ADR1399 reference

[bobAk]

Hello. Has anyone tried to connect adr1399 in series? Like lm399. It’s more interesting how it behaves, since changes in the control loop of the shunt part in comparison with lm399 led to the need to install ~1 µF 5 ohms. In theory, there should be no problems, due to low impedance. Requires ~14 volts with low noise.

[misterpulcri]

Hello,

Apologies, my question is pretty basic but I see a lot of expertise here and no other suitable ADR1366 forum. Happy to be redirected somewhere more appropriate if I am in the wrong place.

I have attempted to build a basic voltage reference using a ADR1399K with +12v supply, 1K3 shunt for 3mA at 12V, 5R1 and 1uF as per the datasheet. Have built this three times now on breadboard and stripboard. Each iteration produces 6.920V instead of the expected 7.05V. At each power on my DVM reads 6.925 and within a few seconds this decreases and stabilises at 6.920V. I have verified that I have just over 3mA flowing through the shunt. Have checked with 2 other hand held DVMs and get the result as my bench DVM. Have checked DVMs against a fairly accurate voltage source

Am I missing something or is my ADR1399 faulty?

Thanks

Sean

[Kleinstein]

There is quite some variations between different units of LM399 and ADR1399. The ref. voltage can vary quite a bit. The specs are 6.75 to 7.3 V, though the extreme values are rare. 7.05 V is just a typical value. So the 6.92 V are perfectly reasonable.

[misterpulcri]

Ah, OK thanks. I can at least stop pulling my hair out.

[Andreas]

At each power on my DVM reads 6.925 and within a few seconds this decreases and stabilises at 6.920V.

Hmm,

Are you Kelvin sensing the voltage at the zener or are you measuring the sum of the zener voltage and the voltage drop of the heater current on the ground connection?
But even without that it needs some time until steady state is reached.

with best regards

Andreas

[bastl_r]

Hi

What did you use to measure the reference voltage? It is advisable to measure it with a measuring device with an input voltage divider in the G-Ohm range. I once had an aha-experience on my Solartron 7066 when I wanted to check the reference voltage before the buffer. The standard 10MOhm of the 34401a was not enough and the voltage dropped a few 10µV...
Whether this also applies to the ADR1399 I cannot say at the moment.

Greetings bastl_r

Translated with DeepL.com (free version)

[misterpulcri]

Hi all and thanks for your replies. I should explain my objective and reason for constructing an ADR1399 based voltage reference. 

My "best" DMM is an ancient Fluke 8050A and I wanted to check the calibration of this device. I need to be able to measure down to 100s of uV which this meter will not do, but I wanted to ensure I was ballpark in the mV range.

I construct Eurorack synthesiser modules which use the 1V/octave system for voltage to pitch  i.e. note C0 = 0v, C1 - 1.0v etc. Each octave is divided into 12 semitones and each semitone is a 0.0833V increment.

Mouser sent a product announcement for the SMT version of the ADR1399 so I investigated and purchased the through hole version naively thinking that if I built the circuit in the datasheet adjusting my shunt value to ensure 3mA for the Zener, 7.05 volts would be present at the reference output and I could gain a degree of confidence in my meter calibration at least down to the mV range.

Before cominng here I did experiment with different shunt values to swing from <3mA up to 5mA and also varied my supply voltage etc and always seeing 6.920V.
I now know that my voltage reference build is fine and is accurate and stable at 6.920 instead of 7.050V 

Thanks again for taking the time to look at my naive newbie experiments, I will most likely purchase a new DMM that gets me into the uV range.

Thanks

Sean

[Kleinstein]

When the layout is right, the ADR1399 has a rather low output impedance (< 0.1 ohm) and should be OK even with a meter with 10 M input impedance.

The ADR1399 is however the totally wrong way to hope for an absolute value of the voltage. There is quite some tolerance in the voltage, but that value is quite stable.
There are other reference chips for a good accurate initial value (e.g. Ref102C or max6126A) : better accuracy, but not as stable.

