### Author Topic: LM399 based 10 V reference  (Read 355655 times)

0 Members and 1 Guest are viewing this topic.

• Super Contributor
• Posts: 1670
• Country:
##### Re: LM399 based 10 V reference
« Reply #350 on: November 25, 2015, 02:58:25 pm »
Don't trust pictures of a single exemplar, better trust the maximum limits.
« Last Edit: November 25, 2015, 07:48:08 pm by branadic »
Metrology Meeting 2020 is canceled. Looking forward for MM2021

#### Andreas

• Super Contributor
• Posts: 2590
• Country:
##### Re: LM399 based 10 V reference
« Reply #351 on: November 25, 2015, 10:19:23 pm »
I suspect that the noise level on an unheated LM329 should be considerably lower than on the LM399 at 90C.

I suspect that they have only "beautified" the diagram.

25 deg -> ~ 300K
85 deg -> ~ 360K
why is sqrt(300K) = 17.3 significant lower than sqrt(360K) = 19
when aproximating the noise as Johnson noise?

with best regards

Andreas

#### Alex Nikitin

• Super Contributor
• Posts: 1021
• Country:
##### Re: LM399 based 10 V reference
« Reply #352 on: November 25, 2015, 11:02:47 pm »
I suspect that the noise level on an unheated LM329 should be considerably lower than on the LM399 at 90C.

I suspect that they have only "beautified" the diagram.

25 deg -> ~ 300K
85 deg -> ~ 360K
why is sqrt(300K) = 17.3 significant lower than sqrt(360K) = 19
when aproximating the noise as Johnson noise?

with best regards

Andreas

Johnson noise calculations won't work here. Just have a look at the noise density graph available in the LM399 datasheet for 25C and 90C - the increase is about 1.5 times (and most likely more in the low 1/f region).

Cheers

Alex

#### Alex Nikitin

• Super Contributor
• Posts: 1021
• Country:
##### Re: LM399 based 10 V reference
« Reply #353 on: November 26, 2015, 11:51:26 am »

You could also use the LM329 (= LM399 without heater) which is cheaper but unfortunately no longer sold in metal can case.

With best regards

Andreas

Hi Andreas,

I've just bought some NOS top grade LM129A from eBay and there are more available.

Cheers

Alex

#### Andreas

• Super Contributor
• Posts: 2590
• Country:
##### Re: LM399 based 10 V reference
« Reply #354 on: November 27, 2015, 08:53:21 pm »

Johnson noise calculations won't work here. Just have a look at the noise density graph available in the LM399 datasheet for 25C and 90C - the increase is about 1.5 times (and most likely more in the low 1/f region).

Cheers

Alex

Hello Alex,

attached 4 pictures of noise measurement of my LM399#20
2 with heater on
2 with heater off.
all other: same conditions. Measured in cake box with additional thermal shielding.
X-axis 1 sec/div, Y-axis 0.5uV/div

So you should be able to tell which are with heater active and which with heater disabled.

$15 for a half LM399 is a bit too pricey for me. With best regards Andreas #### Alex Nikitin • Super Contributor • Posts: 1021 • Country: • Femtoampnut and Tapehead. ##### Re: LM399 based 10 V reference « Reply #355 on: November 27, 2015, 09:44:30 pm » So you should be able to tell which are with heater active and which with heater disabled.$15 for a half LM399 is a bit too pricey for me.

With best regards

Andreas

1) No, I can not tell the difference, sorry! Which means the datasheet is not to be trusted  .

2) It is \$15 for 2 and I've agreed a better price for 10 lots (20 devices in total).

Cheers

Alex

#### Andreas

• Super Contributor
• Posts: 2590
• Country:
##### Re: LM399 based 10 V reference
« Reply #356 on: November 27, 2015, 09:59:36 pm »

1) No, I can not tell the difference, sorry! Which means the datasheet is not to be trusted  .

