Clean voltage reference

I am using a voltage reference part, and I would like to make this voltage as noiseless and temperature stable as possible. I guess my questions are general, but, if it helps, the specific reference part that I've chosen is LM285LPR-2-5 https://www.mouser.com/ds/2/405/slvs023j-206162.pdf . I am biasing the source like the datasheet suggests, and then the voltage is applied to the high impedance input of an opamp.

I am considering a few basic ways to try to fortify this reference to make it as stable and noiseless as possible. I am simple-mindedly thinking that I could put bypass caps on the input and output ends of the part, and then I started thinking that maybe this could actually make the signal worse because I've just added more components that can vary with temperature or other variables. I also know that ceramics can introduce noise with their piezoelectric effect and so I would avoid those.

I was also thinking of providing the voltage to the opamp through a via directly at its pin and then a bypass cap on the pin, so that the via's parasitic inductance could help choke out some noise. I know that idea sounds a little far fetched and the via might be doing nothing depending on the frequency of the noise.

So now, I am not sure if I should proceed to add caps, or just leave it by itself. Does anyone have any advice for me please? If I do put in caps, what is the best cheap capacitor technology for this application (Its not worth it to spend $2 per a cap for me).

There's really nothing you can do to make the reference surpass its specifications (except maybe oven controlling it, but that's pretty extreme). If you need better performance than the datasheet advertises, pick a better part.

Capacitors are nice, but their values varying shouldn't have an affect on the reference output, so long as there's enough capacitance to keep things stable. But piezo effects are definitely a valid concern which must be kept in mind, especially if your circuit will experience vibration or shock. Remember that such vibration can be generated by the circuit itself if there's any high power switching going on (buzzing inductors, etc). If you use ceramic caps, stick to class 1 dielectrics, or use leaded components. Or use tantalum caps instead.

mtwieg said:

There's really nothing you can do to make the reference surpass its specifications (except maybe oven controlling it, but that's pretty extreme). If you need better performance than the datasheet advertises, pick a better part.

Capacitors are nice, but their values varying shouldn't have an affect on the reference output, so long as there's enough capacitance to keep things stable. But piezo effects are definitely a valid concern which must be kept in mind, especially if your circuit will experience vibration or shock. Remember that such vibration can be generated by the circuit itself if there's any high power switching going on (buzzing inductors, etc). If you use ceramic caps, stick to class 1 dielectrics, or use leaded components. Or use tantalum caps instead.

Click to expand...

Thank you for the help! I feared that it was mostly up to the internal circuitry, but I did find this:
http://cds.linear.com/docs/en/application-note/an82f.pdf

And they do offer the idea to use tantalum caps as you suggest to help with the wideband noise.

I also didn't realize that less current can be a better thing because it avoids the power dissipation heating the zener. I had mine biased with a higher current than necessary because I thought that would decrease noise.

Keep in mind, when specifying a reference, you need to keep in mind the difference between accuracy and drift. Many precision systems have references and components with poor initial accuracy, but after calibrating out those effects, the accuracy is limited to long term drift. And conversely, you shouldn't automatically expect a system built with 0.1% initial tolerance components to have stable performance, since those components may have drift far greater than that 0.1%.

So in the case of biasing a reference with different current, as long as the current is always stable, it likely will only cause a constant error which is equivalent to a poor initial tolerance. The temperature drift of the reference will likely not be degraded. That is, once the effects of self heating reach a steady state (seconds, maybe minutes of operation).

For these reasons, precision systems virtually always require a calibration procedure at least once in order to account for initial tolerance issues. After that, only aging, temperature drift, and other external factors are relevant.