LM399 based 10 V reference

[Andreas]

Hello,

each LM399 has its pull-up resistor and will always sink about 1 mA.
So some of the LM399 will sink a bit more than 1mA others somewhat less.
The dynamic resistance of the LM399 zener is around 0.5 - 1 ohms.
so each change of current by 1uA will lead to a around 1uV change of the zener voltage.
But that is ok as long as the changes are small and/or remain constant. (above ~ 0.5mA)

best regards

Andreas

[GigaJoe]

Great, thanks, I completely miss that point

[GigaJoe]

ОК,  question:

I hook 5 of that burned thingi ( LM399), feed from precision V source , toss in thick Styrofoam box, and measure Vout precisely ( in use HP3456A;  temp env. +- 0.3C ) ...
that was APR 28,  now second measure today - MAY 06 ( 8x24  approx 200 hours ) - I got identical numbers, for all 5.

Normally, how long I should wait to get at least one last digit ( 5-th digits after comma) change for LM399.

[pelule]

I would simply calculate:
assuming you are measuring in the 10V range and LM399 volatge is ~7V, thus drift should be at least 2ppm change is 0,000014V
But only, if you assume, the 3456A does not dift at all.
LM399 drift over time is according to LT data sheet typ. 8 ppm / 1khr thus you need to wait typical 250 hours (~11 days)
/PeLuLe

[GigaJoe]

Right, but:
Items seems not a new, so it has some runtime, in addition 6month under +160C add something,  so I'm thinking i found an answer:
https://www.eevblog.com/forum/metrology/lm399-based-10-v-reference/msg507527/#msg507527
and that lead to: I would never find a direction, in a reasonable time, where units are drifting to .....

[zhtoor]

around 6 month under +160 C;  1-st step done, will set a drifting measurement now ...
under board a thick thing are copper wire; board a green pcb, resistors blue 
interesting oxidation of LM's legs, it actually from ebay ( seen as very beginning of leg are OK )

Hello GigaJoe,

great work.
would you suggest "cooking" the LM399's in an inert gas atmosphere like nitrogen/argon to prevent oxidation of metal?

best regards.

-zia

[GigaJoe]

I had cooked resistors,  metal film, a couple days at 170 C - they like burned,  but very stable, immune to soldering and heating. as coating degraded need to recoat it. I use conformal coating a few layers.

[e61_phil]

ОК,  question:

I hook 5 of that burned thingi ( LM399), feed from precision V source , toss in thick Styrofoam box, and measure Vout precisely ( in use HP3456A;  temp env. +- 0.3C ) ...
that was APR 28,  now second measure today - MAY 06 ( 8x24  approx 200 hours ) - I got identical numbers, for all 5.

Normally, how long I should wait to get at least one last digit ( 5-th digits after comma) change for LM399.

You should measure the LM399s against each other. This increases resolution and doesn't depend on the single LM399 in your 3456A.

[neil t]

It's a great hobby to draw pcbs.... 

And one can do so many re-designs:

I fear that the heater current (change) will influence your output voltage.
Why do you refer the ground pin of the 20K resistor to the input and not to the output?

with best regards

Andreas

Thanks for the tip...

I will update.

Looks like you have been having fun in Kicad  my main issue is that the area and layout of the lm399 particularly the partially unshielded trace going from zener + to non inverting input looks remarkably like an E Field probe.

[anymouse]

If you want to parallel LM399, have a look at the voltage reference of the Guildline 4400. This uses eight paralleled LM329 (that's like a LM399 without internal heater).
You can find the schematics of the reference module in an other thread: https://www.eevblog.com/forum/metrology/teardown-voltage-standards/

[branadic]

Did some repeating T.C. measurements (after almost one year) on my 16x LMx99 that are still waiting for further use, as well as low frequency noise measurements with heater on. I could confirm that their T.C. is still within the same region as one year ago. Andreas might forgive me that I use a picture he once sent me with the T.C. of my references included in his diagram.
I'm currently investigating the low noise frequency with heater off. Up to now I can't find any correlation on T.C. and low frequency noise within my measurements.

