Author Topic: LM399 based 10 V reference  (Read 355706 times)

0 Members and 2 Guests are viewing this topic.

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1025 on: August 28, 2019, 08:35:38 pm »
I've been bending the leads a little bit (using 2.54mm pitch socket). Hopefully that stress does not contribute to the popcorn too..
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2590
  • Country: de
Re: LM399 based 10 V reference
« Reply #1026 on: August 29, 2019, 08:53:30 am »
Hello,

I think that is no issue as long as you do not damage the glass seals.
I have here one LM399 which is affected by humidity.
(Seasonal changes of the output voltage).
So I think the hermetical seal is broken.

with best regards

Andreas
 
The following users thanked this post: imo

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1027 on: August 29, 2019, 11:29:43 am »
Interesting.. Pretty hot the 399 package and the humidity has such an effect..  :phew:
 

Offline FriedLogic

  • Regular Contributor
  • *
  • Posts: 88
  • Country: gb
Re: LM399 based 10 V reference
« Reply #1028 on: September 03, 2019, 12:11:07 am »
I've been bending the leads a little bit (using 2.54mm pitch socket). Hopefully that stress does not contribute to the popcorn too..

I got some 2mm pitch sockets (2x2 Preci-Dip 833 Series) for an LM399 ageing box since the spacing is closer to correct and the insertion force looked reasonable. Although unlikely to be a problem, I thought that the less mechanical stress there is the better.
It's not built yet so I can't comment on how well they work (has anyone else tried them for this type of application?)

Even in the few LM399's that I have running I've seen occasional popcorn noise appear or disappear over months, so I'd expect that ageing has some effect, but it could be good or bad.
« Last Edit: September 03, 2019, 12:12:43 am by FriedLogic »
 

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1029 on: September 03, 2019, 09:45:46 am »
I was using the 2.54mm rounded precision ones and my observation had been the touching the 399 while in the socket changed the output voltage by dozens of uVs.
 

Offline Robert763

  • Regular Contributor
  • *
  • Posts: 209
  • Country: gb
Re: LM399 based 10 V reference
« Reply #1030 on: September 05, 2019, 09:09:29 am »
There is a proper socket available that the 399 will fit (0.1" 4 pin square. It's a TE connectivity 8060-1G4
Mouser part No. 571-8060-1G4  Farnell order code 1608447
https://www.te.com/usa-en/product-1-1437504-7.html
Drawing here:
https://www.te.com/commerce/DocumentDelivery/DDEController?Action=showdoc&DocId=Customer+Drawing%7F1437504-3%7F99%7Fpdf%7FEnglish%7FENG_CD_1437504-3_99.pdf%7F1-1437504-7

Robert.

 
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 7509
  • Country: us
    • SiliconValleyGarage
Re: LM399 based 10 V reference
« Reply #1031 on: September 06, 2019, 04:52:01 am »
if anyone needs real lm399's ( not from china ) . i just found about 20 or so in my garage. brand new , NOs , nat semi (old logo )
10$ a piece.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1032 on: September 28, 2019, 11:03:38 am »
What would be the "optimal" value for the RS1..RS4 and CLP1 in case I want to average 4x LM399s?

I plan to use an OPA277 (bip) or OPA2210 (low noise bip), or OPA187 (AZ) there (all are with drift <=0.1uV/C).

PS:

OPA2210  0.4pA/sqrtHz      2.5nV/sqrtHz
OPA277    0.2pA/sqrtHz    12nV/sqrtHz   (fyi - OP07  0.35pA/sqrtHz   11nV/sqrtHz)
OPA187    0.16pA/sqrtHz   20nV/sqrtHz

PPS:

added some calcs, it seems 4x RSx=47k + OP create around 1uVpp (CLP1 not connected).

OPA2120 below should be OPA2210.
« Last Edit: September 28, 2019, 04:20:54 pm by imo »
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 7798
  • Country: de
Re: LM399 based 10 V reference
« Reply #1033 on: September 28, 2019, 03:00:25 pm »
The LM399 has relatively high noise. So it is relatively easy to get get resistors and an OP with lower noise. If filtering is used, it is more the filter function that sets the suitable resistance. To high resistance, bias current drift, leakage and resistor noise set an upper limit. I would consider the 47 K about the upper useful limit, maybe better only 10 K to be conservative with high bias OP.
Without a filter one can go lower - like 1 K if one wants, as there is still not much voltage across the resistors. In most cases I would add at least some filtering, as it's low effort in this position.

The choice of OP also depends on the load to the OPs output - usually no need for something really fancy and the OP07 may be good enough.
 

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1034 on: September 28, 2019, 03:18:15 pm »
When using RSx=100k and CLP1=1uF the BW(-3dB) is 1.6Hz. That lowers the resistor's noise, and let the OPamp noise BW stays the same (??) (say <10Hz due to DMM's sampling period). Sure it is at the edge.

Re noise the OP-07 is similar to the OPA277, the only issue I see with the OP-07 is up to 10x higher TC drift.

PS: fixed RS_CLP1 BW.
« Last Edit: September 28, 2019, 04:22:27 pm by imo »
 

Offline antintedo

  • Contributor
  • Posts: 18
  • Country: pl
Re: LM399 based 10 V reference
« Reply #1035 on: September 28, 2019, 04:03:49 pm »
When using RSx=100k and CLP1=1uF the BW(-3dB) is 1.6Hz.
1-pole filter is far from a brick wall filter, so the effective noise bandwidth is 1.57 * 1.6Hz. Not very important here but good to remember. Also watch for 1/f current and voltage noise of the opamps - parameters are specified at 1kHz but you are most concerned about 0-10Hz region. AD app note MT-048 has some relevant information. To be able to ignore resistor 1/f noise make sure to use good quality metal film or wirewound resistors, crappy thick film ones can be problematic.
 
The following users thanked this post: imo

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 7798
  • Country: de
Re: LM399 based 10 V reference
« Reply #1036 on: September 28, 2019, 04:15:25 pm »
The averaging resistors see rather low voltage. Thus resistor excess noise is not really relevant for these resistors. This is different for the resistor(s) in the 7 to 10 V stage, that see some 7 V.

The 1/f noise can in deed be relevant. Especially the current noise of the BJT based OPs tends to be higher at low frequency.
 

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1037 on: September 28, 2019, 04:25:19 pm »
Therefore the OPA2210 - from DS
Quote
The OPA2210 precision operational amplifier is built on TI's precision superbeta complementary bipolar semiconductor process, which offers ultra-low flicker noise.. SiGe bipolar process..

The Q is how is the current noise at 0.1-10Hz.
A: aprox 22pA p-p (3.3pA rms).
« Last Edit: September 28, 2019, 06:02:09 pm by imo »
 

Offline antintedo

  • Contributor
  • Posts: 18
  • Country: pl
Re: LM399 based 10 V reference
« Reply #1038 on: September 28, 2019, 04:58:18 pm »
Current noise is specified in the datasheet: page 7, figure 4. With 1/f corner frequency and white noise density it is possible to approximately calculate the noise. The app note I mentioned does exactly that.
« Last Edit: September 28, 2019, 05:20:35 pm by antintedo »
 

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1039 on: September 29, 2019, 10:03:11 am »
FYI - the noise analysis of an OP07 and an OPA2210 in a traditional Vref circuit.
The noise of the LM399 not included here.

PS: interestingly, the built-in noise model of the OP07 is pretty optimistic, at least when comparing the OP07 and OPA2210 datasheets :)

PS: added a few popular opamps
« Last Edit: September 29, 2019, 11:01:36 am by imo »
 
The following users thanked this post: antintedo

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1040 on: September 29, 2019, 01:03:52 pm »
A summary of the above noise simulations..

