Actually, I'd rather wonder if phase accurate XO is what really needs the question - does it matter. We don't really hear the phase of the signal, at all, unless there are severe and abrupt changes in the phase response. http://www.ethanwiner.com/phase.html The only thing that we hear in relation to phase, is exactly differences between the two ears.Pan said:
IMO, phase accurate XO is not required, as long as phase is same on both channels and response is smooth. What matters is channel matching and avoiding bad acoustic interactions between drivers at XO. Thats best simply avoided rather than dealt with.
By placing XO into phase sensitive hearing range, we need more care with XO component matching, polar response issues, etc.
Prune,
before you get too excited about square waves, check the link above, and this http://sound.westhost.com/ptd.htm.
Right, this is how XO must be done. The thing that matters is channel matching. I don't think its trivial when your goal is 15microsec precision of phase matching. The less of variables you need to match, the simpler it is.
Phase is important for imaging. Assuming everything else is perfect (i.e. quality binaural recordings with headphones, or with speakers with cross-talk cancellation preprocessing of the sound taking into account the speaker/room setup geometry), then I think phase problems will become a good deal more audible, at least in certain frequency ranges (according to this, phase issus in the low frequencies do not matter).
More interesting discussion of phase, though in another context, is here.
More interesting discussion of phase, though in another context, is here.
Pan said:
I agree with that findings. However phase distortion in the midrange does not affect imaging. It only affects transient response and may lead to sound smearing. A simple way to avoid this is using XO filters with minimal group delay. For example, BW3 (inverted driver polarity) is a good choice for a midrange XO filter target. OTOH, I would avoid the popular LR4 below 1...2 kHz.
Phase itself is not important for imaging. Not a single driver has linear phase response anyway, let alone flat. For imaging, interaural timing is important. Timing has relation with phase, but its not synonymous. XO causes timing changes, so does baffle step. You can adjust timing, and you need only limited precision in that within channel. If that wasn't so, you'd have to position your tweater with micrometer precision.Prune said:
Chasing for perfect phase accuracy is futile, and gives really nothing. You get your square wave only in very precise point in space, and any move from that spot destroys the square wave. Linear phase of XO and acoustic phase response of physical drivers over XO point are very different things. Accounting for that requires XO device thats far from phase accurate.
The very value of the localization article is that it explains the cues human ear is using for imaging, and reasons. Its useful in gaining further understanding what changes are audible. I focused precisely on imaging. The article shows why channel phase match is important below 1.5k and not important above 3k.
What you do want is to avoid acoustic source changing its physical location in phase sensitive range, to keep phase differences between channels as small and linear as possible. It is difficult to achieve unless you keep all the phase sensitive freq range on the same driver.
Proved?? I don't know what they are hearing, but I suspect it is not phase. There are loads of other variables that are interrelated, and if you change one, others change too. Blaming it all on phase distortions is gross oversimplification. Evergreen argument is that phase distortion has occured zillions of times before it ever reaches your speakers. If phase distortion as induced by crossovers in equalizers is prooved audible, then we are doomed, truely.Pan said:
That implies whole signal. In no way does this prove that phase changes within that material are audible unless they are huge and nonlinear.
"Evergreen argument is that phase distortion has occured zillions of times before it ever reaches your speakers. If phase distortion as induced by crossovers in equalizers is prooved audible, then we are doomed, truely."
There are recordings and there are recordings. Sure, many studio productions have a real mess of all types of alterations of the signal, but I tend to prefer purist recordings for acoustic music simply because it sounds more natural and pleasing to me. When I do recordings with my QTC1´s straint into a SBM or PC soundcard, there are no abrubt phase jumps introduced to the signal as x-overs do. The bandpass nature of microphones and electronics is a "no issue" since they are well behaved in a great part of the audiorange. Sure, we have the usual lowpassfilter in PCM, but I´d say (even though audible) gross phase distortion in the midband (as with high order crossovers) is worse.
Also an EQ used carefully does not introduce anywhere near the amount of phase distortion that a 4th order LR filter does.
http://www.ocf.berkeley.edu/~ashon/audio/phase/phaseaud2.htm
/Peter
There are recordings and there are recordings. Sure, many studio productions have a real mess of all types of alterations of the signal, but I tend to prefer purist recordings for acoustic music simply because it sounds more natural and pleasing to me. When I do recordings with my QTC1´s straint into a SBM or PC soundcard, there are no abrubt phase jumps introduced to the signal as x-overs do. The bandpass nature of microphones and electronics is a "no issue" since they are well behaved in a great part of the audiorange. Sure, we have the usual lowpassfilter in PCM, but I´d say (even though audible) gross phase distortion in the midband (as with high order crossovers) is worse.
Also an EQ used carefully does not introduce anywhere near the amount of phase distortion that a 4th order LR filter does.
http://www.ocf.berkeley.edu/~ashon/audio/phase/phaseaud2.htm
/Peter
Good example of wishful thinking. He is under delusion. He quotes paper that shows with scientific method that group delays of upto 8ms are completely inaudible, and then he makes uncontrolled "test" at home to show he can hear smooth submillisecond phase shifts. Of course he does, but it isn't phase shift he hears. He wants to blame it on phase shift, and so he blames. He could aswell blame his room acoustics with changing FR, his setup and his mood, really.
Well, whatever. I'm not against people wasting their time on imagining phase accuracy in highly reverberant room.
So you sugest the following text is something someone just made up under a coffe break?;
"Although not in large numbers, previous research in investigation of the audibility of phase distortion has proven that it is an audible phenomenon. Lipshitz et al. [7] has shown that on suitably chosen signals, even small midrange phase distortion can be clearly audible. Mathes and Miller [8] and Craig and Jeffress [9] showed that a simple two-component tone, consisting of a fundamental and second harmonic, changed in timbre as the phase of the second harmonic was varied relative to the fundamental. The above experiment was replicated by Lipshitz et al., with summed 200 and 400 Hz frequencies, presented double blind via loudspeakers resulting in a 100% accuracy score. An experiment involving polarity inversion of both loudspeaker channels resulted in an audibility confidence rating in excess of 99% with the two-component tone, although the effect was very subtle on music and speech. Cabot et al. [10] tested the audibility of phase shifts in two component octave complexes with fundamental and third-harmonic signals via headphones. The experiment demonstrated that phase shifts of harmonic complexes were detectable."
/Peter
"Although not in large numbers, previous research in investigation of the audibility of phase distortion has proven that it is an audible phenomenon. Lipshitz et al. [7] has shown that on suitably chosen signals, even small midrange phase distortion can be clearly audible. Mathes and Miller [8] and Craig and Jeffress [9] showed that a simple two-component tone, consisting of a fundamental and second harmonic, changed in timbre as the phase of the second harmonic was varied relative to the fundamental. The above experiment was replicated by Lipshitz et al., with summed 200 and 400 Hz frequencies, presented double blind via loudspeakers resulting in a 100% accuracy score. An experiment involving polarity inversion of both loudspeaker channels resulted in an audibility confidence rating in excess of 99% with the two-component tone, although the effect was very subtle on music and speech. Cabot et al. [10] tested the audibility of phase shifts in two component octave complexes with fundamental and third-harmonic signals via headphones. The experiment demonstrated that phase shifts of harmonic complexes were detectable."
/Peter
Also you could check up pcabx.com for a file where you can see if you are able to hear the phase distortion of a 4th LR in mid frequencys. People I trust has reported a very different sound on fingersnapps for example, using this test. Could be done with headphones.
I have never done any tests with audibility of phase distortion but I have always had a feeling of it must be audible. Now I hear and read more and more supporting that.
/Peter
I have never done any tests with audibility of phase distortion but I have always had a feeling of it must be audible. Now I hear and read more and more supporting that.
/Peter
This directly contradicts the Lipshitz findings:
Harwood, H. "Audibility of phase effects in loudspeakers." Wireless World, January 1976, pp.30-32.
I don't have a copy; it's quoted by D. Self here.
Andrew's article also links to Linkwitz's phase discussion, which, ironically, contradicts his position that phase distortion is audible. However, Linkwitz is talking about a 100 Hz crossover, and I've already mentioned that phase is less important in the low end.
In any case, I still have not found contradiction to the findings that phase is important for localization in human hearing. This is why my guess is that phase distortion should be studied with headphones and quality binaural recordings. Perhaps I'll find the time to eventually conduct my own blind tests. It looks like Kreskovsky's crossover is my best bet. Andrew's article mentions it, and he says it has low sensitivity. What does that mean?
