Author Topic: first lm399 10v reference  (Read 8765 times)

0 Members and 1 Guest are viewing this topic.

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
first lm399 10v reference
« on: May 11, 2018, 07:10:49 am »
I'm working on my first (10V) voltage reference, it is built around an LM399 and LT1001. I was using 10ppm resistors. After a lot of noise and drift, the drift seemed to settle more and more now, I'm getting around 10uV stability. This is after running for about a week. This is not battery powered, and done quite hacky for a first prototype. What is your experience with drift and LM399?

My next design is going to use an LT2057, for the op amp, and more stable resistors, to see what kind of results I'll be getting. If you have any feedback, please let me know. My goal is to get a stable and accurate 10.00000V.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14072
  • Country: de
Re: first lm399 10v reference
« Reply #1 on: May 11, 2018, 09:48:44 am »
For the LM399 the LT1001 or a similar precision BJT based OP is likely the better choice than an AZ OP.
If an AZ OP is used, one should consider some RC filtering between the LM399 and OP to reduce the noise spike from the OP effecting the reference. It would also reduce the higher frequency noise a little with a normal OP.

It is only the resistors to set the gain that need to be really stable. The other resistors are way less important.
The adjustment part looks a little odd. The simple series adjustable resistors are usually not the best way. More like 1 selected fixed resistor and than the trimmer in a different circuit, like a divider across the 10 V and a high value resistor at the wiper for a weak coupling to the main divider.
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: first lm399 10v reference
« Reply #2 on: May 11, 2018, 10:20:38 am »
Get rid of expensive trimmer, use fixed resistors only. You missed out completely power decoupling. Have 2.2-10uF tantalum cap + 0.22-0.1uF X7R 1206 cap as close as possible to both LM399 and opamp power pin. Make connection by nice beefy trace for power and caps, better use more than 2 vias each side to reduce ESL. Voodoo slots around 399 are not doing anything much, but you can keep them if you like fancy looks.
I hope your test board had components installed flush, without huge airgap like on your photo. Either way, shielding LM399 from air by non-conductive foam is important, to avoid thermal EMFs variations.

Measure the stability of the raw 7V output first, to make sure LM399 output is stable. After few months of doing that, you will have some confidence than LM399 is settled a bit and then you can start playing with 7V-10V gain stage, which is way more difficult to reach own reference stability level.  :scared:
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 
The following users thanked this post: zhtoor, niner_007

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 893
  • Country: de
  • pushing silicon at work
Re: first lm399 10v reference
« Reply #3 on: May 11, 2018, 10:31:05 am »
Lets call them "Improved pickup loops for magnetic field(tm)"
I'm not a feature, I'm a bug! ARC DG3HDA
 

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
Re: first lm399 10v reference
« Reply #4 on: May 11, 2018, 11:08:58 am »
Get rid of expensive trimmer, use fixed resistors only.
how does that work, wouldn't the coefficient of the trimmer affect the output?  |O

thank you for the feedback
« Last Edit: May 11, 2018, 11:10:31 am by niner_007 »
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: first lm399 10v reference
« Reply #5 on: May 11, 2018, 04:34:38 pm »
I always put 100nF additionally over the zener pins of the LM399 to improve the EMI behaviour.
additionally 100nF + 47uF for the heater.
If you want to trim you should use the trimming scheme
(Figure Standard Cell Replacement) of the data sheet to reduce the influence of the T.C. of the trimmer.

with best regards

Andreas
 

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
Re: first lm399 10v reference
« Reply #6 on: May 11, 2018, 07:34:38 pm »
thank you Andreas, would standard ceramic capacitors work here?
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: first lm399 10v reference
« Reply #7 on: May 11, 2018, 07:54:13 pm »
Hello,

on the heater side yes. (ceramic is ok since the 47uF is in parallel).

on the zener side: it depends.
If you have large vibrations you could see it theoretically
on the output signal. (microphonic effect).
So if you are not shure a film capacitor on the zener side is the safe way.

with best regards

Andreas

 
The following users thanked this post: niner_007

Offline cellularmitosis

  • Supporter
  • ****
  • Posts: 1111
  • Country: us
Re: first lm399 10v reference
« Reply #8 on: May 11, 2018, 08:13:25 pm »
If you have large vibrations you could see it theoretically
on the output signal. (microphonic effect).

I just saw this in a video last night.  https://youtu.be/67M7fsbLUIU?t=21m35s

LTZs: KX FX MX CX PX Frank A9 QX
 

Offline kj7e

  • Frequent Contributor
  • **
  • Posts: 911
  • Country: us
  • Damon Stewart
Re: first lm399 10v reference
« Reply #9 on: May 11, 2018, 08:42:49 pm »
I would be concerned with the daddy long legs construction, lots of air draft and thermal EMF potential.  I would lower the LM399 and leave maybe 5-8mm between the board and the bottom of the hat.  May want to think about some non-conducting closed cell foam around the legs as well (a la 3456A).  The feedback divider resistors are critical on a 7-10v buffer.  You want the most stable and lowest TC parts you can justify.  I would contact Edwin Pettis for some of his PWW parts which are reasonably priced, any value, and a low TC;
https://www.eevblog.com/forum/profile/?u=96921

Calculate the values you need for Rf and Rg, keep it simple and forgo the trimmer section.  This will give you the best stability in the long run.  The value may not be exactly 10.000000v, may end up + or -  some, but that does not matter much.  What matters is short and long term stability.  Looks like you already have a very good DMM (34470A?), your LM399 wont out perform it, so shoot for the best stability, measure it and write it down wherever it lands.

