Repair of Solartron 7081 SN#718

[pigrew]

This thread will detail repair efforts of a Solartron 7081, specifically unit #719. I purchased it from eBay, and it was damaged in shipping. Prior to my acquisition, someone had opened it up and forgot to replace the board mounting screws. This allowed some of the PCBs to float loose in the enclosure, breaking plastic standoffs, the GPIB connector, and VFD (among other things). In addition to these issues, there are age-related faults (capacitors and optoisolators failing).
After initial repairs (replacing capacitors, reconnecting wires, and updating/reprogramming the firmware), the meter turned on and seems reasonable well calibrated (though very noise with the covers off). Major outstanding issues that must be corrected include bad optoisolators and an aged Schaffner power inlet module. I will continue using it headless for the time being.
My plan to create posts in this thread detailing the individual repairs I’m making, and provide links in this post.

In addition, I’ve started looking at the floating supply firmware using the xdasm disassembler (It looks like some work was previously done, but I can’t access it since I don’t have an IDA pro license). I’m posting my work on my github page. I’ll also add some other documents to github, like a table of the modern replacements of the various components I’ve replaced.

Work Completed:

• Replace board standoffs
• Replace all electrolytic capacitors (about 1/3 had failed)
• Change RS-232 to 9600 Baud
• Reattach range switch plunger (switch was damaged in transit)
• Refreshed NVRAM
• Reprogrammed EPROMs with updated version
• Repair GPIB cable
• Replace mains filter
• Replace all optoisolators (PLL is unstable, optoisolators are a red herring; mains line seems noisy, and not triggering well)
• Replace settings-storage battery (replaced with super-cap)
• Replace RTC battery and set date and time
• Replace the VFD
• Replace input terminals with cheaper model (Chinese LEMO clones)
• Apply extended autozero pre-delay firmware patch

Work to do:

• Characterize noise, to see if further repairs are needed
• Test GPIB
• Clean the analog boards? (Lots of rosin left especially unused testpoints, this can’t be good)
• Build cables using the LEMO-style connectors (low quality cables done)

Possible other things:

• Retrofit high impedance nodes with PTFE standoffs
• Thermal insulation around the Zener and thermometer

And for those really interested, there is a plaque on the back describing which modifications have been made. “A” has been scratched off, as have 1 through 10.

[pigrew]

(second post, in case I need it)

[tggzzz]

Replace all optoisolators (PLL is not consistently locking)

Try tweaking C807. Floating logic sheet 5, rear right corner lower board.

[pigrew]

Replace all optoisolators (PLL is not consistently locking)

Try tweaking C807. Floating logic sheet 5, rear right corner lower board.

Thanks for the suggestion. I tweaked it a few days ago. I changed the control voltage from 1.9V to 2.44V, and the PLL was still intermittently unlocked. I ordered the optocouplers, and they were delivered earlier today.

[coromonadalix]

on the eeproms i would put back a new protective tape ... some of them aren't protected at all

[pigrew]

on the eeproms i would put back a new protective tape ... some of them aren't protected at all

Yes, I was worried both about them being erased, bit-rot, and the adhesive foil shorting nets on the board. I pulled the ICs, erased them, flashed newer firmware, and Kapton-taped aluminum foil over the quartz windows.

My progress on replacing the capacitors over the weekend is another story (about 30% of the capacitors I'd consider to have failed). I had been using Kester 331 (organic) flux solder, and was getting the dreaded white residue when trying to clean it with isopropanol. Prior, I had noticed, but couldn't believe, that there was rosin flux left on the board. The cleaning difficulties are due to the mixture of fluxes... it's painful. I almost want to dump the boards into a gallon of heated isopropanol (and then DI water), but I'm worried about water getting into the switches and relays. For now, I've been tediously trying to clean the bottom of the boards. The rosin from the 80s couldn't be doing good things for leakage currents.

[pigrew]

I received my knock-off push-pull plugs (FGG.2B.305.CLAD99Z) from AliExpress, and realized that it's built for much thicker cable (9.2 to 9.9 mm). The sales listing didn't specify which collet it would come with and its photo showed a collet for thinner cable. I'm going to try to buy a thinner collet.

Speaking of which.... what cable to use? My most recent poor-man's idea to to use SuperSpeed USB cable:

Signal	Wire	connector 2B.305 pin
Lo	Black	3
Hi	Red	4
Guard	Drain+shield, isolated from connector housing	1
Lo-sense,High-sense	shielded twisted pair, drain to lo-sense	2,5
(floating)	shielded twisted pair + unshielded pair

My thought is to put high and low next to each other to try to have them have a more similar temperature.