Hello,

the stray from measurement to measurement is relative large on LM399.
Most is due to thermocouple effects on the (KOVAR) legs.

And also the stray from device to device is large.
I have devices with 2-3 uVpp (Like LM399#20)
and others with around 9 uVpp (LM399#2).

Interestingly the LM399 out of a HP34401A also measured 2-3uVpp.

With best regards

Andreas

#### Gyro

• Super Contributor
• Posts: 5862
• Country:
##### Re: LM399 based 10 V reference
« Reply #357 on: November 28, 2015, 07:17:28 pm »
I guess even if the noise level difference is difficult to spot, it's likely that there will be an improvement in long term drift at lower temperature. Personally I find that more of a worry to me due to infrequent opportunities to check calibration.

Chris
Chris

"Victor Meldrew, the Crimson Avenger!"

#### TiN

• Super Contributor
• Posts: 4195
• Country:
• xDevs.com/live - 24/7 lab feed
##### Re: LM399 based 10 V reference
« Reply #358 on: December 25, 2015, 04:06:58 am »
I could not help it, curiosity finally won. Test of LM399 in liquid nitrogen.

Only zener is connected via onboard 3x10.7K resistors. Input power +12VDC from linear Agilent PSU.
Sampling 34970A. Room zener voltage was 6.915VDC, cold voltage - 7.0319VDC. Measured thermocouple Type T temperature -194.72°C.
« Last Edit: December 25, 2015, 04:12:19 am by TiN »
YouTube | Metrology IRC Chat room | Live-cam | Share T&M documentation? Upload! No MB limit, firmwares, photos.

#### SeanB

• Super Contributor
• Posts: 15425
• Country:
##### Re: LM399 based 10 V reference
« Reply #359 on: December 25, 2015, 06:36:19 am »
Now put it back into a drift check, and see if you have reset the drift curve from the thermal shock.

#### TiN

• Super Contributor
• Posts: 4195
• Country:
• xDevs.com/live - 24/7 lab feed
##### Re: LM399 based 10 V reference
« Reply #360 on: December 25, 2015, 10:49:28 am »
I want to get more data, so will back it to ambient on monday. Taking readings every minute. So far it was about 3uV peak-peak but with many "popcorn"-style jumps few tens of uV back and forth. Tried more current also caused bigger jumps. I have CSV for today, will plot data later today.

Also LN2 temp is not that stable, it vary about 0.1K over minutes. Dewar is not pressured, so likely LN2 getting liquid air and oxygen causing temperature to vary.
« Last Edit: December 25, 2015, 10:51:20 am by TiN »
YouTube | Metrology IRC Chat room | Live-cam | Share T&M documentation? Upload! No MB limit, firmwares, photos.

#### Theboel

• Frequent Contributor
• Posts: 267
• Country:
##### Re: LM399 based 10 V reference
« Reply #361 on: December 25, 2015, 12:21:34 pm »
@ TiN
wow You are faster and cooler I been thinking to put LM399 in the freezer with -3 C and turn off the heater to see what happened, just curious because my freezer always on 24/7

#### Dr. Frank

• Super Contributor
• Posts: 1881
• Country:
##### Re: LM399 based 10 V reference
« Reply #362 on: December 25, 2015, 02:51:59 pm »

Also LN2 temp is not that stable, it vary about 0.1K over minutes. Dewar is not pressured, so likely LN2 getting liquid air and oxygen causing temperature to vary.

Liquid N2 normally is quite stable, indeed.

In fact, pure liq-N2 (~77.2K) condenses air, and therefore, the oxygen part of this mixture will slowly raise the temperature, towards the 90K of liquid air. But that takes hours to reach maybe 80K, depending on the volume of the dewar, and the temperature reading will not flicker on minutes basis, instead goes  steadily upwards. Anyhow, you should cover the opening of the dewar with cork, let a small hole anyhow, for the evaporating gas.
Also, changes of atmospheric pressure may slightly change the temperature of the liquid N2, but that also does not happen within minutes.

