Bought several used LM399, how to use it?

[technix]

Just ordered a few used LM399 salvaged from another board. Is it enough to build a transfer standard calibrating my 3.5 digit DMMs with?

Even better, if I am to build an adjustable voltage reference out of it, how many digits can I shoot for?

[Vgkid]

Considering the lm399 is the standard 6.5 digit multimeter voltage reference, you could shoot for 6 digits. Several 7.5 digit meters use it,HP3457, K2001(most likely selected). Considering what level of accuracy you want to shoot for, depends on how much you want to spend. Seeing you are in China, yuou have lots of choices for used/new precision resistors. lm399 datasheet here.
http://www.digchip.com/datasheets/parts/datasheet/321/LM299-pdf.php

[Fungus]

Just ordered a few used LM399 salvaged from another board. Is it enough to build a transfer standard calibrating my 3.5 digit DMMs with?

http://cds.linear.com/docs/en/datasheet/199399fc.pdf

Datasheet calls it a 'Precision Voltage Reference for Multimeters', so the answer is "yes".

Even better, if I am to build an adjustable voltage reference out of it, how many digits can I shoot for?

Datasheet says long term stability is <8ppm, that's 0.00008% - nearly 7 digits.

[kjs]

but initial tolerance is 2% which means you have to calibrate it with something better and calibrated as well.

[daqq]

but initial tolerance is 2% which means you have to calibrate it with something better and calibrated as well.

Calibrate or just measure - if you know the voltage to N decimal places you don't really need exact 10.0000V (in most situations), you will be just as OK with an exact 7.435483V. There are situations where 10 is better, but for 3.5 digit multimeters you are OK.

[technix]

Considering the lm399 is the standard 6.5 digit multimeter voltage reference, you could shoot for 6 digits. Several 7.5 digit meters use it,HP3457, K2001(most likely selected). Considering what level of accuracy you want to shoot for, depends on how much you want to spend. Seeing you are in China, yuou have lots of choices for used/new precision resistors. lm399 datasheet here.
http://www.digchip.com/datasheets/parts/datasheet/321/LM299-pdf.php

So to build an computer adjustable standard, I need ten dozens or so of low tempco precision resistors and ten turn trim pots, 60 DPST reed relays, a precision op amp, 8 74HC595s, a 1602 screen, a keypad and a microcontroller to control it all?

[dom0]

You really don't want trimpots in any kind of precision circuits. Especially not when it's digitally controlled anyway so you can easily calibrate offset, gain and even non-linearity (to a degree).

[technix]

You really don't want trimpots in any kind of precision circuits. Especially not when it's digitally controlled anyway so you can easily calibrate offset, gain and even non-linearity (to a degree).

It is just a Kelvin-Varley divider with relays replacing rotary switches, so some trimming is still required.

[paulie]

IMO depending on precision required, pots are OK for most applications as long as they constitute a small percentage of the resistance range. This generally benefits adjustment ease too.  For example a pot that changes 50% not only contributes to noise and instability but very difficult to adjust ±10ppm.

Also note that a circuit can be much simpler than your list, be it meter or PSU. I've used LM399 with both using little more than a low cost regulator and resistor for the reference part. It is common for internet "experts" to recommend the most expensive and complex solutions when KISS will suffice. Maybe an attempt to show how smart they are or sometimes just like to watch noobs dance.

[Vgkid]

For a simple circuit, take a look at the edc mv106, pretty much an inverting amplifier.
http://www.ko4bb.com/getsimple/index.php?id=download&file=06_Misc_Test_Equipment/EDC/EDC_MV106_DC_Calibrator_Service_Manual.pdf

[Kleinstein]

If you really want a computer controllable voltage, consider a system with  DACs and ADC, similar to LTs Application note an86.

Avoid trimmers if possible, and if really needed take into account there usually poor TK and stability - so the trimmer should only do fine adjustment of something like 1% or less. Typical one uses a fixed resistor divider and the pot in parallel with a large resistor of something like 100 times the resistance from the wiper to the divider.

[Fungus]

but initial tolerance is 2%

Doesn't matter for a "transfer standard". The main thing for transfer standards is stability.

(You have to measure the true output voltage on a calibrated instrument before you can use it, obviously...)

[cellularmitosis]

If you really want a computer controllable voltage, consider a system with  DACs and ADC, similar to LTs Application note an86.

Avoid trimmers if possible, and if really needed take into account there usually poor TK and stability - so the trimmer should only do fine adjustment of something like 1% or less. Typical one uses a fixed resistor divider and the pot in parallel with a large resistor of something like 100 times the resistance from the wiper to the divider.

Indeed, in Analog SEEKrets (the Potentiometer chapter), Leslie Green describes this setup:



And once you've done that, you can replace the pot with a DAC!

http://www.eevblog.com/files/seekPDF.pdf

[technix]

If you really want a computer controllable voltage, consider a system with  DACs and ADC, similar to LTs Application note an86.

Avoid trimmers if possible, and if really needed take into account there usually poor TK and stability - so the trimmer should only do fine adjustment of something like 1% or less. Typical one uses a fixed resistor divider and the pot in parallel with a large resistor of something like 100 times the resistance from the wiper to the divider.

Indeed, in Analog SEEKrets (the Potentiometer chapter), Leslie Green describes this setup:



And once you've done that, you can replace the pot with a DAC!

http://www.eevblog.com/files/seekPDF.pdf

This circuit can be used on adjusting the 10V reference amplifier.

About the main DAC for my digitally-controlled voltage reference, since I may need a little more bits than what is required for 6.5 digits precision, I am thinking about using a relay-operated 24-bit binary Kelvin-Varley divider, driven by a string of TPIC6B595s as the DAC.