I know the wire insulation may have issues with dielectric absorption (and leakage), but I think that the shielding should be reasonable, and the 20 or 22 AWG red and black should be good for current-carrying. Use of copper wire (in physically small connectors) should reduce thermal voltages.

The LEMO connectors have a "test voltage" of 1.60 kVrms body-to-shell, and suggest a max operating voltage of 1/3 or 583 Vrms. Multiply by sqrt(2) to get a 816 VDC (or so) limit.

The proper Fischer connector has a contact to body test voltage of 2.4 kVDC. Dividing by three, we get 800 V, so about the same.

However, knowing that USB cable is probably specified at max 500 V, I should keep it below that, though it'd likely work at much higher.

(Capacitors and optocouplers have been replaced, boards partially cleaned, will fully clean them once I know everything is working, still waiting for the new input sockets. Also, I noticed that in TEST10 mode, a relay switches for every measurement. I'm guessing that's normal.).

[Mickle T.]

Also, I noticed that in TEST10 mode, a relay switches for every measurement. I'm guessing that's normal.).

Yes, it's normal.

[pigrew]

I was able to replace the optocouplers, but it did not help the clock stability. The 10 Hz measured on PCB5/PL501/32, which is the divided down PLL clock, still oscillates between 9.99745 and 10.00261 Hz (or so). The 60 Hz signal PCB/PL501/30 has about 40 us of cycle-to-cycle variation.

I believe that the mains is just too dirty with lots of switching converters (though it's still possible that PCB5/IC834, a LS132 is weak/failing, I've not probed it directly yet.

However, I think that I'm going to just swap in a crystal oscillator, as others have done. Since the 5.24288 is not a common frequency, my plan is to inject 10 Hz or 50 Hz into where the line frequency normally goes. This is due to the "demodulator drive" signal being used directly from the line zero-cross detector. I don't know why they wouldn't just use the divided down clock to drive the chopper.

My option "1" uses a 10 Hz reference to drive the PLL, and my option "2" uses a 50/60 Hz reference, and also overrides the

Another idea would be to try to filter the input AC signal, perhaps by putting a 1 or 2 uF capacitor in parallel with R803.

What does the "PU3" mean on the schematic? Pull-up? Test pad? 

[pigrew]

Based on stories online, I'm always afraid of vintage Schaffner and Rifa line filters. I ordered a new Schaffner to replace the 1985 "FN 372-2/21". I ordered a "FN 372-2/31" from Arrow (free 2-day shipping from the Netherlands, WTF, how do they make money?). USA retailers don't commonly stock the models with 5mm fuses. This "/31" model code is not listed on the Schaffner datasheet. It ended up being a version where the voltage selection wheel selects between 115, 115, 130, or 130 Vac. It also came with quick-connect jumper wires between the IEC input and the switch contacts. Otherwise, it is functionally equivalent. I installed the old wheel in the new filter. A little bit of soldering was required to disconnect the ground wire from the PCB. I replaced the wire (it was too short after cutting it off of the old filter).

After replacement, I tore up the old filter. It seems to contain poly-film capacitors, so perhaps it wouldn't have failed anytime soon.... I'm not sure.

In general, Schaffner seems to use larger inductances in their lower-current filters (providing better filtering), so I suggest not "upgrading" to a higher-current model.

[pigrew]

The "Brain" PCB is specified to have a NiCd battery that powers SRAM which stores calibration data and configuration (and maybe logged data?) (The manual states that the calibration EEPROM should be used an infrequently as possible, so they cache the calibration in the brain's SRAM). The circuit trickle charges the battery (10 mA, reduced current at higher voltages). Over the years, the NiCd battery leaked acid and damaged the PCB. The original battery (3.6V, 110mAh, Varta) is no longer manufactured. NiMH batteries should not be substituted since they are not compatible with trickle charging (according to Panasonic's documentation).

So, I decided to substitute the battery with a EDLC. I choose the AVX SCMT32F755MRBA0. It is a 7.5 F, 5.5 V unbalanced capacitor (unbalanced implies lower leakage current). It stores less energy than the NiCd battery, but should have a longer service life. AVX reps told me that there shouldn't be any special considerations I need to make in order to use the unbalanced device (other than staying within the voltage rating).

After installing the EDLC, I turned on the meter to charge it. The battery charged up to about 4.48 V, but quickly dropped to 4.34 V after turning off the meter. 11 days later, the battery had discharged to 3.74 V, still high enough to retain the SRAM data. I don't know how long the NiCd battery would retain data, but I'm happy with this modification. A drill hole provided a convenient spot for zip-tieing the capacitor.

Next on the agenda is to replace the input connectors, and maybe continue debugging the clock stability issue.

