ESI DC Generator Detector 803A

[trobbins]

I have an ESI DC Generator Detector 803A on the bench, along with a companion 230B guarded bridge.  The 1966 manual that came with it is for the 231C measuring system, and specifically the 230B, however the manual includes the 801 model details.

The on-line short-form description of the 803A indicates a minimum detectable signal of 5uV or better with >500k loading.  That spec is not as good as the 801 (801B), with its 3uV range and 0.75uV sensitivity.  And the 803A generator provides up to a 200V limit, rather than 600V for the 801.

The 803A chassis indicates manufacture circa 1972, and has a reasonable number of changes to the 801 manual's description, although the general operation is the same.  The 803A detector now uses a shielded box DC Amplifier module pn. 13643, and the isolation transformer (Sartron 4365) looks a bit more commercial and tidier than the unit shown in manganin's first post in https://www.eevblog.com/forum/metrology/esi-242d-(sp3632)-resistance-bridge-for-sr104-intercomparison/.

Whilst I'm replaced a cracked front panel mains fuse, I thought it worthwhile asking if anyone has an 803A along with a manual/schematic for it.

Ciao, Tim

[trobbins]

Passed the first operational hurdle.  991.38 ohm on the 230B bridge measured 991.34 ohm 4W on my uncal Keithley K197 (but cross-comparison is within 0.01% confidence).  So I'll start to make some more comprehensive comparisons of the 230B bridge.

The 803A detector side was initially not able to be zeroed.  I firstly checked the 7Vdc power supply, and 14-pin connector contacts, and underneath the main rear cover to reform a 400uF 3V cap across the meter movement terminals, but no change.  Then the metal cover on the DC amp module was removed (not that simple) to inspect the circuitry, which is quite similar to the detector used in ESI's 300PVB.  I checked operation for parts sensitivity to temp and mechanical changes - nothing obvious except for the groupings of 2N2925's - and reflowed some of the pads.  The pcb included another 400uF 3V ecap, so I formed that to 3V - which seemed to recover reasonable operation on the higher ranges, but drift and fickle operation didn't provide much confidence with more sensitive bridge tuning.  So much simpler nulling was achieved using just an Aneng AN8009.  It could take some effort to remove the gremlins from the DC amp, as it could be degraded bjt's, or the chopper relay, so maybe not practical.

The generator section of the 803A appeared to operate ok.  The 803A doesn't include a pre-regulator on the AC feed to the power limit variac.  The unit is internally labelled for 220Vac mains, and with that input it generates up to 1.7V for the 2V range, 20V for the 20V range, and 193V for the 200V range (all no external load), so maybe it was meant for 240Vmains.  The large 15mF 10V Sprague cap reformed fine.

[Conrad Hoffman]

I have one of the ESI bridge systems with the supply/null detector stashed away in storage. Unfortunately the big resistance bridge box is single terminal, not 4-terminal. It's still incredibly accurate after all these years. About all I can offer is that I've had horrible luck with very low voltage caps, like 3V. They almost always go up in value, but with very high dissipation. I replace on sight!

[trobbins]

Yes, I was concerned about the 400uF 3V caps, especially after noting the comment by Ero-Shan in post #30305 https://www.eevblog.com/forum/testgear/test-equipment-anonymous-(tea)-group-therapy-thread/30300/.  Luckily the two such caps were easy to access and confirm low leakage ~1uA at rated voltage, and nominal capacitance. 

It may just be more practical to prepare a detector based on an LTC2054 AZ opamp, such as in the null detector thread on your original design (https://www.eevblog.com/forum/metrology/null-detector-based-on-conrad-hoffman-design-check/msg5480467/#msg5480467, as the powering is already available, assuming I can easily locate a 14-pin connector and can integrate the existing front panel zero and sensitivity pots.

I'm hoping to quickly confirm all the 230B ranges, from 1200 milliohm to 12Gohm FS, are nominally ok, and then to do some more exhaustive testing of known reference resistors that I have confidence in, to sort of confirm if likely accuracy of the 230B is still within 0.005%, and 0.001% for ratio arms.

[Conrad Hoffman]

Does the null detector use a chopper? I vaguely remember hearing it buzz in use.

