Voltech PM3300 power meter

[capt bullshot]

I'm copying a few posts from the TEA thread here, I'd guess its better to find the information here in one place:

So I've got this beast from the bay of evil:



State as delivered: It turns on, but shows all the measured values inaccurate and fluctuating.

[capt bullshot]

So, now there's a new patient on the bench:



For some first tests, I've connected 10V DC / 10V AC to its inputs and found the measured values slowly fluctuating above and below the correct 10.00V reading. All channels voltage and current inputs showed this fluctuation.
OK, time to take it apart, which isn't obvious at the first glance. One has to remove the feet from the bottom (not the back) side, then the whole instrument slides out of its outer cover. Another about ten screws later, one can lift the internal top cover. To the left, three shielded boxes (voltage inputs), below (not visible) another three shielded boxes for the current inputs. Right hand side, the main processing board is buried under the printer drawer and the power supplies. Yes, it's got two of them: One 5V, 12V, -12V, the other 24V for the printer.

First I checked the supply voltages: Look fairly OK, no signs of excessive ripple. Ran the unit off a lab supply just to be sure - no change in the fluctuations.
Next, I had a look at the input modules. Each has two isolated paths: current and voltage, and each has its own isolated supply, powered by some square wave AC through small transformers. So I checked the isolated voltages: One thing is in common - they're all too low. About 9.8V, and fluctuating, where I'd expect 12V, about 4V fluctuating instead of 5V (just an educated guess, don't have a service manual). These are outputs of (apparently) smallish 78L12 style regulators, so I checked their inputs - too low for this kind of regulator.



So far until now, I'll have to come back later to find out how to remove the printer drawer to gain access to the main board and find the source of the square wave AC supplying the input boards. I'd guess there's something bad within this circuit.

[capt bullshot]

Some news from the Voltech PM3300:

I've found the culprits

#1 is that (high grade 1000uF/16V) electrolytic capacitor used to AC couple a half bridge MOSFET stage to drive all the small transformers used to supply the input modules. I found it getting warm and tested its series resistance to about 600mOhm at 40kHz, quite a bit too high. Replaced that one, but the fluctuations weren't gone and isolated supply voltages were somewhat better, but still far from good.



#2 is this pin of the power supply connector. It's the +12V supply line, carrying 1.5 ... 2.0A. Note the slight discoloration of the connector housing, and the pin itself lost most of its plating when I removed it from the housing. There's been about 0.5V ... 1V drop across this contact. After an intermediate fix, the supply voltages with OK with little to no margin, and the fluctuations are gone.





The better fix for this (crap) kind of connector would be to cut it off and solder the wires directly to the pins - maybe I'll apply this later.

Here's a view of the main circuit board (the onboard battery is still good, reading 3.6V quite as new):



Right now, I've connected 10V DC and 1A DC sources to watch the stability. The meter is specified to 0.1% (from reading) accuracy @DC (0.05% AC). Judging from the first results, these specs look quite optimistic. From my own experience I know it's quite brave to specify this kind of wide band power meter developed in the 1990s to this accuracy.

[capt bullshot]

Someone else has experience with this kind of connector:

Ah a shitty SIL connector as mentioned in one of my previous posts yesterday  . I hate the things.

Good find 

SIL is not the problem, It's a crap insulation displacement contact.
Replacing that with a decent quality crimp one (like the white one next to it) should sort it out.
Weird that they used both types. I wonder if they subcontrcted the power supply build.

From my experience, it's the type of contact material that makes the difference. These contacts are (were) available in different materials and plating. Once there's a bit elevated contact resistance due to whatever, the contact heats up and the material degrades more over time. I've seen these type of connectors getting brown and finally molten. Happened not only to these Molex style, but to supposedly better Phoenix too. They were operated well within their max. current specs. Apparently the contacts have gotten better today, since I haven't seen this failure mode any more today.

The power supplies look like common off the shelf units, and use the same type of connector at the other end of the wires.

[capt bullshot]

Gave the PM3300 a fresh calibration, it's quite easy if you have suitable sources.

