Dear all,

I am interested in audio analog filters and I have two questions, one regarding filters approximations and the other one related to signal polarity.

#1: I know in audio signal processing it is important to preserve a constant phase and group delay through the passband. From this, I suppose the best approximations are:

• Bessel
• Butterworth
• Inverse Chebyshev

Am I right ? What are your suggestions about the topic ?

#2: If I would like to input an audio signal to an inverting amplifier, it will sounds equally as with a non inverting amplifier ?

Thank you for your replies.

 

A butterworth squared filter would be better for maintaining phasing across the freq band.

also known as linkwitz riley'

http://en.wikipedia.org/wiki/Linkwitz–Riley_filter

 

#2: Inverting the signal has no effect on the perceived sound. The ear doesn't know or care what the original phase of the sound was.

 

Sallen and Key filters are also popular in audio work.

https://www.google.co.uk/search?hl=.....1ac.1.34.heirloom-hp..0.14.1096.PwkQowuQyUE

 

A butterworth squared filter would be better for maintaining phasing across the freq band.

OK, I will study something more about it.

#2: Inverting the signal has no effect on the perceived sound. The ear doesn't know or care what the original phase of the sound was.

OK that's fine. Thank you.

Sallen and Key filters are also popular in audio work.

Yes but Sallen and Key is just a kind of configuration, not a filter approximation.

You could build a Sallen & Key filter circuit with componets values calculated for a Butterworth or Chebyshev or Bessel approximation, but maybe you already know it.

 

What do you understand about the classification and implementation of filter types under the headings as opposed to classification by order or active/passive or whatever?

Butterworth (please note correct spelling)

Chebyshev

Bessel

Cauer

 

#2: If I would like to input an audio signal to an inverting amplifier, it will sounds equally as with a non inverting amplifier ?

OP amp in non-inverting configuration may sound worse, because of large common-mode voltage present at the input.
http://www.angelfire.com/ab3/mjramp/cm.html
but measurement not always confirm this
http://www.sg-acoustics.ch/analogue_audio/ic_opamps/pdf/opamp_distortion.pdf

 

#2: Inverting the signal has no effect on the perceived sound. The ear doesn't know or care what the original phase of the sound was.

This is apparently not the case.

http://www.diyaudio.com/forums/ever...bsolute-phase-2.html?postid=610921#post610921

w

 

This is apparently not the case.

http://www.diyaudio.com/forums/ever...bsolute-phase-2.html?postid=610921#post610921

If the phase of the various harmonics are changed relative to each other then it's may be possible to hear a difference. But if every thing is just reversed in phase then how can the ear hear that? The two waveforms are identical and the other way you know the difference is by comparing the input to the output, which the ear can't obviously do.

 

Most sub-woofer amps have polarity switches on them. Is this because the ears sensitivity to phase is frequency dependent or simply phase matching with the rest of the speaker system?

Mark

 

What do you understand about the classification and implementation of filter types under the headings as opposed to classification by order or active/passive or whatever?

I am not an expert about physical sound signal processing, but I think I have minimal knowledge about these approximations. Of course I do not remember everything and I am not going to copy and paste from the book I have to here.

You are right about the correct spelling of Butterworth.

 

Perhaps we discuss exactly what we mean by 'inverting the signal'?

I do not see that inverting even a signal that is not symmetrical about the horizontal axis makes any difference to well designed electronics.

However if you can guarantee that the eventual listener receives compressions as compressions and rarefactions as rarefactions as if he were at the original recording then I can see a possible source of difference for largish signals.

I do not think the response of the ear is sufficiently asymmetric to low level signals and very high level signals probably suffer other non linear masking.

So it is all down to ensure that the microphone diaphragm and the loudspeaker cone are going in the same or opposite directions.

I am not an expert about physical sound signal processing, but I think I have minimal knowledge about these approximations. Of course I do not remember everything and I am not going to copy and paste from the book I have to here.

