I have a preamplifier based on an OPA1611 that I wanted to characterize. The amplifier is in a fully enclosed and earthed metal chassis, which has an AC-DC converter in it. The mains cabling goes directly from the connector into the converter's screw terminals via ~5 cm of cabling. The AC "compartment" in the chassis is separated from the remaining low voltage circuit via an aluminum mounting plate (which provides no magn. shielding against low frequency). On the low voltage board, there are some DC/DC converters. The amplifier circuit is eventually supplied with +/-14V after going through a lot of active and passive filtering, both for common-mode with respect to Earth and differential mode. The same can be said about the reference and supply voltage for the ADC. The ADC runs continuously, sampling at 125 kHz. For the data below, blocks of 250 samples are averaged and passed on every 2 ms (sampling rate effectively being 500 Hz). Then, these block averages are collected for a few seconds. The harmonic content doesn't change if the block size is changed.

The preamp input is shorted using on-board switches; the shorting trace encloses a loop of maybe 1 cm². I am taking values from the ADC and FFT them.

I expected to see the usual picture: the amplifier noise, overlaid with a couple of mains harmonics, but instead I found that all the mains harmonics have sidebands with a distance of \$f_{mains}/4=\$12.5 Hz.

enter image description here

A little while later, I tried to capture a faster time trace to have a look at the waveform and obtained this:

enter image description here

Strangely, the waveform now repeats every three mains cycles. And the FFT confirms as much:

enter image description here

The only thing in the chassis that might do something with mains synchronisation is the AC-DC converter. Other than that, I would suspect that subharmonic noise can only come from mains voltage contamination

Is it likely that the AC-DC-converter (e.g. its PFC) can create sub harmonic noise ? If not, what to look for in the environment (an office with a PC and LED lighting)?

  • \$\begingroup\$ What ADC converter you use? Sampling rate? Does it output a stream or some size blocks? How much data you record before doing FFT? \$\endgroup\$
    – Justme
    Oct 21, 2022 at 7:40
  • \$\begingroup\$ @Justme I have added more info at the end of the first paragraph. \$\endgroup\$
    – tobalt
    Oct 21, 2022 at 7:48
  • \$\begingroup\$ How can the FFT go down to a few Hz if the block length is 2ms? Are the blocks concatenated instead of averaged? \$\endgroup\$
    – bobflux
    Oct 21, 2022 at 7:52
  • \$\begingroup\$ @bobflux Then, these block averages are collected for a few seconds. One "block" average is one data point in the final data vector, which contains many such averages. \$\endgroup\$
    – tobalt
    Oct 21, 2022 at 7:53
  • 1
    \$\begingroup\$ Did you rule out the PSU PFC yet? How much load is on it? Can you add or remove load to see if it then uses all mains cycles or less mains cycles? What other devices are on same mains that could explain it (you already mentioned LED lighting, turn lights off etc)? \$\endgroup\$
    – Justme
    Oct 21, 2022 at 10:09

1 Answer 1


As indicated by others, it appears that there is some irregular mains load that skips mains cycles which can be seen in the data.

As almost everything except the power supply and it's active PFC are ruled out, it can be the power supply itself.

To find out if the power supply is the source, add or remove load to the power supply to find if it is the power supply.

This should be enough to see if it stops skipping mains cycles or starts to skip more mains cycles.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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