ADR1001 - Ovenized Voltage Reference System

[iMo]

Andreas, this is how it went during the first week. The capture started after aprox 2 hours of switching both on.
It shows the initial behaviour of my 34401A I mentioned couple of times already - it needs several days to settle after long power off. During the first day it goes up by say +2ppm, then after a few days it settles down.
That is the most probably caused by humidity inside the DMM. That has been tested several times here (with ADR1001#1 kept hot).
PS: median and tema are with DMM TC compensated, 100NPLC, ADR mains powered.

..

..
My 34401A showed 9.999.890 at 23C couple of days back, and now both are switched off again.

[1] Any hint from behind the scene how to track down a humidity affected part in a DMM? My 34401A peaks aprox +3ppm after aprox 1day after longer off, then in 3-4 days it slowly drops to the "expected value". The temperature diff inside the meter against ambient is aprox 18C. The ventilation is rather obstructed as I use an internal thin foam isolation.

I made some small changes in myADR1001#1 (now #1B) 10V reference:
..
The above changes have increased the output voltage by aprox 4-6uV..

[mawyatt]

That is the most probably caused by humidity inside the DMM. That has been tested several times here (with ADR1001#1 kept hot).

Just a crazy thought. Since the 34401A has no fan and no vent holes in the case the air inside is somewhat stagnant with little exchanged from the ambient outside. Wonder if tossing in a few desiccant bags inside the case might improve the humidity sensitivity?

Best

[Kleinstein]

Desiccant bags can slow down humidity changes. However much of the reaction to humidity is already slow for the diffusion in and out of the plastics. So it is already somewhat slow.
The main chance is likely from no using the meter for weeks and than get it warm and dry again.

[mawyatt]

We were thinking more long term instrument humidity sensitivity with the desiccant bags. Example, with instrument powered off, the internal humidity will rise but should be less with the desiccant bags, thus the component/pcb/whatever inside the instrument will absorb less long term since this is starting from a lower baseline internal humidity.

Best

[Andreas]

This is why I run my DMMs 24/7.
It would have been good to start the 34401A 1 week before the measurement.

Since the 34401A has no fan and no vent holes in the case the air inside is somewhat stagnant with little exchanged from the ambient outside.

There are enough holes for humidity. And desiccant does not mean no humidity (only reduction to a ambient temperature dependant level). So I do not believe that it will help much. (And you should exchange the packs at least every week).
Better would be a temperature and (low) humidity controlled environment.

with best regards

Andreas

[Kleinstein]

To keep the meter dry when not in use, one would need not just a desiccant bag, but also seal the meter in something tight (ideally a metal coated foil bag).

[iMo]

Therefore I started the measurement ~6 days after the switching both on. As I wrote my meter has got a thin foam cover inside its box which seals a bit the ventilation. That foam stabilizes the temperature flows inside. Btw., the AHT20 sensor shows here between 22 to 34% humidity (not calibrated yet, hopefully not a fake chip from ali). I will add the BMP280 and AHT20 data into the measurements later on. Waiting on the higher temperatures around 23C in the lab.

[Andreas]

Hello,

I passed 8kHrs of my ADR1001 eval board on 5V output.
I had a one week power off phase on the eval board.
This resulted in a 1 ppm jump which is now restored after 2.5 weeks operation.
Standard deviation over the last 42 days also suffers from the power off phase.

with best regards

Andreas

[iMo]

..my bet the 2.5w is due to the pcb mounting..

[Andreas]

Hello,
I fear you are right. Looks like drying the epoxy of the PCB.
with best regards
Andreas

[iMo]

This is an experiment I made with the ADR1001#1 box - with hot 34401A I switched off the ADR for 24 hours, afterwards you may see the return of its voltage back.
100NPLC, 3.5h lead-in, 24h off, 5h lead-out..
Still my box posts are nickel plated and the HP399 is noisy..
The vertical is 5uV/div.

[faraday]

Looking at EVAL-adr1001 and ADR1001 datasheet  it is  possible to make 10Vref using few mods

[TheSteve]

I have an ADR1001 Eval board running at 10 volts. It is powered from 5 volts still (USB, eek) and will run nicely from a single 18650. The power supplies require some tweaking to make it all work - just a few resistor changes.

[faraday]

I have an ADR1001 Eval board running at 10 volts. It is powered from 5 volts still (USB, eek) and will run nicely from a single 18650. The power supplies require some tweaking to make it all work - just a few resistor changes.

can you post photo of your mod?

[iMo]

BTW. - a hint off the #1744 video on HV diff probe (28:09) - they use o-rings between the pcb and plastic covers around the input voltage dividers (preventing moisture ingress). That might be used with ADR1001 too..

