Comb Filter Effects

[DonH56]

One point to make is that the timing difference between the original and its delayed copy will result in higher comb filtering frequencies which are less harmful since the ERB's are wider. So comb filtering has a time critical window.

Let us take two examples, 1ms and 10ms. At some point, the delay will result in a 180 degree phase shift, which will produce phase cancellation. A 180 phase shift is half a cycle, therefore a full cycle (the period) is at 2ms and 20ms respectively. Because frequency = 1/t, the frequencies affected can be calculated:

• For a 1ms delay, the lowest cancellation point is 1/0.002 = 500Hz. Additional cancellation points will be at 1.5kHz, 2.5kHz, 3.5kHz, etc.
• For a 10ms delay, the lowest cancellation point is 1/0.02 = 50Hz. Additional cancellation points will be at 150Hz, 250Hz, 350Hz, etc.

Thus, a shorter delay will result in comb filtering occurring at high frequencies. As can be seen, a delay even shorter than 1ms may result in comb filtering occurring at mostly inaudible frequencies, for example a 0.1ms delay results in cancellation at 5kHz, 15kHz, 25kHz, 35kHz, etc.

On the other hand, delays of 20ms and above are outside the Haas fusion zone and are perceived as a separate event by the brain, so the delay is heard as space or ambience rather than a change in tonal quality.

@312elements re: diffusers, suggest you read chapter 7.3.2. of Toole "Engineered surfaces and other sound scattering/diffusing devices". I'll quote:

"Some research indicates that the perceptual sum of several mini-reflections is equivalent to a larger single reflection [references given]. This means that the ETC does not convey reliable information about audibility of reflections - replacing a high spike with a collection of smaller ones may sound less different than is visually implied".

^^^ Good stuff; I intentionally stuck with phase and frequency response as I was afraid introducing time response would clutter the basic presentation. That even though my original interest in all this was due to playing with the domain response of my Maggies, which led to investigating why the frequency response was poor, which led to learning about comb filter effects. That was circa 1979 and the first version of this article appeared around 1980~1981, when I used a Fortran program to generate the plots, and the Web wasn't really a thing. Thankfully we've evolved...

 

[bmc0]

One point to make is that the timing difference between the original and its delayed copy will result in higher comb filtering frequencies which are less harmful since the ERB's are wider.

ERB is wider in Hz at higher frequencies but narrower in octaves, so shorter delays cause wider notches if expressed in ERB.

 

[312elements]

One point to make is that the timing difference between the original and its delayed copy will result in higher comb filtering frequencies which are less harmful since the ERB's are wider. So comb filtering has a time critical window.

Let us take two examples, 1ms and 10ms. At some point, the delay will result in a 180 degree phase shift, which will produce phase cancellation. A 180 phase shift is half a cycle, therefore a full cycle (the period) is at 2ms and 20ms respectively. Because frequency = 1/t, the frequencies affected can be calculated:

• For a 1ms delay, the lowest cancellation point is 1/0.002 = 500Hz. Additional cancellation points will be at 1.5kHz, 2.5kHz, 3.5kHz, etc.
• For a 10ms delay, the lowest cancellation point is 1/0.02 = 50Hz. Additional cancellation points will be at 150Hz, 250Hz, 350Hz, etc.

Thus, a shorter delay will result in comb filtering occurring at high frequencies. As can be seen, a delay even shorter than 1ms may result in comb filtering occurring at mostly inaudible frequencies, for example a 0.1ms delay results in cancellation at 5kHz, 15kHz, 25kHz, 35kHz, etc.

On the other hand, delays of 20ms and above are outside the Haas fusion zone and are perceived as a separate event by the brain, so the delay is heard as space or ambience rather than a change in tonal quality.

@312elements re: diffusers, suggest you read chapter 7.3.2. of Toole "Engineered surfaces and other sound scattering/diffusing devices". I'll quote:

"Some research indicates that the perceptual sum of several mini-reflections is equivalent to a larger single reflection [references given]. This means that the ETC does not convey reliable information about audibility of reflections - replacing a high spike with a collection of smaller ones may sound less different than is visually implied".

Keith I appreciate your sharing. I tend do avoid geeking out too much on the technical and let my ears do the decision making. That’s not me trying to imply that measurements and the science here doesn’t matter and that we should all be spending money frivolously on boutique cables and DACs, only that I know that I enjoy my room more with diffusion than without.

That said, am I understanding your quote here correctly? Is it implying that the perception of the comb filtering may in fact be worse than it appears in the sweeps that I’ve run? That’s certainly interesting but would in no way influence my appreciation for what the modified room does for the sound.