Don't get huffy,

I am only trying to clarify your objective, to better offer help

 

Well, I sympathize with your take on the situation, but apparently it's easy to demonstrate by combining a tone and its octave, then switching the phase of the fundamental, when apparently most people can identify A or B with >9/10 reliability.

The other thing comes from physiology, apparently the nerves in the auditory system fire only on the rising pressure wave. This is not that surprising, given that the pulse rate is then identical to the audio frequency. If the ear was incapable of distinguishing between rising and falling peaks, the brain would have to account for 2 nerve impulses for each cycle of sound wave.

w

 

Well, I sympathize with your take on the situation, but apparently it's easy to demonstrate by combining a tone and its octave, then switching the phase of the fundamental, when apparently most people can identify A or B with >9/10 reliability.

So if I understand you correctly you are saying that most people can distinguish between two similar but different signals, since since that is what you will generate by having A = (fundamental + harmonic) and B = (-fundamental + harmonic)

How does that relate to my point about microphones and loudspeakers?

The other thing comes from physiology, apparently the nerves in the auditory system fire only on the rising pressure wave. This is not that surprising, given that the pulse rate is then identical to the audio frequency. If the ear was incapable of distinguishing between rising and falling peaks, the brain would have to account for 2 nerve impulses for each cycle of sound wave.

Thank you for that information. I was not aware of this, but I will look into it.

 

OK studiot, of course I did not want to get huffy, sorry if I made this impression.

My question arise from my interests in signal processing.
Additionally I like music so could be funny building an home made audio equalizer.

As I know, the better response for preserving costant group delay is the Bessel approximation, Butterworth is not so good as Bessel, and the group delay and so the phase shift, increases with the frequency.

I quickly reopened the Analog Devices chapter on analog filter design:

Chapter 8 Analog Filters

where a straight comparision of responses can be done looking at the graphs. Chebyshev with approximately 0 dB ripple seems to be a good choise, same as the Linear Phase with equiripple error.

I would have to look better again on the books I have for more detailed information.

 

For more fun, you should investigate phase shifting circuits as implemented to 'clean-up' audio signals that have been EQ'd. Several aftermarket pro audio examples exist of such equipment. The basic function is one of what sounds 'best to the ear' of the listener. The circuits will divide the signal by freq band(usually just bass and treble) and shift the phase of one with the other to restore whatever 'sound' the listener perceives as being closest to the original audio signal before it was processed by one or more of the filters being discussed. Most of them also offer additional 'effects' as well.

a quick search returned lots of hits, here are a few samples of the equipment.

http://www.fullcompass.com/category/Enhancers-Exciters.html

 

I'm stil not sure where you are going, but here are a couple of extracts from the BurrBrown databook.

 

[QUOTE="studiot]I'm stil not sure where you are going[/QUOTE]

studiot, I have to thank you but I am not asking for a circuit, also because I should read again the chapters on the books I have. I use the book by Van Valkenburg "Analog Filter Design" and "Filtering in the Time and Frequency Domains" by Blinchikoff & Zverev.

I have to say that both are very good books and the concepts are written very clearly.

Regarding audio applications, I was wondering if there are some situations where some designer prefer the xxx approximation because in some situations etc etc etc.

Maybe there is no particular answer for my question, or better said, the better answer is: depends on what are the requirements.

Perhaps building and testing is the best solution.

 

Well, I sympathize with your take on the situation, but apparently it's easy to demonstrate by combining a tone and its octave, then switching the phase of the fundamental, when apparently most people can identify A or B with >9/10 reliability.

True, but that is not the point. The point is about inverting the entire signal, not just one frequency band with respect to another frequency band. Scrambled phases are relatively easy to detect, as are out-of-phase stereo signals when you know what to listen for. Complete inversion is missed by most people, except for trained pros.

ak

 

I agree with AnalogKid and the other users who said the difference it's hard to detect.

I tried with a simple audio editor, playing a track for a few seconds, then I inverted the signal and played again. I cannot distinguish the difference.