Yes, you do not want peaks that upset the feedback loop and are reflected as irregularities with the loop closed. In reality, however, that is easier said than done, because the piezo sensors pick up a lot of rubbish/voicecoil movements you do not always see in a nearfield measurement.
The curves Bolserst was showing are actually, apart from 1 kHz the peak that can be notched out easily, surprisingly clean, enabling some 20dB feedback with a re-
latively simple circuit of just an integrator/6dB lowpass and one notch.
Do not expect to automatically have such luck: you may need some more notching out before you can safely close the loop and adjust to have some 20 dB feedback.
The curves Bolserst was showing are actually, apart from 1 kHz the peak that can be notched out easily, surprisingly clean, enabling some 20dB feedback with a re-
latively simple circuit of just an integrator/6dB lowpass and one notch.
Do not expect to automatically have such luck: you may need some more notching out before you can safely close the loop and adjust to have some 20 dB feedback.
I made some readings and some simulations, and as soon as i have time to put my schematics and simulation results in a readable shape i will share them.
For now, i want to summarize my results:
1- it is very important, in order to grant stability, to greatly oversize the input capacitors of the power amp
2 - the integrator should have a very low corner frequency and a dc gain between 20 and 30 dB (if the output swing of the accelerometer is somewhere near the signal source swing). after all, the phase shift starts to become significant just a little lower than the corner frequency, but if the dc gain is not big enough, you will not be able to use the integrator's output in the frequencies where it's needed.
3 - as a consequence, having an integrator in parallel with a non-filtering gain stage in parallel with a high-pass, and summing the outputs of these three circuits, yields some good results (in simulation at least).
4-it seems to me (i will investigate further) that the notch filter might be unnecessary, and potentially detrimental: at high frequencies the transfer function phase of a speaker is negative. in the lower part of the stopband, the notch pushes the phase even lower. The highpass and the gain stage parallel to the integrator already give a near-to-zero phase shift for frequencies much higher than the integrator's corner frequency, eliminating the high frequency stability issue given by the integrator (somewhere in the few hundreds hertz region).
I am sorry for my poor writing skills, i hope my explanation is understandable!
Cheers
S.
For now, i want to summarize my results:
1- it is very important, in order to grant stability, to greatly oversize the input capacitors of the power amp
2 - the integrator should have a very low corner frequency and a dc gain between 20 and 30 dB (if the output swing of the accelerometer is somewhere near the signal source swing). after all, the phase shift starts to become significant just a little lower than the corner frequency, but if the dc gain is not big enough, you will not be able to use the integrator's output in the frequencies where it's needed.
3 - as a consequence, having an integrator in parallel with a non-filtering gain stage in parallel with a high-pass, and summing the outputs of these three circuits, yields some good results (in simulation at least).
4-it seems to me (i will investigate further) that the notch filter might be unnecessary, and potentially detrimental: at high frequencies the transfer function phase of a speaker is negative. in the lower part of the stopband, the notch pushes the phase even lower. The highpass and the gain stage parallel to the integrator already give a near-to-zero phase shift for frequencies much higher than the integrator's corner frequency, eliminating the high frequency stability issue given by the integrator (somewhere in the few hundreds hertz region).
I am sorry for my poor writing skills, i hope my explanation is understandable!
Cheers
S.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.