A precision voltage calibrator

[mhennessy]

OK - reply sent

It occurs to me that if there is interest, I can extend this service to other members. Send me a PM to register your interest, and I'll think about how to run it. Whatever form it takes, it'll be a "not for profit" service...

[AC/HL]

I bought one of the cheaper ones, same chip & settings but with jumpers rather than the push button cycling. At first I thought mine was faulty too. Set to 10v it read about 2, gradually creeping up. After a bit of experimentation I found that initially only the 5v setting works, but that once kick started at 5, they all work, and seem accurate (I only have 2.5 digit meters, so can't check to 5 decimal places) Odd. Further checks revealed that left to it's own devices it initially creeps to about 4 than jumps to 10, whereupon all ranges work. But... after setting to 2.5 the slow rise from 2 to 4 starts again. It's reproducible over and over, so must be inherent in the chip (mine at least) They do say to leave it to "warm up" in the instructions.

[Skywave]

Bill: I wouldn't go quite so far as to say that that was definitely a faulty unit, (and I'm not saying that you are), but it doesn't say much about the design, does it? Perhaps I'm a bit of a perfectionist, but I would be happy with it: no.

Just out of interest, were there any other components on the pcb, or was there only the AD584?

Al.

[AC/HL]

Three resistors, three capacitors and a LED.
I don't think it's faulty, it just has a quirk in it's operation. It does what I want, checks my meters to an amateur standard, although they weren't suspect to begin with.

[mhennessy]

Originally Posted by AC/HL

I bought one of the cheaper ones, same chip & settings but with jumpers rather than the push button cycling. At first I thought mine was faulty too. Set to 10v it read about 2, gradually creeping up. After a bit of experimentation I found that initially only the 5v setting works, but that once kick started at 5, they all work, and seem accurate (I only have 2.5 digit meters, so can't check to 5 decimal places) Odd. Further checks revealed that left to it's own devices it initially creeps to about 4 than jumps to 10, whereupon all ranges work. But... after setting to 2.5 the slow rise from 2 to 4 starts again. It's reproducible over and over, so must be inherent in the chip (mine at least) They do say to leave it to "warm up" in the instructions.

Now that is most peculiar...

When they talk about "warming up", they're talking about the readings stabilising in terms of fractions of a millivolt. In reality, the chip barely warms up (it only takes 1mA worst-case), and really, it's more about the IC adapting to the surrounding ambient temperature. The tempco is typically 0.003% (300uV) per degree C, so academic on a 2.5 digit meter.

Al's units were (I think) rock-solid, but a few % off. Yours are not stable, and as such, I believe that your unit definitely has a fault. I'm guessing I'm stating the obvious, but I'd start by having a really good look under a loupe or similar, as it sounds like a bad connection or similar, perhaps as a result of transit? Or perhaps there is contamination on the PCB, or a leaking capacitor? Or indeed the chip is faulty - while I'm happy to trust that units coming with a "cal sheet" are tested before dispatch, the others might not be so rigorously examined.

Of all the reports of these that I've read here and elsewhere, I don't think that I've seen this particular behaviour before. This is useful (and interesting) data

[jimmc101]

Most peculiar, I agree with Mark that it's worth checking all the connections.
In addition check for any contamination around pin 5 (strobe) as more than a few uA of leakage to ground can turn the device off.

I think I may have the same module (see post #55) and it certainly does not exhibit this behaviour.

Jim

[AC/HL]

Warm up was my wording, as you say they do quote surrounding ambient temperature, and I wouldn't be able to measure that drift. Interestingly, my printout is identical to Jim's in every detail. The meters are 3.5 digits, not 2.5 as I said, both indicate 10.00.

[jimmc101]

Thanks Bill, that I think that confirms my suspicion that the 'calibration certificate' supplied with this version is a fake.

Jim

[mhennessy]

Should we compile "certificates" to see if that's the case?

In a couple of weeks time, I'll have 1 or 2 of them, so will add details here (along with how well they compare to my meters).

[jimmc101]

I've just added a LM78L15 regulator to my module and it displayed exactly the same behaviour as Bill's; previously I had been using a lab power supply without this problem..

Originally Posted by AC/HL

... At first I thought mine was faulty too. Set to 10v it read about 2, gradually creeping up. After a bit of experimentation I found that initially only the 5v setting works, but that once kick started at 5, they all work, and seem accurate (I only have 2.5 digit meters, so can't check to 5 decimal places) Odd. Further checks revealed that left to it's own devices it initially creeps to about 4 than jumps to 10, whereupon all ranges work. But... after setting to 2.5 the slow rise from 2 to 4 starts again. It's reproducible over and over, so must be inherent in the chip (mine at least) They do say to leave it to "warm up" in the instructions.

Once started it would restart normally after power cycling even after waiting a few minutes, I then tried shorting out the capacitor between pins 6 and 7 (With the power off!) and the slow start behaviour returned every time.
Looking at the datasheet the suggested value for this capacitor is 10 to 100nF, the value in the module was 330nF. I removed this capacitor and replaced it with 22nF and have not had any further problems since.

