Prema 6048 teardown

[Kleinstein]

Not changing the polarity for the null would be a firmware bug, as both polarities can have a different offsets as a combination of ADC and amplifier offset. With just 1 setting one can not measure both.
It may be worth doing at least a fast INL check with the null in autoscale mode.  There is a chance that this could be a relatively easy work around to solve the problem.

[Mickle T.]

Interestingly, the 7.5-digit Prema 6040S has built-in periodic ADC zero correction in both polarities. To do this, all 3 MOSFETs (switching 500, 5k and 50kOhms gain resistors) are opened, and only the leakage current actually flows to the input of the ADC polarity reversal unit.

[branadic]

Removed the parallel resistor to R36, shuffeled of few IC's around, gave the unit a good warmup, then performed nulling on all three units in auto ranging, adjusted 10 V range to the same 10V reference and performed INL test again.
For some odd reason the INL of the P6048b is now smaller than the others, but still visible, while on the other two units it hasn't changed at all.
So I guess the only thing left we can do is to analyze, understand and improve firmware for the 65C02 processor. Martin already started with that.

-branadic-

[Mickle T.]

If you look closely at the nonlinearity graph, you can see that in addition to the discontinuity (the cause of which may indeed lie in the firmware plane), there is a quadratic shape of the curve near zero. My research into the firmware showed no sign of special handling of this case. Therefore, I assume that the reason for the non-linearity is in the hardware.

[Kleinstein]

I had a more detailed look at the fine INL data from this post
https://www.eevblog.com/forum/metrology/prema-6048-teardown/msg4655983/#msg4655983

The attached graph shows the difference from the P6048 to Solatron 7081 (corrected). Vertical is difference in µV and horizontal the voltage in V.
The data seem to be recorded from positive to negative. So very likely the polarity changes at some -45 mV at the point with the jump.
However the data on the positive side of this still show quite some nonlinear effect.  This indicates that there is not just a problem with the offset, but also a more gradual problem with linearity for voltages in the +-100 mV range.  This part looks more like a hardware problem, not a firmware problem.

I would expect a similar shape for the part going from nagetive to positive and thus reading the small voltages with the negative relay setting. Form that it looks like the offset handling is not that bad at all.
With the nonilnearity at low voltage it only looks like the step is caused by an offest problem.

If the problem is with the ADC itself, a possible fix may be adding more offset (e.g. another 100-150 mV and thus some 50-30 K in parallel to R36) this way not use the ADC in the very low voltage range. Depending on how the polarity change is controlled (ADC reading or voltage after the offset is removed) one may still have to force it back to the same voltage and not use the full larger overlap.

Another possible cause of INL may be leakage from the reference transfer for the resistance ranges. One may be able to test if there is significant leakage there and for a test maybe disable &/ remove the link (e.g. remove R37/R38 or K13/K14 of the ADC board and add a link for a dummy 0 ref instead).

[branadic]

I probably wouldn't interprete too much into that as also the S7081, which the measurements are compared to, has an INL error that is worse than a 3458A, something like +/-0.2 ppm in the 10V range. So everything below a certain level could possibly come from the device we are comparing it to.

Edit: Ah, now I see were this arguement is pointing to...

-branadic-

[branadic]

If the problem is with the ADC itself, a possible fix may be adding more offset (e.g. another 100-150 mV and thus some 50-30 K in parallel to R36) this way not use the ADC in the very low voltage range. Depending on how the polarity change is controlled (ADC reading or voltage after the offset is removed) one may still have to force it back to the same voltage and not use the full larger overlap.

Some 30 - 50k in parallel to R36=40k would be quite something, resulting in 17.1k ... 22.2k and the t.c. of the parallel resistor would contribute quite a lot. So in theory it would be better to replace it by a low t.c. 20k instead, wouldn't it?

-branadic-

[Kleinstein]

The R35/R36 part would still have a limited weight. It adds an offset through 750 K while the 20 V range signal comes via 50 K. So a factor of 15 reduced sensitivty.
The parallel resistor would be for testing first. If this turns out to really help with the INL one could look for a better pair of resistors to replace R35 and R36.  There is a good chance that it would imporve on the INL, but not sure.

A possible reason for not so good INL with low voltages could be that in this case the reference is switched on for only a very short time. The OP-amp at the integrator needs some time to settle after the reference is turned on. If turned off before settled this leads to an INL error. Onece settled it would only be a fixed contribution to the offset.

