[Joe Bloggs]

I know that minimum phase EQ is recommended for correcting frequency response (since frequency response distortions are often themselves minimum phase in the first place, correcting with a minimum phase filter often also corrects the phase response), but what about an EQ setting that you know full well isn't a correction?  For some songs I want a really strong bass boost.  I found this "bug" in Poweramp for Android where if you lowered many of the sliders in its EQ (as you might when doing subtractive EQing) you get more and more subbass bass boost--and this was a very tight boost, like a kick in your skull.
 
Now I've EQed all my songs on my phone for correct response with my Philips SHE3580 using foobar2000 and Electri-Q on my computer and the "subbass" EQ setting on Poweramp doesn't work anymore (since there's nothing to correct, I'd have to just lower ALL the sliders, but that somehow changes the frequency response too much).  I tried different parametric EQ settings on Electri-Q on my computer and on Neutron amp on Android but nothing seems to quite provide the tight kick that bugged Poweramp EQ provided.  I could dial the bass quantity up and up but whereas I got this kick in my head with bugged Poweramp, I get this drawn out rumble, like distant thunder, with Neutron for example.  I'm theorizing here that:
 
1. "Bugged" Poweramp is bugged because it somehow doesn't EQ the lowest frequencies, so when you lower many sliders, the lowest frequencies get boosted relative to the rest of the spectrum. (I could for example dial in much the same setting with the sliders higher up the scale and the subbass seems weaker.)
2.  Poweramp may be using a linear phase EQ while I know that Electri-Q is minimum phase, and Neutron may also be minimum phase?
 
Anyway, intuitively it seems that after I'd corrected the frequency and phase response with a minimum phase filter (Electri-Q), boosting the bass artificially with a minimum phase filter would skew the phase response more and more as I boost the bass, but would this be audible?  Whilst if boosting with a linear phase filter, the bass hump should remain centred in the attack although there may be ringing before and after, which may or may not be audible.
 
Anyway I've reached the end of my knowledge here.  Any help from the experts here?

 

[xnor]

It depends on the type of filter, the boost, and the corner frequency. You can even have negative group delay with min. phase filters.
 
Without analyzing what Poweramp does to the signal with all sliders lowered it's hard to tell what's going on.
 
I wouldn't worry about audibility of phase, less than about pre-ringing anyway.

 

[Joe Bloggs]

Quote:

It depends on the type of filter, the boost, and the corner frequency. You can even have negative group delay with min. phase filters.
 
Without analyzing what Poweramp does to the signal with all sliders lowered it's hard to tell what's going on.
 
I wouldn't worry about audibility of phase, less than about pre-ringing anyway.

 
Isn't the phase response of a minimum phase filter deterministic upon the frequency response?  But I don't know anything about the relationship.  Would a bass boost cause bass frequencies to shift later in time relative to other frequencies?

 

[xnor]

Quote:

Isn't the phase response of a minimum phase filter deterministic upon the frequency response? 

Yes (actually: magnitude response, but this and FR are often used synonymously here)
 
Quote:

Would a bass boost cause bass frequencies to shift later in time relative to other frequencies?

As I wrote, it depends. If you use a low shelf filter with a center frequency of 31 Hz, signals above 31 Hz will appear at the filter's output before they're fed into it (negative group delay). Use a peaking aka bell filter and signals around 31 Hz will be delayed quite a bit (possibly hundreds of samples of group delay depending on the filter Q and gain).

 

[EthanWiner]

Quote:

I wouldn't worry about audibility of phase, less than about pre-ringing anyway.

Exactly. Phase shift itself is benign and inaudible in "normal" amounts as you get from amps and preamps, and as used by conventional equalizers. I recently tested a linear phase EQ and the pre-ringing was very noticeable and not in a good way. I'd never use one of those, ever.
 
--Ethan

 

[Joe Bloggs]

Quote:

Yes (actually: magnitude response, but this and FR are often used synonymously here)
 
As I wrote, it depends. If you use a low shelf filter with a center frequency of 31 Hz, signals above 31 Hz will appear at the filter's output before they're fed into it (negative group delay). Use a peaking aka bell filter and signals around 31 Hz will be delayed quite a bit (possibly hundreds of samples of group delay depending on the filter Q and gain).

 
So if I understand you correctly, if I use a low shelf bass boost the bass would be delayed relative to the rest of the frequencies--by how much?  Would it be enough to cause the bass to rumble instead of kick as I mentioned?  I was boosting the bass by up to 14dB relative to what I usually consider "neutral" (where I'd cut the bass).  In Neutron's case I was using a center frequency of about 80Hz and a Q of 1 (which I understand causes a bit of an N-shaped artifact in the magnitude response around the crossover point, but I wanted a sharp crossover), whereas in Electri-Q I was using the Butterworth 24dB/octave low shelf.  Does the type of low shelf filter matter? 
 
Quote:

Exactly. Phase shift itself is benign and inaudible in "normal" amounts as you get from amps and preamps, and as used by conventional equalizers. I recently tested a linear phase EQ and the pre-ringing was very noticeable and not in a good way. I'd never use one of those, ever.
 
--Ethan

 
I know I should use a minimum phase filter if I'm aiming for "correct" response, but...
 
What the hey, there's nothing for it but to experiment.  Can any of you recommend a good free parametric equalizer with a linear phase mode?

 

[xnor]

Low shelf, 80 Hz, Q=1:
 

 

 
200 samples / 44100 Hz = 4.5 milliseconds
 
edit: fixed x-axis labels

 

[EthanWiner]

Quote:

What the hey, there's nothing for it but to experiment.  Can any of you recommend a good free parametric equalizer with a linear phase mode?

Yes, I always urge people to experiment. The best way to hear the failings of a linear phase EQ is with transient sounds like a side stick hit on a snare drum, or calves (Latin percussion instrument). I don't know of any freeware, but I imagine Google will find one if it's out there.
 
--Ethan

 

[BlindInOneEar]

xnor: 
 
Those are interesting charts you posted.  In my experience though 14 db of gain is a lot, far more than I would ever use at home.  Could you post additional charts with more modest amounts of gain, say 3 db or 6 db? 
 
Also, is there a similar delay effect when using cut instead of boost?  That would be interesting to see too.  I actually tend to use cut far more than boost when I equalize.
 
My equalizer consists of the tools JRiver includes in their Media Center.
 
Thanks in advance! 

 
edited to ask an additional question: Does the Q applied effect the delay as well?  Could a less abrupt transition allow you to add in the desired change with less delay? 

 

[xnor]

Quote:

edited to ask an additional question: Does the Q applied effect the delay as well?  Could a less abrupt transition allow you to add in the desired change with less delay? 

 
Yes, of course. Lower Q means smoother change in the magnitude response which also means less phase shift. In the example above changing the Q from 1 to 0.7071 causes the max. group delay to go down to 135 samples (from 206).
 
A gain of 6 dB with a Q of 0.7071 --> max. group delay = 58 samples.
3 dB --> 29 samples.
 
Blauert & Laws (1978) have published thresholds on the audibility of group delay. 500 Hz has a threshold of 3.2 ms = 141 samples (44.1 kHz sample rate).
 
edit: A filter with a cut instead of a boost has the same group delay just inverted (-20 becomes +20 and vice versa).

 

[Chris J]

Yes, I always urge people to experiment. The best way to hear the failings of a linear phase EQ is with transient sounds like a side stick hit on a snare drum, or calves (Latin percussion instrument). I don't know of any freeware, but I imagine Google will find one if it's out there.

--Ethan

CLAVE!

CLAVE!

The latin instrument you speak of is called the clave.
It is two sticks struck together, so if you are trying to simulate an impluse response, it is a good instrument to do so.

pronounced: CLA- vey

 

[EthanWiner]

Quote:

The latin instrument you speak of is called the clave.

Sheesh, I knew that! Damn spell checker. And damn me for not reading more carefully before clicking Submit.

 
This is not the first time that happened, so I just added clave and claves to my User dictionary.
 
--Ethan

 

[Tilpo]

Blauert & Laws (1978) have published thresholds on the audibility of group delay. 500 Hz has a threshold of 3.2 ms = 141 samples (44.1 kHz sample rate).

I just read a section of the book "An Introduction to the Psychology of Hearing" by Brian C.J. Moore. He seems to refer to the same paper, but he states that audible threshold for group delay was at about 400us instead. This was in the range of 1-4kHz, though, so that may explain the difference. However, I would expect detection of phase detection of lower frequencies to be better than for higher frequencies due to a higher degree of phase locking in neural firings.

Did you quote the 3.2ms from the paper itself, or did you use a secondary source?

 

[xnor]

Quote:

I just read a section of the book "An Introduction to the Psychology of Hearing" by Brian C.J. Moore. He seems to refer to the same paper, but he states that audible threshold for group delay was at about 400us instead. This was in the range of 1-4kHz, though, so that may explain the difference. However, I would expect detection of phase detection of lower frequencies to be better than for higher frequencies due to a higher degree of phase locking in neural firings.
Did you quote the 3.2ms from the paper itself, or did you use a secondary source?

That's from the paper. At 2 kHz the threshold is about 1 ms and this threshold is rising in both directions.

 

[Chris J]

Sheesh, I knew that! Damn spell checker. And damn me for not reading more carefully before clicking Submit. :blink:

This is not the first time that happened, so I just added clave and claves to my User dictionary.

--Ethan

Oh well, we seem to have escaped unscathed!
BTW, thanks for all the audio comments in general.