3
\$\begingroup\$

Im working on a project where I need characterise the noise from a power supply so I can then design some filters. The power supply provides 5V, 12V and -12V to a set of analogue sensors (hence the need to characterise the noise and filter).

I am after some advice on which measurements to perform to get a good understanding of the noise. Initially I was thinking of just probing each power line with an oscilloscope set to AC coupling and then looking at the amplitude of the ripple (power is regulated by a flyback converter so I expect some ripple) and look at the amplitude of any other noise present. After thinking some more I was thinking perhaps using a spectrum analyser may also be useful to look at what frequencies are present so I can make sure I have appropriate filtering.

Does anyone know of any other measurements which would be useful for characterising the noise?

\$\endgroup\$
3
  • 1
    \$\begingroup\$ Don't forget to load the supply appropriately. Unloaded measurements are of only so much use. \$\endgroup\$ Jan 9 '15 at 14:41
  • \$\begingroup\$ The oscilloscope idea seems good enough. \$\endgroup\$
    – anrieff
    Jan 9 '15 at 14:45
  • \$\begingroup\$ I'll second the 'scope on AC. A digital 'scope will do an FFT not as good as a spectrum analyzer, but good enough to see the major frequencies. (There is no anti-alias filter on the 'scope FFT so be warned... check spectrum at a few ranges.) \$\endgroup\$ Jan 9 '15 at 17:21
1
\$\begingroup\$

My first reaction is to look at the signal on a scope with AC coupling too.

Details beyond what you can see that way shouldn't be relevant if you design your circuit right. Instead of measuring the frequency spectrum of this supply at this temperature with this load and this input power on this day, just assume it produces broad band noise and deal with it.

Filter the supply according to what the circuit needs. Most analog parts have good power supply rejection ratios, but that only applies up to a certain frequency. Usually a ferrite "chip inductor" followed by a few 10s of µF of ceramic capacitance to ground is good enough to squash the power supply frequencies that cause problems. Sometimes you might need two of those in series, which will usually get rid of crap above 10 kHz or so.

If you can't tolerate long term voltage drift (very low frequency noise), then either fix the circuit so that it doesn't depend on the power supply voltage or somehow regulate the voltage in a analog way to not introduce more noise. This could be as simple as a zener shunt regulator, or a series linear regulator. Sensitive analog circuitry usually requires little power, so shunt and linear regulators can be appropriate. Then you can still follow that with one or two L-C filters as described above immediately in front of each use of that power rail.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks for the information and especially the points about spectrum analysis (it gets expensive when I need to hire one) - the backup is to use a linear regulator for the analogue electronics, but first I want to see if how they perform with simple filtering. I will do the scope measurement and then add a filter in to see what the difference is! Thanks for your help though! \$\endgroup\$ Jan 9 '15 at 16:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.