If you really want to make a trimmer section, have a look at one I did not long ago that is performing very well;
https://www.eevblog.com/forum/metrology/ltz1000-10v-buffer/msg1435737/#msg1435737

Note, I'm not an expert here so take that with a grain of salt.

« Last Edit: May 11, 2018, 09:03:18 pm by kj7e »
 
The following users thanked this post: niner_007

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
Re: first lm399 10v reference
« Reply #10 on: May 11, 2018, 09:19:11 pm »
regarding the long legs, it is just in the prototype, so it's easier for me to take out and since I knew the board had problems
 

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
Re: first lm399 10v reference
« Reply #11 on: May 11, 2018, 09:26:26 pm »
on the zener side: it depends.
and on the zener side, is the capacitor across VCC and ground correct? or or is it across the op-amp's IN+ and ground? thanks :)
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: first lm399 10v reference
« Reply #12 on: May 11, 2018, 09:44:11 pm »
I meant C9 in attached cirquit
 

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
Re: first lm399 10v reference
« Reply #13 on: May 11, 2018, 10:00:02 pm »
thank you Andreas; someone mentioned here that LT2057 might be a problem, would the capacitor be sufficient here to reduce noise?
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 16545
  • Country: us
  • DavidH
Re: first lm399 10v reference
« Reply #14 on: May 11, 2018, 11:53:19 pm »
I would be concerned with the daddy long legs construction, lots of air draft and thermal EMF potential.  I would lower the LM399 and leave maybe 5-8mm between the board and the bottom of the hat.  May want to think about some non-conducting closed cell foam around the legs as well (a la 3456A).

The low noise low frequency performance of the LT1001 and LT2057 also require protecting them and their input networks from thermal EMF.
 

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
Re: first lm399 10v reference
« Reply #15 on: May 12, 2018, 12:57:45 am »
Voodoo slots around 399 are not doing anything much, but you can keep them if you like fancy looks.
I understand about not improving stability, but wouldn't it help the reference stabilize faster, as the heat doesn't have to dissipate thru the board as much?
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: first lm399 10v reference
« Reply #16 on: May 12, 2018, 04:16:11 am »
Hello,

One way to check thermal stability on LM399 would be my "tilting" experiment
https://www.eevblog.com/forum/metrology/lm399-based-10-v-reference/msg360779/#msg360779

Quint-essence: much thermal isolation around the legs of the LM399 increases stability.

For the slots: initially I had larger instability on the device with slots.
With good thermal isolation both LM399 were on par.
But: since I used 2 different devices (with most probably different sensitivity against tilting)
you cannot draw a real conclusion without repeating the experiment with one and the same device in different configurations.

with best regards

Andreas
 
The following users thanked this post: Marco1971

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3221
  • Country: de
Re: first lm399 10v reference
« Reply #17 on: May 12, 2018, 04:25:13 am »
thank you Andreas; someone mentioned here that LT2057 might be a problem, would the capacitor be sufficient here to reduce noise?
Hello,

of course it helps (somewhat) to keep the chopper currents of the LTC2057 away from the (low ohmic 1 ohms) reference.
The main reason why I use it is to keep EMI away from the LM399 and reduce instability.

The low noise low frequency performance of the LT1001 and LT2057 also require protecting them and their input networks from thermal EMF.
If you use plastic packages (LTC2057) they usually have a copper lead frame.
So against the Reference (LM399) with KOVAR legs the thermal EMF is much smaller.
(copper to copper near zero, and copper to solder around 3uV/K against KOVAR to copper with 40uV/K).

with best regards

Andreas
 

Offline Svgeesus

  • Regular Contributor
  • *
  • Posts: 73
  • Country: us
Re: first lm399 10v reference
« Reply #18 on: May 13, 2018, 12:22:15 am »
I would be concerned with the daddy long legs construction, lots of air draft and thermal EMF potential.  I would lower the LM399 and leave maybe 5-8mm between the board and the bottom of the hat.  May want to think about some non-conducting closed cell foam around the legs as well (a la 3456A).

I agree.

I came across this quote by Bob Dobkin:

Quote
If properly insulated, an LM399 can have a ppm/K figure of between 0.1ppm/K to 0.2ppm/K-- the data sheet is *very* conservative.  (Note that the DMM manufacturers are not insulating these other than the plastic insulator that they come in-- they would benefit by better insulating the top and bottom of the LM399 to keep the heat in, and the heater power down)

That also implies that magic slots are no help and indeed actively harmful.

