If the speakers are "good" anechoic (FR, directivity, etc) then room EQ should not be much more than knocking down the peaks below Schroeder at the LP. At least that is a good and safe place to start.
-
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!
Does Phase Distortion/Shift Matter in Audio? (no*)
- Thread starter amirm
- Start date
Thanks, and yep. That's all I've done for many years, electrically.
Otherwise, my 'room tuning' has just been speaker placements, furnishings, a few treatments, etc.......iow, acoustic solutions for acoustic issues.
I did purchase Dirac Live recently to see how it works/what it does.
Ran the full-range stereo setup routine a few times till I got comfortable with it. Never really heard much if any change.
Then i made electrical transfer functions of the processing it had come up with....this is what I was really interested in, much more than what it might do soundwise.
By electrical, I mean the line-level processing, not the speakers' acoustic outputs.
Appeared to leave the left side pretty much alone both mag and phase. And use some maximum phase correction on the right side's low-frequency/sub output, with very little mag correction.
Dirac's way to EQ my particular rig below Schroeder, at LP, I guess.
Anyway, turned it off....maybe to revisit again someday, after I learn a bit more.
Last edited:
- Joined
- May 24, 2016
- Messages
- 644
- Likes
- 2,414
Have you looked at @j_j 's presentation (download the PPT) on Acoustic and Psychoacoustic Issues in Room Correction.
As far as how, one (bigger than I thought) issue is DSP/DRC tool selection. There are so many products on the market, many with little technical explanation on how they work. In my testing of over a dozen of these DSP/DRC tools, most produced OK results with a few going backwards and taking the life out of the bass dynamics. I was surprised by the range of variability as to the quality of the room analysis algorithms used and the filters these tools design and generated. Out of the dozen or so products I tested (with more yet to be tested) only four produced the results that met my design criteria (as explained in the video linked below).
As an audio DSP programmer from the music production world, I come at this a bit differently. Just like how one can model electronics, guitar cabinets, etc., in DSP, one can do the same for a loudspeaker in a room (i.e. frequency and phase, not directivity). In this video, I explain what to correct, what not to correct, per j_j's presentation. Concepts of breaking up the room into acoustic zones and how room correction applied in each zone is different and why. Concepts around frequency dependent windowing and why it is so important in DRC. Think of how the Klippel scanner works; windows out low frequency reflections to get the direct sound, generally (good) DRC DSP does the opposite and windows out mid to high frequency reflections for a semi-anechoic response as we are mainly focused on the low frequency response of the room and loudspeaker).
This is what we need more of. Listening examples. In this video, I compare three DRC FIR filters that have been designed for a stereo triamp system:
- Minimum phase room correction with minimum phase digital crossovers
- Minimum phase room correction with linear phase digital crossovers and driver time alignment
- Minimum phase plus non-minimum phase room correction with linear phase digital crossovers and driver time alignment
I wonder how many folks hear the same as I do?
Unfortunately, there is no easy way to capture the 3D space of the speakers in the room. I do have binaural mics and made speaker comparison recordings before, but I did not have time to set this up and capture with the binaural mics. I am also looking for a better way... suggestions welcome. So one is only getting the (summed) convolved with music digital output with no acoustic space. Even though the differences are audible, it does not convey the in the room sound field differences, subjectively going from an opaque two dimensional sound (i.e. Filterset1) to a well defined 3 dimensional sound field with a solid phantom center image and clear/dynamic sounding bass that sits in the pocket. I.e. Filterset 3.
Anyone have other listening examples?
Last edited:
Thanks for a great post and great links. Do you have a preference for one of these combinations? Do other people that got to listen to this in the room have a consistent preference?