It would be interesting to know what values the other modules use, it looks as if too large a value gives problems with slow rising power supplies.

Jim

Please excuse the bodge, I should have access to an HP33401A tomorrow and needed something in a hurry.

[AC/HL]

After re-soldering the chip, it now starts cleanly at 10.00v, but after using the 2.5v setting still takes a while to creep back up. I'll change that cap and see what happens.

Edit: Success, spot-on and repeatable.

[WaveyDipole]

I've had my Precision Gold for several years, the AVO for a couple and have recently aquired a pair of Thurlby PL230 PSUs. The Precision Gold and AVO give similar readings but not quite the same and the PSUs were pretty close but not quite aligned to the DMM or with each other. To be sure, the differences were not so big as to be a problem for most hobby purposes, but during a recent repair where millivolt accuracy was required I wished that the DMM and PSU had agreed.

In my mind the idea presented here made sense. I needed a starting point to compare against that I could be confident was was accurate enough for practical purposes and at least as accurate as the devices I was comparing against and affordable. If my DMM agreed with this this reliable source, then I could be confident about callibrating the PSUs, and if all devices then agreed with each other then I could have more confidence in using them. The case for purchasing one seemed compelling.

I decided to go down the self-build route so went ahead and purchased an LT1021C. I didn't have the resistors indicated in the diagram, but decided to use what I had to hand, 5 x 6.8k resistors %1 which measured exactly the same on the DMM to provide a range of ouput voltages of 2v, 4v, 6v, 8v, 10v.

As it happens my DMM was a handfull of millivolts out from the 10.00v reference voltage so I adjsted that first and then measured the voltages across the chain:

1. 10.00v
2. 7.99v
3. 5.99v
4. 3.99v
5. 2.00v

Close enough I guess. Just to be sure that the resistor chain wasn't dragging the voltage down I checked the top voltage with and without it connected. The voltage read exactly 10.00v in both cases.

I then checked my AVO. On the 10v DC range it was spot on, but on 50v it was out by a volt or so. I'm not sure whether this is down to the linearity of the meter or some of other factor. Something to investigate another time. I also checked another analogue multimeter which turned out to be spot on with the 10v and 50v range although obviously not millivolt accuate.

Today I callibrated one of the PSUs where again the DC voltage was out by a few millivolts but, as it turns out, other settings were out more significantly. The process went very well and it now agrees with my DMM. At the weekend hopefully I will callibrate the other PSU.

I think building this voltage reference me made sense in my circumstances but it may not make sense for everyone. Its certainly not worth getting too obessed about the odd few millivolts for hobby purposes and if the DMM was brand new then perhaps I would not have needed to build it. I just thought I'd mention it because it was this thread that gave me the idea and applying it in practice seems to have worked for me so the OP deserves the credit for helping me out. I may not need to check the devices for some time again, but its there if I need it again. Other than a little time spent constructing it, the cost was not prohibitive and when it comes to confidence, sourcing components from a reliable component source I think is an advantage.

[jimmc101]

Managed to access a Calibrated Agilent 34401A 6½ Digit multimeter.
Results confirm that the supplied printed calibration certificate (post#55) is more than a little optimistic and the device does not meet the 'L' specification.

Jim

[Skywave]

From Mark Hennessey, I have now received a precision voltage reference unit, which is based on a LM4040D. Its output voltage is stated as 2.503 v. @ 20°C. Thank you, Mark.

Using that item, I checked the d.c. calibration of 3 Fluke DMMs and 2 AVO model 8 meters; results are below.

Fluke 73. Indication 2.471 v.
Fluke 75. Indication 2.471 v.
Fluke 77. Indication 2.471 v.
AVO 8, Mk. 5. Indication 2.47 v.
AVO 8, Mk. 6. Indication 2.48 v.

Obviously, there is some degree of resolution error with the two AVOs. But two significant points stand out:
1. All the Fluke DMMs give the same reading
2. All five readings are below 2.5 v.

The 3 Flukes came into my possession from three different sources. Can it be coincidence that they all give the same (and presumably, incorrect) reading?
Similarly, the 2 AVOs came from different sources.
I find it difficult to believe that the voltage from Mark's calibrator is in error.
But given the above results, how can the above reported discrepancies be explained? I'm sure that there is a simple answer - but right now I'm quite puzzled. Perhaps I'm thinking about this too hard and missing something obvious.

Al.

[mhennessy]

If it helps, the attached photo shows the output of my DIY reference (white box) set to 10V (but it was the same story at all other voltages). The Fluke 287 (far right) is still in calibration. The measurement (2.50274V) was taken with a Keithley 2015THD, which has an extra digit of resolution over the 287, but the two meters agree when compared. For what it's worth, none of the meters in my picture have been adjusted by me, but they all agree to +/- 1 digit.

It appears that your Flukes are reading -1.27%. Certainly not the 3% we were discussing earlier, but still a tad low. But yes, isn't it interesting how they all read the same!