The feedback divider resistors are critical on a 7-10v buffer.  You want the most stable and lowest TC parts you can justify.  I would contact Edwin Pettis for some of his PWW parts which are reasonably priced, any value, and a low TC;
https://www.eevblog.com/forum/profile/?u=96921

Calculate the values you need for Rf and Rg, keep it simple and forgo the trimmer section.  This will give you the best stability in the long run.  The value may not be exactly 10.000000v, may end up + or -  some, but that does not matter much.

I see that sort of advice a lot, and disagree. It might matter or it might not, depending on the application. If you want an ultra-stable 10-ish volts in a box to calibrate a meter or as a comparative reference, the absolute accuracy is largely irrelevant. If you want a highly stable and highly accurate 10 point lots of zeroes reference, such as for a comparator, DAC or ADC, the absolute accuracy is crucial. (The LM399-based circuit I am currently developing falls into the latter category).

If you really want to make a trimmer section, have a look at one I did not long ago that is performing very well;
https://www.eevblog.com/forum/metrology/ltz1000-10v-buffer/msg1435737/#msg1435737

I see that trimmer-adjustable voltage dividers typically use the trimmer in parallel to one of the resistor legs. Unfortunately no online resistor calculator (that I have seen) helps with that sort of calculation. I would love to be able to input the desired dividing ratio, the tolerance and tempco of the resistors, the tolerance on the input voltage, and have it spit out whatever fixed and variable resistors I need for the variable leg so that the range of adjustment is a little larger than the sum of known errors.
 

Offline cellularmitosis

  • Supporter
  • ****
  • Posts: 1111
  • Country: us
Re: first lm399 10v reference
« Reply #19 on: May 13, 2018, 05:25:38 am »


I came across this quote by Bob Dobkin:

Quote
If properly insulated, an LM399 can have a ppm/K figure of between 0.1ppm/K to 0.2ppm/K-- the data sheet is *very* conservative.  (Note that the DMM manufacturers are not insulating these other than the plastic insulator that they come in-- they would benefit by better insulating the top and bottom of the LM399 to keep the heat in, and the heater power down)

That also implies that magic slots are no help and indeed actively harmful


How so?  I draw the opposite conclusion from that quote.  If you are after the best thermal isolation, cut slots and then sandwich between styrofoam.
LTZs: KX FX MX CX PX Frank A9 QX
 

Offline kj7e

  • Frequent Contributor
  • **
  • Posts: 911
  • Country: us
  • Damon Stewart
Re: first lm399 10v reference
« Reply #20 on: May 13, 2018, 05:35:25 am »
^ I would agree with that.  In the end, I think voodoo slots can be effective, at the same time may not be necessary if the board has a very good layout.
 

Offline Svgeesus

  • Regular Contributor
  • *
  • Posts: 73
  • Country: us
Re: first lm399 10v reference
« Reply #21 on: May 13, 2018, 08:05:27 am »
Oh right. I was seeing the slots as cooling slots, but if they are to prevent heat being conducted through the PCB then I can see they might help (once covered).
 
The following users thanked this post: cellularmitosis

Offline niner_007Topic starter

  • Frequent Contributor
  • **
  • Posts: 256
  • Country: us
Re: first lm399 10v reference
« Reply #22 on: May 13, 2018, 08:23:06 am »
this was an interesting thread related to slots https://www.eevblog.com/forum/metrology/slot-holes-around-ltz1000(a)/
 

Offline Svgeesus

  • Regular Contributor
  • *
  • Posts: 73
  • Country: us
Re: first lm399 10v reference
« Reply #23 on: May 13, 2018, 08:22:54 pm »

I came across this quote by Bob Dobkin: ...

I hadn't noted the source of that quote, at the time. I just found it again:
https://xdevs.com/fix/hp3456a/#twk_ref
 

Offline try

  • Regular Contributor
  • *
  • Posts: 112
  • Country: de
  • Metrology from waste
Re: first lm399 10v reference
« Reply #24 on: May 14, 2018, 09:00:38 am »
Hallo Svgeesus,

Unfortunately no online resistor calculator (that I have seen) helps with that sort of calculation. I would love to be able to input the desired dividing ratio, the tolerance and tempco of the resistors, the tolerance on the input voltage, and have it spit out whatever fixed and variable resistors I need for the variable leg so that the range of adjustment is a little larger than the sum of known errors.

the only need I see is for a resistor calculator based on the resistance values you either have or can buy.
That is something you can easily program in excel.
You would use it for the basic divider.
It's no rocket science to shunt one leg and approach the target output voltage.

If you really want to use a trimmer (which I would avoid, see comment from TIN above), the next step is to replace one resistor by another resistor and a trimmer in series.

Regards
try

Added: excel screenshot

« Last Edit: May 14, 2018, 09:25:30 am by try »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf