Author Topic: LM399 based 10 V reference  (Read 532624 times)

0 Members and 2 Guests are viewing this topic.

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2007
  • Country: br
    • CADT Homepage
Re: LM399 based 10 V reference
« Reply #1250 on: June 15, 2022, 10:00:04 am »
If an analog filter for the PWM divider cannot be used, it can be implemented in the digital domain as well. This would be a hybrid design with a resistor divider including a high resolution resistor DAC for adjustment. This gain stage would by slaved to a PWM divider that is wired as precision ratiometer, with a primitive analog filter to make it a sigma-delta ADC (similar to integrator in a precision DVM). With some digital control loop to continuously measure the division ratio and readjust the DAC one should get a ripple free output with the precision of the PWM.

Regards, Dieter
 

Offline iMo

  • Super Contributor
  • ***
  • Posts: 4627
  • Country: nr
  • It's important to try new things..
Re: LM399 based 10 V reference
« Reply #1251 on: June 20, 2022, 06:28:39 pm »
While playing with LTSpice and simulating the "classic" schematics (with an npn buffer transistor after the opamp) I can see the transistor significantly lowers the PSRR of the entire circuit. The typical PSRR values I see are 30-100x only (the higher Vcc the higher PSRR). Is that caused by the transistor amplifying somehow the Vcc noise? I always thought the diodes (after the opamp) and the transistor are within the op-amp loop..
For example - below - the 10mVpp square noise pulses (50kHz) injected into the Vcc translate into 175uVpp at the 10V output..

PS:
With C1=100p and 10mVpp 50kHz square the 10V ripple is 144uVpp.
With C1=100p and 10mVpp 5kHz square the 10V ripple is 133uVpp.

With C1=100n and 10mVpp sine 100Hz the 10V ripple is 62uVpp.
With C1=10n and 10mVpp sine 100Hz the 10V ripple is 6uVpp.
With C1=1n and 10mVpp sine 100Hz the 10V ripple is 0.95uVpp.
With C1=100p and 10mVpp sine 100Hz the 10V ripple is not measurable (nil).

« Last Edit: June 20, 2022, 07:14:53 pm by imo »
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14017
  • Country: de
Re: LM399 based 10 V reference
« Reply #1252 on: June 20, 2022, 06:43:06 pm »
The capacitor C1 slows down the OP loop quite a bit. So it can not compensate for the fast transients. If the relatively large capacitor is needed, one should also filter the supply sufficently to reduce the fast part in the supply ripple.
 
The following users thanked this post: iMo, RoadDog

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2377
  • Country: de
  • Sounds like noise
Re: LM399 based 10 V reference
« Reply #1253 on: June 20, 2022, 07:19:38 pm »
Quote
To be honest i looked into this after reading the 2019 reports of Andreas and branadic on their experiments with PWM gain stages (see above). Meanwhile they seem to have better results, too. As far as i remember they initiated kind of a calibration circle in May, using a PWM.

The reference only uses the modification suggested by Kleinstein (see attachement). It was then monitored with a K2182A @ 1NPLC without anything obvious and afterwards trimmed to a certain value that I can not tell at the moment as the (blind) ring comparison is still ongoing.

-branadic-
« Last Edit: June 21, 2022, 05:28:27 am by branadic »
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 
The following users thanked this post: MiDi

Offline n_haku

  • Contributor
  • Posts: 15
  • Country: ru
Re: LM399 based 10 V reference
« Reply #1254 on: September 25, 2022, 09:40:23 am »
Sorry of sudden question (and hope it's right place):
Though I start interest in voltnutting pretty recently (maybe approx a year) I see a lot of zener's buffer here, in datasheets (like on 1st img) and in DVM's schematics, and almost in all cases zener startup is like on 1st img,  but why not use jfet startup, like in img2? It provide reliable startup and almost no current after, so high PSRR. I see no drawbacks in this setup, but as almost no one incl big companies use it, seems I miss something? Well, I could guess a HF noize feedthru via parasistic capacitance, but resistor behaves same, and this could be reduced with bead.
PS: I ended assembling alpha version of pure DAC reference scaler, it's intended to use w/lm399, but as it's hard to buy it for me now, I use jfet (as in JVR thread). It's ok to post here, or create dedicated thread?
My lord! We need more precision!
 

Offline MiDi

  • Frequent Contributor
  • **
  • Posts: 599
  • Country: ua
Re: LM399 based 10 V reference
« Reply #1255 on: September 25, 2022, 11:00:10 am »
Nothing wrong in using JFET for startup, just additional cost.
With non RRO op amp and op amp V- >= Vz- the startup circuit could even be omitted.
PSRR for startup resistor is dRz * dVsup / (Rs *  Vz), dRz: dynamic impedance of zener, dVsup: change in supply voltage, Rs: startup resistor, Vz: zener voltage.
For Rs = 200k, dRz = 1.5 it is around 1ppm/V (ADR1399 around 0.1ppm/V).
So if the supply is stable to (or noise) better than 100mV, a startup resistor has negligible effect on zener voltage.
 

Online Andreas

  • Super Contributor
  • ***
  • Posts: 3207
  • Country: de
Re: LM399 based 10 V reference
« Reply #1256 on: September 25, 2022, 12:21:51 pm »
Well, I could guess a HF noize feedthru via parasistic capacitance, but resistor behaves same, and this could be reduced with bead.
Hello,

a JFET is much more sensible to RF than a resistor.
I used single JFETs as 3-5V voltage regulators.
But without 100 nF at in and output the JFET likely oscillates.
So your cirquit is missing at least 1 capacitor.

See here:

https://www.eevblog.com/forum/metrology/ad587lw-10v-precision-travel-standard/msg1449488/#msg1449488
https://www.eevblog.com/forum/metrology/ad587lw-10v-precision-travel-standard/?action=dlattach;attach=402886

with best regards

Andreas
 

Offline n_haku

  • Contributor
  • Posts: 15
  • Country: ru
Re: LM399 based 10 V reference
« Reply #1257 on: September 25, 2022, 02:17:43 pm »
Nothing wrong in using JFET for startup, just additional cost.
Well, I had feeling that answer is likely as you described: it definetly makes sense with jfet based reference, as it have pretty high impedance (typ few kOhms), and just useless additional cost for ordinary zener.

I used single JFETs as 3-5V voltage regulators.
What a coincidence, my scaler use jfet as zero-current regulator for pwm microcontroller. Of course here and there capacitors on power lines.  Just checked noise with ACV mode in DVM, seems all ok.
Also, I think this irrelevant to startup, as there will be almost no current. Anyway, thanks for information.
My lord! We need more precision!
 

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2007
  • Country: br
    • CADT Homepage
Re: LM399 based 10 V reference
« Reply #1258 on: September 26, 2022, 02:19:56 pm »
Another cheap and perfect solution is with a second zener, connected from Vcc. Of course for this to work, a stable supply is required - which is a good idea anyway.
Or a voltage divider from Vcc with a diode as gate (2 resistors + diode).

Regards, Dieter
 
The following users thanked this post: MegaVolt, robert.rozee

Offline DavidAzulay5000

  • Regular Contributor
  • *
  • Posts: 76
  • Country: il
Re: LM399 based 10 V reference
« Reply #1259 on: March 22, 2023, 10:55:07 am »
my 10 volt ref
base on LM399A /LT3042 /LTC2057

« Last Edit: March 22, 2023, 03:09:03 pm by DavidAzulay5000 »
You don't need pants for the victory dance
Cuz Baboon's better than Weasel
 

Offline DavidAzulay5000

  • Regular Contributor
  • *
  • Posts: 76
  • Country: il
Re: LM399 based 10 V reference
« Reply #1260 on: March 25, 2023, 08:28:33 am »
Can you be interested in precision cookies?

48 hours in a row
with a voltage of 10.000.00
the LM399A zener output voltage 7.086.15V
the lt3042 output volt 15.000.2V
Total current consumption of the circuit 27 milliamps.

During the transitional season in Israel
that at night it is relatively cold 12 degrees and in the day 23 degrees
The change ranges from
10.000.00
to
10.000.03
This LM399A has gone through about 500 working hours
« Last Edit: March 25, 2023, 09:09:21 am by DavidAzulay5000 »
You don't need pants for the victory dance
Cuz Baboon's better than Weasel
 
The following users thanked this post: felixd

Online jorgemef

  • Regular Contributor
  • *
  • Posts: 105
  • Country: pt
Re: LM399 based 10 V reference
« Reply #1261 on: January 16, 2024, 10:16:38 am »
Hello,

I am planning to build two of these boards. Since is not just pennies invested I would like to ask your advice.

1 - For the Opamp I have 3 options. LT1001, LT1012 and LT1037 (and OP07). Which one is best for this application?
2 - Should VOffset compensation be used or not used for the Opamps?
3 - Is it better for dual layer or single layer? With or without ground planes? Remove copper area around LM399? Remove FRP as well for betther thermal isolation?
4 - Power filtering, on board or off board? I am planning to add the layout for it if in future I want to add some permanent low noise DC source but at the moment was planning to feed it with some X20 parkside battery. :)



Cheers,
Jorge
 

Offline iMo

  • Super Contributor
  • ***
  • Posts: 4627
  • Country: nr
  • It's important to try new things..
Re: LM399 based 10 V reference
« Reply #1262 on: January 16, 2024, 11:02:06 am »
All those opamps are usable, imho, and "I would.."
1. not use the voffs compensation,
2. .. btw ADI recommends as less copper as possible beneath the 399 (to limit the thermal flow)
3. add 1u ser 5ohm at the 399 zener (as the option for the 1399)
4. use no ground planes (use tracks in the star config)
5. not put the voltage regulators on the board (they will change their temperature based on the input voltage change)
6. use a current limiter made of a npn transistor (Q2 with 1k in the base) and a shunt resistor, ie 27ohm for 20-25mA limit (none R2)
7. wire the C2 to the opamp's output (and something like 4n7-10n)
8. not use R12 (or optional on the pcb)
9. ..btw the R11 shall allow some 1mA for 399 and 3mA for 1399
10. add 1-10u ser 5ohm at the output
11. add a 12V TVS diode at the output.
« Last Edit: January 16, 2024, 11:34:56 am by iMo »
 
The following users thanked this post: Mickle T., ch_scr, jorgemef

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14017
  • Country: de
Re: LM399 based 10 V reference
« Reply #1263 on: January 16, 2024, 11:59:16 am »
For the OP amps suitable candidates are the LT1001, OP07 and newer  ADA4077, OPA202 or OPA205 or OPA207.
The LT1037 is not at all working as it is not stable at such a low gain.
The LT1012 is a super low bias type and a bit high in voltage noise - not a total fail, but still not a good choice.
The precision OP-amps are usually good enough with the offset and not need to trim - the trimmers are an additional possible point of failure and drift source.
Offset from the OP-amps would be seen as part of the reference voltage.

The capacitor C2 in the circuit is at the wrong place. There should be a capacitor in direct feedback at the OP instead (between pins 7 and 2).

For the circuit a single layer would be enough, but the PCBs from China a 2 layers anyway. So one can use 2 layers. A ground plane is good for fast circuits that have possible EMI isssues. A star ground is the more suitable way for a precision more DC circuit.  One can still fill areas with ground, but a start ground should have priority.

The area around the LM399 should have not too much copper and may be relatively thin traces to the LM399 to reduce the heater power a little. Extra PCB cut outs are more overkill.

The supply parts as shown in the PDF is bad - the LM399 does not like when the negative side of the heater is not connected and thus gets lifted higher than the ref. ground.
The logical and simpler solution is to use only a positive supply and have the heater neg. side at GND, like the reference part, just with separate links to the center ground.
For the voltae regulator one might consider using more than just 12 V for the heater - ideally as much as avialable from the battery (e.g. 15 V with a low drop regulator). That should be about the lower end for a 5 cell Li battery pack.  There is usually little need for extra supply filtering. The battery is essentially noise free. The main other alternative would be based on a classical transformer - so one could include a rectifier and buffer cap if one considers this.

I would consider to also have a 2nd OP-amp to buffer the 7 V voltage, so to have a 2nd output that is not effected by resistor drift.
Another small addition would be space for a RC series element (e.g. 1 µF + 5 ohm) parallel to the zener so that one has the option to use an ADR1399 as a upgraded version for lower noise.


The use of 8 equal resistors for the 7 to 10 V step makes sense especially if the resistors are from a resistor array. Chances are one would have space to allow both options: a resistor array like NOMCT (SO16) or TDP (DIP16) and the seprate resistors as currently planed. Depending on the resistors used one may not need the copper for TC compensation.
 
The following users thanked this post: jorgemef

Offline Eraldo

  • Regular Contributor
  • *
  • Posts: 64
  • Country: it
Re: LM399 based 10 V reference
« Reply #1264 on: January 17, 2024, 09:23:21 pm »
I have been thinking, what about using a boost converter at the input with a 7815 in series. Would the noise be too high to be usable for the reference?

The biggest problem would be high frequencu noise but a simple filter would do the job. Also some of these converters offer quite good noise values in the single digit mV rms range. You could also put a metal can on top of it

Anyone tried it before or any reason this would be a bad idea
 

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14017
  • Country: de
Re: LM399 based 10 V reference
« Reply #1265 on: January 17, 2024, 09:30:12 pm »
With enough filtering one could use a boost converter, but it still need effort and care with the layout. It can be done, but the return is not that great. One can as well start with a higher voltage from the transformer / battery. Ideally this would be in a way (e.g. with diodes) to allow seamless change from one to the other. The power consumption of the reference is not very low ( e.g. 200 mW range) and one would thus anyway want a relatively large battery.
 

Offline Wolfgang

  • Super Contributor
  • ***
  • Posts: 1772
  • Country: de
  • Its great if it finally works !
    • Electronic Projects for Fun
Re: LM399 based 10 V reference
« Reply #1266 on: January 17, 2024, 10:20:28 pm »
With enough filtering one could use a boost converter, but it still need effort and care with the layout. It can be done, but the return is not that great. One can as well start with a higher voltage from the transformer / battery. Ideally this would be in a way (e.g. with diodes) to allow seamless change from one to the other. The power consumption of the reference is not very low ( e.g. 200 mW range) and one would thus anyway want a relatively large battery.

What can be tried are converters that support limiting switching speeds in order to reduce RFI (they more or less create a rectangle with slow ramps). I think TI has a chip for that.
 

Offline iMo

  • Super Contributor
  • ***
  • Posts: 4627
  • Country: nr
  • It's important to try new things..
Re: LM399 based 10 V reference
« Reply #1267 on: January 18, 2024, 09:49:07 am »
..and perhaps get rid of the negative supply.. Single 15V would be ok, the heater current will be higher.

What Kleinstein suggests above with two power inputs isolated with the diodes is important - I've been using that in my ADR reference - I have the internal linear source (trafo+723) and 2 posts for an external battery. Below a simplified wiring. I even put a resistor in parallel with the isolation diode such I can peek the voltage inside (the input of the 723) and also trickle charge the battery a little bit (not using trickle charging much). 99% of time it runs off the trafo.

While on the shelf it runs off the trafo, when moving it to LABs or measuring it is powered off the batteries (5x 18650). You may wire the batteries anytime on/off..

At the output I've been using a C-L-C filter as well (RF suppression) - like 1n ceramics and 3-4 turns bifilar on a small FT43 toroid.
« Last Edit: January 18, 2024, 10:00:34 am by iMo »
 

Online jorgemef

  • Regular Contributor
  • *
  • Posts: 105
  • Country: pt
Re: LM399 based 10 V reference
« Reply #1268 on: January 31, 2024, 11:57:00 am »
I have redone my circuit and implemented it.
I am seing some fluctuation with temperature which I find strange. I traced it to the Zenner but I cannot tell if is leading or lead as is bootstraped to the 10v. The gain resistors are all from same batch so TCR influence from them should be minimal or none.

Is it coming from the zenner, and is it an acceptable level?

Adding the charts on the zenner, buffering opamp (gain1) and 10V output.

R3 and D2 are not populated neither R11/R15.

Cheers,
Jorge
 
The following users thanked this post: ch_scr, iMo

Online Kleinstein

  • Super Contributor
  • ***
  • Posts: 14017
  • Country: de
Re: LM399 based 10 V reference
« Reply #1269 on: January 31, 2024, 12:11:42 pm »
The voltage still looks a bit noisy and the temperature effect is also not normal. Some of the noise can be from the meter, but I would not expect that much (depends on the settings).

The layout is not perfect in seprating the ground for the heater and reference. So some of the heater current can couple to the reference.
 
The following users thanked this post: jorgemef

Offline Gyro

  • Super Contributor
  • ***
  • Posts: 9392
  • Country: gb
Re: LM399 based 10 V reference
« Reply #1270 on: January 31, 2024, 01:05:49 pm »
I notice that you have uses SRBP for the PCB rather than FR4. SRBP has poor dimensional stability with temperature. I don't know if this is going to introduce additional temperature related stresses on the SMDs and opamp legs?
Best Regards, Chris

"Victor Meldrew, the Crimson Avenger!"
 
The following users thanked this post: jorgemef

Offline iMo

  • Super Contributor
  • ***
  • Posts: 4627
  • Country: nr
  • It's important to try new things..
Re: LM399 based 10 V reference
« Reply #1271 on: January 31, 2024, 01:12:18 pm »
You should see something like 3-5uVpp noise directly at the zener in quiet periods, imho (always use the same NPLC).
The "ratio TC" of the divider causes the 10V follows the temperature as well.
Mind the zener itself is usually equal or better than +/- 3.5uV/C.
Your "ratio TC" of your divider has to be similar or lower.
The same batch guaranties nothing.. You may calculate the TC of your divider when you plot the 10V voltage against temperature in the excel and you make the linear trend, you get the number (do the "scatter graph"). Do the same for the 7V output.. Subtract TC_7V from TC_10V then..
Also try to compute the standard dev of your 7V output, it helps with the noise comparison against your future improvements..

PS: below an example of TC (+1.28uV/C)

PPS: you may also post your data in a .csv or excel file, and we may have look at it..
« Last Edit: January 31, 2024, 05:22:50 pm by iMo »
 
The following users thanked this post: jorgemef

Online dietert1

  • Super Contributor
  • ***
  • Posts: 2007
  • Country: br
    • CADT Homepage
Re: LM399 based 10 V reference
« Reply #1272 on: January 31, 2024, 07:24:54 pm »
If a temperature variation of 29 - 27.9 = 1.1 K gives 57 - 37 = 20 uv this is about 3 ppm/K of 7V after the amplifier, more than expected for a LM399. For the 10 V output the TC is more like 6 ppm/K. Maybe the meter used to determine these graphs exhibits some TC to start with.
The difference between 3 and 6 ppm/K is certainly caused by the voltage divider. What is the TC spec of those resistors?

Regards, Dieter
 

Offline iMo

  • Super Contributor
  • ***
  • Posts: 4627
  • Country: nr
  • It's important to try new things..
Re: LM399 based 10 V reference
« Reply #1273 on: January 31, 2024, 08:21:10 pm »
..and what is your voltage regulator there? The resistors values around it are quite large..
 

Online jorgemef

  • Regular Contributor
  • *
  • Posts: 105
  • Country: pt
Re: LM399 based 10 V reference
« Reply #1274 on: January 31, 2024, 08:28:27 pm »
If a temperature variation of 29 - 27.9 = 1.1 K gives 57 - 37 = 20 uv this is about 3 ppm/K of 7V after the amplifier, more than expected for a LM399. For the 10 V output the TC is more like 6 ppm/K. Maybe the meter used to determine these graphs exhibits some TC to start with.
The difference between 3 and 6 ppm/K is certainly caused by the voltage divider. What is the TC spec of those resistors?

Regards, Dieter

1/2W 207. As per seller specs is about +-10ppm/C. https://pt.aliexpress.com/item/1005006320806775.html

So you mean the circuit is performing as per component specs, and I need to look for strategy to reduce the circuit TC like adding copper resistance on the lower side of the feedback network (and other resistor on the upper side to maintain the 10v)?
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf