Ref81

[branadic]

Hello,

while Ref80 is more or less a raw burried zener, Ref81 is the more compatible competitor to ADR1001 in terms of functional capabilities.
Source of images: https://www.ti.com/tool/de-de/REF81EVM





Nevertheless, specs for noise of the zener core slightly increased, as given by the preliminary but yet incomplete datasheet available on the TI website. For Ref80 it is specified to 0.12 ppmpk-pk (912 nVpk-pk) and 0.15 ppmpk-pk (1.14 µVpk-pk) for Ref81. That is at least similar to LTZ1000 specifications with 1.2 µVpk-pk, but still worse compared to ADR1000 (900 nVpk-pk) or ADR1001 (600 nVpk-pk).
It just so happened that I got my hands on a Ref81EVM, populated with an engineering sample of this new device (marked with X). Unfortunately, the binding posts attached to the board are Pomona 3670, so the nickel plated brass version. Hence why the first thing to do was to replace them by proper TBP1 copper binding posts that I had available thanks to the group buy and forum member pipelie.
For the last ~100 h I have monitored the raw zener output with a 3458A and 24 V from a lab power supply applied to both LDOs for heater (LM317) and zener (TPS7A49), drawing about 1.4 W of power after thermal stabilization within the cardboard it came in.
We currently don't have any specifications for the long-term drift due to the incomplete datasheet, but can only compare to the specs of Ref80, which states typ. 3 ppm for the first 1000 hours and 0.3 ppm for 1000 - 5000 hours.
The reference specimen on this board lost about 14 ppm within the first 22 hours, another 2-3 ppm on the second day and ended up drifting by -18 ppm in the first 100 hours and currently reads 7.58588 V.
I'm not judging the results, as initial drift is to be expected, I just present them as they are and keep going observing how that drift developes over time.

-branadic-

[branadic]

An update, first 200 hours of contiuous operation.

-branadic-

[branadic]

An update, first 300 hours of continuous operation.

-branadic-

[MiDi]

Any explanation for 250h and 300h sudden jumps?

[CalibrationGuy]

Now that the drift is much lower and will continue to get lower as time passes, do you have an aged reference like a 732 for comparison to continue to monitor the drift, both raw and 2x configuration? The 3458a will probably not be as useful in the sub 2ppm range.

Also, what is your opinion of this device vs. the ADR offering?

TomG.

[branadic]

Any explanation for 250h and 300h sudden jumps?

I can neither explain the sudden jumps nor the two level output that showed up since ~245 h.

Also, what is your opinion of this device vs. the ADR offering?

I won't judge nor will I draw any conclusions, I leave that up to you.

-branadic-

[Doctorandus_P]

Any explanation for 250h and 300h sudden jumps?

The dip @140h correlates to a dip in humidity. the "ramp" at 190h also has rapid humidity changes and both the 255h and 300h are close to humidity changes too.

This looks like enough correlation to become suspicious.

[branadic]

I don't fully agree with the explaination.

-branadic-

[rikkitikkitavi]

I agree, the changes in humidity is small considering the temperature swings is 23+-1C or  so.

There is a delay between humidity and impact on chip aswell, otherwise Minitabing the data could give some insight to anny correlation.

I think it is just the chip still adjusting to its new life and some measurement artifacts (noise?)

Bur who knows?

I am mostly impressed you measure ppm level changes. Lucky I am not a voltnut :)But this is one of the most interesting subforums on Eevblog. The dedication to the task...

[branadic]

About 1600 hours of continuous operation.

-branadic-

[miro123]

Thanks for sharing this information. I have a few questions, and I apologize if these have already been thoroughly explained elsewhere in the forum, as I couldn't readily locate the specific details.

1. What is the confidence interval for the presented data?
2. Could you clarify the type of data shown in the graph? Specifically, does the graph represent raw data, or has any aggregation or filtering been applied?
3. Could you provide more technical background on the setup used? This would include details such as specific settings, aperture time, and data rate.
Thank you,
Miro

[branadic]

The readings are taken at the Ref output of the evalboard using a 3458A @ NPLC100, that performs ACAL every morning at the same time. The evalboard is sitting inside the cardboard it came in and is powered by a HMP2030.

-branadic-

[branadic]

In the past weeks we had issues with the A.C. in the lab and outside temperatures went up to 40 °C. It is fixed now and lab temperature stabilizes again. So the Ref81 measurements are superimposed by temperature changes. However, meanwhile the reference seem to have stabilized.

-branadic-

[miro123]

Thanks for sharing.
My interpretation of the graph. After 2000Hours the 7V stabilizes and we are looking at daily HP3458 ACALs.

[guenthert]

Superficially looking at the graph, I get the impression that the output voltage seems fairly independent of average ambient temperature, but a bit sensitive to large, sudden temperature changes.  Some mechanical stress perhaps? Or Seebeck?

[branadic]

For ~600 h the A.C. in our lab is working again. Interestingly, since then also humidity seems stable. You might notice a small jump at time = 3500 h, that was me cleaning the mesh filter of 3458A, which not only improved internal temperature to stable 36 - 37 °C, but also improved measurement stability of the reference.
Today I switches cables to the 5 V output and started monitoring that. Unfortunately, the evalboard does not allow to easily set the reference to 10 V, the board designer was smart enough to connect pin 9 (BUFF_INV) and pin 12 (BUFF_OUT) underneath the chip on the top layer.  So in order to modify the board for 10 V, the reference chip has to be removed from the board.

-branadic-

[iMo]

3 pages long DS..

[branadic]

Roughly 200 hours of the 5 V output.

-branadic-

[branadic]

I stopped logging the 5 V output today, removed the reference from the board, removed the aforementioned trace underneath the package, resoldered the chip and modified the output for 10 V.
At the moment I'm logging the Zener voltage again to see what impact it has that I was messing with the reference. Once that is clear, I'll switch to the 10 V output.

-branadic-

[CalibrationGuy]

That is an interesting experiment. I am very curious to see how desoldering and resoldering a chip after it has aged for a while will affect its output. I am following your efforts with great interest.

Best,
TomG.

[branadic]

Here are the first 120 hours after my solder job, which had quite the impact.

-branadic-

[branadic]

While I was enjoying my time at the CPEM2026 in Madrid/Spain, meeting some of the guys you might have heard of or read about, Ref81 had time to drift further.

-branadic-