Harwood, H. "Audibility of phase effects in loudspeakers." Wireless World, January 1976, pp.30-32.
I don't have a copy; it's quoted by D. Self here.
Andrew's article also links to Linkwitz's phase discussion, which, ironically, contradicts his position that phase distortion is audible. However, Linkwitz is talking about a 100 Hz crossover, and I've already mentioned that phase is less important in the low end.
In any case, I still have not found contradiction to the findings that phase is important for localization in human hearing. This is why my guess is that phase distortion should be studied with headphones and quality binaural recordings. Perhaps I'll find the time to eventually conduct my own blind tests. It looks like Kreskovsky's crossover is my best bet. Andrew's article mentions it, and he says it has low sensitivity. What does that mean?
Studying phase using headphones does not show how we perceive phase problems in the real world. We don't hear things in only one ear, sound is heard by both ears and the temporal and level differences create the picture we hear. The interaction between the speakers signals when it reaches our ears is the issue.
Again, remember that phase and timing between the ears is one thing, phase distortion as introduced by a crossover is another thing. The audibility of crossover distortion can be checked with headphones to avoid certain issues with a stereo speaker set up.
Wether further effects can show up when listening to loudspeaekrs, I don´t know, probably. Though I believe that phase distortion from crossovers is audible, I don´t think it necessarily have so much with imaging and localization to do, but more about how it sounds and how it affects listening fatigue and such things.
That said, the most natural soundstage, imaging and localization I have heard when listening to loudspeakers, has been from 1st or low order filters with good attention to room acoustics.
/Peter
Wether further effects can show up when listening to loudspeaekrs, I don´t know, probably. Though I believe that phase distortion from crossovers is audible, I don´t think it necessarily have so much with imaging and localization to do, but more about how it sounds and how it affects listening fatigue and such things.
That said, the most natural soundstage, imaging and localization I have heard when listening to loudspeakers, has been from 1st or low order filters with good attention to room acoustics.
/Peter
How can you simulate the phase relationship of multiple sources with headphones? The woofer and tweeter are usually not concentric and this is where a lot of the problem arises. Two points creating the same freq but not in phase alignment with eachother will cause the image to shift.
The crossover distortion is an electrical property while we are looking at the end acoustic result.
The crossover distortion is an electrical property while we are looking at the end acoustic result.
Did you read the actual paper? This excerpt is just mentioning results of previous work, its NOT result the quoted Koya paper reached. Andrew is blatantly misquoting Koya! "Hidden agenda" alarms are ringing. Read the results: http://www.music.miami.edu/programs/mue/Research/dkoya/chapter_5/chapter_5.htm and notice that content that even remotely exceeded random chance correlation (result >8) was for 70Hz sawtooth and impulse with 4ms group delay distortion. For rest, 8ms was random chance - no correlation. Note how audibility results dropped dramatically when going from headphones to speakers.Pan said:
Yes, I do not trust Lipshitz results, as he uses waveforms suitably chosen, where suitably means discontinuous signals introducing heavy spectral content that isn't intended. People do mistakes. I have conducted ABX on the famous 200+400 Hz signal with phase changes, there is _nothing_ to be heard of difference. What Lipshitz heard was some other distortion.
The Koya research established that only gross group delay distortions are audible, where gross means over 4-8ms.
And thats the point. What we hear is not phase, but too bad group delays. Phase is attached to frequency, group delay is not. Thats what you can possibly hear in steps and pulses. Thats your "smearing". Your hearing is able to distinguish temporal events, but that resolution is limited, fundamentally.
Consider this. 360 degrees phase shift is exactly 1 period of a sinewave. Its a delay of 1 period. Before 1 period of any frequency component has passed, there is even no way to tell that such frequency is present in the signal. Its undeterminate yet. Now, for any higher frequency F at least 2 periods must pass before F/2 is even defined. Your temporal resolution of frequency is fundamentally limited by its period. For 20Hz signal, you simply have no means to detect its occurance timing with better than 50ms precision.
Now consider LR4 with cutoff at 300Hz. It has 1.5ms group delay at 300Hz. Thats half the period. And it has 1.5ms delay at 20Hz too. Thats such a small fraction of the possible temporal resolution at 20Hz that it is irrelevant. For higher frequencies group delay falls faster than temporal precision grows - ie. temporal "smearing" is never larger than 1 half of period - at XO frequency. And thats the same for any XO frequency - largest group delay is 1 half period at XO, and is smaller for any other frequency.
