Elektor Voltage Current Calibrator Project

[Martin72]

Update:
No update, I´m a lazy boy...
Or better, time is not enough, but it is not forgotten.
Several projects are currently in the pipeline:
- Voltage Reference, installation in the housing and ageing before it is sent in for calibration.
- 1kHz ultra low distortion oscillator:
Installation in the housing, testing.
- Arduino Curve Tracer:
Finish assembling, then test.
- ENOB test environment:
Filter boards designed, still need to be assembled and tested.
- New HMI for Hioki 5Mhz LCR:
So far only collected ideas

And a lot more....
As I said, lazy boy.

[KungFuJosh]

On the bright side, you'll probably have your SDS3000XHD by the time you finish all that. 🤣🤣

[Martin72]

Do you know the foot massage scene in pulp fiction?
The final punchline where Vincent Vega had Julius by the balls and how Julius reacted to it ?
I´m Julius....

[KungFuJosh]

Nah, I don't remember. Maybe I'll watch the movie again and get back to you. I could probably watch it a couple times before.......................... 😉

[Martin72]

Months later...
The operation was a success. I had to disconnect the display from the board and build both circuit boards separately, then connect them with wires.
Pictures will follow later.
I would leave the reference switched on permanently for a month, but I still wonder if that's necessary at all.
The MAX6305 is in the “oven” and is kept at a constant temperature....does it still need to be pre-aged?

[KungFuJosh]

Wouldn't hurt, right? Plus you can test the reliability of everything before you inevitably pay somebody to calibrate it.

[Kleinstein]

Some burn in for the reference would definitely help. The pre-aging is not for the TC part, but for things like chip internal stress and early electronic effects. The oven may speed up the process a little, but a month is not really that long.

Another effect that can come in is humidity and PCB stress. With the heat there will be PCB stress and the reduced RH will cause some slow settling. The humidity effect can be quite slow with a timescale of days to weeks. For ultimate stability it would make sense to have the critical ref. part power 24/7. It is still the question if than a more classic heated reference like LM399 would have made more sense.

[Martin72]

The whole thing is very much based on the AD reference board.
If all this turns out to be unsatisfactory, you can always switch to the “original”.
But I hope it goes well, because it was expensive enough as it was.

[Kleinstein]

The design looks OK for short term stability, noise and linearity. However it is not a really good long term reference. Within this limitation it can still be working well.

[Martin72]

We'll see what it can and can't do.
I had just turned it on and changed the value until the Brymen displayed 2.5V - that's the offset that you have to compensate for afterwards.
However, I will only do that next month or the month after, until then the reference is allowed to run non-stop for a month together with the SDM3065X.
I first wanted to take it to work and connect it to a UPS, maybe I'll do that too.
In just under 2 months, we will have a calibration of all multimeters in-house. If the lab should bring their 8.5-digit meter, I'll ask them if they could leave me alone with it for an hour.

[mawyatt]

Consider cycling the power on and off in the early stages of long term operation & testing. Use a rate that allows things to fully warm up then fully cool down to ambient. This relieves some of the mechanical stresses from the PCB build and mounting/cases assembly, and also taxes the components more than just constant operation at a fixed temperature.

This was often implemented in aerospace/military equipment as temperature cycling. Recall a survival radio where this test uncovered a flaw where the temp cycling would cause the moisture in the potting compound to condense out on the surface of an alumina substrate. The moisture would collect salts on the substrate surface and osmosis would pull the moisture under a decoupling chip cap which had Vcc across it. Dendrites formed under the chip cap and eventually shorted Vcc, causing a radio failure. Took awhile to figure this out, as at first the chip cap was thought to have failed, only to find after removal they were fine and the short was gone as the dendrites were sucked back into the molten solder upon chip removal. Eventually the problem was found and source was traced to moisture contaminated potting compound.

This cost a bundle with recall/rework and even worse when a user needed the radio for survival comms it could fail 

Senior management tried to hide the problem

Best,

[dazz1]

Hi
I have updated the BOM with working links and current prices.
The cost is driven by just two parts, the:

• LCD and,
• 20bit DAC.

The BOM doesn't include the heater option.
I have not investigated whether a programmer is needed for the LCD.

[Andreas]

The MAX6305 is in the “oven” and is kept at a constant temperature....does it still need to be pre-aged?