The Fluke 5100 does all the AC voltage ranges, and AC current up to 2A. As the meter has current ranges up to 30Arms, its calibration requires sources up to 30Arms. I've got a DIY transwhatever amp that does up to 8Arms, good enough for the 10Arms range. A 1mOhm shunt (measured to 1.0049mOhm @DC) connected to the 34401 gives the current reading here. I've intended to use the 30A secondary transformer with my AC source for the largest ranges, but as the AC source is a voltage (not current) source, it's not stable enough, so I skipped this.



The result (100V and 1A @100Hz generated by two Fluke 5100 with ext. source from the Tek AFG) is at least is within the spec now:




So far, it's done. I'll check the accuracy and stability later again.

[capt bullshot]

I've continued testing the PM3300 a bit:

I did the calibration (adjustment to be correct) when the unit was warmed up for more than two hours.
Next day, freshly turned on, the measured values are about 0.5% off, and slowly crawl upwards until they eventually reach their expected values.

So this is quite a bit of thermal drift happening here for an unit that is specified to 0.1% accuracy. I shall dig into it again, and figure out whether this is by design or still some faulty part.

[capt bullshot]

So it appears the bad thermal drift is by design.

The device uses a Sipex SP7800 ADC (looks like a slightly different clone of the Burr Brown ADS7800). The only thing mentioned about it's reference voltage is a small block "Internal Ref" in the block diagram (Sipex). Nothing else, no value / accuracy / drift / whatever specified for this internal reference.

I've set up a small test:
- Stable 10V DC applied to the input
- Measure internal 5V supply
- Record internal 5V supply voltage and PM3300 reading of the applied 10V DC.

Result:
The 10V reading of the instrument drifts the same direction and amout as the internal 5V supply does while the unit warms up.
So the internal "reference" of the SP7800 ADC apparently is just a fake and the ADC uses it's supply voltage as reference. This particular 5V supply is derived from a simple 78L05 regulator, these are anything but temperature or long term stable.

WTF?
This is a multiple 1000$ instrument back then, specified for better than 0.1% accuracy and they couldn't even do a simple reference voltage right?

[ch_scr]

Sure enough, pin 3 on ADS7800 is +2V Reference Output. Same pin 3 on SP7800 says "Not internally connected". Is the ADC socketed, may it have been substituted previously?
Are there components attached to pin 3 on the PCB?

[capt bullshot]

The ADC is soldered, and by the looks pristine. There's a electrolytic capacitor connected to Pin 3, as Burr Brown states in their data sheet.

BTW. Burr Brown doesn't specify much about that internal reference, except it's typically 2V.

There's one interesting aspect: As I wrote, the 10V reading and 5V supply drifts the same direction -> 5V rising leads to 10V reading rising. Wouldn't one expect the 10V reading to fall, if the ADC reference is directly derived from that 5V rising?
Drift direction with AC input is the same, so I'd rule out input amplifier offset drifting.

[Robert763]

You may have a faulty ADC. The datasheet says power supply sensitivity is +_ 1 LSB so it does not use the +5V as a reference directly.
If there is -15V on pin 22 I'd suggest replacing it with a ADS7800.

[capt bullshot]

No, it's the very same at all channels. So I'd have 6 faulty ADCs?

But the drifting direction (5V and 10V reading drifts the same direction) points to somewhere else but the reference.
I'd expect the 10V reading falling if the reference is rising proportionally to the 5V supply.
A very interesting fact is: 10V reading rises the same percentage as 5V supply rises, and it doesn't matter if I apply DC or AC to the input.

As I'm just at the beginning of the investigations, I might find some other cause ...  E.g. there's 4 voltages (+/- 12V / +/- 5V) regulated by 78L / 79L on each channel, and their input voltage is very marginal to provide for their specified dropout.

Yes, it looks like the original design used the ADS7800 and someone changed the BOM to Sipex SP7800 at some point. ADS7800 is still available, but pricing is slightly prohibitive to give it just a try for 6 channels at 93EUR per piece.

[DC1MC]

No, it's the very same at all channels. So I'd have 6 faulty ADCs?

But the drifting direction (5V and 10V reading drifts the same direction) points to somewhere else but the reference.
I'd expect the 10V reading falling if the reference is rising proportionally to the 5V supply.
A very interesting fact is: 10V reading rises the same percentage as 5V supply rises, and it doesn't matter if I apply DC or AC to the input.

