Hysteresis with LM399 and ADR1399

[NaxFM]

Hi everyone, I'm trying to develop ad ultra stable reference using the ADR1399, I've been studying circuits for weeks, even those using the LTZ1000, and I'm getting everything figured out, also thanks to countless simulations (I can't believe how useful LTSpice is)
However, there's one thing that bothers me: hysteresis effects. I know that to have an extremely stable voltage one must keep the device powered on like the famous Fluke 732 standard, but that would be very inconvenient.
I was wondering if you happen to have any information on the hysteresis effects with the LM399 or the ADR1399, especially the effects of repeated power cycles on the long and short term stability of the zener.
I have already ordered an evaluation board to do some experiments, but i really don't have the equipment right now to start counting ppms. All I have is a calibrated 34401A, but soon i'll buy a DMM7510 or even a 3458A if I find a good deal.

[Andreas]

Hello,

cannot tell for the ADR device since I have no measurements up to now.
Especially if you want to use the SMD version which might have a influence from the PCB due to humidity swelling.

My LM399 have rather low hysteresis during normal operation at room temperature.
Even a power outage of ~10 days (see around day 2840 in the diagram) does not change much.

The larger problem is to sort out the devices with "popcorn noise". (See the dark blue device in the diagram).

with best regards

Andreas

[NaxFM]

Hello,

cannot tell for the ADR device since I have no measurements up to now.
Especially if you want to use the SMD version which might have a influence from the PCB due to humidity swelling.

My LM399 have rather low hysteresis during normal operation at room temperature.
Even a power outage of ~10 days (see around day 2840 in the diagram) does not change much.

The larger problem is to sort out the devices with "popcorn noise". (See the dark blue device in the diagram).

with best regards

Andreas

Thank you! This was exactly what I was looking for.
It looks good, except for the blue one which is really bad.
Interesting to see that some of the little spikes occur at the same time for all the devices.
Do you have any explanation for that?

[Andreas]

Hello,

no measurement instrument is perfect.
So probably some common mode effect like popcorn noise / (not completely compensated) temperature drift on the measurement instrument.
Standard deviation of the instrument used (ADC13) is around 0.25 ppm for a 7V reference.

with best regards

Andreas

[Chance92]

Hello,

cannot tell for the ADR device since I have no measurements up to now.
Especially if you want to use the SMD version which might have a influence from the PCB due to humidity swelling.

My LM399 have rather low hysteresis during normal operation at room temperature.
Even a power outage of ~10 days (see around day 2840 in the diagram) does not change much.

The larger problem is to sort out the devices with "popcorn noise". (See the dark blue device in the diagram).

with best regards

Andreas

Hi Andreas. Are these measurement taken automatically?

[Andreas]

Hello,

yes, with the help of a analog (differential) multiplexer.
https://www.eevblog.com/forum/metrology/measurements-on-emf-error-of-switches/msg1254204/#msg1254204

with best regards

Andreas

[MegaVolt]

but soon i'll buy a DMM7510 or even a 3458A if I find a good deal.

https://www.eevblog.com/forum/buysellwanted/fs-keithley-dmm7510-complete-with-cal-data/

[rf-design]

My guess is that reason for reference voltgae hysteresis is a stored stress in the dia attach glue.

https://www.eevblog.com/forum/metrology/ultra-precision-reference-ltz1000/msg408645/#msg408645

The stress will impact the bandgap voltage of the NPNs. This lead to a change in forward voltage and is well known. The effect on the buried zener and the regulation circuit is most likely that either the sum of NPN-Vbe plus zener is modified or the temperature setpoint and indirect over the remain temp-coeffcient the reference is changed.

A thermal isolated mounting with radiated instead of conducted heat transfer would eliminate this effect. But this is not cost effective.

Another solution to instead of thermal cycle over 20-80K after poweron is to live with an operating point where you do no hit the zero tempco and instead calibrate. Then temp cycling is only ambient which you can low-power record in standby to judge the accuracy. The heater set circuit is nothing less that a 70s non-compute solution to overcome the temp-effects.