LT1021CMH based mini-reference

[Alex Nikitin]

Just looking at the LT1021 datasheet, I hadn't realised there was an LT1021-7 which does not include internal scaling resistors (or trim pin). The thing that surprised me is that its long term drift is quoted at a mere 7ppm/kh typ versus 15ppm for the other variants, marginally lower than the LT LM199 spec (8ppm/kh). Presumably this is for the H package only.

I'm just curious that it's the lowest drift I've seen quoted for any IC reference. It's also quieter than the LM199 and with the benefit of a buffered output. Ok, the TC is higher than the LM199 but much lower than an unheated LM329. It looks as if it could work very well in a low temperature oven. Anybody tried it?

You need to be careful about datasheet figures, it is not something to rely on from my experience, especially in regard to the noise, tempco and long term stability. At best it is a rough guide. My examples of LM1021CMH are relatively noisy (noisier than a good LM399), with a rather large tempco (~10-15ppm/C) though the long term stability and hysteresis after one month appear to be very good. I have chosen this chip for a travelling reference mostly because it provides 10V with internal (i.e. placed in a hermetic case) resistors, so all critical bits (the reference, the step-up network and the resistor reference/divider) are in a hermetic packaging. OK, it is relatively noisy with 1ppm p-p of a pop-corn "jumps" but the long term stability appears to better than 1ppm/month (how much better only the time will tell) and it proved so far to be quite insensitive to travelling and temperature cycling from 5C to 40C.

I have ten LM129AH in a hermetic packaging, perhaps I'll build the next reference with that one (or 2-3 in parallel, or maybe 3 in series for ~20V main reference voltage) and see how that will work. Another option is my JVR metal can JFET based reference, which appear to be very stable (less than 5ppm drift over more than a year, essentially less than I could measure reliably) and has a low tempco around room temperature. It does rely upon some good resistors though and doesn't have a particularly low noise (~0.5ppm p-p).

Cheers

Alex

[Gyro]

You need to be careful about datasheet figures, it is not something to rely on from my experience, especially in regard to the noise, tempco and long term stability.

Oh absolutely. It just struck me as strange that LT would go to the 'trouble' of stating half the typical long term drift for the 7V version rather than just sticking with the more or less universal 15ppm/khr. It effectively implies that [Edit: the ratio drift of] the on chip resistive divider used on the other variants contributes typ 8ppm/khr, not something you would expect them to want to draw attention to.

I was wondering how your JFET reference was doing, I hadn't seen an update for quite a while. It sounds way better than your initial expectations.

[Alex Nikitin]

The reference is on its way to one of the forum members in Netherlands, anybody else would like to be next to use/measure it?

Cheers

Alex

[SvanGool]

The reference is on its way to one of the forum members in Netherlands

That is me and I just picked it up today. 

Two questions for Alex:

• Did you calibrate at 23°C or another temperature?
  
• Is R5 (200k) a low tc resistor, I saw it has quite some impact when temperature changes?

[Alex Nikitin]

The reference is on its way to one of the forum members in Netherlands

That is me and I just picked it up today. 

Two questions for Alex:

• Did you calibrate at 23°C or another temperature?
  
• Is R5 (200k) a low tc resistor, I saw it has quite some impact when temperature changes?

1) Yes, 23-24C ambient. It should be written on the label.

2) 200K is MELF 0.1% , ± 15ppm/°C . It should not have any noticeable influence on the output  v temperature. The temperature compensating diodes are right next to it though.

Cheers

Alex

[SvanGool]

I received and unpacked it, thanks a lot Alex !

I did some first experiments with the traveling standard with my HP 34401A (close to 1 year cal), Alex used a HP 3458A 002 for measuring the travelling standard, before sending it.

Questions popping up, I am trying to imagine the possibilities of this travelling standard and I'm still learning:

• What does it mean if my HP 34401A is displaying 10.00038k (23C) when the travelling reference has 10.00045k on the label? Is this HP 34401A 10k range at least 10ppm (HP 3458A 002 accuracy) + 25ppm (0.0020% + 0.0005%, HP 34401A 24h accuracy) + 3ppm (travelling standard 1ppm noise + 2ppm 1C temp. accuracy) + 7 ppm (difference in meter readings) = +45/-31 ppm accurate  for the next 24 hours?
• Would it mean, that if I would calibrate (assuming the travelling standard was stable since its first check with the HP 3458A 002) the HP 34401A with the travelling standards 10K and enter 10.00045 as calibration value, in that range, that the HP 34401A would have a minimum accuracy, for that 10k range, of 10ppm (HP 3458A 002 accuracy) + 25ppm (0.0020% + 0.0005%, HP 34401A 24h accuracy) + 3ppm (travelling standard 1ppm noise + 2ppm 1C temp. accuracy)= 38 ppm for the next 24 hours, in stead of the current 110 ppm (HP 34401A 1 Year 23C+/-5C accuracy)-25 ppm(24h 23C+/-1C ) = 85 ppm more worse minimum accuracy I have now?
• What would this mean for the HP 34401A 90 days and 1 year accuracy after 24 hours and 90 days, is it correct that e.g. the 1 year accuracy would become 110+10+3=123 ppm in stead of the 110ppm when I would officially re-calibrate the unit? So the price of using the traveling standard (in stead of officially calibrating) would be 13ppm for the involved ranges, but the calibration times would be reset again, for those ranges.

Of course (for now when we don't know the exact behavior yet of the traveling standard) I would have to do the above exercise with my own LTC1021 voltage reference, VHP202Z 10K000 0.1% resistor and VHP202Z 1K0000 0.05% resistor, that I would measure simultaneously and would use those ones for calibrating when Alex has verified the standard completely again, on return. To mitigate the risk that I would ruin my current calibration, for those ranges, when the standard seems not to be stable upon return, assuming that my own references and meter have small aging drifts meanwhile.

[ap]

Without having verified the detail ppm levels you state for each of the meters and cases, your logic is right. If you have a travelling standard with a certain total uncertainty (its cal value uncertainty, its possible temperature drift error, its aging related drift, anything else such as e.g. noise from surrounding; you need to establish all that), then you can use this reference to calibrate your 34401, and going from there, t0 starts again for your 34401. The uncertainty of that cal resistor adds to the time-depending uncertainties of the 34401.
Be carefull when establishing the basic uncertainty of the resistor (you say 10ppm). The 3458A opt 002, btw, is no better doing resistance measurements than a standard 3458A. The 1 year 3458A uncertainty would be 10+0.5+3ppm absolute, if a standard cal was done as per cal manual of Keysight. They do have calibrations that are more precise than the 3ppm though. But if not done at Keysight, it may as well be worse. I have seen people calibrating a 3458A with a Fluke 5500...

[lars]

Hello Alex,

I like your small box. Very clever use of the 10+1.111k with the switch! Even if I prefer NTC compensation,  diodes also works very well . 
Your idea to send it around to get statistics is very good I think. Guess I should PM you.

Hope the discussion on uncertainty will give good answers. That is one of the most interesting things for me. Having no volt nut friends nearby or access to sufficient good calibrated equipment getting low uncertainty is a problem for me (and many others).

In December I think I wrote a long post in the Calibration club thread but hesitated to send it. TiN´s questions (provocation) about uncertainty really is important. Using GUM (Guide to Uncertainty of Measurements) for DIY is probably as useful as for my professional work, but how to get statistics???

As far as I understand from my professional work: When I send eg a 34401 for calibration the uncertainty for the next 1 year will be the specification assuming the calibration lab says "pass"! An interesting discussion I have had with myself for my hobby is what say 34401 10k resistance that is off 7ppm at latest calibration and the last years have been say 10, 4, 15ppm. Can I conclude something?? (Of course I have characterized the aging, temperature and humidity sensitivity (takes long time)

Lars

[Kleinstein]

With many parts the drift follows a simple curve, often getting slower the older the part and moving in one direction. With some parts (e.g. voltage ref. chips) it is even known from experience in which direction and about as fast they usually drift.  With high end parts, there is however the problem, that the readings during calibration also has some uncertainty and this may be more like random scattering (if it was a known trend, they should have corrected it). So you instrument history reflects both the instrument aging and the scattering and maybe drift of the calibration lab(s).

