[time-nuts] How to measure regulator noise?
bruce.griffiths at xtra.co.nz
Sun Jan 6 18:07:07 EST 2008
John Ackermann N8UR wrote:
> I'm experimenting with some low noise regulators, and want to do some
> meaningful measurements to compare them (and hopefully compare to
> published specs).
> I have an HP 3561A FFT analyzer, which can probably be beaten these days
> by a sound card, but has the advantage of absolute calibration and GPIB
> to dump data. I also have a very low noise 40dB audio amplifier that
> covers <10 Hz to >100 kHz.
> I've tried just looking at the regulator output with the analyzer, with
> and without the amp, but am not sure what I'm seeing. In particular,
> even with a capacitor to block the DC, the analyzer indicates
> "Overrange" at a reference level below -5 or so dBV. That puts most of
> the noise spectrum below the bottom of the screen.
> Any suggestions on an appropriate test configuration, and on how to get
> readings that directly relate to the root-Hertz value shown by regulator
> data sheets?
To convert the readings to dbV/rtHz, you have to know the system gain
and the noise bandwidth of the analyser.
One method of calibration is to use a white noise source of accurately
known power spectral density.
The simplest way to build such a noise source is to use a low noise FET
input amplifier to amplify the Johnson noise of a 150k metal film
resistor (50nV/rtHz) to the required level.
You can achieve a gain of around 10X with a FET opamp with a 3dB
bandwidth of 100kHz or so. Additional amplifiers can be used to increase
the output noise level.
The lower frequency (3dB increase in noise spectral density) limit
occurs when the FET amplifier flicker noise is 50nV/rtHz.
More information about the time-nuts