If I recall correctly, Greisinger's discussion was not about inter-aural but intra-aural. Maybe both?
I will have to read Zwicker's section again, but I am more inclined to music as a realistic test signal. One can get mislead using un-real signals for testing.
Since the phase distortion of a crossover would be the same as an excess group delay around the crossover, the work that Brian Moore did on the audibility of group delay would be pertinent. He did not find it significant for music in a real listening room setting at normal listening levels, but admitted that it could be audible under some other conditions like higher SPLs or less room reflections.
It sounds like it might be audible under certain conditions but that it is not a "night and day" thing as some have claimed it to be. Sounds like it is rather subtle at best. Still might be worth looking into.
Dumptruck - I agree, I will not do a search for others, but if I know of a study off the top of my head I will reference it. That's what I was looking for, not asking you to do a search for me. Like the Moore paper, it came to me after I thought about the topic for a minute.
You have to control these things or you just get back noise. If it is worth doing it is worth doing right. If I do it, it will be done right. At any rate, it sounds like there may already be enough data to put this to rest. I'm not interested in studying things that make so little difference. There are lots of things that make a big difference. Those are worth the time spent.
Hi Art
This is pretty much where I am coming out on this as well. Except for the few who hear "major improvements" the rest don't seem to hear much at all. But that's usually the case. There are always those who hear things that the rest of us cannot. What I always find amazing is that almost universally these claims are state as "obvious" to even the most casual listener. It seems a lot like "The lady doth protest too much, methinks."
Art,
Thanks for taking the time to post actual measures and your listening impressions. Much appreciated.
Re: “Although one can argue that linear phase "sounds better", from a frequency response standpoint, it does not appear to be measurably better.”
My interest is in the time domain. Are you able to post an impulse response from your measures?
What I am looking for is if anyone has measured a linear phase corrected speaker system that has a similar impulse response to the ones I have posted. So far, it appears no one has.
This article clearly shows that linear phase XO’s sum to a perfect Dirac pulse while minimum phase does not: http://files.computeraudiophile.com/2013/1202/XOWhitePaper.pdf
That, plus time aligning the drivers, and correcting excess phase, appears to produce the most accurate image that I have heard. The research I have done shows that other folks, independent of myself, have come to similar conclusions.
For example, see attached impulse response of another person’s system comparing Audessey (blue) and Acourate (gold). Given the magnitude of difference between the two impulse responses, how can there not be an audible difference?
I asked for this persons “subjective” opinion that went along with his measure: “It is just incredibly good. The clarity is remarkable with such crystal clear separation between the various instruments. Ultimately decent recordings are just so much more musical than before and by that I mean enjoyable to listen to. I don't know exactly which changes contribute to that quality but I like anyway. The only downside is that some recordings are exposed badly in comparison, quite annoying really!” from: http://www.hometheatershack.com/forums/rew-forum/73135-measuring-results-convolution.html#post683849
Another set of objective measures and subjective opinion: http://digitalroomcorrection.hk/http___www.digitalroomcorrection.hk_/Welcome.html
While I have not validated this research, a compendium of information that seems to agree with what others have measured: http://www.linkwitzlab.com/Attributes_Of_Linear_Phase_Loudspeakers.pdf
Again, all saying similar things. So are we imagining this listening experience?
That’s why I think it would be interesting to see anyone else’s measures and impressions of linear phase versus minimum phase, from a time domain perspective and seeing if the measured results correlate with listening impressions.
I am not a product shill. I have no vested interest in any of these commercial DRC products or the “open source” free DRC software I linked to earlier. Nor do I have any products to sell.
I worked 10 years as a professional recording/mixing engineer and have worked in some very well acoustically designed recording facilities. If anyone spends more than 2 seconds reading any of the articles I have posted here will see that I use objective measures backed up with proper ABX testing. I ain’t no audiofool. I am looking at science to assist me in achieving the most accurate playback system I can achieve with the knowledge and tools I have at my disposal, which has led me here.
My goal is to share what I have measured and observed with others and learn from others. So, if one has the capability to design and implement both linear phase and minimum phase filters, with driver time alignment and excess phase correction, please post your measurements and listening impressions and we will see if there is enough of a correlation to come to some conclusion.