Your sensor is not of best quality, and very sensitive. A PT100 will deliver much more stable readings.

Frank
« Last Edit: December 25, 2015, 02:59:42 pm by Dr. Frank »

#### TiN

• Super Contributor
• Posts: 4195
• Country:
• xDevs.com/live - 24/7 lab feed
##### Re: LM399 based 10 V reference
« Reply #363 on: December 25, 2015, 03:38:54 pm »
Here is the data, over 7 hr span. This is rather experiment to see if it's worth to spend any time on this direction, or forget it. As there are ways to get these temperatures for 24/7 without pouring LN2, if there are visible benefits of doing so. Will see back monday how it behave, after which I will leave same module to log two more days with ambient temperature, so we can compare both datasets.

Dewar is Air Liquide TR26 with ~6 liters of LN2, not pressurized.

Those jumps are ~2ppm each. Perhaps due nitrogen cavitation and bubbles?

CSV with all the wobbles.
« Last Edit: December 25, 2015, 04:10:56 pm by TiN »
YouTube | Metrology IRC Chat room | Live-cam | Share T&M documentation? Upload! No MB limit, firmwares, photos.

#### Dr. Frank

• Super Contributor
• Posts: 1881
• Country:
##### Re: LM399 based 10 V reference
« Reply #364 on: December 25, 2015, 04:09:55 pm »
You should not leave this dewar open! That's very important!
Otherwise, Air and humidity will enter the vessel.
If you put a cork or similar on top, not tight, of course, the evaporating gas causes a slight overpressure, which will prevent any gases entering  from outside.

26l of liquid nitrogen.. That's a lot of stuff. The temperature should be rock stable.
Not pressurized simply means that it is always under atmospheric pressure.

The fluid boils a little bit, but no 'cavity' will be inside..

Also test a better thermometer, that will give directly stable results.

#### TiN

• Super Contributor
• Posts: 4195
• Country:
• xDevs.com/live - 24/7 lab feed
##### Re: LM399 based 10 V reference
« Reply #365 on: December 25, 2015, 04:15:32 pm »
There is hardfoam cap on it, so it's not fully open. I have pressure valve device for it, but seal O-ring is broken so I did not use it. Perhaps I can repeat test next weekends, but will need bring few of my Keithley's to lab. There is nothing better than 34970 atm

Also there is actually about 5-6L of nitrogen in it only as for now.
YouTube | Metrology IRC Chat room | Live-cam | Share T&M documentation? Upload! No MB limit, firmwares, photos.

#### Dr. Frank

• Super Contributor
• Posts: 1881
• Country:
##### Re: LM399 based 10 V reference
« Reply #366 on: December 25, 2015, 05:10:17 pm »
Thermo elements usually have very low thermoelectric voltages at lower temperature.
This type T has -15uV/K only.
The fluctuations you see, are equivalent to 3uV only.
That easily explains these pseudo temp measurement changes.

I used a special Kondo effect thermo couple (*), with much higher sensitivity down to 4.2K.
It showed such dips also, afair.
So I replaced it with a high grade PT100 down to about  25K, and a special diode thermometer down to 1.4K, which both gave much more stable measurements.

It is not necessary to pressurize liq. N2, otherwise you want to create lower temperatures by pumping on the cryogenic liquid.

Frank

(*) NiCr/FeAu, I think
« Last Edit: December 25, 2015, 05:12:06 pm by Dr. Frank »

#### TiN

• Super Contributor
• Posts: 4195
• Country:
• xDevs.com/live - 24/7 lab feed
##### Re: LM399 based 10 V reference
« Reply #367 on: December 25, 2015, 05:20:23 pm »
Have Honeywell HEL-705 1k RTDs so can use those.
YouTube | Metrology IRC Chat room | Live-cam | Share T&M documentation? Upload! No MB limit, firmwares, photos.