PS: added Vn_out with RS=10k

PPS:
BTW - the chopping frequencies:
                  Datasheet           Simulation peaking at
LTC1050     2.5kHz                2.5kHz
LTC1052     330Hz                 140Hz (model without external 100n caps)
LTC1250     5kHz                   none
LTC2057     30kHz                 30kHz
ADA4522    4.8MHz/800kHz   none (sim up to 10MHz)
« Last Edit: September 29, 2019, 04:20:13 pm by imo »
 
The following users thanked this post: blackdog, TiN, ManateeMafia, Andreas, antintedo

Offline imo

  • Super Contributor
  • ***
  • Posts: 2653
  • Country: 00
Re: LM399 based 10 V reference
« Reply #1041 on: December 23, 2019, 12:32:38 pm »
Fyi - what happens when you add resistance of pcb traces (35um copper, l=length, w=width, includes TC) into the LTspice simulation :)
An example where gnd and power traces are 50mm long with 0.8mm width.
« Last Edit: December 23, 2019, 12:39:22 pm by imo »
 
The following users thanked this post: eplpwr

Online blackdog

  • Frequent Contributor
  • **
  • Posts: 611
  • Country: nl
  • Please stop pushing bullshit...
Re: LM399 based 10 V reference
« Reply #1042 on: December 27, 2019, 10:50:58 pm »
Hi,

Maybe I'll start my own topic, but I'm working on this circuit right now.
The circuit will be placed in a small oven that will be kept at about 42 degrees.

The heaters of the LM399AH will get their own "floating" power supply.
And these power supplies will be connected at one point, that is in the schematic by means of R1.

The output buffer is not in the oven but on a separate circuit board.
If there is enough interest, I can create a new topic about this 10V Reference.

The schematic


Shoot @ it!  ;)

Kind regards,
Bram
“Two things are infinite, the universe and human stupidity, and I am not yet completely sure about the universe.”
 

Offline guenthert

  • Frequent Contributor
  • **
  • Posts: 409
  • Country: us
Re: LM399 based 10 V reference
« Reply #1043 on: December 28, 2019, 12:28:00 am »
[..]
Shoot @ it!  ;)
[..]
  I'd appreciate some stated goal, some specifications to be met, to ground the discussion and to evaluate, whether such specifications can (and once realized) have been met.  This is even more needed, if a not-so-obvious design is chosen (here: why put the LM399, which has its own heater, in an oven?  The LM399 was developed to make ovens or temperature compensation unnecessary).

  Said that, even w/o such stated goals, I can say that you're in hot water there as from the schematics it looks like low on the heater (pin 4 on the LM399) might be on a higher potential than low on the "zener" (pin 2 on the LM399).  The data-sheet says that's a no-no.
 

Offline Andreas

  • Super Contributor
  • ***
  • Posts: 2590
  • Country: de
Re: LM399 based 10 V reference
« Reply #1044 on: December 28, 2019, 09:08:36 am »
the heater (pin 4 on the LM399) might be on a higher potential than low on the "zener" (pin 2 on the LM399).
Not if the heater supply is really "floating".

(here: why put the LM399, which has its own heater, in an oven?  The LM399 was developed to make ovens or temperature compensation unnecessary).
I guess the oven is only for the 7->10V transfer resistors.
And of course it also reduces the Seebeck effect on the Kovar leads of the LM399.

Shoot @ it!  ;)
Id like to hear some of the design decisions / reasons.
Why use a LT1010 (150mA) as low ohmic buffer if you use 50 Ohms series resistors.
(The ADA4522 with  ~80 mA would be sufficient).

- I would use decoupling capacitors parallel to the heater. (47uF + 100nF see also data sheet).
- also 100 nF across the zeners (to reduce EMI).
- D7 (12V) is not clear to me together with 15V supply

with best regards

Andreas
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 7798
  • Country: de
Re: LM399 based 10 V reference
« Reply #1045 on: December 28, 2019, 10:04:41 am »
There is some sense in using an extra buffer, to avoid the heating inside the AD4522 from the load current. However the load current should normally be low, so a normal OP (e.g. good old NE5534) would be good enough.

The zener D7 also confused me at first  it is just for start up. Later on it would see some 5 V and thus not conduct.

Having the LM399 also in the oven is a two sided thing. It helps a little with the Kovar leads - though not very much as the LM399 is internally still much hotter and the average temperature at the other side does not matter that much as the wires are likely reasonably well matched.
The down-side of the LM399 in the oven is that this is a significant heat source. So the oven temperature needs to be higher or more conduction to the outside. This would also need more power to the heater. The parts that could really profit from a stable temperature are the resistors.

