Author Topic: Yet another (LM399) volt-reference  (Read 5818 times)

0 Members and 1 Guest are viewing this topic.

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Yet another (LM399) volt-reference
« on: December 15, 2014, 08:14:06 pm »
I have built a basic LM399 reference.
I never really felt like ordering 100's of euros worth of resistors, so right now I intend to use the zener-voltage directly, measured with a high-impedance meter.

R1, R4, R5 and R6 are 15 ppm 0.1%, U1 can be something reasonable, I'm using OP07 or OPA27 mostly right now.
 

Offline Seekonk

  • Super Contributor
  • ***
  • Posts: 1831
  • Country: us
Re: Yet another (LM399) volt-reference
« Reply #1 on: December 15, 2014, 08:53:56 pm »
Somehow, I feel like I'm missing something.
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #2 on: December 15, 2014, 08:59:21 pm »
uhu?

You make a bunch of them (6), but them in a box with a rotary switch so you can easily compare the individual drift.
Each ref-board is small enough to be shipped as a letter.
Apparently LM399 does not complain about being powered down.

(and btw, I did find 18 399s in the trash the other month...)
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2410
  • Country: de
Re: Yet another (LM399) volt-reference
« Reply #3 on: December 16, 2014, 06:29:51 am »
I have built a basic LM399 reference.
I never really felt like ordering 100's of euros worth of resistors, so right now I intend to use the zener-voltage directly, measured with a high-impedance meter.

R1, R4, R5 and R6 are 15 ppm 0.1%, U1 can be something reasonable, I'm using OP07 or OPA27 mostly right now.
Hello,

so I understand that you are using the 7V output only as it is the only high stability point in the cirquit.

Did you build the cirquit already?
I think for stability (oscillations) it would be better to have some
integration capacitor from output of OP to the negative input.

With best regards

Andreas
 

Offline Seekonk

  • Super Contributor
  • ***
  • Posts: 1831
  • Country: us
Re: Yet another (LM399) volt-reference
« Reply #4 on: December 16, 2014, 09:18:32 am »
For those who want to argues such things, I was giving it a bump up before it was lost to oblivion in the foggy time warp.  This was mentioned in the other thread, how does it start.  It seems you are hoping for all those things you want to design out.  A bump resistor could be added across the FET and it would have no further effect.

Happened to look on ebay ans found gobs of 399's that were pull outs.  Makes me wonder what to source of all these were and what kind if issues they have.
« Last Edit: December 16, 2014, 09:48:20 am by Seekonk »
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #5 on: December 16, 2014, 09:40:46 am »
I'm taking measurements of them right now, layout, photos and data later.

This circuit will self-start! Don't worry!

Andreas: if the op is too fast there will be need for a compensation-cap, I have a foot-print for it on the board, with op07 and opa27 there seems to be no need, I smoke-tested with ca3140, and I think I mounted the compensation-cap on the first 2 boards
(3140's natural state is oscillating... it can be hacked into an opamp  >:D )
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #6 on: December 16, 2014, 07:33:12 pm »
Photo of the board, a small travel-case I made and the drift since last night (measured with a Keithley 192)

Andreas: No, I did not mount any compensation-caps on these boards, there is one on the prototype but not these (the foot-print is still there).
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2410
  • Country: de
Re: Yet another (LM399) volt-reference
« Reply #7 on: December 16, 2014, 08:09:14 pm »
Photo of the board, a small travel-case I made and the drift since last night (measured with a Keithley 192)

So I think you will have to do some "run in" phase until the reference stabilizes before you can use it.
The question is where the noise comes from.
Is it the Keithley or your reference? Can you improve the noise by thermal shielding of the LM399?
Usually the LM399 noise is typical around 4uVpp (some bad samples have more)
so it should be not visible with your instrument (10uV) resolution.
But of course this depends also on power supply quality.

With best regards

Andreas
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #8 on: December 16, 2014, 08:45:15 pm »
The Keithley is rather noisy, this is measured without
any filtering, one 100 ms every minute, with my 7075
the noise is less than 10 uV at 1 s integration (one count flickering).

The 399's come from power-supplys for a XPS, or something like
that, has been running 24/7 for 10 years, I rather believe
it's the keithley that is drifting...

 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1716
  • Country: de
Re: Yet another (LM399) volt-reference
« Reply #9 on: December 17, 2014, 07:56:40 am »
.. And .. are you still riding to Germany next Saturday?

.. the volt-nuts party is already fixed for Saturday evening...

Frank
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #10 on: December 17, 2014, 09:05:00 am »
No, all of Saturday I will on the road, about as for to Zwickau as to Frankfurt, ETA about 2300 if traffic is good...

But, if you want, 3 of my Refs could participate! If you PM me your address and I mail them before 1800 today they will probably make it.
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 806
  • Country: de
  • pushing silicon at work
Re: Yet another (LM399) volt-reference
« Reply #11 on: December 18, 2014, 06:51:40 am »
Re Party: I dropped out, a nasty little virus took me out of order and then my wife sequentially, so need saturday to compensate for things that didn't happen.
I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1716
  • Country: de
Re: Yet another (LM399) volt-reference
« Reply #12 on: December 18, 2014, 10:50:05 am »
Ok, well, that's a pity.. We'll have a 1 on 1 party only, then.

Hendrik, get better again!

Frank
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 806
  • Country: de
  • pushing silicon at work
Re: Yet another (LM399) volt-reference
« Reply #13 on: December 18, 2014, 01:09:45 pm »
Thank you, the upside is the bug leaves as abrupt as it appears. 24 hrs feeling sick, then rapid getting well.

But still bad for the party this time. But such events should happen on a regular base, right ? :)

I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #14 on: December 18, 2014, 10:08:05 pm »
Now I've figured out a way to better measure the noise and drift, both params will be lower than shown in the picture, any drift in both curves will be the DMM, and there's still some math to get to the ref-drift/noise. I will collect data over night (Christmas?) and come back on that.
Anyway, noise is less than 5 uV p-p.
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #15 on: December 24, 2014, 06:43:26 pm »
Now some more measurements over a few days, tie-scale is in hours and the magenta-trace is the noise of the voltmeter, the other thre traces should be dominated by the noise and drift of either of the three references:

 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2410
  • Country: de
Re: Yet another (LM399) volt-reference
« Reply #16 on: January 07, 2015, 08:14:34 pm »
Hello,

for me it is unclear what the y-axis is.
relative error (ppm) or absolute drift in uV.

With best regards

Andreas
 

Offline wiss

  • Frequent Contributor
  • **
  • Posts: 485
  • Country: ch
Re: Yet another (LM399) volt-reference
« Reply #17 on: January 07, 2015, 08:37:03 pm »
There are 3 references, vr1, vr2, vr3.
There are 4 measurements vm1 = vr1 - vr3, vm2 = vr2 - vr1, vm3 = vr3 - vr2 and a short.

In the last graph I plotted
( vm1 - vm2 ) / 2 =  vr1 - 1/2(vr3 - vr2)
( vm2 - vm3 ) / 2 =  vr2 - 1/2(vr1 - vr3)
( vm3 - vm1 ) / 2 =  vr3 - 1/2(vr2 - vr1)
zero

versus hours.

Ofcourse, the average of each voltage was subtracted before plotting. ..
« Last Edit: January 07, 2015, 10:24:32 pm by wiss »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf