ADR1399 - New/Old Voltage Reference

[MikeP]

 I believe that many people are already aware of the existence of an improved version of the LM399. It can be bought!
 Therefore, I ask the respected community to evaluate the possibility and expediency of using a ADR1399 instead LM399 in existing devices.
 Thanks.

[bob91343]

I read the data sheet and am confused regarding its application.

First, the absolute accuracy of the zener voltage is poor.  So it has to be measured with an accurate voltmeter before use.

Second, the data sheet shows a shunt resistor in series with the zener.  No clue provided how to select said resistor.  The zener current is variously specified at 3or 13 mA.

How to use this in a voltage calibrator?

[ocw]

It was nice to see the ADR1399 finally show up at Arrow today for small quantity purchases.  Not as nice was the $30.14/each price there as compared the $12.28 price if you purchased 18 of them direct from Analog.  Arrow had 30 of them this morning.

[Cerebus]

I read the data sheet and am confused regarding its application.

First, the absolute accuracy of the zener voltage is poor.  So it has to be measured with an accurate voltmeter before use.

Stability is the feature here. Initial accuracy isn't necessarily a particularly useful characteristic if it's going to change with time or with temperature.

Second, the data sheet shows a shunt resistor in series with the zener.  No clue provided how to select said resistor.  The zener current is variously specified at 3or 13 mA.

Ohms law? As to how to pick the current, try looking at figure 3 on the datasheet, that should make it abundantly clear that any constant current on the flat portion ought to be acceptable. In practice, one would put it inside a closed loop with an op amp where the zener voltage would provide the reference for a constant current source.

How to use this in a voltage calibrator?

That's a much more involved question.

[NaxFM]

First, the absolute accuracy of the zener voltage is poor.  So it has to be measured with an accurate voltmeter before use.

Even the mighty LTZ1000 used in the best multimeter in the world has very poor accuracy, however it's extremely stable.
It doesn't matter if the output is 7.034523V or 7.245321V, what really matter is that if the output at one given moment is 7.245321 it STAYS precisely at that value no matter what.
Once we know the voltage is stable, we can compare it with calibrated voltage standards to get the exact number.

Every multimeter and voltage reference works this way, you can't just built it and expect to be accurate, at most you can expect to be precise (that is it does't change its value)

[NaxFM]

Few days ago i read the datasheet and it seemed very nice, expecially for the much reduced dynamic impedance with respect to the LM399, this way it can tolerate much higher zener current variations.
It seems better than the LM399 in every way and from just the datasheet it even seems it can rival the LTZ1000 while being way cheaper

[MikeP]

 My idea is not understood. Obviously, the new product is significantly superior to its predecessor. Yes.
 But. The question is this: we install an aged new reference source in a certain multimeter, then we calibrate it... What is the result?

[Kleinstein]

The main difference is that ADR1399 is considerably lower (about half, maybe a little better) noise than the old LM399. In addition it need an extra RC network to improve higher frequency stability - so one may need a little more than just a replacemant of the chip.

It depends on the instrument if lower reference noise will be noted. In something like an HP3478 one would hardly notice a change. Similar a 34401 may not improve that much. In meter like the KS 34465 the difference could be quite a bit, like half way towards the 34470 for the noise.
Depending on the meter grade the LM399 used may be selected for lower than typical noise and maybe low drift. Chances are the ADR1399 could still be an alternative to such a selected chip.

[NaxFM]

My idea is not understood. Obviously, the new product is significantly superior to its predecessor. Yes.
 But. The question is this: we install an aged new reference source in a certain multimeter, then we calibrate it... What is the result?

We can't tell.
In theory it will be better than the LM399, in practice, according to the datasheet, it can even oscillate and get worse accuracy than a 5 dollar chinese multimeter...
I assume that if we don't consider the oscillation problem, a multimeter with this reference should have less noise and be less suscettible to temperature variations

[Kleinstein]

Temperature wise I would consider it a close call. Often the rest of the DMM would be more important than the reference.
For the long term dirft I would call it too early for a definitive answer, probably not a big different, but we don't really know yet. Time will tell how the new reference is in 5 years.

[MikeP]

 Thanks Mr. Kleinstein. I am well aware that 399 is not the only noise-determining link. And I see differences in the noise of multimeters with identical reference voltage sources.
 Why should we expect stability issues and why do we need an additional correction?

[Kleinstein]

For the long term stability we just don't know for sure. It looks OK so far, but it just takes time to get a definitive answer. A the reference already runs quite hot there is no easy accelerated aging and such tests are never as reliable as real word data.
For the temperature stability the LM399 is already quite good and not the limiting part in many DMMs.
The stability against oscillation is noted in the data-sheet. The extra RC element is nearly needed and definitely a good idea. The LM399 would probably also like such an RC element, just gets away with it a little better.

There is one more point to change: the LM1399 needs a high minimum current and thus may need a changes resistor.