I’m thinking of building a high precision voltage reference. Anybody else interested in making it into a group project?
The ultimate voltage reference is the LTZ1000 from Linear, however, in quantities of 10, it costs about $43USD each. The specs are just crazy: 1.2uV peak-peak noise, 2uV / SQRT( kHours ), 0.05%ppm/C.
So that I don’t waste money on expensive chips while I learn, I’m thinking of using something along the lines of:
Intersil’s ISL21090x, 7ppm/C, 20ppm/time, 0.03% Vout initial accuracy, $3, $12 for evaluation board
TI’s LM4030, 10ppm/C, 40ppm/time, 0.05% Vout, $1
Maxim’s MAX6126 3ppm/C, 20ppm/1000 Hours, 0.02% Vout initial, 1.3uVp-p, 60nV/SQRT(Hz) (can be improved based on the circuit), It has separate Vout sense lines for remote sensing.$6.70
Of course there are others, and I would be willing to explore any other references available.
Initially I plan to make it powered by a 9V battery for additional noise and portability.
The expensive part of an LTZ1000(A) reference is not even the reference itself, but the precision resistors required to maintain the tempco in the surrounding network. If you want to get serious about this, TiN has developed a very nice board he’s calling KX Reference with a lengthy discussion on design issues around the chip, precision references in general, full BOM ($300 USD per!, mostly resistors) and gerbers.
LM399 is a good compromise, still offering extremely good stability but not requiring the fancy resistor network. There’s some discussion of a cheaper alternative of TiN’s board at his site too, though less fleshed out. This option is only about $25 in parts.
Do you want precision or accuracy? The Linear super-zeners for example don’t have good accuracy at all, but are very stable. You will need an excellent measurement setup as well to characterize them if you plan to use them as transfer standards. What’s your end-goal here? I know it sucks as someone who likes to build stuff, but your best bet might be to just buy a reference someone else has characterized.
The Voltage Standards products have all moved to using the LT1021 reference, presumably for good reason. Its specs are comparable to the ones you list above. REF5050 is another strong contender. Many volt-nuts consider the older 10V designs like AD587 superior as they are available in ceramic packages which are much less susceptible to humidity, but also much more expensive.
This is a rabbit hole ;). Beware if you don’t have specific goals in mind. If you do build something, I have some uncalibrated-but-trustworthy precision multimeters and I might be able to help you characterize what you built in exchange for one of your references ;).
Have a voltage source to use to calibrate my oscilloscope, multi-meters, battery chargers (I would consider 4 digit precision and 0.1% accuracy to be a success)
I definitevely do not want to embark in that boat. $300 is far too much for my purposes. In fact I’m hoping to end up with a BOM in the <$50 range.
I think that one of the self-contained parts (max6126, ref5050, ISL21090) will simplify the design a lot and reduce the cost of external components. Granted, it won’t be as good at the ltz1000, but I think I should be able to meet my expectations.
we’ll see what I end up with