#### Dr. Frank

• Super Contributor
• Posts: 1881
• Country:
##### Re: LM399 based 10 V reference
« Reply #368 on: December 25, 2015, 05:30:12 pm »
Have Honeywell HEL-705 1k RTDs so can use those.

So the noise on the LM399 output should have nothing to do with any changes of the N2 temperature.
Maybe the silicon does not like temperature shocks.

#### Kleinstein

• Super Contributor
• Posts: 7788
• Country:
##### Re: LM399 based 10 V reference
« Reply #369 on: December 25, 2015, 05:54:39 pm »
I don't think it's the temperature that cause the 10 µV steps. The steps are rather uniform an look quite fast. So it's more an chip internal thing similar to popcorn noise.

I don't think a low temperature stabilized reference is really practical, as this also causes quite some trouble with humidity and even condensation. One nice thing with a heated reference is that relatve humudity is low (e.g. less than about 50% for 10 degrees over ambient). So I would definitely prefer something like 40-50 C so well above ambient, but not as high as the oroginal LM399.

Also the  way to low noise is using a rather large reference or several in parallel, thus quite some heat source.

#### Alex Nikitin

• Super Contributor
• Posts: 1021
• Country:
##### Re: LM399 based 10 V reference
« Reply #370 on: December 25, 2015, 06:17:46 pm »
So I would definitely prefer something like 40-50 C so well above ambient, but not as high as the oroginal LM399.

I've bought some LM129A zeners to try for that very reason. One interesting side of using the LM129/329 is that (unlike the LM399) it is graded by tempco, with the "A" grade is under 10ppm/C from -55C to +125C and the "D" grade is under 100ppm/C from 0C to +70C. In the LM399 all that range is randomly spread.

Cheers

Alex

#### Alex Nikitin

• Super Contributor
• Posts: 1021
• Country:
##### Re: LM399 based 10 V reference
« Reply #371 on: December 25, 2015, 10:03:51 pm »
Hmm, interesting, thank you. I was also thinking that higher noise of the 399 is due to the low Zener current. In the JFET based reference it is not a problem though. I already have results close to the buried Zener in the noise and tempco, if the long term stability is good, with a simple oven it would be an interesting alternative approach.

Cheers

Alex

• Super Contributor
• Posts: 1670
• Country:
##### Re: LM399 based 10 V reference
« Reply #372 on: December 27, 2015, 05:09:42 pm »
Somewhat off-topic, but I found LM329 spice model, which means a model of LM399 without the heater:

http://www.diyaudio.com/forums/power-supplies/236257-lm329-ltspice-model-needed.html

Attached you find the Subcircuit (.sub) and the Symbol (.asy) for LTSpice.

There is another Spice Model simulating additional parameters: http://www.ece.utah.edu/~ece2280/reg_zener.lib
« Last Edit: December 27, 2015, 05:12:56 pm by branadic »
Metrology Meeting 2020 is canceled. Looking forward for MM2021

#### Macbeth

• Super Contributor
• Posts: 2562
• Country:
##### Re: LM399 based 10 V reference
« Reply #373 on: December 27, 2015, 06:33:29 pm »
Thanks, that reminds me that I have the other half of my rubbish attempt at drawing up those schematics in LTSpice.

Find attached my heater schematic. I'm sure it is laughable. Perhaps someone who knows what they are doing can combine the models and make a proper netlist?

#### Andreas

• Super Contributor
• Posts: 2590
• Country:
##### Re: LM399 based 10 V reference
« Reply #374 on: January 06, 2016, 11:02:22 pm »
Hello,

again a ageing chart of 2 LM399 references.
now after 2 years (more than 730 days) 24/7 operation.

And probably the last time. Since I may need the multiplexer
channels for two brand new LTZ1000A references soon.

https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/msg796829/#msg796829
https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/msg705039/#msg705039
https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/msg566721/#msg566721

with best regards

Andreas

Smf