I don't think one is needed, we both know the purpose of proper ammount of smoothing.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
Audibility of group delay at low frequencies
- Thread starter Kvalsvoll
- Start date
I don't think one is needed, we both know the purpose of proper ammount of smoothing.
OP
- Thread Starter
- #42
100ms only. It was difficult.
OP
- Thread Starter
- #43
Did you listen to speakers or headphones?
Oh, in sighted headphone listening, the 100 was very easy to tell.
To speakers. I needed to crank it up to produce app 85-90dB at my LP to be able to hear the difference.
@Kvalsvoll , @amirm , does my subjective description of difference matches your perception?
OP
- Thread Starter
- #46
I say a 100% match. Once I reach a volume loud enough on speakers, it is easy to hear, the original hits harder and more precise, more realism.
But it was also possible to hear on headphones, so it is not just tactile punch - sound physically felt on the body - hearing is also affected.
One aspect worth noticing is that even the horrible delay of 100ms does not destroy sound quality in a way that makes it worthless to listen to. At low volumes, if you can not compare directly, the difference will be hard to notice.
Yes, the 100 msec file had anemic/different bass. The 20 msec was close to original but I thought the difference was there as well. I did not get a chance to do a blind test though.
I thought the levels were also different but did not investigate that much.
OP
- Thread Starter
- #48
It is not impossible that you can hear the difference in frequency response on the 100ms, there is a broad dip peaking at -1.2dB around 60Hz, and that can be audible.
This frequency response error should have been corrected, but it takes time and effort, so I chose just to do the article and publish the samples as they were.
I take it you listened the samples on headphones? If that is so, is there any chance you can listen it on your speakers as well?
P.S. I checked the levels in Audacity and they seem to be the same.
KSTR
Major Contributor
That's my impression as well and it underlines that group delay / phase issues are basically of minor significance, yet not completely irrelevant. It's the icing on the cake so to say. IMO it should *not* be a primary design goal to strive for linear phase at the cost of other parameters of potentially higher relevance. Trimming a well designed multiway speaker to linear phase is trivial these days anyway, with DSP power available in any PC etc.
Finally, getting a tolerable LF response from the room is most likely far more important to begin with.
dasdoing
Major Contributor
I would prefer hearing this with a more natural original. the original allready has stranges delays in the bass
tuga
Major Contributor
Where (on which menu) do you control the length of the window?
It's under /Tools/IR Windows
But I think that example would better be done with applying FDW than reducing Right window, so you may want to try to apply FDW of 3 cycles to show realistic GD. If you have less reflections in your room you can try raising # of cycles, but I'd say width of 3 cycles gives a good measure of GD.
dasdoing
Major Contributor
on reflections influence GD I have this.
same speaker, same room. green: no treatment, violet: after treatment (only the speaker position is a little different afair).
no smoothing/windowing
same speaker, same room. green: no treatment, violet: after treatment (only the speaker position is a little different afair).
no smoothing/windowing
Seems like the group delay for after treatment has less deviation from 0 ms. By the way why does group delay goes to negative? Is 0 ms just reference not absolute?
dasdoing
Major Contributor
Don't all the graphs have negative values? I don't realy understand fully this graphs in a tecnical way.
but if you generate minimum phase version, the exessive graph is all in the positive.
I use exessive phase GD plot to get the delay right on my sub which I included after these graphs
Last edited:
OP
- Thread Starter
- #57
GD is just the derivative of phase, which means it will be a negative number when phase turns in the negative direction.
Pio2001
Senior Member
The opposite : it is negative when phase rises, and positive when phase falls.
Physically, it represents the position where the maximum amplitude of a short frequency burst is shifted.
If you have a resonant room mode, the amplitude is going to rise as and when the sound bounces from a wall to the other, until it reaches a maximum. This maximum occurs later that in the original signal. The group delay is positive.
If you have an anti-resonant room mode that absorbs the sound, then, as the sound bounces from a wall to the other, the amplitude at the listening position starts decreasing as soon as the walls start playing against the speaker, even if the original signal goes on increasing for a short while.
The maximum amplitude was reached sooner that on the original. The group delay is negative.
Pio2001
Senior Member
Here is an artificial example of group delay caused by a one band parametric EQ.
Top : linear phase parametric bass boost. Group delay = 0. Some pre-ringing is visible.
Bottom : minimal phase parametric bass boost. Group delay = 70 ms. No pre-ringing.
The positive group delay shifts the position of the maximum of the bass burst to the right.
Negative group delay occurs when the process cancels the trail of the burst. Only the left side part remains. The maximum is thus shifted to the left.
Top : linear phase parametric bass boost. Group delay = 0. Some pre-ringing is visible.
Bottom : minimal phase parametric bass boost. Group delay = 70 ms. No pre-ringing.
The positive group delay shifts the position of the maximum of the bass burst to the right.
Negative group delay occurs when the process cancels the trail of the burst. Only the left side part remains. The maximum is thus shifted to the left.
Pio2001
Senior Member
Oh I forgot.
You can listen to it :
You can listen to it :
Similar threads