[dazz1]

I could post the KiCad, Gerber and BOM file for the circuit I build if that fits your need. But they are pretty messy since that was my first design in KiCad.
I would need at least a few days to clean up the most important errors.

Hi
I like your design.  Simple single purpose, no unnecessary complexity.
If you haven't already, could you clean and post your files please?
Are there any changes or improvements you would make?

[Andreas]

Hello,

PCB #3 with a new set of ADR1399 after first 500 Hrs (+~15 hours for first bug fixing of the new software of the multiplexer)
Multiplexer is now with bistable relays.
not shown: additional thermal isolation + die cast housing for the whole pcb.

LS8-package has date code 2043
TO-46 metal can date code 2316

It seems that LS8 package needs longer time and more drift to settle than the TO-46 package which seems to stabilize after ~1 week.
Measurements are still without T.C. compensation since I want to wait with T.C. measurement until the units are sufficient stable.

with best regards

Andreas

[ivo]

It seems that LS8 package needs longer time and more drift to settle than the TO-46 package which seems to stabilize after ~1 week.
Measurements are still without T.C. compensation since I want to wait with T.C. measurement until the units are sufficient stable.

Can you share what circuit scheme you intend to use to remove... 0.2ppm tempco?

Have you looked if the LCC or TO46 package have any difference in 1/f noise?

[Andreas]

Hello,

T.C. will be only calculated out (mathematically) for the ageing drift.

Up to now I have only noise figures from LS8 packages on PCB#1

https://www.eevblog.com/forum/metrology/adr1399-reference/msg4445218/#msg4445218
https://www.eevblog.com/forum/metrology/adr1399-reference/msg4453792/#msg4453792

so obviously  a good idea to make some measurements on PCB#2 + PCB#3

with best regards

Andreas

[CalibrationGuy]

Andreas, I've been inspired by your efforts. I am working on aging boards. Here's a picture of my ADR1399 farm ready for planting!

TomG.

[CalibrationGuy]

The boards will go in this oven for a few months. Anyone have any suggestions as to optimum aging temperature? I was thinking 40/45 degrees C. Thoughts?

TomG.

[Kleinstein]

Aging should be at a higher temperature. Under normal operation the temperature is regulated to some 90 C. So that or a little higher would be the logical temperature for aging. This could be done by using the chip internal heaters.

[CalibrationGuy]

I see that I miscommunicated my intentions. I understand the operating temp is higher than the temperature I selected for the oven. What I meant to say was I will keep the parts at a temperature significantly higher than ambient so I can more accurately assess drift without ambient lab temperature being one of the variables.

TomG.

PS - Thanks for taking the time to reply.

[laichh]

I see that I miscommunicated my intentions. I understand the operating temp is higher than the temperature I selected for the oven. What I meant to say was I will keep the parts at a temperature significantly higher than ambient so I can more accurately assess drift without ambient lab temperature being one of the variables.

TomG.

PS - Thanks for taking the time to reply.

Hi TomG,

Take a look at the Design Note 229 by LT (now ADI) https://www.analog.com/media/en/technical-documentation/design-notes/dn229f.pdf.

[CalibrationGuy]

I see that I miscommunicated my intentions. I understand the operating temp is higher than the temperature I selected for the oven. What I meant to say was I will keep the parts at a temperature significantly higher than ambient so I can more accurately assess drift without ambient lab temperature being one of the variables.

TomG.

PS - Thanks for taking the time to reply.

Hi TomG,

Take a look at the Design Note 229 by LT (now ADI) https://www.analog.com/media/en/technical-documentation/design-notes/dn229f.pdf.

Yes, I've read this paper, thanks. I'm going to say this once more. I am aging the parts normally, not at elevated temperatures in order to accelerate aging. The parts are going in the oven because I want a very stable thermal environment to measure initial drift accurately. I am also placing boards with different parts and configurations with the 7 to 10 volt gain stage in the oven to see if their performance is the same in a constant temperature environment.