Hello,

my 2 MAX6350 samples (in DIP8) were ageing ~10ppm/year for the first years.
(now some years later they are at ~5ppm/year).

https://www.eevblog.com/forum/metrology/how-to-age-a-max6350-voltage-reference-ic/msg3370012/#msg3370012

The SO-8 package should be more sensitive to humidity changes (influence from the EPOXY PCB). Here heating may help to keep it more constant.
If you really want a long term stable system you will need a reference in a CERDIP or Metal CAN case.

with best regards

Andreas

[Martin72]

In principle, I would “only” have to exchange the reference and leave the rest as it is.
I would be interested to know what happened to the project:
https://bastelblog.runlevel3.de/en/measurement/ltz1000a-1ppm-precision-voltage-reference/

[dazz1]

Hi
Looking at the Elektor Reference design, the 20bit DAC looks to be a weak link.
The very expensive high spec version i accurate to 1.5bits, so it is really a 18.5bit DAC
The expensive lower spec version is 4.5bits, so really a 15.5bit DAC.

It would be possible to correct those errors using a 7.5 digit voltmeter and measuring the output voltage of every input value of each individual DAC.  Only a million or so.  Easy to automate. 
It would then be possible to create a lookup table that corrected every value.  Ideally it should be possible to get 0.5bit accuracy. 
That method would require the source code (to modify) and changes to the Elektor design.  Not really feasible.

If you had a 7.5 digit meter, you wouldn't need the Elektor voltage reference.  So as much as I would like to build the Elektor Reference, I don't think I ever will.

[Kleinstein]

Things are not that easy with checking to a DMM. The DAC is fast, but the DMMs usually need a few seconds to get there full nominal resolution. In addition one may have to add some extra points for the full scale (for the reference) and maybe zero to allow for drift there.
So 1 million points for a full test would some weeks. A problem is that the INL can be temperature dependent too. Things are likely better than spec at the ideal temperature where the DAC was adjustend / trimmed.

It may be feasible to check at lest the major bit steps, like the upper 4 or 6 bits with way less points and a managable memory. The internal correntions can already complicate things, as the steps may be fussed out a bit.

[mawyatt]

One could do a check and adjustment at the popular voltages such as 0.1, 0.5, 1, 2, 2.5, 5, 7.5, 10 Volts or whatever makes sense.

Best,

[Martin72]

The very expensive high spec version i accurate to 1.5bits, so it is really a 18.5bit DAC
The expensive lower spec version is 4.5bits, so really a 15.5bit DAC.

The AD chips used are all B-grade, i.e. the higher-quality variants.

If you had a 7.5 digit meter, you wouldn't need the Elektor voltage reference.

At 10ppm, this was never an issue.
If all goes well, you have a voltage available that is accurate to the 5th place.
You can calculate the multimeter resolution up to which this will suffice.
What is special about this reference is not necessarily the 5 decimal places or that you can set any value from 0V-10V.
Rather, that you can change the polarity and that a fairly accurate current of up to 40mA is available.
I was primarily interested in the technology behind it, how to build and calibrate something like this, and what is necessary to achieve stability.
Of course it is expensive, almost €600 to date, but I have already paid significantly more in my life for things I learned the hard way.

[dazz1]

Things are not that easy with checking to a DMM. The DAC is fast, but the DMMs usually need a few seconds to get there full nominal resolution. In addition one may have to add some extra points for the full scale (for the reference) and maybe zero to allow for drift there.
So 1 million points for a full test would some weeks. A problem is that the INL can be temperature dependent too. Things are likely better than spec at the ideal temperature where the DAC was adjustend / trimmed.

It may be feasible to check at lest the major bit steps, like the upper 4 or 6 bits with way less points and a managable memory. The internal correntions can already complicate things, as the steps may be fussed out a bit.

A DAC is a voltage divider so measuring the input voltage (occasionally) and the output voltage  would eliminate voltage drift errors.
It would be necessary to automate the DAC calibration.  It wouldn't really matter if it took weeks, or even months to check every value.
The correction look up table would only require 5 bits max (for low spec DAC) for each value to store the corrections.  Likely interpolation would further reduce memory requirements.

 
The whole discussion is moot unless someone would take on the task of writing the software given that the source code is not available to amend (at least I don't think it is).

[dazz1]

Of course it is expensive, almost €600 to date, but I have already paid significantly more in my life for things I learned the hard way.

Wow, €600 is almost 2x the current cost.  I like Elektor projects, but not enough to pay €600 or even €300 for this one.   It is not that it isn't worth that much money.  The software alone would cost much more to produce and the designer is not getting anything in return for releasing his project.    It is that I don't have a problem that this project would solve. 

I'd still like to build one.

[KungFuJosh]

I don't have a problem that this project would solve. 

I'd still like to build one.

Same. 🤣