[trobbins]

Yes the detectors in the 300PVB and 803A appear to use the same unmarked chopper, which is visible in the DC amp photo in post #2.  I guess it would be one of the risks in chasing a cause for poor detector operation, but I'm not sure if they 'degrade' as a failure mode, and if such degradation could be repaired such as with contact spray.  The metal enclosure of the chopper doesn't look too inviting to open up, although it appears to be somewhat easy to unsolder its shielded cable connections and slide out.

The chopper isn't noisy, but the special isolation transformer in the generator section certainly is when the variac is turned up, and I might have to see what can be done to tame that.

[trobbins]

To update information on this ESI resistance measuring system - it is the 232A system, with short-form description attached, along with a scan of the 230B Instruction manual.

My post #3 had incorrect accuracy specs, as this 230B has a general 0.01% accuracy.

I used the system to trim two Fluke 19k999 0.01% reference resistors with matched opposing tempco ( https://www.eevblog.com/forum/metrology/fluke-895a-19-999k-precision-pairs/msg4385353/#msg4385353 ) as a 1:1 divider.  I added 7ppm (0.144 ohm) to one side to get equal resistance measurements from the 20k ends to a centre tap, with a resistance resolution of about 0.5ppm at about 20C.  The 20k pair has nominal opposing 0.5ppm/C tempco markings, so I may repeat this at 23C, but for the moment it's fine.

I then used the Fluke pair as a 1:1 divider to check the 230B range 5 accuracy with its 100k:100k divider.  The 230B circuit is attached below, along with a description of the connections used.  The 230B internal link resistance from Unknown 1 to top of range divider is included in the 1 ohm.  The link from Unknown 2 to DET 1 was measured at 30 milliohm (with Dekastat at 0 setting), so the internal link resistance from DET 1 to bottom of range divider should be much less than 30 milliohm.  Matched, short, low resistance connections were used to the Fluke divider, and when making Fluke resistance measurements, so should be <1ppm consequence.  The resulting bridge null was away from balance by +10uV and -14uV, which was about half of the additional imbalance contributed by shorting the 7ppm padding resistance added to the Fluke pair, so it appears that the 230B range 5 configuration is within about 5ppm of balance, and able to support the overall 0.01% tolerance assuming the Dekastat is still good (which seems to be the situation).  And increasing the Fluke pair's enclosure temp by 4C showed <3ppm change in bridge balance, so that was also reassuring.

[Kleinstein]

Yes, I was concerned about the 400uF 3V caps, especially after noting the comment by Ero-Shan in post #30305 https://www.eevblog.com/forum/testgear/test-equipment-anonymous-(tea)-group-therapy-thread/30300/.  Luckily the two such caps were easy to access and confirm low leakage ~1uA at rated voltage, and nominal capacitance. 

It may just be more practical to prepare a detector based on an LTC2054 AZ opamp, such as in the null detector thread on your original design (https://www.eevblog.com/forum/metrology/null-detector-based-on-conrad-hoffman-design-check/msg5480467/#msg5480467, as the powering is already available, assuming I can easily locate a 14-pin connector and can integrate the existing front panel zero and sensitivity pots.

I'm hoping to quickly confirm all the 230B ranges, from 1200 milliohm to 12Gohm FS, are nominally ok, and then to do some more exhaustive testing of known reference resistors that I have confidence in, to sort of confirm if likely accuracy of the 230B is still within 0.005%, and 0.001% for ratio arms.

For the large resistor range, one may consider 2 separate null detectors. One for the lower resistors (e.g. up to some 100 K), that could use an OPA387 and one for the higher resistors, that could use the LTC2054 / max4238 or maybe a non AZ amplifier like OPA145. One still has to check the isolation in the supply and maybe consider a battery for the null detector and have it as a fully separate project.

[trobbins]

Ta. At the moment I am getting by with just an hand-held Aneng AN8009, with min reading of 5uV, but 1uV resolution, which has so far provided adequate capability for what I've needed to resolve.  I'll get through some more testing and see how I go - just building up to test a vintage Sullivan Volt Box ( https://www.eevblog.com/forum/metrology/hw-sullivan-volt-ratio-box-from-1964/msg4531616/#msg4531616 ).