[Kleinstein]

A plastic cover is more like a short time solution, to avoid the daily variation, maybe working over a few weeks. For the seasonal variations it would likely not be good enough. The OP-ring would need smooth surfaces and the thin plasics is usually somewhat permeable to water.

[rikkitikkitavi]

Maybe not so much moisture stabilisation as dirt and moisture protection as it might impact insulation, CMRR and so due to the protection over the divider net? Anyhow it is not stabilising.
Screwing a large plastic cover at the corners with orings at a flat surface , well it is a race between permeability of moisture in the cover/PCB or just sneaking through the o ring  seal.

For diffusion barriers aluminium tape is pretty good when applied overlapping.

[TheSteve]

As requested here is a picture of my ADR1001 eval board modified to provide 10 volts out.

The changes made are mostly as per the datasheet example of a 10 volt ref.

TSET is left open to set ~ 70 degrees
A heater current limit resistor of 10 ohms was added to reduce the initial maximum current to not overload the boost supply specification.

The power supply changes required to run the unit off of USB 5 volts or a single LiPo etc are:

R1 changed from 2K to 680 ohms (increased the maximum current to ~ 250 mAh, this could maybe be ignored but I like the ADR never starving for current with a lower start voltage).
R2 changed from 59K to 51K (increases the linear regulator voltage in the LTM8048 to ~ 12.9 volts)
R3 changed from 110K to 120K (increases the linear regulator voltage of the LT3045 to ~ 12 volts)
R6 changed from 2.94K to 2.7K (increases the switching regulator voltage in the LTM8048 to ~ 13.8 volts)

The 5 volt input(or less, mine works fine down to 3 volts) gets boosted by the switching supply(LTM8048) to ~ 13.8 volts, this feeds the linear reg in the switching supply(LTM8048) which outputs 12.9 volts. The 12.9 volts feeds the dedicated linear reg(LT3045) which provides 12 volts to the ADR1001(the ballpark minimum for a 10 volts output).
You could play with these values for more or less headroom. I found the above values worked well and are standard easy to get values.

[iMo]

As requested here is a picture of my ADR1001 eval board modified to provide 10 volts out..

Have you ever measured the output impedance of the internal buffer in the AEZ version? The engineering sample had 1.7ohm, afaik.

https://www.eevblog.com/forum/metrology/adr1001-ovenized-voltage-reference-system/msg4580656/#msg4580656

[TheSteve]

So far I've only done some basic monitoring of the output with a 3458A and a 34470A.

[Andreas]

Hello,

I passed 9kHrs of my ADR1001 eval board on 5V output.

See also:
https://www.eevblog.com/forum/metrology/adr1001-ovenized-voltage-reference-system/msg6234067/#msg6234067

Standard deviation over the last 42 days has now also calmed down to previous level.

with best regards

Andreas

[Elman_tarrah]

Hi imo,

I am interested in your ADR1001 10V reference project.

I would like to ask if it would be possible for you to share the latest full schematic of the board you built.

I tried to recreate the schematic from your posts, but since the design has been upgraded several times, I became a bit confused about which version is the latest.

Thank you in advance for your help.

[iMo]

..I would like to ask if it would be possible for you to share the latest full schematic of the board you built.

Welcome to the club! Below the schematics.

The power source not shown.
I've been using an LM723 based 15V source (with a mains transformer). After the rectifier there is 20.6V on a 2200uF/35V capacitor. An LM317 or LM7815 based PSU may work fine (or a similar).

Another power input is for 5x18650 cells (21V) through an 1N4001 diode when powered from battery. The cathode of the 1N4001 wired to the 2200uF/35V capacitor.

The temperature sensor not shown. I've been using an LM35 sensor (10mV/degC).
I would not connect the sensor (a chip or a NTC resistor) to the ground or Vcc of the reference. Let the sensor float.

I do not have the 10ohm in the heater wired. The datasheet has it there..

The RISET (aprox 109ohm DIY manganin wire wound in my case) is for setting of the 10.000V at the output. Its "10V setting sensitivity" is aprox -300uV per "+1ohm of the RISET's change" (the sensitivity may vary per sample).

Wima foil red bricks like 3u3 for 50V (and the other values perhaps too) may have large leakage at 40-50degC. So doublecheck the leakage at ambient and at higher temperatures.

Opamp - other AZ/choppers may work fine. Perhaps jfet or bipolar one with small TC as well.

ADR1001 package soldering - I've been using copper wires (a spider web or trampoline style), somehow thick (perhaps 0.2mm), next time I will use even thinner one. Also a flex pcb may work nice.

C2 and R8 - yep, the R8 there is after the C2 - I did it because of a post from past where it was indicated the OPA189 has got higher noise with a capacitor wired directly on its input.. Not sure 100% it is so, but I put the R8 there. With none AZ/choppers I would not populate the R8.