Best is to enclose the whole thing in insulation. Long term drift till warm, then stable, which if you ovenise the entire reference circuit, and keep that at 40C, makes it much more stable as well.

Testing as a single value only guarantees accuracy at that specific value. For true accuracy you need to check over a range of values so you can also check linearity.

@@JaenEngineering Would help if it had multiple calibration points. single point cals are never the greatest for accuracy.

Analog Devices/Burr Brown with batteries

Slight asterisk - it's often better to "limit the influence" of a pot by using a series resistor from its wiper to limit the current driving it can steer an adjustment with. Especially if its top rail is already from a stable reference. This is because yes, the pot has a large TC, but usually inside of the pot - *both sides of it will have relatively the same TC*. So both sides will drift the same way. That's not true if you add padding resistors on either side.

@@ivolol Padding resistors ideally you want the pot in the middle of range, and thus nulling out TC to a great extent. The padding resistors it is much cheaper to get ones with low TC so the TC of the pot contributes less, though if you are going to ovenise the complete reference, including the opamp and trimmer, you get better overall drift. Military I used to use resistor boxes to set offsets and nulling on the bench, then used fixed resistors to set the actual value, often meaning I was using the ratio to get both to be close to an E96 value in use, and if they were more than 2:1 ratio that opamp get changed out for excessive drift. I changed lots of SFC2741T opamps, as they had drifted offset past 10mV over time. Next most common were 5474's, they did not like being used for input data recovery, logically fine just input threshold drift outside of TTL levels.

@@SeanBZA Ye Olde 'SOT' Select On Test !

@@andymouse No drift either, once set and new opamp it never came back. At least not for that channel, the rest and the power supply though....

@@SeanBZA :)

Better still Batteries.

This is something that fascinates me, I'm on the search for the truth but what I have found so far suggests in favour of the pour ! and that's where my money is !

Besides what he said in the video you mean?

4 layer, everyone gets good grounds then

@@charlesdorval394 Kind of. I mean compared with a star topology, is ground plane a drawback? That's the point that's not quite clear...

@@nashaut7635 Oh okay, sorry I misunderstood what you meant

Kicad

The legend that is 'KiCad'

It's a single *point* ground where each conductor or wire arrives separately. So that the current on one circuit element doesn't affect the others. Same thing can be done for non-Grounds too, when you need (for example) a single point voltage reference. It's somewhat related to 4-wire measurements, so you can look that up for more explanation.

One thing you *can* know is that the LM399 or ADR1399 (if you are buying new, better to buy the latter) is likely to drift extremely little after you leave them turned on for a week. So for everything but crazy precision instruments you know that the voltage you measured from it will not have changed at all in a year or two years from when you last measured it.

What you could do though is check whether your cheap voltmeters drift relative to a fairly precise voltage reference. This assumes that the circuit around the LM399 is properly designed. But for normal people mostly not really worth it.

@@ivolol For economic-practical reasons I bought a voltage reference based on the AD584 already "calibrated" by the Chinese, with a report of dubious validity, but it was the best I could do. But since I would like to build my own I can read up on the ADR1399. Thank you.

@@sebastian_harnisch I haven't tried the voltage reference "toy" I bought yet, because I'm way behind on making the voltmeter, but I think it's a good idea. The voltmeter is analogue therefore the precision is relative. I only need it to calibrate the various trimmers. Thanks

@@leonardochiruzzi7642 I also bought this device from Amazon and it was spot on but I was worried before it arrived!