Thanks, Mitch
Attachments
Hello,
Mitchba :
" That, plus time aligning the drivers, and correcting excess phase, appears to produce the most accurate image that I have heard."
I agree.
Curiosly, abx comparing linéar phase vs minimum phase with convpare / rephase ect does not work well with musical program on headphones, but sound stage definition is very clearly improved with near/ mid field speakers test.
crd
Mitchba :
" That, plus time aligning the drivers, and correcting excess phase, appears to produce the most accurate image that I have heard."
I agree.
Curiosly, abx comparing linéar phase vs minimum phase with convpare / rephase ect does not work well with musical program on headphones, but sound stage definition is very clearly improved with near/ mid field speakers test.
crd
An externally hosted image should be here but it was not working when we last tested it.
Mitch,
if you talk about FIR linear phase vs IIR/FIR minimum phase you have to take some more things into account:
Minimum phase filters can't be very steep, or group delays goes up and WILL get audible. Because of this you'll get interferences over a broad frequency band. Also, the lower band driver is delayed, thus the main lobe is lifted upwards.
These changes are audible, while the audibility of phase distortion is questionable.
if you talk about FIR linear phase vs IIR/FIR minimum phase you have to take some more things into account:
Minimum phase filters can't be very steep, or group delays goes up and WILL get audible. Because of this you'll get interferences over a broad frequency band. Also, the lower band driver is delayed, thus the main lobe is lifted upwards.
These changes are audible, while the audibility of phase distortion is questionable.
Griesinger is talking about intra-aural: http://www.davidgriesinger.com/Acoustics_Today/intro1_convolved_2.mp3
But inter-aural is equaly important for localization.
Yes, I remember that paper and I think those results are in his book as well. Seems like all the scientific studies say that there is only a small effect at best, but all the non-blind subjective ones say otherwise. Nothing new there.
That's what I thought. So where does your inter-aural claim come from? I don't doubt it, just that you didn't say. Actually it was Griesinger's intra-aural discussion that got my attention since I had understood most imaging to come from inter-aural effects. Phase linearity would not be a factor in inter-aural as long as the two speaker were well matched. But in the above paper David shows how with intra-aural effects phase can be a factor. It was quite revolutionary in my mind.
Markus - Thank you very much!! We can always count on you to have the references if they exist. If you read the conclusions you will see that they are in stark contrast to the subjective claims of audibility made here. I'll go with the scientific studies and stick with IIR filters.
Just a hypothesis based on localization cues available to our hearing.
Have there been studies about audibility of inter-aural phase distortions when signals are presented over a standard stereo triangle?
I always wanted to test this with my Realiser but haven't had the time yet. The Realiser allows to create a virtual multichannel speaker setup from one single real speaker. It would be easy to introduce phase distortion just to one virtual speaker.
A perfect 2-way with minimum-phase crossover at 1kHz will have phase shifts similar to the red (24dB/oct) or green (48dB/oct) curves I posted page 29.
This can be "solved" with simple (ie short) FIR filter in front of an existing loudspeaker, and you would get the blue curve (and further linearization with longer FIR filters could also let you linearize the HP filter of the reflexbox, but that would of course led to preringing, which audibility is just as much discussed as phase distortion itself
)The goal (at least the one I have with rePhase) is to correct the source (the speaker itself) not the system (speaker+room). With phase linearization the source's impulse is tighter, and that is the point I wanted to make. If you only correct the source (based on anechoic measurement and theoretical data) the way it will interact with the room (reflexions) will stay unchanged. This kind of correction is not location specific.
If anything, having a "tighter" impulse (almost dirac-alike) from the source will make measurement analysis (manual or automated...) easier.
Last edited:
But to me this is precisely what a good loudspeaker should do - mine do. I design for a compact impulse response that yields a flat frequency response across a controlled (but fairly narrow) range of speaker angles. My point is and always has been that a good loudspeaker design will not benefit from "room EQ" above Schroeder. Poor designs certainly will.
PS. The correction is only "not location specific" if the system is CD.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.