Author Topic: LM399 based 10 V reference  (Read 330506 times)

0 Members and 2 Guests are viewing this topic.

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1509
  • Country: de
Re: LM399 based 10 V reference
« Reply #375 on: January 07, 2016, 08:26:37 am »
Thanks for sharing Andreas. For my understanding, the blue curve measuring the LTZ#2 indicates the drift and noise of the ADC's reference? Which reference is used for this reference?
Thus, the remaining error of the ADC's reference is still within the curves for the LM399s? I'm asking because there seems to be some common mode behavior between the blue and green curves but also a much smaller fraction between green and red curves.


Maybe you can explain the diagram in more detail? Thanks.
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2447
  • Country: de
Re: LM399 based 10 V reference
« Reply #376 on: January 07, 2016, 11:57:29 am »
Hello branadic,

LTZ#2 was intended to show the drift of the ADC reference.
LTZ#2 itself has most probably a drift of around -2 ppm / year.
But since the calibration standards that I use to calibrate LTZ#2
also drift it will need some years until I know the whole truth.

By the way ADC13 drifts around the same amount as LTZ#1.
LTZ#1 calibration values show a lower drift than LTZ#2.
(so I should have used this as reference, but who knows in advance).

The common mode effects are partly due to environment
humidity + temperature. Temperature is compensated within ADC.
But nothing is perfect. And over the time also the contacts of the
measurement setup get loose and have to be moved/tightened again.

Attached the ageing curves of ADC13 compared to my
"stable" standards over the same time window.

The conclusions from my side are:

you will need minimum 200 days (half a year) run in time
 until you can start qualifying the LM399.

In this case I would prefer LM399 CH7 which shows the higher stability.
(less standard deviation)

With best regards

Andreas
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1752
  • Country: de
Re: LM399 based 10 V reference
« Reply #377 on: January 07, 2016, 02:13:08 pm »
Hello branadic,

LTZ#2 was intended to show the drift of the ADC reference.
LTZ#2 itself has most probably a drift of around -2 ppm / year.
But since the calibration standards that I use to calibrate LTZ#2
also drift it will need some years until I know the whole truth.

...
With best regards

Andreas

Will we three be able to check that coming Saturday?

Frank
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2447
  • Country: de
Re: LM399 based 10 V reference
« Reply #378 on: January 07, 2016, 04:07:21 pm »
I hope at least getting 2 new 8.5 digit measurement points on the diagrams  8)
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1752
  • Country: de
Re: LM399 based 10 V reference
« Reply #379 on: January 07, 2016, 06:56:02 pm »
I hope at least getting 2 new 8.5 digit measurement points on the diagrams  8)

Andreas, certainly. But this was a two-fold question:

..Will we three be able to check that coming Saturday?

Frank

Anyhow, I'll carry my new 34465A also..

« Last Edit: January 07, 2016, 06:58:31 pm by Dr. Frank »
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1509
  • Country: de
Re: LM399 based 10 V reference
« Reply #380 on: January 07, 2016, 07:00:51 pm »
Hi Frank,

I'm with you on saturday, together with Keithley 2002 (last Cal in 11/2015) and my Prema 5017 SC (not Cal'ed yet). What else I come up with, well, let's see.
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Online Dr. Frank

  • Super Contributor
  • ***
  • Posts: 1752
  • Country: de
Re: LM399 based 10 V reference
« Reply #381 on: January 07, 2016, 07:22:51 pm »
Hi Frank,

I'm with you on saturday, together with Keithley 2002 (last Cal in 11/2015) and my Prema 5017 SC (not Cal'ed yet). What else I come up with, well, let's see.

GREAT!   :-+ :clap: :-+ :clap: :-+ :clap:

Then let's have a contest.
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 814
  • Country: de
  • pushing silicon at work
Re: LM399 based 10 V reference
« Reply #382 on: January 08, 2016, 08:46:43 am »
Klingt als hättet Ihr was vor! :)

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

Offline bingo600

  • Super Contributor
  • ***
  • Posts: 1417
  • Country: dk
Re: LM399 based 10 V reference
« Reply #383 on: January 08, 2016, 09:06:09 pm »
Klingt als hättet Ihr was vor! :)

Agreed ....

What are you guyzz up to  :popcorn:

/Bingo
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1509
  • Country: de
Re: LM399 based 10 V reference
« Reply #384 on: January 11, 2016, 11:33:04 am »
Nothing fancy, only a small comparison between meters and voltage references :)
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Offline bingo600

  • Super Contributor
  • ***
  • Posts: 1417
  • Country: dk
Re: LM399 based 10 V reference
« Reply #385 on: January 11, 2016, 07:48:36 pm »
Nice gear.

What's the "Red 7-Seg" one ?
A Prema ?