Loading effects are unlikely to be responsible - I did quantify those in a PM - but it's easy to check that your Flukes are still 10M. But having said that, did you check them all at once? Sometimes, the MOVs fail and leak heavily - if you have all 3 in parallel, it's possible that 1 leaking MOV would bring down all readings together.

It might be worth double-checking the battery voltage, as anything less than ~5V will cause the output to fall. The battery was s/h, but still measured >9V under load. But then I left it powered for most of last week before taking the final measurement. Though I'd expect a failing battery to result in changing voltages...

The plot thickens

[Skywave]

The battery reads approx. 9 v. under load. The DMMs and the AVOs were tested one at a time.

The Flukes came my way from different sources. One was from a firm that was having a clear-out; one was from an auction and the other one . . . can't remember! They were all acquired separately over about a 15 year period. I suppose that it is theoretically possible that at one point in their lives they were all (incorrectly) re-calibrated in the same certified calibration house - but that does seem pretty unlikely. What's more likely is that they've never been re-calibrated since manufactured. But then that raises the extremely unlikely possibility that all three were not calibrated properly (or at all) at Fluke.

I do wonder if the initial calibration was done with a voltage much higher than 2.5 v. - say 25 v., for example. And from that, is it possible that by design - or by the usual ageing effect - the calibration is correct at that higher voltage, but inaccurate at 2.5 v. ? What are the likely chances that the MOVs in all three Flukes are defective - and to the same extent - and thus causing the discrepancy?
And all of that fails to take into account why both AVOs indicate less that 2.5 v., albeit by not a lot. (Of course, I am talking about analogue instruments here).

Although for all practical purposes, the implied errors are of little consequence, right now, I'm somewhere on the 'irritation / fascination' axis with all this!

Al.

[Skywave]

Edit to the above.

Soon after having written the above, I recalled that I do own an Iso-Tech small hand-held DMM: it's model IDM17. (For various reasons, this is kept away from my workbench. It's rarely used, so I overlooked it). I bought this from new during 1992, (approx.). When connected to the calibrator, it indicated 2.497 v. That, of course, is 2.50 v. to 2 decimal places. And the calibrator voltage is stated as 2.503 v. - and that is also 2.50 v. to 2 decimal places.

Obviously, that result is clearly highly relevant, even if it fails to shine any light on the mysterious Fluke readings.

Al. / March 20, '16 //

[mhennessy]

Originally Posted by Skywave

The battery reads approx. 9 v. under load. The DMMs and the AVOs were tested one at a time.

OK, that rules out my thoughts. Good

Originally Posted by Skywave

I do wonder if the initial calibration was done with a voltage much higher than 2.5 v. - say 25 v., for example. And from that, is it possible that by design - or by the usual ageing effect - the calibration is correct at that higher voltage, but inaccurate at 2.5 v. ?

The calibration adjustment procedure is described in the manual - they use 3.000V. Once adjusted, everything else should be within spec - there are no other adjustments.

The (expensive!) thick-film divider network ensures good accuracy between different ranges, and excellent long-term stability. It's one of the key things you're paying for with decent meters.

Originally Posted by Skywave

What are the likely chances that the MOVs in all three Flukes are defective - and to the same extent - and thus causing the discrepancy?

Zero. The only possibility was the scenario of all 3 connected together with one (or more) pulling down the output to the observed value, but that's not the case here. But to check, it's perfectly valid to connect two meters together with one set to Volts and the other set to Ohms.

Originally Posted by Skywave

And all of that fails to take into account why both AVOs indicate less that 2.5 v., albeit by not a lot. (Of course, I am talking about analogue instruments here).

Rightly or wrongly, I'm actually ignoring all the readings from the AVOs

Ultimately, with the best will in the world, I'm not sure you'll ever get to the bottom of this particular mystery. The question now is simple: do you adjust all 3 of your Flukes to 2.503V, or do you seek a second opinion from someone else with an known-accurate meter or reference? If it's the latter, I'm not offended in the least. Quite the opposite, in fact - perhaps something might be happening in transit? Rigour is fun

[Oldtestgear]

I have been reading this thread with great interest but have not commented until now.
I have a 30 year old Fluke 77 which I check fairly regularly & it has always been accurate. Like all equipment it could drift due to ageing or environmental factors but to date it has been perfectly acceptable for what it is.
There is only one way for Skywave to be clear what is happening & get his Flukes tested on all ranges using a decent calibrator. I am happy to do a full comparison using my Datron 4705 & Keysight 34461A DMM in the hope that this might shed some light on the problem. The Datron has not been calibrated for years (by an independent cal. house) but can be referenced against the Keysight which is still within the calibration period.
Perhaps a trip to Torquay might give help? Anyway, if you are interested please let me know.

[Skywave]

Oldtestgear:

Thank you: your generous offer is much appreciated. I don't drive a car these days: my wife is effectively my driver. I will need to arrange travel dates, etc., with her: I do not anticipate problems in that regard.
I'll be in contact with you, via a P.M., as soon as I can.

Al. / March 21 //