# More PSU sensitivity¶

I have previously measured the frequency sensitivity of the +20{V} supply for the lamp, now it was to measure the non-lamp +/- 20{V} supplies.

The good news is that they’re not horribly sensitive, the bad news is that they are sensitive enough that it matters.

I pulled the A15 power supply board, and supplied the +20{V} lamp, +20{V} non-lamp and -20{V} rails from my HP6626A PSU.

I measured the frequency over 4000 {s} periods for ±25 {mV} and ±50 {mV}, and used linear regression to figure out if I had a signal or not.

## +20{V} non-lamp rail¶

The correlation coefficient is 0.4, which means there is a sensitivity but it is not high enough to give us any confidence in the slope of the line.

But just to get an idea where we are, we’ll run with the numbers anyway:

If the sensitivity is 9e-12 and our stability goal is 1e-14, the +20{V} non-lamp supply should be stable and noisefree to better than 1/900th of a volt - aka. a millivolt.

I don’t trust that result, because when I look at the time series, it looks like my 4000 second period has been too short to capture the effect:

Eyeballing this curve I suspect the sensitivity may be on the order of 2e-12{s/s} for a 0.1{V} change - about twice previous result.

That points to a PSU stability requirement on the 500 {µV} domain.

## -20{V} rail¶

A correlation coefficient of only 0.2 gives no reason to think there is a signal.

But again, looking at the actual data indicates the timelag is longer than the 4000 {s} periods I used:

Here my eyeball estimate is about the same as for the +20 {V} rail, about 2e-12{s/s} for a 0.1{V} change.

## Conclusion¶

Previously the measurement of the frequency sensitivity of the +20{V} supply for the lamp found a regulation goal of 666{µV} for a 1e-14{s/s} frequency goal.

Even though these measurements suffers from a lot of noise, they seem to indicate that the non-lamp 20{V} rails are in the same ball-pack.

Since there are three rails in total, they probably need to be regulated to better than 200{µV} each, to keep the total PSU contribution under 1e-14{s/s}.

phk