voltage standard/reference plans

[croyleje]

Hello Everyone

I have been looking online for plans for a voltage reference that I can use to check a few meters I have picked off ebay.  I was planning on just using a simple MAX6126 voltage reference looks like I can get one with 0.02% tolerance for digikey.  Anyone have any ideas on how I can check the accuracy/calibration of these meters there both in pretty good shape just want to give them a check.  I am sure there are lots of better ways to do this my biggest problem is that the best meter I have that I know is accurate is an Agilent 1233a which is a 4 digit 6600 count and the meters I got from ebay are both benchtop 6 1/2 digit hp34401a any ideas?  Any help would be great thanks.

Jason

[robrenz]

Read all of this thread
https://www.eevblog.com/forum/general-chat/incoming-stuff-voltage-reference/msg102595/#msg102595

[alm]

My 2 cents: Get a stable source (stable for 5 minutes is sufficient), ignore initial accuracy. Initial accuracy will most likely not be sufficient to test the accuracy of a 6.5 digit meter anyway (the expensive references used in the 34401a might have an initial accuracy of 1%). Compare the three meters. If the 34401a meters agree within their expect tolerance and agree with the 1233a, then it's very likely that they're in 1 year spec on that range. Especially if all three meters came from a different source. It's actually quite rare for them to need adjustment during yearly calibration; they're known to be quite stable. Traceable calibration down to the 30ppm or so of basic DC accuracy will cost serious money (Joe Geller's SVR might be the one exception that gets close). 0.02% is almost an order of magnitude worse than the meters. This method will also let you test more than one range / function.

[T4P]

Martin introduced these to me : I only primarily looked for tempco stability as there's no point being all that accurate if it drifts like a madman (You can correct inaccuracies but it's hard to keep track of constant drifts)
ADR34 series

[saturation]

Yes, folks seem to think tolerance is important, it isn't once the item is calibrated.  What matters is stability.  Temperature is a common foe, but when dealing with uV and less, other issues include aging, and more environmental variables like pressure and humidity [ if not on the reference, on the entire circuit assembly its built on] .  Then there are things like the act of doing the measurement, e.g. touching the leads as in the video, lead to issues.  It'll help to review the free "Keithley low level measurements handbook" for a compendium of techniques.

[PA4TIM]

Be awared, the uV virus is a nasty one. I builded several references and have several commercial ones. Stability is the main issue.
While builiding:
-Keep everything clean. No fingerprints on the pcb and reference
-Avoid mechanical stress, for instance mounting it dead bug of cutting a U form out the pcb
-Take in account the seebeck effect, so compensate points with extra solder joints if needed
-Download the keithley book
-More references parallel reduce noise
-Avoid 10 turn trimpots, and i f you use them, make sure theire "span" is very small. For instance one whole turn is a uV. They behave " digital" and are mechanical unstabe. I made one that  adjust over 1mV, if I tapped on the cabinet, or lifted it and put it on the table if jumped like 40 to 100 uV. -Replaced the trimpot by low tempco resistors and the trimpot over a 10 ohm resistor.
-Wait with fine adjustment after the burning in. Let it on 24/7 for a few months to stabilize and then do the fine adjudment
-Use low tempco components and low leakage, zero drift/offset opamps like the lt1052 chopper or very good opamp. Metalfilm resistors. Look at the tempco, not accuracy. You want hem stable over time and temp. Not spot on at 25 degrees with a 1000 ppm/C tempco. ( but often low tempco once have good accuracy)
-ovenize it if possible.

http://www.pa4tim.nl/?p=2531 some of my experiments.

But most important part, you will never know how good it is unless you have it calibrated. I have one 5,5,  two 6,5 and two 7,5 digit meters. The two 6,5 are calibrated one year ago, one 7,5 is too. They differ 2 to 3 uV ( after about 6 hours warming up) they also have a tempco so measuring at 20 degrees will give other results as at 30 degrees. i'm lucky to have a Guildline standardcel cabinet with known history. The Prema was during he last calibration 0ppm in the 3V range so I measured the cells. Did the same with the two calibrated 6,5 digits. The difference was less as 2 uV and comprnsated for temperature it was sub uV. So the next decade or so I will be close enough to the " golden" Volt for amateur use.