Author Topic: LM399 based 10 V reference  (Read 318965 times)

0 Members and 1 Guest are viewing this topic.

Offline SimonSatCom

  • Contributor
  • Posts: 12
  • Country: dk
Re: LM399 based 10 V reference
« Reply #100 on: February 16, 2014, 12:11:27 pm »
Andreas are you referring to this board?

http://www.maxmcarter.com/vref/

And btw...does anyone have any pictures or schematics of the high precision version of the HP3458A reference board?

- Simon
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #101 on: February 16, 2014, 06:32:25 pm »
Yes, thats it.

I do not believe that the high precision version differs from a standard version.
Except that the pre-ageing of the reference at the facility has been either done for a longer time
or by selecting the better boards.

With best regards

Andreas
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #102 on: March 09, 2014, 08:54:36 pm »
Hello,

now after about 1500 hours the LM399 CH6 + CH7 seem to stabilize.
So I plan to repeat the tilt measurements.

With best regards

Andreas

 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1495
  • Country: de
Re: LM399 based 10 V reference
« Reply #103 on: March 09, 2014, 09:23:51 pm »
What a bummer, Dave would say. LM399 CH7 is drifting incredibly and hasn't stopped yet even if it seems to settle to smaller drift rates.
Is it just a variance or is there a correlation to the lead lengths? Your guess is as good as mine, but would be interesting to investigate :)
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: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #104 on: March 09, 2014, 09:40:15 pm »
Just to clarify:

lead length is the same (short leads) for CH6 + CH7.

CH7 uses a slotted board area.
CH6 uses non slotted board.

But I do not think that this has directly something to do with ageing drift.

with best regards

Andreas
 

Offline babysitter

  • Frequent Contributor
  • **
  • Posts: 808
  • Country: de
  • pushing silicon at work
Re: LM399 based 10 V reference
« Reply #105 on: March 10, 2014, 08:14:25 am »
Let me take the role of the "Bild" Newspaper:

SLOTS IN PCB INCREASE AGING DRIFT
Effect discovered by amateur scientist contradicts prior scientific work considering slots to be advantageous. We all gonna die!


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

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1495
  • Country: de
Re: LM399 based 10 V reference
« Reply #106 on: March 14, 2014, 07:51:50 am »
Quote
Let me take the role of the "Bild" Newspaper:

SLOTS IN PCB INCREASE AGING DRIFT
Effect discovered by amateur scientist contradicts prior scientific work considering slots to be advantageous. We all gonna die!

 >:D Don't, there are people out there that could believe in that, see below. One exemplar is not of statistical relevance, thats for sure.

Quote
So, it seems that the slots are counter productive in this use case.  Maybe, it would be better to let your LM399(s) be on a separate PCB from the rest of the circuitry.  That would make the whole LM399 assembly easier to insulate.

No, there is no significant number of examples that proof such an argument. It was just a loud thought of mine, but this larger drift of one exemplar could be nearly everthing and is not constrained by an effect of slots in the board. Maybe I should delete my thought in my previous post? Damn.  :palm:
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 macfly

  • Contributor
  • Posts: 30
  • Country: de
Re: LM399 based 10 V reference
« Reply #107 on: April 06, 2014, 10:29:22 am »
Has anyone an information about the temperature stability of the thermal regulator inside the LM399 ?
This is  a possible question to Bob Dobkin ...

Regards,

macfly
Genius is one percent inspiration, ninety-nine percent perspiration (Thomas Alva Edison 1903)
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #108 on: April 06, 2014, 07:34:53 pm »
Hello macfly,

not so good as in the LTZ1000.

you can see this in my tilting experiment.
https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/msg360779/#msg360779

Further the stability is dependant on heater supply voltage.

You can get good results with stabilized heater voltage and good thermal management.

With best regards

Andreas
 

Offline macfly

  • Contributor
  • Posts: 30
  • Country: de
Re: LM399 based 10 V reference
« Reply #109 on: April 06, 2014, 08:20:24 pm »
Hello macfly,

not so good as in the LTZ1000.

you can see this in my tilting experiment.
https://www.eevblog.com/forum/projects/lm399-based-10-v-reference/msg360779/#msg360779

Further the stability is dependant on heater supply voltage.