The heavy filtering after the reference can have a tricky side: To get the high capacitance one would likely need to use polyester caps and these show quite some DA. I have not done the math and don't know the exact DA numbers for the caps, but from some point on a low frequency filter will need considerable settling time. Not because of the filter RC time constant but the DA settling over the time of several hours. The typical time scale for the popcorn noise of the LM399 is still longer than the filter time constant - so it would not really help with the popcorn noise.
 

Online blackdog

  • Frequent Contributor
  • **
  • Posts: 611
  • Country: nl
  • Please stop pushing bullshit...
Re: LM399 based 10 V reference
« Reply #1046 on: December 28, 2019, 11:14:36 am »
Hi,

I'm going to explain why the schematic is structured like this.

But first this...
I still had a handful of good parts lying around and another one of my component test ovens, so I thought about building a 10V reference with it.
I just want to do the best I can with the parts I have here.
.
The setup shown here will probably only be built once, so I don't take into account the parts costs and labour as this is unimportant to me.

About a month is here in a test box 2x LM399AH to age and from these then two are selected for low noise and absence of popcorn noise.
Andreas, thanks, i forgot the heater capacitor in the next version of the schematic is show it wil be there.

R4 and R5 mix the reference voltage of the two LM399AH IC's and these resistors are also used to create the low pass filter together with C1 to C6.
R6 helps to protect the +input of the opamp a little with the charge of the capacitors when switching off.

The opamp + buffer is besides a 1.4x amplifier also a low pass filter.
This is mainly meant to keep the noise of the amplifier and the buffer low.

The measurement of the noise without already being optimized was about 1uV RMS measured with my Audio Precision measuring set at a 22KHz bandwidth.
Of course this was using the 400Hz high pass filter.
This was necessary because despite the decent build on a breadboard en some shielding the 230V 50Hz field is still present.

This measurement was done in the test setup with 1x LM329 as zener.
But the noise tests were meant to find out which opamp works best with the precision resistors I have at my disposal.

Yesterday I tested three opamps, the LT1007AC, the OPA140 and the ADA4522 which is currently doing the best in my test setup.

Andreas, you're asking for the extra buffer if the ADA4522 can deliver peak currents of up to 80mA?
I don't like that this precision opamp is directly connected to the output.
My starting point was also, that the 10V output can be abused.  :box:

D7 and R20 is used for startup, this was not necessary for an ADA4522, but it was sometimes necessary for the OPA140 and LT1007.
When power on this circuit, it pulls the input of the LT1010 buffer positive for a moment.
In normal operation D7and R20 do nothing, a little leakage current if there is one, is controlled by the opamp loopgain, normaly that wil be no problem.
I'm going to do some more tests on these parts, to see if these two components can be connected directly to the +15V output, then they're completely out of the loop.

Q1 limits the output current, and in my test circuit this is a little less than 35mA short circuit current.

The problem I run into at the moment, is that I am against the noise floor of my Audio Precision Analyser...
This is about 0.8uV RMS in 22KHz bandwith.

A simple extra preamplifier is not possible.
This should be a symetrical preamplifier with low noise, high common mode suppression and a not no too low input impedance.
I am thinking about how I am going to build it.

So, this schematic isn't ready yet, but I'm satisfied with the output buffer.
It's solid and relieves the opamp of eny load and overload.

I also tested some of my transistor buffers after the opamp, but that wasn't a good plan as far as the noise behavior of the circuit is concerned.

This needs to be tested further:
R10 + C8 and C10 with a small series resistor of say 10-Ohm and a 1uF capacitor directly over the output.
This to keep enough bandwidth to control the noise behavior of the LT1010 buffer.

The 2nd order filtering of the opamp + buffer is only meant to keep the noise of this circuit as low as possible, it is not meant to filter the noise out of the 2x LM399.
R4 +R5 and the 6x 10uF have a much lowerfilter corner.

So, this is not a circuit to become a better than an LTZ1000 reference, that was not the starting point.
But to build as good a circuit as possible with the present parts.
By building the temperature sensitive parts in a oven, I have almost no more problems with the Seebeck effect.

These are de components for R4 and R5, R7, R8 and R13, R14 and one of the 10uF capacitors.



When I've done some more tests, I'll show you the results here.

Thanks for your remarks!

