor pair current mirrors are a good idea in the first place. At least the current sources also ensure proper startup vs. the zener diode used in the original circuit.
Another option could be to use 3 NPN of a "monolithic quad" for the sink part, more easily servo'd by a e.g. TLV431 and use the additional output to mirror the current up, to get around the 2 additional opa's of the proposed "active" version.

[ch_scr]

Here's what I've arrived at. A few warts still remain, the tempco compensation is unsymmetric (as in the original, and not possible to adjust for overall linear tempco), the wideband output noise would profit from integrated output capacitance / fixed load of a few mA.
Also since the zener current is dominated by the "sense input bias current compensation" source now, there might be noise conversion through the zener impedance. Measured 2.2uV std Dev @1000 samples.
On the upside it works from 15V for 10V out, and I've made an excel sheet to work out all the "moving parts". The latter did work an I've hit 10.000x volt without trim.
I've measured the output impedance, it did droop 5uV from 0 to -5mA, then returned to the initial value at -10mA. No "thermal effect" on the droop was observed, likely courtesy of the external pass transistors. (I've wired up an AD588 for +-5V in a similar 4 wire scheme, but without external transistors - it did droop 10uV at 7mA, half of that comes&goes with a "thermal delay")
I've put the "6.3 to 10V" network (and the centering one as well, but of lesser importance in the "single supply" configuration) on little pcb's. Thinking it might isolate them from board stress & thermals. Definitely easier to build/measure by itself like this. Did go the "measure individually and script optimize" route on the networks, managed to hit the ratios within a few ppm. Measurement error on the resistors seems to dominate, not shift from soldering as initially expected. Have a script read them from a Keithley 2100 with 5 reading average, that avg was a notable improvement.
All in all pretty happy with it. Since the B341 transistor array isn't exactly common I don't think it's too usefull to share the files, but would be happy to on request.
I've used an "individual quad" array for the Wilson current mirror (think it's appropriate there), but didn't find a in-production "monolithic quad" with reasonable price to replace the B341/CA3045...