You can get good results with stabilized heater voltage and good thermal management.

With best regards

Andreas

Hello Andreas,

very interesting, but not really an answer of my question. But I asked with some thought's unspoken.
Yesterday, I tested my 5 pieces of LM399H in a manner, I have not read anywhere:

1. Supply the zener with ~1mA  - but do NOT supply the heater.
2. Measure the output voltage
3. Put the heater on, wait for thermal balance.
4. Then re-adjust the zener current, to get rid of the zener's dynamic resistance.
5. Measure ouput voltage again.

With these two measurements you may approximately calculate the temperature coeffcient of the device.
If we assume ~90°C working temperature with a stability of +/-0.1 degrees Celsius, the best of my LM399
has a temperature coefficient of only ~0.06ppm/°C.
Now you understand my first question about the stability of the regulation?
I was very astonished about these result.   

Regards,

macfly
Genius is one percent inspiration, ninety-nine percent perspiration (Thomas Alva Edison 1903)
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #110 on: April 06, 2014, 09:34:32 pm »

Now you understand my first question about the stability of the regulation?


Hello macfly

Not really. (how are the exact measurement conditions and what do you mean when speaking of temperature coefficient).

The unheated reference should be in the around 50ppm/K range. (Similar to the LTZ1000).
The heated reference near room temperature is much better.

On my LM399#2 i measured 8uV maximum voltage difference (including hysteresis + noise) in a range of 8-46 degrees celsius environment temperature (measured outside the card box). Note: the voltage (right y-axis) is measured as difference in mV between 2 LM399 references. One at room temperature the other within a "thermal chamber".

So the temperature "coefficient" according to "box method" would be 0.008mV / 6860 mV / (38 K) = +/- 0.015ppm/K.
The temperature "slope" is around -0.07ppm/K as best fit straight line.

But most of the measured error is probably due to thermocouples at the solder joints.
Since I get similar results with a LTZ1000 based reference.

With best regards

Andreas





 

Offline macfly

  • Contributor
  • Posts: 30
  • Country: de
Re: LM399 based 10 V reference
« Reply #111 on: April 07, 2014, 05:48:23 pm »

Now you understand my first question about the stability of the regulation?


Hello macfly

Not really. (how are the exact measurement conditions and what do you mean when speaking of temperature coefficient).

The unheated reference should be in the around 50ppm/K range. (Similar to the LTZ1000).
The heated reference near room temperature is much better.



Hello Andreas,

attached you will find my measurement results. What me surprised is, that the tempco (calculated out of the Vz @ 20°C and Vz @ 90°C)
is more then ten times lower than the 50 ppm/K. With my guess of a regulator stability of 0,1K, one of the LM399 has nearly the same tempco
as a LTZ1000 !

Regards,

macfly

Genius is one percent inspiration, ninety-nine percent perspiration (Thomas Alva Edison 1903)
 

Offline Mickle T.

  • Frequent Contributor
  • **
  • Posts: 342
  • Country: ru
Re: LM399 based 10 V reference
« Reply #112 on: April 07, 2014, 07:33:09 pm »
I have a very different results of the LM399 TempCo measurements.

 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #113 on: April 07, 2014, 08:07:32 pm »
Hello macfly,

the results are too good to believe.
That brings me to the idea to select the references according to raw tempco.

I have looked up my measurements: unfortunately I did not document the values carefully.
But the difference of 2 of my references is at least 4-5mV unheated / heated.

@macfly + Mickle:
Did you use National Semiconductor parts or LT parts?
On my side all are National Semiconductor parts up to now.

@macfly:
Did you kelvin sense of the zener or is the heater current affecting the zener voltage?

With best regards

Andreas
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2517
  • Country: us
Re: LM399 based 10 V reference
« Reply #114 on: April 07, 2014, 09:33:03 pm »
It seems that it will be an even better reason for me to follow this thread, my 6.5 digit meter had a 200uv offset at 10 v.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline Mickle T.

