mantrid wrote: ↑Tue Jul 30, 2019 11:11 am
I may have found the issue: The power saving 3-state signal contains frequency remnants from the modulation waveform. I should be able to fix this and will send and update in the next days.
Was premature: The weak low frequency components in the spectrum came from a to small window for the Fourier analysis. Due to the low duty cycle a wider window has to be used in order to have sufficient non-zero samples.
I performed some analysis and measurements:
Spectra
The spectrum of the power saving 3-state signal (it's not rectangular) has its maximum between 600 Hz and 1500 Hz. Frequency components below 600 Hz are less than -20db compared to peak and there are no frequency components below 175/2 Hz.
As written earlier, the 2-phase spectrum has a maximum at 175 Hz and strong peaks at 3 times, 5 times, ... this frequency. The spectrum of the 3-state signal is better. I can resort something such that the 2-phase spectrum looks similar to the 3-phase spectrum.
cfs6t08p wrote: ↑Mon Jul 29, 2019 5:13 pm
If I run these files through a 300Hz high-pass filter they improve massively, they still feel "busier" than the original but the uncomfortable sensations are all gone.
Means that you filter something that does not come from the sound files.
Measurement
I also attached an oscilloscope to my estim box. The signals look as desired. The 2-channel signal is a little bit deformed due to the my high pass filter (60% at 175 Hz).
Conclusion
If the spiky sensation disappear with a high-pass filter, either the sound card or amplifier seem to generate low frequency components. That can be caused by some kind of equalization using a to narrow window/filter. I would also expect this effect with continuous signals, but it would be much weaker.
As mentioned in the previous post: Very low frequencies can be dangerous. If the transformer lets them through you have to filter them.