Scanner/Multiplexers for voltage references

[Kleinstein]

This is an interesting construction to get low thermal EMF reed relais.

The contact block does not look that impressive: I kind of miss more area to thermally couple the incoming wires to the metal block.

[Andreas]

Hello,

the effort is absolutely necessary.

The wires through glass are most probably out of Covar with a high thermal EMF against copper.
And the wires are not closely neighboured.

with best regards

Andreas

[dietert1]

Meanwhile i started using the low thermal scanner i described above, with two 3457A meters connected to the output. I am monitoring one of the 10 V references in 30 V range and voltage differences in 10 mV range. Today i found that avoiding transient overvoltage to the meter can make a difference. This means:
- not using autorange
- setting the meter for a higher voltage range before switching the MUX
- switching the MUX before setting the meter for a lower voltage range

When measuring 10 V the average TC of the two meters is about 0.6 ppm/K. One thing i still need to study is the guard of the HP 3457A. It does have a cage around the floating part but no separate guard terminal. Maybe i can disconnect the cage from the low terminal and add a guard terminal.

Regards, Dieter

[dietert1]

Here i have some results. The diagrams show daily averages since June, 24. As one scanner round takes about 40 seconds, each average contains about 2270 * 2 = 4540 measurements of each voltage.

The setup includes three experimental voltage references: Two with LTFLUs, one with 2x JFET. The best agreement is between the two LTFLUs, the diagram is +/- 0.1 ppm now! Biggest day to day step observed was 0.06 ppm. The drift  of 17.2 nV/day between the LTFLUs continues as expected from an exponential relaxation fit i showed some time ago in the LTFLU thread. It predicts 18.1 nV/day.

The 2x JFET JVR i made in February still isn't tired of drifting. 120 nV/day amounts to 4.4 ppm/year. Noise appears to be low, though.

Regards, Dieter

Edit: 120 nV/day * 365 days/a gives 44 uV/a, i.e. 4.4 ppm of 10 V. Not as bad.

[branadic]

I was struggeling for quite some time, but I have now ordered G6AK-274P-STLT-US-DC12 to replace the original G6AK-234P-ST-US-DC12 in my Prema2080. Reason for the decision was that the unit already has everything I want: case, display, GPIB interface, 40-channels, ...
So we will see if that is improving things further and fingers crossed it does. I'm not expecting too much, but you never know.

-branadic-

[branadic]

Yesterday the new relays (G6AK-274P-STLT-US-12) as well as a desoldering station arrived, so I spent the whole evening to desolder all of the original relays (G6AK-234P-ST-12) and cleaned the board (first IPA, afterwards Ethanol).
This morning, after some coffee break, I've soldered all the new relays in place and cleaned the back of the board. The unit is now back together, working and warming up. Hopefully I can manage to start a measurement on it and see what has changed. I'm not expecting any miracles though.

-branadic-

[Anders Petersson]

... cleaned the board (first IPA, afterwards Ethanol).

If you forgive an off-topic question, what's the advantage of combining the two cleaning agents? A basic Google search didn't help.
IPA is readily available but access to pure ethanol is regulated so perhaps some denatured version is acceptable?

[branadic]

IPA disolves most of the dirt, but sometimes leaves a white film on the top when evaporating. Ethanol cleans the board much better, at least that is my experience. Pure ethanol can be bought at the drug store, at least here in Germany.

-branadic-

[e61_phil]

I also experienced this white film after cleaning with IPA.

To investigate if it really came from the IPA, I filled a lot of IPA in a test tube and let it evaporate completely. There was nothing left in the end (no white film).

It seems that this film isn't directly from the IPA.

[CDN_Torsten]

I've always assumed that the white film remaining after an IPA wash was caused by small amounts of water in the IPA (which reacts with the flux)....but I'm not 100% certain.

[branadic]

Why is there white residue when I clean my PCB?

White residue is generally a symptom of ineffective PCB cleaning. Common conductive flux residues from the soldering process can include various unreacted activators, binders, rheology components, and saponifiers. Among these are numerous iterations of acids (abietic, adipic, succinic among others), highly basic ingredients (amino compounds), and even constituents found in “soaps” such as phosphate and sulfate ions. When a cleaner does not fully dissolve all the constituents, or the cleaner is not allowed to flow off the PCB, the remaining solvent can evaporate off and leave behind residue that is either white or like water spots.

Source: https://www.chemtronics.com/why-is-there-white-residue-when-i-clean-my-pcb

However, I only have that issue from time to time with IPA, not with Actenone or Ethanol. As Acetone can melt some thermal plastics I try to avoid it on such boards, but go IPA and Ethanol instead. Works best for me.

-branadic-

[branadic]

So here is the measurement of all the 40 channels, one after the other. Some improvement, but not as good as a Dataproof. Some of the channels are quite close to each other though, thus by selecting the proper channels the influence of the scanner can be negligible .

-branadic-

[maat]

I figured the white residue are salts left from whatever acid there is in the flux, therefore IPA does not dissolve it. When I wash the boards by hand I typically use destilled water first to get those water soluble salts and then I throw the boards in IPA and let them dry (in the liquid). Then I give them a good brush. Works every time. What works even better is a good ultrasonic cleaner at 130 kHz and 60 °C. Don't go for the cheap chinese rubbish. Been there, done that.

But you have to be careful when using an ultrasonic cleaner, because some parts just don't like that. ALPS rotary encoder for example. I don't know what kills them, but they die. Every time so far...

[ramon]

I remove the white residue just by rubbing it with a cotton rag  (for small areas a few cotton buds also works, but is not as good). The trick is to do it 10 or 15 seconds after applying IPA (maybe because the PCB still has some water content and helps it to adhere into the cotton rag?. )

I guess that this trick just removes the visible white residue but doesn't remove it 100%. Probably not good for ultra high impedance or femtoampere currents. In that case, what branadic proposed would work best.

[dietert1]

So here is the measurement of all the 40 channels, one after the other. Some improvement, but not as good as a Dataproof. ...

For scanning voltage references low leakage is a pointless exercise (and it doesn't improve on low thermal EMF). Why don't you try to understand the solution i proposed and demonstrated. Some people successfully copy things even without understanding all the details. Better than buying old stuff of unknown origin. I mean exactly with relays you don't want the old ones that already executed millions of cycles.

Regards, Dieter

[branadic]

Why don't you try to understand the solution i proposed and demonstrated. Some people successfully copy things even without understanding all the details.

Don't get me wrong dieter1, but the way you communicate is really weird and strange. You act like you know it all and I'm pretty sure that is not the case.

Better than buying old stuff of unknown origin. I mean exactly with relays you don't want the old ones that already executed millions of cycles.

Carefully reading what I wrote before reveals, that the relays have all been replaced, so they are in new condition directly from a valid distributor. And you don't seem to understand, that I like all my gear with the very same interface and as all of my equipment has GPIB interface, I like the scanner to have that too, so that all of it can be connected to the same bus, without multiple different cables for different interfaces running all over the place. If you for yourself have a different view on that, that is your very own opinion and I leave it like that. Please accept that other people have different preferences. Thanks.

e61_phil suggested to run the scanner in 4 wire mode and to use the sense lines only, as routing is more even, so I will give that a try next. Thanks for that suggestion.

-branadic-

[Kleinstein]

The problem with the prema board is that there relatively hot driver part is relatively close to the relays.
Even if old, the scanners are often not used for that many cycles and if so, often just a few channels.

I think it is mainly the software / interface part that makes it attractive to reuse the old PCB and not start from scratch, like Dieters design, which looks better and also has better measured results.

If the scanner is 4 wire (e.g. 2 sets with 2 poles), one could chose a better suited set of contacts, and even mix the 2 sets.
Chances are much of the thermal EMF effect is relatively stable. So it would not hurt measurements of drift and noise vey much.