Kind regards,
Bram
« Last Edit: December 31, 2019, 09:10:54 am by blackdog »
“Two things are infinite, the universe and human stupidity, and I am not yet completely sure about the universe.”
 
The following users thanked this post: Andreas

Online blackdog

  • Frequent Contributor
  • **
  • Posts: 611
  • Country: nl
  • Please stop pushing bullshit...
Re: LM399 based 10 V reference
« Reply #1047 on: December 28, 2019, 01:00:31 pm »
Hi Kleinstein,  ;)

About the buffer, yes you can use a NE5534a as a buffer if its compensated well.
But i think i have more then 100 pieces of all kind of buffer IC's, so why not use some of them...
And the LT1010 is a great buffer.

The oven is not initially intended to help the LM399, but of course it does help to make the effects of ambient temperature even smaller.
As far as my knowledge of the documentation on the LM199 zener is concerned, a constant ambient temperature helps to increase stability.
Also some extra shielding helps the stability because this limits the power of the heater, so if the space allows it in my oven,
there will be some insulation around the two LM399 ICs.
The difference between the LM399 oven temperature and de main oven temperature is big, normaly > 40C.

The main reason for the oven is to keep the opamp circuit as stable as possible.

The difference between the oven in the LM399 and the main oven is big enough for a good control of the oven in the LM399.
Because I keep the buffer IC and the output load resistor outside the oven, I have enough control capacity left for the main oven at about 30C LAB temperature.
With some simple insulation, the main oven consumes about 1.5-Watt.

Filtering
Concerning the filtering of the zener noise, only the passive filter and putting the two LM399 ICs in parallel really helps by the low corner of 0.3Hz.
I wanted to do something different this time, several times I used a bootstrap electronic capacitor filter.
Such a filter(bootstrap) is also in my main LAB reference with 4x an LT1021.

I'm going to do another test with the 6x 10uF capacitor, but the current test has been done with the same series of capacitors only of a higher voltage and high capacity, these only don't fit in the oven.

Starting up the oven naturally takes time to reach the equlibrium.
The slow reaching of the maximum charge of the capacitors is therefore not such a problem.

Popcorn!
By selecting the LM399 IC's I hope to have as little popcorn as possible.

Kind regards,
Bram

« Last Edit: December 29, 2019, 09:07:51 am by blackdog »
“Two things are infinite, the universe and human stupidity, and I am not yet completely sure about the universe.”
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 7798
  • Country: de
Re: LM399 based 10 V reference
« Reply #1048 on: December 28, 2019, 02:56:10 pm »
Compared to a relatively large extra oven the capacitors may in deed be fast, though slower than one may think.
With a dielectric absorption in the 0.5% range the equivalent circuit for a 100 µF capacitor has something like 0.5 µF with a series resistor in the GOhms range in parallel. With a low RC cross over this extra part can become a problem, as 1 Gohms and 10 K from the references give an effect in the 10 ppm order of magnitude. So settling may take a few of the Gohm * 0.5 µF (  = 500 second range) time constants. So it could take up to about 1 hours for the DA to finally settle to less than ppm level of effect. The resistance value in the equivalent circuit depends on the DA and may change with temperature.

Chances are one could still be just OK with the 60 µF, but one should keep this effect in mind if the filter cross over gets lower than 0.3 Hz. At some point (maybe from 0.1 Hz) on one would have to wait for the DA (and thus more like hours) and not just the filter RC (minutes).
 

Online blackdog

  • Frequent Contributor
  • **
  • Posts: 611
  • Country: nl
  • Please stop pushing bullshit...
Re: LM399 based 10 V reference
« Reply #1049 on: December 28, 2019, 03:29:49 pm »
Hi Kleinstein,

I measured one of the 10uF WIMA MKS 2 capacitors on my Philips PM6303 Bridge.
F = 1KHz

C   = 9,9uF
Q   = 240
D   = 0.004
RP = 3.8K
RS = 0.07-Ohm
Z    = 16.08-Ohm
Cp = 9.898K
Cs = 9.898K

Kind regards,
Bram
« Last Edit: December 29, 2019, 09:08:18 am by blackdog »
“Two things are infinite, the universe and human stupidity, and I am not yet completely sure about the universe.”
 
The following users thanked this post: 3roomlab, maat


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf