[volt-nuts] Low-cost voltage reference questions

Lars Walenius lars.walenius at hotmail.com
Wed Nov 25 13:33:56 EST 2015

Mostly a reply to Charles but also to Ian

First of all I should say I prefer 10V reference ICs over 2.5, 5, 7V etc references as I think a decent hobby standard should be 10V to be easily comparable to a Fluke 732 or other 10V stds. If you have for example a 7V IC you have an extra problem to amplify it. So far I have seen no good stable amplifier down to a few ppm except if you use very expensive hermetic resistors. I am experimenting with the LT5400-3 to get a gain of 1.4 together with a zero drift opamp. After four months it still looks good but I have only one amp so far so too little statistics. 

I think specifications are a good starting point but not so much more. Practical tests are really needed.

As you understand my favorites are ceramic AD587 and metal canned REF102 just because I have tested them and found them stable over time, humidity and temperature (after compensation)

I have designed in quite a lot of references in my job but normally not to the low ppm level. Favorites have been MAX6350 in SMD and AD688 in SMD and long ago in ceramic but I have used all sorts from AD, LT and TI. Mostly it has been for IO systems.

I don´t have a MAX6350MJA and normally I would like to get IC´s from Digikey, Mouser or Farnell. Even in my professional job I have found it worthwhile to check if they are available at least on any of these distributors. So a question is if anybody knows anything more about the MAX6350MJA especially have tested some samples?

For MAX6350 in SMD I have done tests over a couple of years and the problem is the humidity sensitivity, I see about 10-15ppm just for a seasonal variation of 30%RH. A note: sometimes humidity sensitivity cheats people to believe it is temperature hysteresis as the package dries during heat up. The drift of two MAX6350CSA over 5 years have been 25ppm and quite constant. I have also tested just one MAX674 10V ref in SMD that was even worse with about 35ppm seasonal variations and 65ppm over 5 years.

One test I haven´t done with any other IC is to short circuit it. I did this a couple of years ago with a MAX6350CSA by accident while testing a couple of years ago. The short was only during a couple of minutes. First it went down 50ppm but  during a couple of months went back so it was 30ppm low and now have stayed so for two years compared to another MAX6350 in the same box. 

I haven´t tested the AD588KQ. I found it on Digikey but it was really expensive so I will probably not test it (like the VRE310 that I haven´t tested either). What I don´t like with the AD588 is the dual supplies needed. This is the same for the AD688 that I have done some long term tests on. The ceramic is good with drifts over 5 years up to 10ppm. What I have seen a couple of times are jumps over 2-4ppm that hasn´t gone back the worst I have seen once is 20ppm!). Another problem with the AD688 is the large internal heating that I think makes it more difficult to temperature compensate. It also has a longer warm-up. My only long time tested AD688 in SO-16 have had seasonal variations of about 10ppm for 30%RH and a drift of 40ppm over 5 years.

The new LT refs in LS8 packages I have just started to test so I have no real experience more than that they seems to have more temperature hysteresis (or humidity? ). My ceramic AD587 normally have 0.5-1ppm over a temperature scan of 15C but the LT1236LS8 I have tested have been up to 5ppm hysteresis. Also LT warns for humidity sensitivity if not special mounting is used. Seems as a drawback.

Why I think the REF102CM is worse than the AD587LQ is that the first years drift was +-10ppm and also power-off for a month after a long power on is much worse. On the REF102CM I have seen 5ppm but not more than a ppm on the AD587LQ.

About AD587 and REF102 in plastic packages my experience is that the AD587 has both less humidity sensitivities (10-15 versus 15-20ppm over 30%RH) and lower long term drift (up to 30ppm versus up to 80ppm over 5 years) compared to the REF102.

On this EEVblog page you can see four of my REF102CM compared to an AD581 over the last 14 years.


Noise from voltrefs used for calibrating voltmeters is an interesting subject that might have its own


Från: Ian Johnston
Skickat: ‎onsdag‎ den ‎25‎ ‎november‎ ‎2015 ‎12‎:‎28
Till: Discussion of precise voltage measurement

Hi all,
Just a note:
On my own project, I had used the MAX6350 in the prototype and the major 
downside for me was the thermal hysterysis it had. Then again, I had mounted 
a small crystal heater on top of it (set to 32degC +/- 0.1degC)......and I 
was pushing the limits as I was trying to resolve down to 38uV on a 0-10Vdc 
range using the 6350 as the 18bit DAC's 5V ref.

-----Original Message-----
From: Charles Steinmetz <csteinmetz at yandex.com>
To: Discussion of precise voltage measurement <volt-nuts at febo.com>
Date: Wed, 25 Nov 2015 06:17:07 -0500
Subject: Re: [volt-nuts] Low-cost voltage reference questions

Andreas wrote:

>Hello Lars,  or course you are right: the AD587 (10V) and AD586 (5V) 
>are the best (unheated) chips with hermetically package today.

I'm curious why you think the AD587/586 are better than other 
hermetically packaged references with better drift and noise 
specifications -- in particular,


(see attached table).  (Note that the hermetic version of the MAX6350 
has a 2.5ppm/C drift spec, not 1ppm/C as listed in the table for a 
different package.)

The 587 has a better long-term drift spec than the 6360 and the 6655 
(but note that the 587 is 3x worse than the REF102 in this 
regard).  To my mind, the better tempco and noise of the 6350 and 
6655 outweigh the better long-term drift of the 587 in an overall ranking.

Best regards,


[Sent by MDaemon Mail Server at IanJohnston.com]
volt-nuts mailing list -- volt-nuts at febo.com
To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/volt-nuts
and follow the instructions there.

More information about the volt-nuts mailing list