/Bingo
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1509
  • Country: de
Re: LM399 based 10 V reference
« Reply #386 on: January 11, 2016, 07:52:56 pm »
Yes, a Prema 5000 and beneath a Prema 5017 SC.
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Offline F64098

  • Contributor
  • Posts: 38
  • Country: de
Re: LM399 based 10 V reference
« Reply #387 on: January 14, 2016, 08:12:06 pm »
Hello everybody,

after reading the whole thread and some other threads and articles, i finished my first revision of a LM399H based reference. Or is it AD587LQ based?
At the end, i don't really know.
Few weeks ago i soldered a 100% copy of an SVR-T on a small chinese breadboard and since that time, i got 10,000.00V, drifting between 9,999.99 and 10,000.01V, depending of the room temperature.
I used simple 25ppm resistors and a 100 ppm trimmer with 200 ohms for the trimming of the AD 587.
Quick and dirty T.C. compensation is done with a 100k thermistor in series with a 220k fixed resistor (50 or 100ppm, don't know).

So i decided to use the combination of the LM399 and the AD587 as described in Joe Gellers patent, to add the stabilities of both parts.
I tried my best to avoid the known constructive errors, but i'm not sure, if i should use a ground plane on both layers for a better heat distribution
or if i should omit it for a better temperature isolation. 
Please have a look at the pictures of both layers and tell me your opinions.

Thanks in advance

Frank
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1509
  • Country: de
Re: LM399 based 10 V reference
« Reply #388 on: January 14, 2016, 10:35:50 pm »
So you want to use SMD resistors? Do you plan to use some thermal stabilisation for them?
As you can read in this thread I've finally used a pickaback board with the gain setting SMD resistors (5ppm/K) and a crystal heater on top of them, to avoid temperature changes. Seems to work quite well.

However, I would spend at least some buffer amp for the LMx99.
Fluke 8050A | Prema 5000 | Prema 5017 SC | Advantest R6581D | GenRad 1434-G | Datron 4000A | Tek 2465A | VNWA2.x with TCXO upgrade and access to: Keysight 3458A, Keithley 2002, Prema 5017 SC, 34401A, 34410A, Keithley 2182A, HDO6054, Keysight 53230A and other goodies at work
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2447
  • Country: de
Re: LM399 based 10 V reference
« Reply #389 on: January 14, 2016, 11:05:00 pm »
Hello,

I would thermally decouple the 85 deg LM399 from the room temperature NTC and AD587.

You have anyway at minimum 2-3 deg C temperature difference between NTC and the chip of the AD587.
The LM399 will worse that if thermally coupled.

The AD587 has a bad PSRR. (10ppm/V) So I would not put the 15V stabilisation too far away from the cirquit.

With best regards

Andreas
« Last Edit: January 15, 2016, 07:07:44 am by Andreas »
 

Offline F64098

  • Contributor
  • Posts: 38
  • Country: de
Re: LM399 based 10 V reference
« Reply #390 on: January 15, 2016, 07:30:49 pm »
So you want to use SMD resistors?

Yes, and i have to use 0805, because my favorite vendor Reichelt sells the complete E96 row only in 0805.
Next step would be 5ppm resistors. Where do you buy them in germany?

Quote
As you can read in this thread I've finally used a pickaback board with the gain setting SMD resistors (5ppm/K) and a crystal heater on top of them, to avoid temperature changes. Seems to work quite well.

Maybe the next edition has a temperature stabilizer on board, with the heating resistors placed on the "right" positions.
But it's not easy to get a high precision temperature stabilizer and then find out the right positions for the heaters and the sensor.

Quote
However, I would spend at least some buffer amp for the LMx99.

In this configuration, the AD587 is the "buffer" for the LM399H.

Best regards

Frank
 

Offline F64098

  • Contributor
  • Posts: 38
  • Country: de
Re: LM399 based 10 V reference
« Reply #391 on: January 15, 2016, 07:43:52 pm »
I would thermally decouple the 85 deg LM399 from the room temperature NTC and AD587.

In my opinion, there is enough decoupling to avoid hot spots and enough coupling to reach a kind of thermal balance for the whole unit. I will place the unit in a TEKO tin-box and fill it with molten paraffine. This little box will be isolated with minimum 25-30mm foam and placed in another box, maybe heated with a simple PTC-heater or a heating-foil to ~40°C

Quote
You have anyway at minimum 2-3 deg C temperature difference between NTC and the chip of the AD587.

The parrafine should avoid such temperature differences.

Quote
The AD587 has a bad PSRR. (10ppm/V) So I would not put the 15V stabilisation too far away from the cirquit.

I already finished a LM317 tracking preregulator with 2 LM329 in series for voltage stabilization as a small module.
It will be placed in the outer box.

