Precision voltage reference/supply for measurements

Hi everyone,

I'm trying to log some datas with the best possible resolution (external 16bits ADC on the way).

Problem is : I need a very precise/stable/low-noise (or do I) source of voltage to both be a reference and feed the pressure sensors.

So, OK, in this case, the sensors are ratiometric and thus if they are supplied with the voltage reference, the measurement should always be good but I'd like to educate myself on the subject and have a versatile set-up that I can use with any +5V sensors. :slight_smile:

I've been looking at LM317 that I'd very precisely calibrate but is it stable over time (reasonable of course... 1-2 years)?

I also have a LM336 precision zener but it won't cover the 80mA total need of my pressure sensors...

Does any of you guys know how this problem is tackled in state-of-the-art equipments?

Thanks for the help!

Marc

P.S : and Happy New Year!

Lots to choose from here:-
Voltage references

I don't have a good feel for the accuracy & stability of these things...

I'm trying to log some datas with the best possible resolution (external 16bits ADC on the way).

So, you want an error of less than one bit? That's 1/65,535 or about 15ppm.

A 16-bit audio ADC should have less than 1 bit of "audio noise", but I wouldn't be surprised of the long-term accuracy (drift) is not that good.

Problem is : I need a very precise/stable/low-noise (or do I) source of voltage to both be a reference and feed the pressure sensors....

...I also have a LM336 precision zener but it won't cover the 80mA total need of my pressure sensors...

Typically your reference is simply a reference, and you'd need an "amplifier" or voltage regulator built-around the reference if you need significant power or current.

If your sensors and your ADC both use the same reference, drift may not be a problem. Some ADCs & DACs have a reference pin that you can use for other things.

I've been looking at LM317 that I'd very precisely calibrate but is it stable over time (reasonable of course... 1-2 years)?

Drift is likely to be thermally related and you'll probably see nearly the worst case over a couple of hours (assuming you get the full worst-case temperature range of those hours). But, the nature of randomness/noise/drift is that you are never sure you've seen the worst case.

In laboratory or production applications, multmeters, oscilloscopes, and other measurement equipment is usually calibrated once or twice a year.