312elements
Active Member
Hey Mitch, thanks for posting the link. I actually got a lot more out of your other video, “
Understanding the State of the Art of Digital Room Correction”
For what it’s worth, and I doubt that I’m alone in this, but I really appreciated the step by step process and the suggestions you made regarding rules for what to do and how to do it. Like most, I assume, I’m far more concerned with getting the most out of my system and enjoying the music than understanding the theory behind it. I think for the majority of us, we learn a little bit of theory as we go and it becomes cumulative knowledge. Most tutorials I’ve watched and a lot of the experts on the internet seem to be more interested in flexing their knowledge than actually helping people. A suggestion, if I may be so bold… should you ever create another video on another software program, create a 15 minute version featuring your process with whatever the latest and greatest software tool is. Maybe something like HLC? Forget about teaching anyone anything other than your process in said video. It’s not that the excess information isn’t important, it’s just not the right time to overwhelm people with theory. I think that there are a lot of people that would invest their time and money in better room correction if they didn’t have the obstacle of knowledge placed between their speakers and better sound. Again, the why is important but less important than the how. At least when we’re talking about those of us who do this as a hobby. Once we understand the how, we can always come to forums like this and ask why. Thanks again for the great video. It was enough to make me think this is worth trying to figure out.
Last edited:
Hi Mitch, thanks for the reply. I think you are probably the most experienced person I know of, in terms of one who works with room corrections.
Yes, I'm familiar with the j.j. presentation you linked. I just reread it, and other than some of the details in the second Sequence of Operations section, it's stuff I've absorbed pretty well I think. Breaking the room into zones makes easy sense. So does frequency dependent windowing/solutions...certainly if one is trying to use global type corrections.
Since I avoid global corrections like the covid, never had to worry about that one Lol
What I'm really looking for, is not the best DSP/DRC tool out there, not in terms of its automated capability.
I look for the best tools in terms of manual capability that let me explore and make roll-my-own tunings/corrections.
If anything, I'm after the best indoor measurements i can make. Particularly with subs/bass. I know how hard it is for FFT/IFT to give good repeatable low freq data....I have a hard time believing the legitimacy of most low freq measurement post i see. Anyway, gonna read Sequence of Operations again.
Biggest issue, when I look at myself as honestly as i can, I realize I just need to get serious about even wanting room tuning.
I'm am about as far away from a seated sweet-spot listener as one can get, my speakers on subs are all on castor wheels, and they are constantly changing as I DIY experiment....and.....i still hear the most glorious sound I've ever heard from outdoor setups ! Hard to get the room thingy motivation going.
But I digress badly...this is a thread about phase.
As to locating the locations of speakers, there's a paper in the AES convention record by "Johnston and Fezjo" (not sure which order), maybe with some other authors, too. It's been a while, but it is possible to capture such information and get proper timings pretty easily. Even the tiny room correction in Windows 7, 10, and 11 (dig deep deep deep into the audio tools in the driver level, you'll find it) will do an extraordinarily good job of time EQ, btw. The interface, last time I tried it, was busted. You need stop and start the audio server in order to sort that, sometimes. If it refuses to get a stable solution, that's the problem.
Yes, Serge Smirnov and I wrote that one.
Yes, Serge Smirnov and I wrote that one.
One metric I've been wondering if phase correction of a speaker and room combo can improve on is measured clarity. I've gotten mixed results with my attempts so far. Nothing outstanding by any measure, so my general conclusion is DSP can't do much for two speaker system clarity.
Here's a paper that suggests that room EQ on two speaker systems, including correcting for excess phase can work at close listening distances, like 1 meter, to preserve low frequency information, not so well at further listening distances. Makes sense that the problems becomes increasingly complex as you move away from the speakers.
They jump straight from 1 meter to 4 meters, so there maybe varying degrees of success in between.
Art Noxon has been promoting the MATT signal for years. It's a modulated signal, as they are using in this paper. It's a listenable test. If you can hear each pulse at all frequencies clearly delineated in a large room at a few meters listening distance, without any weird sounding blurry areas, that's really impressive. I've been looking at MATT recordings in various people's rooms for a few years now. The rooms I see that are nearly devoid of problem areas are always nearfield, well treated, and sometimes use DSP too. Never larger rooms at further listening distances. There are decent larger rooms when they've put a lot of effort in to acoustics. I have also seen some surround systems that use room corretion involving all the speakers get marginally higher clarity scores than when in pure 2 speaker mode.
If we're not looking for improved clarity with the phase manipulation, what else might we expect to hear? Naturalness, even if it isn't clear?
Here's a paper that suggests that room EQ on two speaker systems, including correcting for excess phase can work at close listening distances, like 1 meter, to preserve low frequency information, not so well at further listening distances. Makes sense that the problems becomes increasingly complex as you move away from the speakers.
They jump straight from 1 meter to 4 meters, so there maybe varying degrees of success in between.
Art Noxon has been promoting the MATT signal for years. It's a modulated signal, as they are using in this paper. It's a listenable test. If you can hear each pulse at all frequencies clearly delineated in a large room at a few meters listening distance, without any weird sounding blurry areas, that's really impressive. I've been looking at MATT recordings in various people's rooms for a few years now. The rooms I see that are nearly devoid of problem areas are always nearfield, well treated, and sometimes use DSP too. Never larger rooms at further listening distances. There are decent larger rooms when they've put a lot of effort in to acoustics. I have also seen some surround systems that use room corretion involving all the speakers get marginally higher clarity scores than when in pure 2 speaker mode.
If we're not looking for improved clarity with the phase manipulation, what else might we expect to hear? Naturalness, even if it isn't clear?
restorer-john
Grand Contributor
Yamaha engineers achieved this in the early-mid 80s. (DSP-1 release 1985).
I used to have a lot of information from Yamaha on the array and data capture technology, but it's gone now (somewhere in my storage).
All I can find now is a picture of one of the later 4 microphone arrays used for later model venue data capture. (circa 1990)
The soundfield data of the room could be captured and then the soundfield data of the room+speakers (using the same impulses) and the difference would become the soundfield of the speakers.
Andysu
Major Contributor
- Joined
- Dec 7, 2019
- Messages
- 3,063
- Likes
- 1,600
yamaha dsp 100 , i have two of them was going to be used for top secret project , but anyway and also have another idea use , not what thinking but more less the same but its complex to set-up and due to health may never get done , but they make great use for
I have been rolling my own room DSP using just REW and MConvolverEZ with Audio Hijack on my Mac. I watched your video and have done my best to get follow your guidelines as well as I can. I was not getting any interesting success at first. This was because I didn't reallize that the convolver I was using at the start only worked in mono, so when I fed it a stereo impulse it just ignored one of the channels.
Now that I've got stereo convolution working and I've largely matched the phase and frequency response of both speakers at the listening position for the bass and midbass/lower midrange, I can say it's quite an impressive improvement! I've also corrected for some gnarly late arrivals/early cancels that were causing peak level to jump to 20ms delay in the left channel between about 100 and 200. The price of that is some pre-echo and more overall delay. I'm up to about 40ms total delay with that fix. I'd say it's totally worth it. Not only is the imaging greatly improved, the clarity is better.
A little pre echo to reduce a nasty post echo is a big net gain in my opinion.
I've also tried pushing the phase to linear just to hear for myself what the problems are. I don't hear anything that bothers me. It sounds really tight and clear. Maybe with more listening I'll detect some ramp up effects before big impacts like bass drum hits. I've been listening to those kinds of things and I don't hear it yet. I can make a filter to target minimum phase but I'm in no hurry.
So in a nutshell, this room correction technology has real audible goodness potential beyond simply taking down bass peaks. The improvement in imaging is not subtle if your room needs some help in that regard. I had no idea the imaging could be improved so much by matching those lower frequency areas.
Keith_W
Major Contributor
Hi @j_j I have a question for you. Since my initial confusion earlier in this thread about Amir asserting that phase is inaudible, I have had some time to do some reading and experimenting to clear up my own understanding about the subject.
First question, is this statement correct: "Phase differences between left and right speakers are audible because it affects the ITD. We are not sensitive to phase changes that occur symmetrically between left and right channels".