The OP77 is an obsolete pretty good BJT based OP-amp (could well be selected ones for the 6048), so replacing it is a somewhat risky step.  With the OPA202 there would be a possible upgrade with higher speed (about 2x) and also lower 1/ f noise.  With possibly some additional gain inside the Prema chip it is a bit tricky to replace the integrator OP-amp and expecially go much faster. A faster OP-amp is also not necessary faster settling. Even the parasitic capacitance between the amplifier PCB/supply and the ADC PCB can effect (in a good or bad way) the settling as some of the capacitance is at the integrator input. Chances are R29 is there to make it more a good way.  Because of this measuring the settling may not work, as the scope (also just the ground) would effect the result. The settling part in question is also tiny and not sure one would see it:  Something like 1 mV for some 2 µs would be enough to cause an 1 ppm range INL error. So a measurement would only be to get an idea on how short the pulsed get.

[branadic]

I can report that 47k in parallel to R36=40k leads to 100mV offset. Nulling however leads to Error4, which means "Error during offset measurement. The offset which is present at the input sockets is too large."

So this approach is no approriate way to get down INL error.

-branadic-

[Kleinstein]

Not allowing a much higher offset / shift is unfortunate. It would have been a relatively easy fix.
As a last resort, one could try it in combination with a modified ROM, that accepts the higher offset. Still extra effort on understanding and changing the code (probably just a constant).
I don't see a principle reason why the ADC should not have enough headroom - actually surpring that the 2 V range is only 2 V and not 5 V.

Replacing the OP77 may or may not work.  To judge the chances a bit better,  it would help to know the frequency of the reference switching and thus how short the pulses get, when at low voltage.
It could be pretty hard to see the settling part of the integrator with a scope: If there is no additional gain in the prema chip it should be around 0,8 mV  (dI / (2*pi*GBW*C_int) ) steps and possibly a bit tricky triggering so that the averaging mode in the scope can be tricky. The extra capacitance from the cope would also not help.

From the photos it looks like the OP-amps are mostly in sockets. So a replacement may not be such a scary thing.
The open loop response of the op77 does not look special - so not extra much phase reserve.  The OPA277 and OPA202 as 2 possible substitutes look like they both have more phase reserve. So there is a good chance for faster settling. The OPA277 may even be available as a DIP version for a direct drop in, but with more LF noise (may be a relevant part of the overall noise).

[Rax]

Hey, branadic and all, great thread and work on here!

I may have just gotten one of these - currently in transit and I really have no idea yet of its condition (other than "it turns on..." argh). When I clicked the BIN button, I felt like I'm signing on my voltnut certificate!

It looks like this exploration is still ongoing (I'm still getting up to date with it all). branadic' last post is almost two months old - I am aware of at least one other project he's been on, as that's also of high interest to me - so maybe there was just not enough time for progress. Is an update in order?

I do not really right now know if/what I can contribute more data from my sample, but I'll definitely be happy to.

[branadic]

No progress on my side and I'm not tempted to spend more time on it at the moment. Unfortunately, essele left the community and is selling his unit, the only one who could have contributed. So yeah, I gave up on it I guess.

-branadic-

[Rax]

No progress on my side and I'm not tempted to spend more time on it at the moment. Unfortunately, essele left the community and is selling his unit, the only one who could have contributed. So yeah, I gave up on it I guess.

-branadic-

The obvious question is - who are they selling their unit to, and is the receiver interested in furthering this? Disregarding me for a second here. Not looking for an answer to this, just to set a tone here of looking at the future.

Birdseye view on this, this was a hair over 10uV offset at the 0V crossing point, is it not? Which may be as close to .5ppm (+/- of range) to be borderline compliant. 

[Rax]

I wonder of notes #2 and #3 (p.2/1) in the manual gives an indication on this observed behavior ("values in each case +/- 1 digit and after offset correction. [...] +/-0.0005% m. rdg. must be added during 100ms after signal change"). Maybe this also relates to Kleinstein's point on settling time.

[Kleinstein]

The 100 ms settling time are for the settling of the input amplifier - though this looks a bit excessive. It may include a little DA effect that can be relatively slow. Esepcially small changes in the input voltage should get away with less waiting. AFAIK the observed INL error was with enough waiting times and thus not extra error from settling.

