Take this subjective opinion for the nothing that it is probably worth, but linearizing phase seemed most significant in sound to me in the upper bass. In the upper midrange and above, it makes for impressive looking impulse or square wave responses, but didn't seem to make a terribly significant difference to the sound. But that was changing from a phase response that wasn't too bad to begin with, to one with linear phase within about 15 degrees, so maybe not even a valid worthless subjective opinion!
Honestly, I have no idea. Historically phase was considered a non-issue, mostly because the ear above about 700 Hz is not phase sensitive - so how could phase matter? Then I read the paper by Griesinger and it made a lot of sense. It was completely consistent with what I know about hearing and it implied that phase in the region of 700-7 kHz was important to perception. What is the tolerance? Its too new a concept to me to sat what this is. I only know that the speakers that I have come to like for imaging etc. do tend to have flat phase. I have always appreciated that a compact impulse response was a good thing for perception, but I did not recognize that this also equates to a flat phase from 700 - 7 kHz.
This point was really driven home to me when I looked at my own speakers, which I believe are as good as any that I have ever heard, which did indeed have a very flat phase in this frequency range. Again, could be totally coincidental, and it is quite contrary to my previous beliefs, but the evidence is mounting.
If I still did psychoacoustic studies, this is an area that I would definitely study.
This point was really driven home to me when I looked at my own speakers, which I believe are as good as any that I have ever heard, which did indeed have a very flat phase in this frequency range. Again, could be totally coincidental, and it is quite contrary to my previous beliefs, but the evidence is mounting.
If I still did psychoacoustic studies, this is an area that I would definitely study.
What would you say the frequency range is? Sounds like it is below 700 Hz, which doesn't fit with my hypothesis. I really do think that some subjective testing is warranted.
I'd estimate between about 80Hz and 500Hz or so (could be wrong, though, as this is based on the instruments that seemed to sound more realistic, and those all have overtones that might be causing real effects....). Another confounding this is that room modes are in play there. Listening is done pretty close to the speakers to reduce that, and the FIR eq was done with mic very close to the speaker. Justifying the reported valuation of this kind of subjective report.
Having spent last night down in the cave listening to phase differences on a lot of different recordings, I'll go with Bill's frequency range.
My cave speakers are simple open baffle 2-way with 35' to the wall behind them. The smoothed phase looks exactly like it should, a 2nd order crossover at 700Hz and a medium Q box-like HP at 35Hz. Using rePhase to correct this brings the phase close to flat, instead of falling thru 720 degrees. I can turn this on or off at the push of a button.
The subjective change is subtle on my system, but audible. With phase flattened, the sound seems to tighten up and pull forward and center. The most noticeable area being the lower mids. Overall a little tighter, cleaner and more monophonic. That can be good or bad, I haven't decided.
My cave speakers are simple open baffle 2-way with 35' to the wall behind them. The smoothed phase looks exactly like it should, a 2nd order crossover at 700Hz and a medium Q box-like HP at 35Hz. Using rePhase to correct this brings the phase close to flat, instead of falling thru 720 degrees. I can turn this on or off at the push of a button.
The subjective change is subtle on my system, but audible. With phase flattened, the sound seems to tighten up and pull forward and center. The most noticeable area being the lower mids. Overall a little tighter, cleaner and more monophonic. That can be good or bad, I haven't decided.
I may have said this wrong, but my point was that it is the responsibility of a patent holder to do due diligence with respect to their patent. If it can be shown that they have not been diligent in protecting their rights then it can be argued that they have relinquished them.
I hold some 30 patents and have testified in numerous patent cases, but I am not a patent attorney.
I hold some 30 patents and have testified in numerous patent cases, but I am not a patent attorney.
Yes, it can be confusing what actually gives the "right", "correct", "realistic", subjective impression.
For instance, how far out of phase does the attack of base guitar - lots of signal over 1khz- have to be before it starts sounding "off"?
Phase might not matter with music that's small ensemble or with instruments with simple tonal structures, but it might matter with big ensembles or instruments like pianos.
Looking at it from the other end, how 'far off the reservation' do folk go using phase when making their mixes?
Hi All,
My opinions on linear phase;
It depends on where you're coming from, I am a professional acoustician and sound designer, I design large scale (1000-50.000 audience) sound systems / room acoustics and sound recording/mastering studios and I have experimented a lot with linear phase.