  • Frequent Contributor
  • **
  • Posts: 342
  • Country: ru
Re: LM399 based 10 V reference
« Reply #115 on: April 08, 2014, 04:48:51 am »
Parts from National Semiconductor with the same date code, kelvin connection.
 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2517
  • Country: us
Re: LM399 based 10 V reference
« Reply #116 on: April 08, 2014, 06:04:44 am »
What are the differences in the H vs. AH specced parts?
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #117 on: April 08, 2014, 06:40:39 am »

That brings me to the idea to select the references according to raw tempco.


Hello,

when thinking twice the better idea would be to test the tempco without heater around 90 degrees.
There is a large possibility that the tempco is not linear but either a parabolic or s-shaped curve (due to compensation of the zener with a Vbe).
So it can be that at 25 degrees and 90 degrees the voltage is accidently the same but in worst case with a large drift around 90 degrees.

It seems that it will be an even better reason for me to follow this thread, my 6.5 digit meter had a 200uv offset at 10 v.

200 uV = 20 ppm is below the 35 ppm/year as in the specification of a good 6.5 digit meter. (Which one is it?)
If it includes the first year drift I would not bother about.
Later the drift should go down to 1-2 ppm / year for a good instrument.

What are the differences in the H vs. AH specced parts?
The A-type is selected for a smaller (overall) tempco and smaller initial tolerance (National Semiconductor only). See datasheet.

With best regards

Andreas

 

Offline Vgkid

  • Super Contributor
  • ***
  • Posts: 2517
  • Country: us
Re: LM399 based 10 V reference
« Reply #118 on: April 08, 2014, 03:53:55 pm »
I should have said that there was a 200uv offset between the +\- readings on the 10 v scale, I had to use a 9v battery to get a stable enough reading.
My meter is a Schlumberger Solartron 7065, mine is about 33 years old.
If you own any North Hills Electronics gear, message me. L&N Fan
 

Offline macfly

  • Contributor
  • Posts: 30
  • Country: de
Re: LM399 based 10 V reference
« Reply #119 on: April 08, 2014, 04:48:14 pm »
Hello macfly,

the results are too good to believe.
That brings me to the idea to select the references according to raw tempco.

I have looked up my measurements: unfortunately I did not document the values carefully.
But the difference of 2 of my references is at least 4-5mV unheated / heated.

@macfly + Mickle:
Did you use National Semiconductor parts or LT parts?
On my side all are National Semiconductor parts up to now.

@macfly:
Did you kelvin sense of the zener or is the heater current affecting the zener voltage?

With best regards

Andreas

Hello Andreas,

1. ... That brings me to the idea to select the references according to raw tempco -> No - this was my idea  8)
2. ... the results are too good to believe -> yes, they are too good - and  that was the reason for my post.
3. ... Did you use National Semiconductor parts or LT parts? -> I used NS parts.
4. ... Did you kelvin sense of the zener or is the heater current affecting the zener voltage? -> For sure.

At weekend I will repeat the measurement.

@Mickle
Thanks for your measurement results - they are just like that, what I expected.

Regards,
macfly
Genius is one percent inspiration, ninety-nine percent perspiration (Thomas Alva Edison 1903)
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #120 on: April 08, 2014, 09:16:13 pm »
Hello,

I have now modified my LM399 ageing board with 2 jumpers to switch off the heaters.
There are 16 LM399 from National Semiconductors with 2 different datecodes (12 + 4)
Room temperature was 23 deg C so with around 2 K self heating of the 7mW zener I guess 25 deg for the cold value.

The measured values cold + hot are as follows:

With best regards

Andreas



 

Offline Galaxyrise

  • Frequent Contributor
  • **
  • Posts: 526
  • Country: us
Re: LM399 based 10 V reference
« Reply #121 on: April 09, 2014, 03:19:55 am »
Interesting that the lower TC modules tend to be < 6.9V.  Makes me wonder if the others would have a lower temp Co at a lower current.
I am but an egg
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2426
  • Country: de
Re: LM399 based 10 V reference
« Reply #122 on: April 09, 2014, 04:32:21 am »
Interesting that the lower TC modules tend to be < 6.9V.  Makes me wonder if the others would have a lower temp Co at a lower current.