For a meter send in to calibration, one has to distinguish between an calibration with adjustment and a pure check / reading. You only get the tighter short time specs if there was an adjustment or if you use the actual readings during calibration for correction. The direct reading from the display are not getting better from just a calibration that says still in specs (which?) without an adjustment.

With some instruments (especially high end ones) it is also that the quality level of calibration can make a difference - not every calibration will bring you back to the stated short time specs.

[SvanGool]

I finished my measurements on Alex's LT1021CMH based "Travelling reference":

Equipment used:

• 6.5 Digit multimeter HP34401A:
• Last calibration: Dec 2014 (16 months over 1 year calibration accuracy)
• Settings: 10GOhm DCV input imp., 100PLC, manual range, 4-wire resistance measurement
• Specified accuracy DCV  10.00000 V:           40 ppm (1 yr)
• Specified accuracy DCV  1.000000 V:           47 ppm (1 yr)
• Specified accuracy Res. 10.00000 KOhm:  110 ppm (1 yr)
• Specified accuracy Res. 1.000000 KOhm:  100 ppm (1 yr)
• Oregon temperature meter TWR800: accuracy 0.5°C
• Racal Dana 9232 Lineair Laboratory Power Supply:
• Output voltage: 15.16 V
• Load regulation: <0.01%
• Ripple and noise: < 1 mV
• Output impedance: <5 mOhm @ 1 kHz

Result:

• 10 VDC:    09.99999 V (@23.3 °C), HP3458A calibration: 10.00000 V, difference -1 PPM
•   1 VDC:    1.000033 V (@23.3 °C), HP3458A calibration:   1.00004 V, difference -2 .. -12 PPM
• 10 KOhm: 10.00037 K (@23.5 °C), HP3458A calibration: 10.00045 K, difference -8 PPM
•   1 KOhm: 1.000093 K (@23.5 °C). HP3458A calibration:   1.00010 K, difference -2 .. -12 PPM

Recommendations:

• Also specify the calibration for a lower temperature (e.g. 20°C), it was difficult to reach 23 °C in the cold North  .
• Assess why it was not possible to get a stable 6.5 digit reading for the 1.000000 VDC and 1.000000 KOhm ranges on the HP3458A (on the HP34401A on "slow 6 digit", it was no problem).
• Consider to include wiring, to exclude measurement error addition by the wiring.

Conclusion:

if Alex verifies the calibration still to be intact on return of the reference at his place, it proves that the reference is behaving excellent and I can conclude that the accuracy of my equipment would be within 124 ppm or less, for these ranges again for the next year, if I would calibrate it now (see also my previous posting).

I find this a very good result and it looks very promising for achieving the goal of this travelling standard.
Thanks a lot Alex, very good job!

Large fast picture of measurements.

[Alex Nikitin]

Thank you, good results!

Cheers

Alex

[Squantor]

Hello Alex,

I am measuring the traveling standard as we speak. I am doing measurement runs but I am yet to bring my logging facility up.

I am testing with the following requipment: HP34401A, Two Keithley 2700's and a Keithley 2000. Too bad all their calibration statuses are unknown. So I have no idea what my contribution will be, maybe just academic interest . I am measuring everything in my server room where it is a balmy 26 degrees centigrade, this is measured with a Klockis Ikea multifunctional clock, alarm, timer, temperature measurement thingy.

More information coming up.

[The Soulman]

How is the reference doing?
I'm interested in the drift caused by shipping, temperature changes, etc.
Only Alex can tell...

[Squantor]

I was really busy at work and other private matters. While measuring with my hp34401 calibration date unknown I noticed at 23C it was spot on with the reference! Very impressive! The cables that SvanGool made worked very well too. Did some quick tests with thermal EMF's with my recently repaired hp419a. It was about 200nV but not logged it for a longer period. I wanted to compare it to a battery of keithley 2002's that we have lying around but I could not find time due to overtime I have to put in these weeks.

I will arrange with SvanGool to return the reference and first work on my own measurement setup to enable automated measuring and make a few extra references based on this design. Then if Alex is willing reship it to me to compare and measure again, maybe with the keithley 2002's from work.

[SvanGool]

Hello,

I got the "Alex" LTC1021CMH reference back from Squantor and measured again with the same setup (@23.4 °C, 15.16V input, HP34401A 10GOhm DCV input imp., 100PLC, manual range, 4-wire resistance measurement)
as here: https://www.eevblog.com/forum/metrology/lt1021cmh-based-mini-reference/msg1172332/#msg1172332 after three hours warming up and same cabling.

The results, 5 weeks after the first measurement:

• 10 VDC:    10.00001 V, +2 PPM on first measurement,         +1 PPM HP3458A
•   1 VDC:    1.000033 V,   0 PPM on first measurement, -2 .. -12 PPM HP3458A
• 10 KOhm: 10.00032 K,  -5 PPM on first measurement,         -13 PPM HP3458A
•   1 KOhm: 1.000089 K,  -4 PPM on first measurement, -6 .. -16 PPM HP3458A

Still not bad at all.

I will now send the unit back to Alex in the UK, who will get his (re-calibrated ?) HP3458A also back this week 

[The Soulman]

And? 

[Alex Nikitin]

And? 

The reference is not back to me yet.

Cheers

Alex

[SvanGool]

But it will be this week 

I kept it to check also a HP3457A, I just repaired, which took a bit longer then expected, the last results:

• HP34401A (last calibration Dec 2014):
• @24.1 °C: 10.00005 V
• @24.1 °C: 1.000041 V
• @24.1 °C: 10.00023 K
• @24.1 °C: 1.000081 K
• HP3457A (last calibration Dec 2013):
• @23.7 °C: 10.00023 V
• @23.7 °C: 1.000047 V
• @23.8 °C: 10.00047 K
• @23.8 °C: 1.000084 K

Comment: the HP3457A has a nonadjustable 10 MOhm input impedance on the >=10V range and >10GOhm on the <10V range. On the HP34401A I used the >10 GOhm setting on all ranges.
I'm curious for your HP3458A results.

[The Soulman]

Thanks, looks like this little reference is holding up pretty well.
I'm also interested to see the results on the 3458, but is has to cross the pond first...

[Alex Nikitin]

Comment: the HP3457A has a nonadjustable 10 MOhm input impedance on the >=10V range and >10GOhm on the <10V range. On the HP34401A I used the >10 GOhm setting on all ranges.
I'm curious for your HP3458A results.

Thank you. In this case the input impedance at 10V does not matter, as the output impedance at 10V output is a fraction of an Ohm, and at 1V output the output impedance 1K, however both meters are at a high impedance input at that range.

Cheers

Alex

[Alex Nikitin]

Yesterday I've received the reference back and today took it to the lab and after 2h warm-up measured it with the freshly calibrated (9 May 2017) HP3458A Opt 002 (after ACAL ALL):
Here are the results @24C:

10.000043V
1.000044V
10.00045K
1.000096K

The wiring offset was less than 0.5uV. The popcorn noise "jumps" in the voltage are about 0.5ppm, as before.

It looks like the reference drifted about +4ppm on 10V and 1V, with no visible drift on the resistance values (less than 1ppm) over the time period of about four months and some international travel  .

The calibration results for the HP3458A show +1ppm error on 10V and 1V DC gain.

Cheers

Alex

[Alex Nikitin]

Well, the 1010B reference is now available on loan to the forum members in the UK and Europe, the only conditions are to pay postage and to return the reference in a good state in 2-3 weeks time.

The reference provides stable 10V, 1V, 10K and 1K - measured with better than 10ppm accuracy. In the UK I can provide a good 24V DC wall-wart supply, otherwise a reasonable quality 15-30V DC supply is required.

Cheers

Alex

P.S. Update - the reference is booked for the next two weeks or so. Please PM me if you would like to join the queue  .

[Alex Nikitin]

Another update after the reference has traveled to Italy and back. Here are the measurements results, using the same HP3458A Opt 002 after ACAL. The lab temperature was 24.5C.

10.000040V
1.000043V
10.00044K
1.000095K

The reference is again available.

Cheers

Alex

[Alex Nikitin]

Here is a 3 hour run today in the lab for 10V output. 1ppm/div vertical scale

Cheers

Alex

[Alex Nikitin]

The reference is booked for another two weeks.

Cheers

Alex