As I'm just at the beginning of the investigations, I might find some other cause ...  E.g. there's 4 voltages (+/- 12V / +/- 5V) regulated by 78L / 79L on each channel, and their input voltage is very marginal to provide for their specified dropout.

Yes, it looks like the original design used the ADS7800 and someone changed the BOM to Sipex SP7800 at some point. ADS7800 is still available, but pricing is slightly prohibitive to give it just a try for 6 channels at 93EUR per piece.

You can get last 6 pcs. from this Chinese seller, as ugly as they look they're definetly not fake, but pulls, and price won't destroy your wallet:
https://www.ebay.com/itm/1X-ADS7800JP-12-Bit-3ms-Sampling-ANALOG-TO-DIGITAL-CONVERTER-ADS7800-/120939937547

Cheers,
DC1MC

[capt bullshot]

You can get last 6 pcs. from this Chinese seller, as ugly as they look they're definetly not fake, but pulls, and price won't destroy your wallet:
https://www.ebay.com/itm/1X-ADS7800JP-12-Bit-3ms-Sampling-ANALOG-TO-DIGITAL-CONVERTER-ADS7800-/120939937547

Guess I'm not that patient to wait for them to be delivered
So for now I'll continue searching for more hints to the root cause

[capt bullshot]

Next result:

ADC and internal reference are fine, my first guess was wrong.

Measured the ADC input voltage at SP7800, pin 2:
it's drifting, matching the 5V and read out value drift. ADC input voltage gets larger over time.

Injected some artificial drift to the 5V supply: shows no influence to ADC input voltage, reading stays stable

There's some DC offset with input zero at the ADC pin - inverted the input 10V polarity: ADC input voltage inverts, keeps drifting over temperature, absolute value gets larger over time

So is there some gain over temperature drift within the input amplifier?

The input amplifier is a multi-stage amp (2x OPA27, 3x OPA37 and a bunch of 'HC4053 muxes). Now I'd really like to have the schematic available to make some guesses where the drift occurs.

Still looks like a bad design induced issue to me, as all channels behave the same (no, I didn't check each SP7800 input, just one, I'm extrapolating here from previous observations).

[capt bullshot]

One step further:

Made some more tests to rule out the influence of the 78L  / 79L voltage regulators onto the readings. Couldn't find a correlation between artificially induced supply voltage fluctuations and readings.

So next, I've cooled and heated various ICs at the channel input amplifier - got stuck at this TLE2072:



This is the last stage of the input amplifier circuitry, driving the ADC input from Pin 7 (one can measure 2.5V @ 10V input). Pin 6 and 5 measure zero, so it's an inverting configuration obviously. Pin 1 is the output of the other OpAmp inside this package, about  0.882V there.
Amplitude @ Pin 1 stays relatively constant, while amplitude @ Pin 6 varies by 0.4% with temperature changes (carefully applying freezer spray and heat gun) and correlates with changes in readings.

I don't understand yet why, I don't see components around that IC which are supposed to have that large temperature dependence ...
I'd rule out these small ceramic caps bodged across resistors, since it doesn't matter if there's DC or AC applied, amplitude changes anyway.

Here's a picture of the voltage input channel:

[capt bullshot]

Investigating further showed this resistor has quite a temperature coefficient (colour code 13.3k, 1%). Crude heat gun testing showed 0.3% deviation over some 70...100 °C which seems plausible for a 50ppm or 100ppm type. The other gain setting resistor (4k7, green body) showed less deviation.

There's another one of the same value and type, all other resistors are different value and have a 6 band color code (green body). This one has a different body colour (light blue) and only five bands. I'd guess the others (green body) have a better tempco than this one.



Looks like someone has messed up the inventory of precision resistors at Voltech back then?

'nough for today, will try to find a better replacement and do more checks to confirm tomorrow.

[capt bullshot]

Ordered some precision resistors from Mouser to replace these offenders. As there's always two resistors involved to set the gain of an OpAmp circuit, I've located all the gain setting resistors around the TLE2072 and ordered more resistors of the same type. Since I don't know which make / spec the original resistors are, I've decided to replace all gain setting resistors for these two particular amplifier stages. If you want stable gain over temperature, you should use resistors with low and similar or matched tempco. Easiest way would to use resistors of the same manufacturer / type.