My findings are;
-With a selection of my reference tracks (which I have listened to 1000's of times in 20 years) I score 100% in abx testing in the frequency range 20Hz-5kHz IF the environmental circumstances are good.
With good I mean: (known)headphones, well-designed studio's (with virtuallly no standing waves), very large spaces (25.000m3 and larger or outdoor).
-In these situations there is virtually no room influence in the low frequencies and I find it very easy to detect linear or non-linear phase, especially in the 20-250Hz range.
I have to admit that I am very much trained on this, if you listen to this on a daily base my experience is that once you know what to listen for, you will hear it, the same as with clipping amps, limiting power supplies etc.
I am always very skeptical about subjective testing on a large group of people and drawing conclusions on this regarding these things, I feel that it is not legit to claim that things don't matter if "the majority" of people don't hear it. My girlfriend (who is a psychologist) certainly doesn't hear the differences UNLESS I point to her what to listen to, than she scores maybe 70% in abx testing. So in my opinion subjective testing is valid, but only if you test on trained listeners (unless you are designing the next crappy mp3/sonos devices for the masses).
What I have noticed with linear phase in a studio environment is it really helps hearing early reflections patterns from the live room (so where to put your microphones), and balancing mixes with percussive instruments. (linear phase in the 80-500Hz range makes hand percussion sound "pok" instead of "pak"), but also with things like piano. It makes sense that the more percussive sounds benefit more from linear phase of course.
What I have noticed with large scale PA systems that are linear phase in the 250-5000Hz range (the PA world never does linear phase in the sub-lows because of the latency) is that , besides snare drums and percussion, vocal intelligibility really benefits from linear phase, even to the point that you can take down speech vocal level by as much as 2-3dB without loosing definition, this is a big deal in PA-world where 3dB represent a lot of speakers.
What I also have noticed is that a system should ALWAYS be causal, by which I mean that the total wideband system phase response should follow the frequency response, i.e. if your subs drop of 24dB/ocatve @30Hz, your phase should follow this behavior, so it must also wrap 360 degrees. If you make the Hpf response of your system linear phase and not causal anymore it starts to sound really strange and "wrong", I also score 100% on this in abx testing.
To learn what to listen for I can recommend listening to music with hand percussion (congas bongos, snare drums etc.)
(I always use Paul Simon, Hearts and Bones, the last 20 seconds of the track)
with 8th order allpass filters digitally inserted (if you don't have DSP you can process your audio files with it in a free audio editor such as audacity). This would make the same phase response as a 16th order lowpass filter, which is excessive.
You can try the filters first on 250Hz (this is around where the fundamentals of most percussion live), it should be easy to to notice that the percussion does "piew" instead of "pok" because there is severe time-smearing. Once you score 100% on this you can try 4th order and 2nd order allpass filters and different frequencies. Try it on headphonse first and once you hear it switch to speakers, since your listening environment will make it harder to hear.
Disclaimer; sometimes it is better to not hear things, since it can make you enjoy music less...
my 2ct
Kees
One other, non-related, thing: I am trying to improve my English writing which I know is not too good. I would really appreciate it if the native English speakers could point out where I am using strange sentences, so pointers like "you don't say this like this, but like this" .
Thanks.
My opinions on linear phase;
It depends on where you're coming from, I am a professional acoustician and sound designer, I design large scale (1000-50.000 audience) sound systems / room acoustics and sound recording/mastering studios and I have experimented a lot with linear phase.
My findings are;
-With a selection of my reference tracks (which I have listened to 1000's of times in 20 years) I score 100% in abx testing in the frequency range 20Hz-5kHz IF the environmental circumstances are good.
With good I mean: (known)headphones, well-designed studio's (with virtuallly no standing waves), very large spaces (25.000m3 and larger or outdoor).
-In these situations there is virtually no room influence in the low frequencies and I find it very easy to detect linear or non-linear phase, especially in the 20-250Hz range.
I have to admit that I am very much trained on this, if you listen to this on a daily base my experience is that once you know what to listen for, you will hear it, the same as with clipping amps, limiting power supplies etc.
I am always very skeptical about subjective testing on a large group of people and drawing conclusions on this regarding these things, I feel that it is not legit to claim that things don't matter if "the majority" of people don't hear it. My girlfriend (who is a psychologist) certainly doesn't hear the differences UNLESS I point to her what to listen to, than she scores maybe 70% in abx testing. So in my opinion subjective testing is valid, but only if you test on trained listeners (unless you are designing the next crappy mp3/sonos devices for the masses).