The paper I posted shows that in midrange resolution of phase _difference between ears_ is orders of magnitude higher than temporal resolution. Thats possibly source of confusion and also audible differences due to mismatch of channels.
Infact, the sensitivity is shown to be so incredibly high (13us), that it might be used to argue that 44.1K sampling rate (22.7us) is inadequate to reconstruct that precisely.
It shows. You should do them. With headphones, then with speakers. Eliminating sighted bias. Eliminating other more important distortions first, or at least not closing your eyes on them.Pan said:
http://www.pcabx.com/technical/LR-300-3K/index.htm
Thats a good source for revelation. Take the samples, ABX them - 20 tries without seeing score. I won't even try scoring ABX, as it is obvious from few tries that there is nothing to be heard.
I had similar feelings too. Until I couldn't ABX few completely different looking waveforms under any conditions at home. ABX is revelation. I couldn't believe how much bias affects hearing until I got proof in my face.
Now I hear and read more and more that $8000K cable is more "fluid" than $2000K cable, too. Its a quality of tests that matters, not quantity.
Proper listening tests to prove audibility of phase alone are difficult to make. There is too much false positives due to secondary effects. ALL references to audibility of phase distortion infact admit that the audibility is very subtle even in face of quite large distortions. That alone should make you very cautious - phase shifts cause so severe waveform changes that you face the core question: either phase change is always immediately audible without effort as pitch or timbre, or it has no or very little correlation with audibility at all. Evidence for the latter is overwhelming.
See what John Atkinson has to say about phase:
http://www.stereophile.com/reference/100/index3.html
Thats what Kevin Voecks has to say about this:
http://www.hometheaterhifi.com/volume_11_2/feature-interview-kevin-voecks-4-2004.html
under "Do you feel that time-coherence is an important property of a loudspeaker"?
Note, thats a man who has conducted 1000+ blind listening tests in unechoic chamber and actually develops speakers.
Check the waveforms attached. There are 4 wavs, 200Hz main harmonic with added 400Hz harmonic at 0, 90, 180, -90 degrees phase shift. They all look different, even asymmetric. I can't hear any difference. You could run them in sequence looping, and they would sound as one. I'm using AKG-K401 headphones.
wimm,
thanks for the zip, I´ll try it soon with my HD600.
Regarding phase distoriton being subtle, that´s possible and likely, but I think all sources of errors should be attacked. It´s not necessary to enjoy a rig, but it´s the only way to make progress... eliminating all (as much as possible) alterations to the twodimensional signal. With a "perfect" replica of time and amplitude we can concentrate on teh enormous problem with power response and room acoustic induced "distortion".
You talk almost exclusively about sinewaves while I believe that it is mostly the transient distortion and ringing that is problematic with high order analog filters.
About Kevin Voecks. I auditioned the studio (or saloon) and then went home to a friends studio where a "phase linear" speaker designed by me was playing, no competition. Allthough I like the Gem, much due to the freedom from phase turns in the midrange I believe. Sure, different electronics and rooms, but still, in the end I always prefer low order filter with nice time behaviour.
I will not argue more until I have prooved I can nail phase distortion down in blind tests though ;-).
Thanks for a good debate.
/Peter
thanks for the zip, I´ll try it soon with my HD600.
Regarding phase distoriton being subtle, that´s possible and likely, but I think all sources of errors should be attacked. It´s not necessary to enjoy a rig, but it´s the only way to make progress... eliminating all (as much as possible) alterations to the twodimensional signal. With a "perfect" replica of time and amplitude we can concentrate on teh enormous problem with power response and room acoustic induced "distortion".
You talk almost exclusively about sinewaves while I believe that it is mostly the transient distortion and ringing that is problematic with high order analog filters.
About Kevin Voecks. I auditioned the studio (or saloon) and then went home to a friends studio where a "phase linear" speaker designed by me was playing, no competition. Allthough I like the Gem, much due to the freedom from phase turns in the midrange I believe. Sure, different electronics and rooms, but still, in the end I always prefer low order filter with nice time behaviour.
I will not argue more until I have prooved I can nail phase distortion down in blind tests though ;-).
Thanks for a good debate.
/Peter
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.