So this is what I odered, TE H8 series (15ppm/°C):
https://www.mouser.de/datasheet/2/418/5/ENG_DS_1773230_H-731295.pdf
Values 13k3, 4k7, 7k32

to replace these:



Here's a picture of one input module (note the current / voltage channel looks nearly identical). An interesting detail: They've used two transformers back to back to achieve better isolation and less coupling.



Here's a picture of the current shunts

[capt bullshot]

So I've got the resistors delivered today. Replaced them as described above in one channel (voltage and current CH1).

Result: CH1 is still drifting with temperature / unit warm up. It's slightly better than before (about 0.3% vs. 0.4%).

So my conclusion for now: I haven't found the root cause for the temperature drift. I still believe it's by design, since all channels show the same behaviour and after two hour warm up the measured values are within spec.

[capt bullshot]

Turns out the resistors I've ordered have dissimilar and opposite direction tempco for the 13.3k and 4k7 / 7k32 parts.
For one amp, the (non-inverting) gain is set by 13k3 / 7k32, for the other one (inverting) by 13k3 / 4k7.

So now, go figure what happens if 13k3 has positive tempco while the others have negative. Yes, you're correct, the gain drifts with temperature.
They aren't that bad at all, within spec (roughly +12ppm/°C the 13k3, and -3ppm/°C for the 4k7)

I'm kind of annoyed now, as these weren't cheap, I had to wait a week for them to be delivered and they don't fix the problem.

Interestingly, the other gain stages don't show that problem, looks like they were just happy with the tempcos of their gain setting resistors.

[capt bullshot]

Searched through my stockpile of resistors. Found a combination of two 1k5 and a 10k2 that makes up a 13k2 resistor at a single digit TC.

I've tested that combination (bodge + precision resistors) as gain setting resistors in that small test setup, showing nearly constant gain over temperature.
Then, I replaced the 13k3 precision resistors by that bodge in one channel, resulting in a slightly different gain (as one would expect by the change from 13k3 to 13k2). A quick calibration test showed this is within the range the unit accepts.

Now I ran another two hours unit warm up, and found the modified channel drifting about 1/3 of the previous value (0.1%). Looks good, so I decided to modify all left over channels.

Next step will be a full calibration (adjustment) of the warmed up unit and then watch the drifting behaviour of all inputs.

[capt bullshot]

Now testing the Voltech 3300 power meter:

It's working, thermal / warm up drift is better than before my last modification, but there's still an issue that I'm hunting after.
Here's the setup:

I'm using the Fluke calibrator the generate the input for voltage measurement.



Then this Kepco BOP to generate a constant AC current fed into this transformer.




The shunt (1mOhm) and homemade x10 precision amplifier to monitor the transformers secondary




The output:



One needs the dual channel AFG to generate two reference signals of same frequency and variable phase relation. One is fed into the Fluke 5100A as ext. Osc. Signal, the other one into the Kepco BOP. To measure the voltage across the shunt (30mV @ 30A), I'm using an HPAK34401A. There's a nice trap for beginners with that: I you apply 1A -> 1mV across the shunt, it's about 1% of the measuring range, and most multimeters have a significant error at this level. So I had to build the x10 amplifier to cope with that. The reason is: The maximum current out of the Fluke is 2A, so I had to calibrate the shunt at this level to transfer the current measurement to the transformer / Kepco setup.

[capt bullshot]

Now let's go for some fun with TEA: Testing the Voltech 3300 power meter.

It's working, thermal / warm up drift is better than before my last modification, but there's still an issue that I'm hunting after.

Bingo. Found the issue. It's all about the same story - don't let yourself be distracted when doing serious work - I've replaced the wrong components at the last board mod. 
Interestingly, it just worked with the wrong components replaced - until the unit got warm enough. Now one knows what the bodged small ceramic cap across the 13k3 resistor is good for: To keep the amplifier from ringing / oscillating at elevated temperatures.



BTW. I've replaced the components just above the 7k32 and 13k3 resistors (4k7 and 8k2) - similar enough, the channel still worked and passed cal.