What I have noticed with linear phase in a studio environment is it really helps hearing early reflections patterns from the live room (so where to put your microphones), and balancing mixes with percussive instruments. (linear phase in the 80-500Hz range makes hand percussion sound "pok" instead of "pak"), but also with things like piano. It makes sense that the more percussive sounds benefit more from linear phase of course.
What I have noticed with large scale PA systems that are linear phase in the 250-5000Hz range (the PA world never does linear phase in the sub-lows because of the latency) is that , besides snare drums and percussion, vocal intelligibility really benefits from linear phase, even to the point that you can take down speech vocal level by as much as 2-3dB without loosing definition, this is a big deal in PA-world where 3dB represent a lot of speakers.
What I also have noticed is that a system should ALWAYS be causal, by which I mean that the total wideband system phase response should follow the frequency response, i.e. if your subs drop of 24dB/ocatve @30Hz, your phase should follow this behavior, so it must also wrap 360 degrees. If you make the Hpf response of your system linear phase and not causal anymore it starts to sound really strange and "wrong", I also score 100% on this in abx testing.
To learn what to listen for I can recommend listening to music with hand percussion (congas bongos, snare drums etc.)
(I always use Paul Simon, Hearts and Bones, the last 20 seconds of the track)
with 8th order allpass filters digitally inserted (if you don't have DSP you can process your audio files with it in a free audio editor such as audacity). This would make the same phase response as a 16th order lowpass filter, which is excessive.
You can try the filters first on 250Hz (this is around where the fundamentals of most percussion live), it should be easy to to notice that the percussion does "piew" instead of "pok" because there is severe time-smearing. Once you score 100% on this you can try 4th order and 2nd order allpass filters and different frequencies. Try it on headphonse first and once you hear it switch to speakers, since your listening environment will make it harder to hear.
Disclaimer; sometimes it is better to not hear things, since it can make you enjoy music less...
my 2ct
Kees
One other, non-related, thing: I am trying to improve my English writing which I know is not too good. I would really appreciate it if the native English speakers could point out where I am using strange sentences, so pointers like "you don't say this like this, but like this" .
Thanks.
Last edited:
Thanks for posting this, Kees,
That might explain my preferred correction, while not a Horn system it does touch on the phase discussion going on here. Hope Nate approves this deviation of his horn topic, once again.
I have line arrays that I correct with FIR filters. I have lots of parameter choices and can obviously go for a linear phase or minimum phase behaviour.
I first tried this:
Generated minimum phase plot included, this is a 4 cycle frequency dependant window.
And later on I changed the phase behaviour to:
I expected the first plot (linear phase) to sound better, but I actually prefer the second one. It feels much more natural. It basically works on everything I throw at it.
I also recognise much of the rest of your response. You can learn to listen for these subtle differences. Once you learn to hear it, it becomes harder to unlearn. Depending on the recording the results can sound beautiful though. I experience it as more realism on a lot of instruments. And when you have this kind of phase behaviour it seems the midrange even feels more powerful.
Can't help you with sentences and other linguistic things as I make a mess of that myself.
That might explain my preferred correction, while not a Horn system it does touch on the phase discussion going on here. Hope Nate approves this deviation of his horn topic, once again
.I have line arrays that I correct with FIR filters. I have lots of parameter choices and can obviously go for a linear phase or minimum phase behaviour.
I first tried this:
Generated minimum phase plot included, this is a 4 cycle frequency dependant window.
And later on I changed the phase behaviour to:
I expected the first plot (linear phase) to sound better, but I actually prefer the second one. It feels much more natural. It basically works on everything I throw at it.
I also recognise much of the rest of your response. You can learn to listen for these subtle differences. Once you learn to hear it, it becomes harder to unlearn. Depending on the recording the results can sound beautiful though. I experience it as more realism on a lot of instruments. And when you have this kind of phase behaviour it seems the midrange even feels more powerful.
Can't help you with sentences and other linguistic things as I make a mess of that myself.
Kees,
Thanks for sharing your views and expertise on listening to linear phase and its importance. Well done - a great gem of info in this post.
You must be kidding asking about your English!? It's better than most native speakers ability to write. Not kidding.
Love the bit about how to listen for "pok" vs "pal" sounds.
Flat phase is sort of like transient perfect. Once you hear it you can't undo your learning and things don't sound right anymore unless you get it.
Thanks for sharing your views and expertise on listening to linear phase and its importance. Well done - a great gem of info in this post.
You must be kidding asking about your English!? It's better than most native speakers ability to write. Not kidding.
Love the bit about how to listen for "pok" vs "pal" sounds.
Flat phase is sort of like transient perfect. Once you hear it you can't undo your learning and things don't sound right anymore unless you get it.
Linear phase discussion is certainly welcome and is something I've been looking into.
Kees - your English looks pretty good to me. Having never learned another language it's impressive to me that those of you from abroad can carry on technical discussions here with clarity and are willing to do so. Thank you for that.
Kees - your English looks pretty good to me. Having never learned another language it's impressive to me that those of you from abroad can carry on technical discussions here with clarity and are willing to do so. Thank you for that.
A couple of points; Your English is very good
How does one make any room with no standing waves? As having studied this problem for decades ALL rooms have standing waves. Just because you do not have parallel walls does not means that you don't have resonances and hence standing waves. Are you maybe saying something else.
In any good psychoacoustic test one should define a very precise procedure that attempts to minimize the effect of lack of training. But that said I should also point out that Sean Olive at JBL has found that non-trained listeners arrive at the same results as trained listeners, they just take more trials. And from a scientific point of view one does not do tests to find the exception, but to find the norm.
Finally, in my own tests I have seen situations where trained listeners performed worse than untrained listener. They try and "second guess" the tests and end up just providing unreliable data.
As log as we are on linear phase, I have a couple of issues that I can't resolve in my own mind. Everyone here seems to be saying that they can hear linear phase in the LFs but not at higher ones. This agrees with tests done by Prof. Brian Moore for NXT (JAES publication) on the audibility of group delay where he found that under his conditions this was not audible. This paper is of note to me because in it he states that group delay is know to become more audible at higher SPLs. This was the first time that I had heard this.
Now, on the other side is Dr. David Griesinger who claims that phase would be audible an important for imaging from 700 - 7 kHz. His argument is persuasive. But still in stark contrast to the Moore study.
Its a dilemma to me as all these results seem to contradict one another.
PS. linear phase also means zero group delay.
PPS. If one does have flat LF sound field then would not the phase be flatter than otherwise. So for LF in the modal region doesn't saying that one has linear phase mean that one has also smoothed out the LF response. At LFs don't they have to be the same thing? (Not so at HF, of course) So is not linearizing the phase just soothing out the response?
PPPS. Could not "learning to listen" to certain effects be, in essence, just convincing yourself that you hear something that isn't actually there. Kees claims to do this in double blind tests, and yes that works, but he has not described the tests.
Now, on the other side is Dr. David Griesinger who claims that phase would be audible an important for imaging from 700 - 7 kHz. His argument is persuasive. But still in stark contrast to the Moore study.
Its a dilemma to me as all these results seem to contradict one another.
PS. linear phase also means zero group delay.
PPS. If one does have flat LF sound field then would not the phase be flatter than otherwise. So for LF in the modal region doesn't saying that one has linear phase mean that one has also smoothed out the LF response. At LFs don't they have to be the same thing? (Not so at HF, of course) So is not linearizing the phase just soothing out the response?
PPPS. Could not "learning to listen" to certain effects be, in essence, just convincing yourself that you hear something that isn't actually there. Kees claims to do this in double blind tests, and yes that works, but he has not described the tests.
Last edited:
I find this point very interesting. Are you sure that you really hear the phase response and not some artifacts of the FIR filtering (e.g. pre-ringing)?
Agreed on finding the norm. And also what's said here about 'trained' vs 'untrained' listeners can be explained more, for anyone interested, in a lecture given by Dr. Toole. It's on YouTube. Just do a little search and you'll find it.
Edit
I believe this is the one
https://www.youtube.com/watch?v=zrpUDuUtxPM
Edit Edit
I'm just starting to listen to this interview of Dr. Sean Olive. I am willing to bet it speaks to the point. By the time I'm done I won't be able to edit this in so here it is.
https://www.youtube.com/watch?v=6NgkP8aABO4
Last edited:
I missed this the first time around. I must be missing something because "linear phase" is not "causal". It seems to me you are saying saying contradictory things. Could you explain?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.