TomG.

[Andreas]

Hello,

to cancel out environment temperatures it is usually sufficient to keep temperature around 10 deg C above the max environment temperature. (including self heating of your PCB).

Better would be to keep the environment exact to 25 deg C (according to the datasheet conditions).
The question is: how temperature stable (hysteresis) is the oven between heating phases, and how much thermal mass do you need around your references to keep the environment temperature really stable?
And how temperature stable is your measurement instrument. Near room temperature the T.C. is between -0.1 and -0.3 ppm/K on the 4 metal can samples of PCB #2.
My strategy is different: make the first 1000 hours run in at room temperature (and log the temperature at the bottom of the die cast housing). At the moment I am determing the T.C. of PCB#3 to compensate (mathematically) for it.

with best regards

Andreas

[Andreas]

Hello,

PCB#3 now passed 1 kHr. (Still not temperature compensated).
See also:
https://www.eevblog.com/forum/metrology/adr1399-reference/msg5703775/#msg5703775

Unfortunately the temperature was changeing a lot the last days.

At the moment I am doing T.C. measurements so that I will be able to give a better picture how the ageing drift really is.

with best regards

Andreas

[CalibrationGuy]

Andreas, how are you measuring the voltage for those data points? I am planning on using our recently calibrated 732c as a reference to normalize the data. The 732c will be on channel one while the DUTs will be on the other channels. I can then calculate the drift of the DAQ and remove its drift from the final results.

TomG.

[Andreas]

Hello,

The plan with the 732c looks good.

In case of the ADR1399 boards I am using the average of my 3 most aged 24 bit ADCs. (ADC13, ADC15, ADC16)
Each ADC has a LTC2400 with a (selected) AD586LQ reference and a NTC as temperature sensor.
INL and 3rd order T.C. are calculated out.

Example for ADC see here:
https://www.eevblog.com/forum/metrology/t-c-hysteresis-measurements-on-brand-new-lt1027dcls8-5-voltage-reference/msg1224922/#msg1224922

And of course each ADC has a precision capacitive 2:1 voltage divider (LT1043) which extends the range up to 10V.
On ADR1399 board #3 the bypass input is used to monitor stability of the measurement by a pre-aged ADR1000 reference.

Stability of the ADCs in 10V range is better than my 34401A and better than my (temperature compensated) K2000.
https://www.eevblog.com/forum/metrology/stability-of-5v-references-compared-to-ad586lq/msg5257890/#msg5257890
https://www.eevblog.com/forum/metrology/project-pimp-a-keithley-2000/msg1106839/#msg1106839

with best regards

Andreas

[CalibrationGuy]

I read the posts from those links, very interesting reading. You've obviously put a ton of effort into this. I am curious about several things, but one thing interests me the most. You are using the AD586LQ reference chip. How did you select the best ones? What was your primary reference for the 10V absolute value and for measuring drift? What did you use for the comparison?

Best Regards,

TomG.

[Andreas]

Hello,

the AD586 selection criteria is explained in one of the links of previous message.
My primary reference is a 7V LTZ1000 (#4) which is compared against different calibrators or fresh calibrated DMMs to get traceable values.

with best regards

Andreas

[Andreas]

Hello,

I had switched off PCB#1 and PCB#2 for 3 weeks.
on PCB#2 the metal can packages restored the old voltage after 8 hours of warmup phase.
The LS8 packages on PCB#2 showed 0.5 - 1 ppm shift after warm up.
A short power outage of some minutes on day 460 restored the old values.

On PCB#1 the LS8 packages showed similar behaviour of 0.5 - 2 ppm shift after 3 weeks power outage.
But one of the LS8 packages also showed a permanent 1 ppm shift after the short power outage. (day 625)

So obviously the LS8 package has more hysteresis after long power off periods than the metal can package.

with best regards

Andreas