Hello,

very good eyes !!!

so the "ideal" voltage on my references seems to be around 6875 mV.

In the ageing cirquit I use simple 6K8 Metal film resistors (#4-11) or 10K + 22K in parallel (#12-19)
along with a 15V low drop regulator.

with best regards
 

Offline macfly

  • Contributor
  • Posts: 30
  • Country: de
Re: LM399 based 10 V reference
« Reply #123 on: April 09, 2014, 07:21:15 am »
Hello Andreas,

thanks for yur measurement results. This nearly confirms my measurements.
And as Galaxyrise observed, my LM399 with the lowest TC has the lowest Vz.

Next step to do is changing the zener current and looking what happens ....

Regards,

macfly
 
« Last Edit: April 09, 2014, 10:23:07 am by macfly »
Genius is one percent inspiration, ninety-nine percent perspiration (Thomas Alva Edison 1903)
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 1495
  • Country: de
Re: LM399 based 10 V reference
« Reply #124 on: April 09, 2014, 06:38:55 pm »
Somewhat off topic but interesting for all of us. I'm currently looking for precision resistors and here is what I found up to now:

Distributor   Resistor type   Value   Toleranz   TC   Stability   Price
HBE-Shop   Y14535K00000V9L   5K   0,005%   ± 0,05 ppm/°C   to ± 0.005 % at 70 °C, 2000 h at rated power   34,94 €
HBE-Shop   Y145320K0000V9L   20k   0,005%   ± 0,05 ppm/°C   to ± 0.005 % at 70 °C, 2000 h at rated power   37,81 €
HBE-Shop   Y00625K00000T0L   5K   0,010%   ± 1 ppm/°C   ± 0.005 % at 70 °C, 2000 h at rated power   18,44 €
HBE-Shop   Y006220K0000T9L   20k   0,010%   ± 1 ppm/°C   ± 0.005 % at 70 °C, 2000 h at rated power   not in stock
HBE-Shop   Y07855K00000T9L   5K   0,010%   ± 2 ppm/°C    to ± 0.005 % at 70 °C, 2000 h at rated power   20,53 €
HBE-Shop   Y078520K0000T9L   20k   0,010%   ± 2 ppm/°C    to ± 0.005 % at 70 °C, 2000 h at rated power   21,14 €
                  
Burster   1142   5k   0,010%   ± 2 ppm/°C   0.01% over years   23,13 €
Burster   1142   20k   0,010%   ± 2 ppm/°C   0.01% over years   23,13 €
                  
Rohpoint   8G16D 5K   5k   0,100%   ± 3 ppm/°C   ±35ppm/10,000hours ±50ppm/26,000 hours   4,52 €
Rohpoint   8G16D 20K   20k   0,100%   ± 3 ppm/°C   ±35ppm/10,000hours ±50ppm/26,000 hours   5,30 €
Rohpoint   8G16A 5K   5k   0,010%   ± 3 ppm/°C   ±35ppm/10,000hours ±50ppm/26,000 hours   6,80 €
Rohpoint   8G16A 20K   20k   0,010%   ± 3 ppm/°C   ±35ppm/10,000hours ±50ppm/26,000 hours   7,95 €


The hermetically Vishay types seem to be unobtainium and have long waiting time. Most of the Vishay parts seem to be not in normal stock, a simple of the shelf (these are my words because its standard datasheet value and not custom made) DSMZ divider needs at least 3 months to arrive. Are you guys at Vishay kidding me? For a proof of concept this is much to long and inadmissible. The price is for sure fair:

1-24 pc.   25-49 pc.   50-99 pc.
21,05 €   15,30 €   12,40 €

Someone with other and better resisitor sources out there?
« Last Edit: April 09, 2014, 09:07:00 pm by branadic »
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
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf