I'd like to know how voltage reference stability is obtained in products like 6.5 digit multimeters that use LM399 as a voltage reference in spite of using two gain resistors. Let's assume an ideal LM399 and ideal op-amp and ideal current source (so ignore the 200k and 5k resistors.) It would seem to me that even at 5ppm, the output would be very sensitive to temperature (relative to a real LM399.) So then how can this work? Is it the case that the resistor temperatures do not fluctuate much because they are near the temperature controlled heater? Or are they measuring the temperature with something like an LM35 (and matched tempcos on the resistors) and calibrating in software?

Even at 5ppm/C and only 0.5C change in temperature (leaving out the trimpot), I'm calculating a range of 10.170 to 10.167 (gain ranging from 1.4634 to 1.4629) which would seem horrible for such an instrument. Do resistors effect gain in the way that I think they do?

I've looked at schematics for a couple such multimeters now and they all use at least two or three discrete resistors and usually aim for an stable output voltage around +/-10V or maybe +/-12V, so is the "Portable Calibrator" a reasonable approximation of a real application?

LM399 portable calibrator

  • \$\begingroup\$ what does the title of your question mean? \$\endgroup\$
    – jsotola
    Feb 1, 2018 at 1:47
  • \$\begingroup\$ I'm sorry, I didn't finish typing it apparently. I wonder if I can edit that. \$\endgroup\$
    – Anthony
    Feb 1, 2018 at 1:51
  • \$\begingroup\$ fixed. "Don't the temperature coefficients of the resistors in the LM399/LM199 "Portable Calibrator" significantly effect the output stability?" \$\endgroup\$
    – Anthony
    Feb 1, 2018 at 1:52
  • \$\begingroup\$ Is the answer that they just match the tempcos precisely? Since it's just a ratio, if both resistors increase equally with temperature then everything is fine... \$\endgroup\$
    – Anthony
    Feb 1, 2018 at 2:03
  • \$\begingroup\$ I don't follow your math, 3 mV deviation due to 0.5 degC, 5ppm of 10V is 50 uV \$\endgroup\$
    – sstobbe
    Feb 1, 2018 at 2:06

3 Answers 3


Well if the LM399 has a 0.5 ppm/'C spec 0~75'C that's not good enough for 6.5 digit DMM with 1 ppm accuracy. However the LTZ1000 is 10x better at 0.05 ppm/'C.

It is already thermally heated inside with thermal feedback. Normally better stability can obtained with a double thermal oven servo over the chip just as done in better OCXO's.

Getting laser trimmed Resistor ratios to 1 ppm is a harder task.

Never assume the accuracy is the same as the resolution. Sometimes the extra resolution in the short term is what is needed.

Here is a 7.5 digit meter with 50 ppm accuracy ( only !) enter image description here

Here's a test that compared references enter image description here Keysight 34498A ... innards of a 6.5 digit DMM enter image description here

Keysight's Truevolt 34465A DMM, the 1-year specification applies for temperatures ± 2 °C of the calibration temperature and with self-calibration achieves an accuracy in the 10V range within 24h if 10 ppm of reading and 4ppm of range.

The only other Keysight DMM with ACAL is the high-end 3458A with its high-end price; a bestselling 8.5-digit DMM enter image description here

  • \$\begingroup\$ "Never assume the accuracy is the same as the resolution." That is the key here as well as my trivial arithmetic slip-up. \$\endgroup\$
    – Anthony
    Feb 1, 2018 at 2:46

The ratio of the 8.8K resistor and the 19K + trimpot have a strong effect on the output voltage. The others are not very critical.

The gain of the circuit is approximately 1.43, depending on the exact voltage the Zener diode happens to be. So a 1ppm change in one of the resistors affects the output voltage by 0.43/1.43 = 0.3ppm.

If you use Z foil 0.2ppm/K resistors you will get +/-0.12ppm/K tempco from the resistor ratio. They cost something like $15-$20 per resistor and are not easily available in a wide range of values.

This is one of the problems with these references- the accuracy of the voltage source is not that great so you need to add resistors that can significantly affect the output voltage.

I would suggest using a nominal 1.5:1 ratio and trimming that a bit with fixed precision resistors and the last little bit with a trimpot (wirewound has lower tempco but lower resolution). Or live with the error and correct it later in the digital domain. You can also correct for the drift to some degree by compensation. Or ovenize the whole mess in addition to the buried zener reference but that's typically messier than it sounds and can lead to hysteresis errors.


That is a good reference circuit. Don't forget you can buy .1% tolerance resistors for a cheap price now days, including SMD package.

If you need extreme accuracy over time the TLZ1000 suggested by Tony Stewart is the best on the market. Price is about $30 USD for 50 ppb accuracy and $50 USD for 10 ppb accuracy. The manufacturer suggest shielding the leads from any blowing air, as it causes 10 uV noise on the output lead.

The TLZ1000 is good enough for laboratory calibration equipment up to 9 digits, or possibly better. Be careful of trim pots as they have a drift of 200 ppm C. Use the smallest value trim pot possible, so its overall drift is absorbed by series and/or parallel resistors.

  • \$\begingroup\$ Yes the LTZ is the bee's knees but thermal calibration and auto zero are essential. \$\endgroup\$ Feb 1, 2018 at 2:39
  • \$\begingroup\$ @TonyStewart.EEsince'75. Yes it is. +1 for the pretty pictures. I hate how some of you guys seem to have them at the ready no matter the subject. \$\endgroup\$
    – user105652
    Feb 1, 2018 at 2:42
  • \$\begingroup\$ I just search for them ... TLZ shud be LTZ google.ca/… \$\endgroup\$ Feb 1, 2018 at 2:44

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.