onatirec
Mechanical
- Aug 7, 2020
- 91
I've been looking at noise and vibrations from a small-medium, 3-phase AC motor and have found peaks at multiples of 4kHz (8k, 12k, 16k)… They show up in a spectrogram just as I would expect PWM signals and their corresponding harmonics to appear.
The PWM is set at 16kHz, and these peaks show up at/around 4k, 8k, etc even when changing the PWM frequency to other values (I've been constrained by multiples of two).
I finally got around to putting a scope on one of the phases and it seems that there are similar artifacts in the phase voltage.
I'm curious as to why these subharmonics are present. All of the sources I've been able to find only mention subharmonics when the ratio of fundamental electric frequency to PWM carrier is on the high side. Here the fe is around 350Hz vs 16kHz for the PWM.
I'm attaching a picture of the scope signal with overlaid FFT (in red). The high peak is at 16kHz and one can see the additional peaks to the left.
I had originally posted in the acoustics forum here: A spectrogram of the acoustic noise is in the OP there and the 4k, 8k, 12k vertical lines can be seen for more context.
Thanks a bunch.
The PWM is set at 16kHz, and these peaks show up at/around 4k, 8k, etc even when changing the PWM frequency to other values (I've been constrained by multiples of two).
I finally got around to putting a scope on one of the phases and it seems that there are similar artifacts in the phase voltage.
I'm curious as to why these subharmonics are present. All of the sources I've been able to find only mention subharmonics when the ratio of fundamental electric frequency to PWM carrier is on the high side. Here the fe is around 350Hz vs 16kHz for the PWM.
I'm attaching a picture of the scope signal with overlaid FFT (in red). The high peak is at 16kHz and one can see the additional peaks to the left.
I had originally posted in the acoustics forum here: A spectrogram of the acoustic noise is in the OP there and the 4k, 8k, 12k vertical lines can be seen for more context.
Thanks a bunch.