-branadic-

[GigaJoe]

just an update:
may 29 -ON june 24 - OFF for LM drifting measure:
in ppm, in the end of period:
1: -2
2: -47 (weird jump)
3: -3
4: -2
5: -5
6: -1
7: -2
8: i blew it up ...
it can be multi-meter drift, i use only one unit (3456) to measure periodically turn it on every week, and env temp very stable +18.7 (basement). at least 4 candidates ... As i assumed, do it in parallel doesn't increase the precision, as all of then will drift in the same direction with similar speed .....
 

[Insatman]

just an update:
may 29 -ON june 24 - OFF for LM drifting measure:
in ppm, in the end of period:
1: -2
2: -47 (weird jump)
3: -3
4: -2
5: -5
6: -1
7: -2
8: i blew it up ...
it can be multi-meter drift, i use only one unit (3456) to measure periodically turn it on every week, and env temp very stable +18.7 (basement). at least 4 candidates ... As i assumed, do it in parallel doesn't increase the precision, as all of then will drift in the same direction with similar speed .....

My HP3456A takes about 48 hours to stabilize after being off for more than a couple of hours.   I suggest you just leave it on for several weeks and see if your data changes.  I leave mine on 24/7, but periodic power outages here in Philippines give me plenty of opportunity to see the effect.   After a couple of hours on time the unit appears stable but over the next two days it will drift by ~20uV compared to my other 4 meters (2x HP34401, HP3457A and FL8505).   

[borghese]

Four years ago Bob Dobkin said at "Diligent Mind":
"The LM399's do not age when they are turned off and have almost no hysteresis-- so keeping the reference *off* until a few hours before you need to use it [and/or calibrate it] is the best way to keep the long-term drift minimised."
Cheers

[Insatman]

Four years ago Bob Dobkin said at "Diligent Mind":
"The LM399's do not age when they are turned off and have almost no hysteresis-- so keeping the reference *off* until a few hours before you need to use it [and/or calibrate it] is the best way to keep the long-term drift minimised."
Cheers

I'm not sure I believe it, but even if it is true there is the thermal equilibrium of the LM399 and any other elements to consider.   It may take more than a couple of hours. 

I also have my own observations which show a small but significant change after a couple of hours warmup. 

[borghese]

I do not know the truth, but now I quote the whole sentence as from «Reply # 635 on: March 13, 2014, 08:02:10 AM»(now deleted):

6) After telling him that I wanted a voltage transfer device, he said a better way might be to use [at least 6] LM399's in parallel [like the Bob Pease idea].  He said that the burn-in procedure would be to operate these in an oven set to 125C for 2 weeks, which would be equivalent to 1000's of hours of normal operation.  Any LM399's that are drifting too much after that can be replaced [i.e., you burn-in more than you need, and select the best units for the array].  The LM399 is much more sensitive to board stress than the LTZ [because the LTZ has a special mechanical arrangement in the die mount]-- so the LM399 should be mounted off of the PCB a little bit to allow for this.  The long term stability of the array of LM399's will be directly related to the power required to run the heater-- and this can be minimized with insulation-- the more the better!  The LM399's do not age when they are turned off, and have almost no hysteresis-- so keeping the reference *off* until a few hours before you need to use it [and/or calibrate it] is the best way to keep the long-term drift minimized.

7) Note that for LM399-based designs, the slots in the PC board [plus a lot of insulation top and bottom] make sense-- Bob said that the less power the heater requires, then the more stable the output voltage will be.  So, in this case, the slots [plus insulation] are helping with this.  Oh-- and he also said that the LM399 should be run at about 1mA of Zener current for best stability.  The more stable you can make the Zener current, the more stable will be the output voltage.  He said that there is about 1uV of voltage change for 1uA of current change.

[branadic]

Using slots around LM399 and massive insulating of the reference at least works for my LM399 ff. quite well.
After a few years the reference is also pretty stable, beside the popcorn noise issue. Attached is a 150h measurement with LT1763 preregulation. Time to put the preregulator inside the case.