[branadic]

I think it is mainly the software / interface part that makes it attractive to reuse the old PCB and not start from scratch, like Dieters design, which looks better and also has better measured results.

If the scanner is 4 wire (e.g. 2 sets with 2 poles), one could chose a better suited set of contacts, and even mix the 2 sets.
Chances are much of the thermal EMF effect is relatively stable. So it would not hurt measurements of drift and noise vey much.

It works better, agree, but only after some iterations and its form follows function (ugly Hammond case). If that works for him, fine.

But experience is not about simply copy&paste, but about making experiements yourself, taking some traps and improve where you spotted an issue and it's hobby to see, were the limits of improvements are, even on old test gear. That way and only that way you can develope a gut feeling, not by reading and simply copying. And noone should explain himself, why not to simply copy some others design, but share results and also drawbacks during the way.

Otherwise we could all wait until someone else made a solution or buy what is already out there, stop hobby and buy what others came up with. Nothing I like though.

-branadic-

[dietert1]

My remark about old relays was a response to your mention of dataproof scanners. Somewhere in this forum you can find a report of TiNs attempt to fix an old dataproof scanner.

Regards, Dieter

[branadic]

As suggested I've measured TEMF of the Prema 2080 in 4W mode, first the sense channels and as a consequence also the force channels. Results attached.

-branadic-

[dietert1]

branadic was also wrong with his comment on the aluminum enclosure i used, it is no Hammond.

That is one of the details i did not write down until now. The case is a RND 455-00396 IP65 enclosure. It was selected as it is moisture proof (rubber seal). And its wall thickness is such that it serves as a "temperature equalizer". I am using it inside an outer case made from MDF boards plus 20 mm of styrofoam isolation. The MDF box is almost air tight, too. There are no connectors but pigtails long enough to pick up input signals at their origin. The output cable directly feeds the DVMs.
That RND aluminum case is available for € 12 at Distrelec, order code 300-64-617. For € 20 they also have a painted version.

Regards, Dieter

[guenthert]

    There are different approaches.  I was a bit surprised seeing a national lab using rotary switches actuated by stepper motors in what superficially looks like a home-made robot kit 

    They do put quite some effort in assuring low thermal EMF across the switches, a thick copper base and the bane of home construction, Cd based solder: https://www.bipm.org/documents/20126/27085544/bipm+publication-ID-2250.pdf/0b48f272-169c-c175-72df-6aa670c4b763?version=1.7&download=true

[dietert1]

Above i recommended trying a TI ADS1256 with its built-in muliplexer.
Meanwhile i learned that there is an improved part ADS1263 and today there are first results with a raster board test circuit including a MAX14483 SPI isolator and +/- 2.5 V voltage regulators for the analog front end. The setup includes another board with a linear +/- 3.3 V isolated power supply and a STM32F429I-DISCO.
I put the ADS1263 board into a plastic bag to reduce air draft. The on-chip scanner exhibits thermal offsets with a standard deviation of 45 nV over several channels with low thermal shorts. The offsets are very stable with a standard deviation of 9 nV, so the scanner is already useful to this level. May improve yet when i put the board into a constant temperature metal enclosure similar to the HPM7717.

Regards, Dieter

[Kleinstein]

The standard CMOS MUX chips also behave reasonable as a low voltage MUX. They have low power which makes it realtively easy to get small temperature gradients.  A problem is however that they have little protection, so ESD or to high voltage (outside the supplies) may damage them. Especially old types are supposed to be relatively susceptible to latch-up.

[dietert1]

The ADS1263 is much more than a MUX. Next step is to read the on-chip temperature sensor and see how it correlates with the residual offsets. Maybe it's possible to calibrate away large parts of the offsets

The numbers above are with the PGA at G=2. Then, using the on-chip 2.5 V reference i have an input range of +/- 1.25 V. A 1 V test voltage divided from the +2.5 V supply gets measured with a standard deviation of about 2 or 3 ppm. The on-chip reference is supposed to be good for 1 ppm.

Regards, DIeter