Electronics > Projects, Designs, and Technical Stuff
Audio equalizer design help with phase shifts and filtering
(1/1)
mgk39:
Hey, this is my first post on here! So my goal is to design an audio equalizer which I can actually use to play music. I already have an audio amplifier designed that's working quite well (better than I expected actually). The amplifier is 2 differential stages followed by a class AB output stage. I now want to design an audio equalizer so I can tune the bass, mid range, and upper range frequencies of the audio. I want to do this by designing 3 band-pass filters (which are actually made from a high-pass filter followed by a low-pass filter). I then want to sum the outputs of these 3 filters in a mixer (basically a summing op-amp). The problem I'm foreseeing is that each band-pass filter will have a different phase response, and so the 3 outputs will not necessarily be in-phase, which they need to be to be properly summed by the mixer. How do I approach this? Am I approaching the equalizer design correctly, i.e. signal -> 2-stage differential amplifier -> band-pass filters -> mixer/summing junction -> output stage?
Also the output of pre-amp is differential, so I would actually need 6 band-pass filters, 3 for the positive output and 3 for the negative output. I can't connect the negative to the filter ground because I'm using dual rail op-amps whose rails need to be referenced to a constant ground. Unless I could get around this somehow? Maybe I could put the rails in series with the negative output and use the negative output as a ground, that way the rails stay at a constant voltage difference? I could avoid this by filtering the input signal first (which is referenced to ground), but I've read that it's better to filter the pre-amp outputs to be able to work with smaller signals. Is this true? Or should I avoid this altogether and consider going with passive filters instead? Though that still raises the question of filtering the signal or the pre-amp output, and still doesn't solve my phase problem. Any thoughts or advice greatly appreciated!
Benta:
The first thing you need to do is to read about Linkwitz-Riley filters. It concerns filters that are phase coherent at the transition frequency. The original application was loudspeaker crossover filters, but can probably be used in your case as well.
I attach a sketch of how the principle of an active LP/BP/HP looks.
It's a combination of low pass and all pass filters with summing amplifiers. The circuit has phase coherent outputs over frequency that can be summed without problems.
LP 4. = 4th order low pass
AP 2. = 2nd order all pass
Important to note is, that the 4th order LP sections do not have "normal" 4th order responses, but 2nd order squared, meaning two identical 2nd order sections with the same cut off frequency and response cascaded. Otherwise the LP and AP phase responses will not track.
The LP filters can be Bessel, Butterworth, whatever.
Have fun :)
EDIT: you can achieve the same functionality with "normal" LP/HP filter topologies, but it's somewhat challenging to calculate/synthesize.
mgk39:
--- Quote from: Benta on May 13, 2020, 08:25:00 pm ---The first thing you need to do is to read about Linkwitz-Riley filters. It concerns filters that are phase coherent at the transition frequency. The original application was loudspeaker crossover filters, but can probably be used in your case as well.
I attach a sketch of how the principle of an active LP/BP/HP looks.
It's a combination of low pass and all pass filters with summing amplifiers. The circuit has phase coherent outputs over frequency that can be summed without problems.
LP 4. = 4th order low pass
AP 2. = 2nd order all pass
Important to note is, that the 4th order LP sections do not have "normal" 4th order responses, but 2nd order squared, meaning two identical 2nd order sections with the same cut off frequency and response cascaded. Otherwise the LP and AP phase responses will not track.
The LP filters can be Bessel, Butterworth, whatever.
Have fun :)
EDIT: you can achieve the same functionality with "normal" LP/HP filter topologies, but it's somewhat challenging to calculate/synthesize.
--- End quote ---
Hey thanks! Let me preface this with saying I'm not that knowledgeable in this area, I had to re-read most of the chapter on filters from my college textbook. I can't really find that much on linkwitz-riley filters. From what I have found my understanding is that it's just an LP and HP filter cascaded with their -3dB points lined up. This creates a BP filter with a -6dB gain. I tested this out in LTspice and the phase response is the same over the bandpass frequency range, but not flat. I'm pretty sure I want a flat phase response correct? Because I want the the signals coming from both bandpass filters to be in phase over their entire respective BP filter ranges. So if I have an audio signal with a component at the lower cutoff frequency of BP filter 1 and another component at the upper cutoff of BP filter 2 I still want them both to be in phase. Thanks for taking the time to draw a sketch! I have a few questions about it. Is the phase response of the AP the same as the LP? If so how do they cancel out? How do the phase responses of the LP4 and AP2 compare to one another and how does this achieve "phase correction"? The way I'm understanding it is that each time you subtract the AP output it phase shifts by 180 degrees right? I guess I'm having a little trouble visualizing the effect of subtracting the AP output on the overall phase response. LP - AP gives me HP. And an HP cascaded with an LP gives me the BP filter. An HP cascaded with an AP is still an HP but with a different phase response. And then how does then subtracting that from the BP give me another HP filter? Also I'm assuming that this is not perfect and for there to be some kind of phase distortion? I'm gonna take a look at the section on filters in the art of electronics pdf I have, but any suggestions on other resources where I can read more about this to get a better understanding? Once again, thanks so much!
Benta:
First, you won't achieve flat phase response with analog filters. For that you'll need to move into the digital domain (FIR filters).
Second: an AP filter has double the phase shift of an LP filter. This is the reason for pairing a 4th order (or rather 2nd order squared) LP with a 2nd order AP.
Navigation
[0] Message Index
Go to full version