-branadic-

[Insatman]

I do not know the truth, but now I quote the whole sentence as from «Reply # 635 on: March 13, 2014, 08:02:10 AM»(now deleted):

6) After telling him that I wanted a voltage transfer device, he said a better way might be to use [at least 6] LM399's in parallel [like the Bob Pease idea].  He said that the burn-in procedure would be to operate these in an oven set to 125C for 2 weeks, which would be equivalent to 1000's of hours of normal operation.  Any LM399's that are drifting too much after that can be replaced [i.e., you burn-in more than you need, and select the best units for the array].  The LM399 is much more sensitive to board stress than the LTZ [because the LTZ has a special mechanical arrangement in the die mount]-- so the LM399 should be mounted off of the PCB a little bit to allow for this.  The long term stability of the array of LM399's will be directly related to the power required to run the heater-- and this can be minimized with insulation-- the more the better!  The LM399's do not age when they are turned off, and have almost no hysteresis-- so keeping the reference *off* until a few hours before you need to use it [and/or calibrate it] is the best way to keep the long-term drift minimized.

7) Note that for LM399-based designs, the slots in the PC board [plus a lot of insulation top and bottom] make sense-- Bob said that the less power the heater requires, then the more stable the output voltage will be.  So, in this case, the slots [plus insulation] are helping with this.  Oh-- and he also said that the LM399 should be run at about 1mA of Zener current for best stability.  The more stable you can make the Zener current, the more stable will be the output voltage.  He said that there is about 1uV of voltage change for 1uA of current change.

Interesting.   The datasheet for the LM399 suggests a 7.5K resistor with a 15V supply, which gives 2mA rather than the one Bob suggested.   I have built several LM399 and LM199 references this way.  I use an LT3080 as a pre-regulator.  I have built 2 boards with 3x LM199 in parallel that were purchased used from Ebay.  These have performed well.  Some data from the first one has been submitted to this forum previously.
https://www.eevblog.com/forum/metrology/data-on-various-voltage-reference-ics-over-time/msg1479949/#msg1479949.  These boards have an op-amp which scales the output to 10V.  This does likely increase the drift/tempco somewhat.

I have not used slots in my previous board designs, but I do use insulation.  My LM199/399 devices are surrounded by a small hollow styrofoam cube.  The entire board including the pre-regulator and output scaling op-amp (~2.5x2") are then typically housed in an insulated box. 

thanks for the feedback...I always enjoy learning.

[AG7CK]

Quote: " ... datasheet for the LM399 suggests a 7.5K resistor with a 15V supply, which gives 2mA ..."

That's an old trap I guess we have all gone into here or there or somewhere else: 15 Volt over 7.5 kOhm is 2 mA, but 15 Volt minus the 7.x Volt over the LM399 gives - Yes, a tad more than 1 mA through 7.5 kOhm.

[Insatman]

Quote: " ... datasheet for the LM399 suggests a 7.5K resistor with a 15V supply, which gives 2mA ..."

That's an old trap I guess we have all gone into here or there or somewhere else: 15 Volt over 7.5 kOhm is 2 mA, but 15 Volt minus the 7.x Volt over the LM399 gives - Yes, a tad more than 1 mA through 7.5 kOhm.

to quote Homer Simpson "Doahhhh..."  Of course you are correct

[borghese]

The HP in the HP3456 multimeter uses a current of 1.3mA - 1.4mA.
Too bad that as isolation use only the cup made of valox and it is known that moving it varies the voltage probably because it varies the thermal gradient

[Insatman]

The HP in the HP3456 multimeter uses a current of 1.3mA - 1.4mA.
Too bad that as isolation use only the cup made of valox and it is known that moving it varies the voltage probably because it varies the thermal gradient

I added some foam insulation around the voltage reference in my HP3456.  Not much, only about 1/8" as the clearance to the shield isn't much.  I did a full calibration afterward which includes adjusting the reference.  The unit is very stable once it's been on for a couple of days. 

[JimmyJo]

Can I join the club now?
Single LM399H (non A), Vishay NOMCA 7V-10V gain resistor, LT1001.
Reference board is shielded from drafts, but the 3ft of pomona minigrabbers are not.


Questions and suggestions welcome.

[TiN]

Interesting that you used NOMCA's. Perhaps you can share board photo for us?
I got some NOMCA and NOMC networks too for tests, but did not get around to actually do it yet.

[JimmyJo]

This is really a hack job compared to your work, TiN.
On paper, NOMCA just seemed like a relatively low cost way to obtain closely matched resistors.  Both in terms of tempco and aging.
Two 8x 1K NOMCA were glued back to back, then wires soldered, jumping between the two packages to connect all 16 resistors in series.  A section of 10 was picked at random to become the gain setting resistor.

A more cost effective design would use the leftover NOMCA resistors for the 3k current setting resistor.