The settling in the ADC that could effect the INL is settling of the integrator in response the the reference switching as part of the SD ADC function. At low voltage (or slightly negative, before switching the polarity) this could include some pretty short pulses down to about 1-2 µs. The question is if the relatively slow OP77 at the integrator is fast enough to get reasonable settling in this time.
The way the ref. switching is done, the input impedance of the integrator changes and incomplete settling would effect the result in a nonlinear way. How fast the Integrator settles depends on the details of the OP-amp and the part inside the custom chip (e.g. capacitive loading) - so hard to tell.

[Rax]

The question is if the relatively slow OP77 at the integrator is fast enough to get reasonable settling in this time.

Thank you, Kleinstein. I am adding a couple of OPA277s to my next parts order. As it seems all these are on sockets, it seems a fairly convenient experiment.

It seems there's a couple of flavors at my typical supplier, depending on how tight the specs are on the input. Would I want the tighter spec? (P vs. PA suffix). Not sure how critical this aspect would be.

[Kleinstein]

The main difference for the grades of the OPA277 would be the offset voltage (does not really matter as long as it is stable) and the amount of testing. So I see no real need for the higher grade version, as the other parameters are not that critical.
For the settling part (hopefully fixing much of the INL problem) the OPA277 could be an easy test. For low noise an OPA202 with an adaper board from SO8 or SOT23-5 to DIP could be better (less 1/f noise).

[Rax]

Does anyone have some insight on if/how R56 is supposed to be adjusted?
The resistor on pin 11 (per 74HC4046 datasheet) seems to set the frequency range of the VCO, but I couldn't find any indication in the manual as to how and when is it supposed to be adjusted. 

[dietert1]

The 4046 doesn't require any precision adjustment. When the PLL locks, this is the adjustment. If the VCO input is stable in the 0.9 V to Vcc-0.9 V range, it is locked. When unlocked the VCO control voltage will be outside this range or oscillating.

Regards, Dieter

[Rax]

The 4046 doesn't require any precision adjustment. When the PLL locks, this is the adjustment. If the VCO input is stable in the 0.9 V to Vcc-0.9 V range, it is locked. When unlocked the VCO control voltage will be outside this range or oscillating.

Regards, Dieter

Thank you, Dieter, but that doesn't explain why the control and how to adjust it (or I don't see it), or when necessary. There's no trace of direction on it in the SM.

[Kleinstein]

The trimmer abjusts the capture range of the PLL. An adjustment may be needed when switching from 50 Hz to 60 Hz mains frequency.
The trimmer will effect the PLL test signal / test point, that needs to be stable and well inside the valid range (e.g. 1-4 V).
To test if the trimmer setting is good one may have to also check a new start up, as the catch range can be smaller than the hold range.
The exact trimmer setting should still not be that critcal.

[Rax]

I got my unit today. The best news about it is - it's not nearly as ugly as I thought it'd be, judging by the pics online. Score!!

But more seriously, it's actually in better shape than I thought, given what I paid for it. It seems to be measuring in the ballpark in most ranges - still running through the gamut - except current (which doesn't seem to work, though a big grain of salt on that as I've not had enough time to assess).

I enclose pics measuring various sources I have, including the FX which is still in my custody (about to be shipped back).

Inside, the unit seems extremely clean, no work prior or butchery detected, which is great.

(running into some issues with pics - how do you guys handle this, usually? I've seen tons of high res pics attached to posts!?)

[bdunham7]

(running into some issues with pics - how do you guys handle this, usually? I've seen tons of high res pics attached to posts!?)

You need to use a photo-editing program to get them down to 500k  in size (for 10 photos/post), or up to 4M if you only post one photo.  You still should be able to keep a pretty good resolution and quality if you use .jpg or .png. and just the right settings.  Then getting them to appear as full-sized rather than a thumbnail, that's an additional step.

[Rax]

Pics.

[Kleinstein]

But more seriously, it's actually in better shape than I thought, given what I paid for it. It seems to be measuring in the ballpark in most ranges - still running through the gamut - except current (which doesn't seem to work, though a big grain of salt on that as I've not had enough time to assess).

The 6048 has very few resistors in the gain path: the 200 mV/2 V and 20 V mainly have 1 pair of canned resistors and the LTZ1000 reference to produce drift. So I would expect it to hold the calibration very good, even without ACAL. The 2 V range may be the more stable range, as here it is the ratio of 2 equal value resistors.

For the current ranges there is the obvious fuse that can be bad and also relay contacts could be an issue. The terminals for current are separate, so a bit different than with most other meters. If there should be an issue this should be relatively easy to fix, though getting a relacement relay or shunt can be tricky (or ugly with a different form factor).