With best regards

Frank

 

Offline lars

  • Regular Contributor
  • *
  • Posts: 132
  • Country: se
Re: LM399 based 10 V reference
« Reply #392 on: January 16, 2016, 07:41:55 am »
For the bad PSRR at DC of the AD587 I would say Analog Devices is very conservative. For my SVR´s with AD587LQ I have I have measured 0.1-0.3ppm per Volt around 15V in. I just did a quick check for 6pcs of AD587, 2pcs REF102 and 2pcs LT1021 and all were 0-3ppm for a quick change of 15 to 20 to 15Volt in. The REF102 and LT1021 was not better than the AD587 even if the spec says so.

Lars
 

Offline lars

  • Regular Contributor
  • *
  • Posts: 132
  • Country: se
Re: LM399 based 10 V reference
« Reply #393 on: January 16, 2016, 07:49:11 am »
Hello Frank

Will be interesting to see how the result for the LM399+AD587 combination will work. I have thought of testing this but with the SVR-T I think some of the benefits vanished. I think Joe did the combination to get away with the worse temperature coefficient of the AD587LQ compared to the LM399.

Lars
 

Offline F64098

  • Contributor
  • Posts: 38
  • Country: de
Re: LM399 based 10 V reference
« Reply #394 on: January 17, 2016, 02:43:26 pm »
Will be interesting to see how the result for the LM399+AD587 combination will work.
I'm also a little bit curios...

Quote
I have thought of testing this but with the SVR-T I think some of the benefits vanished.
Indeed! My breadboarded and not perfectly compensated unit runs stable since 7 weeks, varying only 15-20µV over the changing of room temperature.

Quote
I think Joe did the combination to get away with the worse temperature coefficient of the AD587LQ compared to the LM399.
Yes, but i think, most of the TC is produced by the resistors and the amp and not by the unheated zener.
Maybe there will be only a small effect to the total TC.

Frank
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2447
  • Country: de
Re: LM399 based 10 V reference
« Reply #395 on: January 17, 2016, 05:09:12 pm »
Hello,

I am missing the cirquit diagram for your PCB.

One problem might be to set the current stable (1mA within 0.1%) through the LM399 zener.
The AD587 has a 4K source resistance at the NR pin.
The voltage at this pin is most probably around 7V-7.5V with some tolerance.
(Just guessing from the +3/-1% trimming range).

For the resistor network:
from the datasheet the T.C. is trimmed in some way.
So either the zener current is adjusted to "zero T.C. current"
or the resistor network itself is done non-linear to compensate for the T.C. of the zener.
Would be interesting how different AD587 behave for the 2 halves of cirquit.

With best regards

Andreas
 

Offline Galaxyrise

  • Frequent Contributor
  • **
  • Posts: 526
  • Country: us
Re: LM399 based 10 V reference
« Reply #396 on: January 17, 2016, 05:13:01 pm »
Indeed! My breadboarded and not perfectly compensated unit runs stable since 7 weeks, varying only 15-20µV over the changing of room temperature.
That could easily be variations in the DMM if you haven't compensated for it
I am but an egg
 

Offline F64098

  • Contributor
  • Posts: 38
  • Country: de
Re: LM399 based 10 V reference
« Reply #397 on: January 17, 2016, 06:36:48 pm »
I am missing the cirquit diagram for your PCB.

Mine is too big for posting.
It's like http://patentimages.storage.googleapis.com/US7382179B2/US07382179-20080603-D00000.png
with
R3 = 3k
R4 = 820R
R7 = 3k01
R8 = 200R
R9 = 7k87

I added a 1µF for noise reduction on #101 and the TC compensation resistors from the SVR-T.

Quote
One problem might be to set the current stable (1mA within 0.1%) through the LM399 zener.
All resistors are 0,1% with 25ppm.
 
Quote
The AD587 has a 4K source resistance at the NR pin.
The internal source will be nearly total suppressed by the LM399.

Quote
Would be interesting how different AD587 behave for the 2 halves of cirquit.
I bought several LM399 and AD587, but they are still on their way to my work bench.
As a fresh infected volt-nut my stock is already virgin.

With best regards

Frank
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 6665
  • Country: de
Re: LM399 based 10 V reference
« Reply #398 on: January 17, 2016, 08:38:00 pm »
Looks like a really clever circuit to use the amplifier and stable resistors in the AD587 for the LM399 reference. The trimming part may need adjustment for individual units. I would have used less current here, just to keep selfheatung lower - but thats a minor detail.
 

Offline F64098

  • Contributor
  • Posts: 38
  • Country: de
Re: LM399 based 10 V reference
« Reply #399 on: January 18, 2016, 01:11:41 pm »
I would have used less current here, just to keep selfheatung lower - but thats a minor detail.

I have taken the values from the SVR-T which are proven as good for the perfect stability of these units.
But you are right, there are some trim circuits out in the web with much higher values and e.g. a 1 megohm resistor between the trimmer and the trim input.
I will get enough PCBs to test all variations.

Frank
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf