# Voltage reference vs voltage regulator

I need to bias some op amp inputs with a voltage. I’m in the 10-60 hz range. I need the bias voltage to be as low noise as possible, like -155 dbv . Would a voltage reference, voltage regulator, or even zener diode work the best here? I’m confused as to which part would work the best. And do I need to drive the bias through a voltage buffer or can I put it right onto the inputs?

• How stable must this be, as temperature changes? – analogsystemsrf May 17 '19 at 4:41
• "as good as possible" how much liquid helium can you afford? – Jasen May 17 '19 at 5:07
• Sounds like an XY problem to me. – Andy aka May 17 '19 at 11:49

A voltage regulator is meant to keep voltage consistant across widely varying output current. A voltage reference keeps it consistant across a range of other conditions but is meant to operate without a load on it. You want a voltage reference.

Your goal of -155dBv, given -160dBv is 10^-8 below 1volt (peak-peak?) is 10 nanoVolts.

We know, using Vnoise = sqrt(K * T /C)

that 10 picoFarads as the Cap, in a low-pass-filter with any value of R, will produce 20 microVolts RMS total integrated noise.

You want to be 2,000X lower noise than 10pF will produce.

Thus your capacitor must be 10pF * ( 2,000 * 2,000) = 40 uFarad. Try a 47uF.

Pick your R as you wish. Just use at least 40uF in the LPF.

==================

The OP wants at least -155, which is close to -154, which is 6dB more voltage than the 40uF produces. 6dB, or 2x more voltage, allows 4x small cap, at 10uF.

• How about thinfilm cap? – Damien May 17 '19 at 5:48
• That's a neat trick with the capacitor value, presumably it involves the resistor having a defined noise_power/rtHz regardless of R. I've not seen a derivation of that result anywhere. Is there a published derivation you could link to? – Neil_UK May 17 '19 at 6:08
• I was exploring the relationship of R and C and the total integrated random thermal (resistor) noise, and attempting to come up with a nomograph to summarize. A very experienced Aps Engr looked briefly at my efforts, and advised "That is just the total noise of a sampled-capacitor," and YES he was right. @Neal_UK: have not found one yet. – analogsystemsrf May 17 '19 at 7:37
• This is the ideal noise case for a thin film resistor, but a Zener or any voltage ref diode may be worse so you choose an ultra low noise Vref that is ultra-stable, I believe. – Tony Stewart Sunnyskyguy EE75 May 17 '19 at 17:39
• @ Sunnyskyguy Is the excess noise (in non-thin-film Rs) a shot-noise proportional to sqrt(current) thru the resistor? – analogsystemsrf May 18 '19 at 11:15