[RobK_NL]

What does the "PU3" mean on the schematic?

Those "PU" thingies designate connections to IC825.

[borghese]

My option "1" uses a 10 Hz reference to drive the PLL, and my option "2" uses a 50/60 Hz reference, and also overrides the

Have you got any improvements using external 50Hz?
Out of topic, do you have a drawing of the calibration key or maybe one to sell?

[tggzzz]

My option "1" uses a 10 Hz reference to drive the PLL, and my option "2" uses a 50/60 Hz reference, and also overrides the

Have you got any improvements using external 50Hz?
Out of topic, do you have a drawing of the calibration key or maybe one to sell?

I wouldn't expect improvement with an external 50Hz, but I would expect improvement with something tracking the mains frequency (i.e. nearly but not exactly 50Hz).

The user https://www.ebay.co.uk/usr/bielli-int sometimes sells cal keys. I have one, it turns in the lock, but I haven't used it in anger

[pigrew]

My option "1" uses a 10 Hz reference to drive the PLL, and my option "2" uses a 50/60 Hz reference, and also overrides the

Have you got any improvements using external 50Hz?

I've not yet gotten a chance to try external frequency references (and my meter isn't yet stable enough to even see the difference. It's sitting disassembled on a table.).

The PLL output oscillates low and high by a little bit over a period of about a minute, similar to the integration period for 8 1/2 digit mode. I am worried that if the integration "glugs" are slightly longer during part of the conversion and shorter during other parts (in a non-predictable way), the result will be effected. I think it's a balance between removing the 60 Hz effects, and stability of the frequency. My understanding is that most meters just measure the 50/60/400 Hz once upon startup, and then synthesize the 50/60/400 Hz thereafter (instead of using a PLL to track the mains frequency).

My question about the two options is that I don't know if the "demodulator drive" is important to be accurate. I saw it's used to drive the chopper amp at the line frequency, so it shouldn't need to be super accurate? I saw others had directly injected a (roughly) 10.48 or 5.24 MHz clock (the exact frequency XO is not an off-the-shelf component), but it seems easier to just inject 50 or 60 Hz, and let the existing PLL multiply it, accepting a little extra phase noise. This assumes that my measured frequency oscillation is due to the mains zero-crossing circuit and not the PLL itself.

I wouldn't expect improvement with an external 50Hz, but I would expect improvement with something tracking the mains frequency (i.e. nearly but not exactly 50Hz).

It's a balance of the stability of the forcing frequency versus line frequency rejection. My gut feeling is that it's better to have a stable frequency, and accept a small offset from the mains frequency.

Out of topic, do you have a drawing of the calibration key or maybe one to sell?

I do not have a calibration key. I've been running my meter with the front panel removed (and the calibration key disconnected). I found the user bielli-int on eBay posted a photo of a key. I'll attach it here. I also read that you can by the whole lock assembly for $20 or so (IIRC), and it'll come with a key. Search for "(honeywell,aml) 20 (key,lock)" on eBay. I believe that the proper keys are labeled "601" (based on other eevBlog posts).
 

[pigrew]

The next step of my repair is to replace the input connectors, so that I can measure external signals.

I received the LEMO-style sockets, and have replaced the rear connector. I used a dremel-tool to cut an extra flat in the socket, so that the red mark is in the "up" position.

I still have the covers off the unit, so I won't post any exact numbers/noise results. Measured resistances are fairly accurate. Using an old crappy 1.3m long coaxcable, I measured about 300 MOhm input resistance in kohm mode. In true resistance mode, it was hitting the limits.

Unfortunately, the measured voltages seemed way off, about 3000 ppm low. This could be due to dirty PCBs (I need to give them a bath, first desoldering the front/rear input switch, I think everything else is water-safe). Another possibility is aged/damaged components, which means I may need to just recalibrate it. I saw one person reporting a leaky integration capacitor, it may be worth investigating that, too. I think this would mean that the measured voltage would be a function of the integration period selected? Or is there a better way to test the capacitor?

Oh... one other thing... My meter came with a front-panel of a binding-post breakout box. I may try to re-use it. I wish it had come with the box itself... Cry. Needing a fuse is weird; maybe it was a current-shunt box?

[Kosmic]

Oh... one other thing... My meter came with a front-panel of a binding-post breakout box. I may try to re-use it. I wish it had come with the box itself... Cry. Needing a fuse is weird; maybe it was a current-shunt box?

Look like the backpanel of a solartron 7150.

https://holzleitner.com/el/solartron-7150/filter/_dsc1585a.jpg

[Kleinstein]

The voltage reading low by something like 3000 ppm could be due to a problem with the reference: the reference current has an odd fine adjustment via DAC and onto-coupler. The onto-couplers are known to sometimes fail and failing OC could shift the reference a little. So it might be worth checking the OCs (read the logic levels on both sides - they are supposed to be constant anyway). Because of the constant settings it is rather odd to have the OCs here - just jumpers would have done the job as well.

It would need quite some dirt to cause a reading that is 3000 ppm off.

Slight variations in the PLL frequency are not that critical, as the ratio of the positive to negative reference will not change much with time. Still a lot of jitter here can contribute to the overall noise. So the frequency should be stable as good as the simple PLL allows. It sometimes only needs adjusting the trimmer at the VCO.

Having a fixed but stable frequency can result in the low beat frequency to appear in the result. I am not sure how much amplitude will be there.

[borghese]

Unfortunately, the measured voltages seemed way off, about 3000 ppm low.

If the output of IC304 (LM741) is -10V this shift low the zener voltage of several mV; in a corner of the CS there is a label with the value of the DAC; you can check the programmed value by sending the commands with the calibration key in the CAL position:
Output, RS232, ON
CALIBRATE, DUMP

[pigrew]

Out of topic, do you have a drawing of the calibration key or maybe one to sell?

I ordered a new lock from "neat_stuff_from_ohio" on eBay. It was $10 after shipping for me. It looks like he ships to Europe, but perhaps not Slovenia. Shipping rates to you would be around $25 through the USPS (PM me if you need to use me as a proxy to purchase them).... The key that it came with works in the Solartron's calibration keyway. I'm attaching a better photo to this post. It seems to be an Ilco HV61 (made in Canada).

As far as my meter, I've left it on my bench for the last week. I seem to have damaged part of the DAC circuit; the DAC's buffered output is very noisy. I ordered a replacement DAC, and it should arrive by Monday. Expect an update in a week or so.

[Kleinstein]

A noisy output from the DAC would be an unusual defect for the DAC. It might point to a problem with the digital inputs - thus possibly the opto-couplers being right at the edge giving invalid / "digital" signals.

[pigrew]

After replacing the DAC, the signals became more stable. Perhaps I zapped it while trying to probe something... not sure. Regardless, the DAC's output is pretty stable now (34401A shows output fluctuates only on the last digit... so perhaps DMM noise).

Using an oscilloscope on the 10V reference node and the current-reference voltage, shows large, fast spikes, presumably from the chopper amplifier. I guess that could be normal, though I don't like it too much. The attached image is the 10V testpoint.

(Youtube video of oscilloscope and also how I probed it, the 7081's case is completely removed)

The reference uses a BC107 and BCY70 transistors(common base) to buffer the reference voltages. I wonder if they're getting old, causing the spikes to be larger than they should be.... I plan to substitute a pair of BC550 and BC560 transistors, as I can't think of a good way to test the transistors in-circuit. The collector currents are approximately 10 mA, dissipating ~40 mW). (I'm about to order the parts, to be delivered Tuesday, I only have 2n3904 & 2n3906 on hand here, BC550/BC560 should be lower noise).


EDIT: Another idea to stabilize the reference voltages would be to bias the transistors a bit hotter by bodging in a 1k resistor between +10V and GND, and another between GND and -10V. I need to measure the DMM noise before and after, to see if these things make any difference.

EDIT2: Readings are about 250 ppm low.

[Kosmic]

FYI, I'm seeing exactly the same type of noise on the 10V ref of both of my Solartron 7061. One is 2.4v P-P the other 1.9v P-P. This is probably normal.

[Kleinstein]

The Spikes on the reference voltage are not good at all. This should not be normal at all - maybe 1/100 the size might be acceptable though I would prefer < 1/1000 and thus barely visible.

The 2N3904/6 should be OK as a replacement. The transistors are inside a loop and there noise is thus not important. A slightly more powerful replacement (e.g. 2N2222 or similar) replacement might help a little, if the original types are more difficult to get.

The main purpose of the extra transistors should be to keep the power loss away from the OPs and maybe slightly lower impedance to current spikes.
Despite the spikes I somewhat doubt the transistors are broken.

The reason for these spikes could be a problem with the reference circuit, but also some point (e.g. cross conduction of TR201,TR202 in the ADC) causing excessive current spikes. So it might be worth checking of the spikes correspond to some control signals like the +-GLUG of chopper signal.

Another point to check would be the supply the transistors and also the OPs in the reference circuit get - the spikes might come from here too.
Failing R308/R309 (getting too high a value) could be a problem too.

[Kosmic]

Hmmm I will spend some time investigating the 10V ref then.

I'll let you know if I find something.