Second question, is there a phase equivalent of ERB's where our ears are more sensitive to interaural phase differences at certain frequency bands? I am guessing the answer should be "yes". But if so - I am also guessing it would be the opposite of ERB. ERB's show we are more discerning of volume differences at lower frequencies, but my guess is that we are more sensitive to phase differences at higher frequencies. The reason I suspect this is because we use higher frequencies for sound localization. I am aware of the other factors that affect localization - the direction dependent frequency response modification done by our ears, pinnas, etc., integration with the vestibular system, proprioception from the neck, and ILD.
First question, is this statement correct: "Phase differences between left and right speakers are audible because it affects the ITD. We are not sensitive to phase changes that occur symmetrically between left and right channels".
Second question, is there a phase equivalent of ERB's where our ears are more sensitive to interaural phase differences at certain frequency bands? I am guessing the answer should be "yes". But if so - I am also guessing it would be the opposite of ERB. ERB's show we are more discerning of volume differences at lower frequencies, but my guess is that we are more sensitive to phase differences at higher frequencies. The reason I suspect this is because we use higher frequencies for sound localization. I am aware of the other factors that affect localization - the direction dependent frequency response modification done by our ears, pinnas, etc., integration with the vestibular system, proprioception from the neck, and ILD.
Well, with the right input, about 15 degrees change inside an ERB can be detected (this after subtracting the constant-delay part, which is important to understand) even with mono signals.
Up to about 1000Hz phase difference between channels is extremely detectable as direct imaging components. That starts to get a bit lesser above 500Yz, and that sensitivity is gone by about 2000Hz or so. For signals with flat envelopes above about 2000, and surely above 4kHz, phase does not matter.
For signals with sharp onsets, however, the same 10 to 20 microsecond time offsets, due to phase or anything else, *IS* audible. No sharp onset, no sensitivity. Pulsed bandpassed sequences, then audible, sometimes annoyingly so, but not always understood as actual imaging. Effects are audible in both speakers and headphones, but sound totally different. I'd be tempted to say that using speakers might be more sensitive, but that's not something I have hard numbers on.
Keith_W
Major Contributor
OK. If I am reading you correctly, achieving phase symmetry between left and right is more important between 500Hz - 2kHz.
The reason I ask is because I am experimenting with a reverse all pass filter to correct asymmetric phase with my speakers. I was wondering where I should apply it, since a global reverse AP filter is almost impossible to design. Your insight has been very helpful, thank you!
How about "snare drum".
No, it's MOST important up to 2kHz. DC to 2kHz. Above that it's less important but still matters. How much depends on the audio content.
Edited to add: NEAR DC. Of course there is no "phase" at DC.
Last edited:
@j_j
Do you consider this older report current with your understanding of things?
" THE ROLE OF THE PINNA IN HUMAN LOCALISATION"
Was described as "Here is a great article that describes how the human ear's pinna can discern phase shifts in the single microsecond range (giving us ability to locate direction with only one ear). Work funded by US military in 1964 and performed by Tufts Univ. This implies we hear amplitude to 20khz but phase to 300kHz. Hence, amps should be designed to be phase flat out to 300kHz for best stereo image and spatial soundstage." by another fella @ diyaudio.com
Do you consider this older report current with your understanding of things?
" THE ROLE OF THE PINNA IN HUMAN LOCALISATION"
Was described as "Here is a great article that describes how the human ear's pinna can discern phase shifts in the single microsecond range (giving us ability to locate direction with only one ear). Work funded by US military in 1964 and performed by Tufts Univ. This implies we hear amplitude to 20khz but phase to 300kHz. Hence, amps should be designed to be phase flat out to 300kHz for best stereo image and spatial soundstage." by another fella @ diyaudio.com
Well, except it in no way says we can "hear phase to 300kHz", how well one can hear phase is a function of both sampling rate AND SNR. I even have a talk somewhere on this.
Similar threads