As we’ve all likely learned over the years, sound reproduction, especially when it comes to designing speakers is a game of compromises and things like proper engineering, measurements and DSP have solved a lot of problems and advanced the science here, but there are always compromises to me made, especially once you get your gear into your room and have to deal with “room issues” .

The problem I was trying to solve in my asymmetrical room was the centering of the phantom image and recreating a balanced soundstage. The speaker to the right of my chair is 3’ from a wall boundary and the speaker to the left of my chair was about 6’ from a cluttered bookshelf. The speakers themselves are about 6’ apart. The reinforcement from the wall boundary on the right was not just skewing the center image but also made the soundstage feel wider on the left and narrow on the right. My first attempt involved absorbing first reflections on both sides. I employed an absorber on a stand (to the left) equidistant to an absorber on the wall on the right. This did help to mitigate the balance problem but created a somewhat sterile and boring experience. I had some scrap MDF laying around so made a quick diffuser to handle the first reflection and preferred it to the absorption solution. Then room treatments kind of evolved from that until I reached my current solution of employing multiple diffusers throughout the room and turning my drop ceiling into a giant bass trap. I found what works for me in my unique scenario. Any negatives in the form of measurement anomalies or otherwise pale in comparison to what the wall reinforcing the sound was doing to my listening experience.

 

[Keith_W]

ERB is wider in Hz at higher frequencies but narrower in octaves, so shorter delays cause wider notches if expressed in ERB.
View attachment 388375

Thank you for that insight. I presume that green = 0.1ms, blue = 1ms, and purple = 10ms?

Not that I don't believe you, but I will have to simulate that myself to see.

That said, am I understanding your quote here correctly? Is it implying that the perception of the comb filtering may in fact be worse than it appears in the sweeps that I’ve run? That’s certainly interesting but would in no way influence my appreciation for what the modified room does for the sound.

Not really, Toole said the opposite. He said that adding a diffuser might improve the appearance of the ETC, but the audible differences are slight. If I have misinterpreted him I am sure someone on ASR would take great delight in straightening me out

If you were to deliberately introduce comb filtering into your music signal (via a VST) and listen with headphones, you will easily hear the consequences. It sounds oddly hollow and has a weird metallic sheen. Yet, comb filtering is normal in listening rooms. I don't think my listening room sounds hollow or metallic. There is a dramatic difference between comb filtering embedded in the signal, and acoustic comb filtering which reaches your ears! Why is there such a difference?

If you were to ask Ethan Winer, he would say that our brain plasticity has adapted to the sound of the comb filtering of listening rooms, and very few of us have experienced lack of comb filtering in a proper acoustically treated control room. He says as much in one of his articles. Now I do respect Ethan, but he does have a vested interest in promoting acoustic products (he makes and sells them) and he is directly contradicting Toole, who straight out says that room comb filtering is not as audible as the measurements would indicate (again chapter 7). I am prepared to entertain Ethan's proposal, but it is always unwise to dismiss Toole.

Unfortunately, Toole does not provide a good explanation as to why room comb filtering is less audible. I have my own theory of room comb filtering. Be warned, this is purely my conjecture and it is unsupported by any science I am aware of! I suspect that it is because comb filtering is position dependent. A microphone measures comb filtering at one point in space, but our heads move around in a soundfield. Thus we are exposed to hundreds of little comb filters in the upper frequencies which continuously change as we move our heads - in other words, it is uncorrelated. On the other hand, comb filtering which is embedded in the signal remains correlated at every position in the room. I think this is why our brains latch on to it.

I don't know. I need to spend more time thinking about it. @DonH56 would love to hear your thoughts.

 

[312elements]

Thank you for that insight. I presume that green = 0.1ms, blue = 1ms, and purple = 10ms?

Not that I don't believe you, but I will have to simulate that myself to see.



Not really, Toole said the opposite. He said that adding a diffuser might improve the appearance of the ETC, but the audible differences are slight. If I have misinterpreted him I am sure someone on ASR would take great delight in straightening me out

If you were to deliberately introduce comb filtering into your music signal (via a VST) and listen with headphones, you will easily hear the consequences. It sounds oddly hollow and has a weird metallic sheen. Yet, comb filtering is normal in listening rooms. I don't think my listening room sounds hollow or metallic. There is a dramatic difference between comb filtering embedded in the signal, and acoustic comb filtering which reaches your ears! Why is there such a difference?

If you were to ask Ethan Winer, he would say that our brain plasticity has adapted to the sound of the comb filtering of listening rooms, and very few of us have experienced lack of comb filtering in a proper acoustically treated control room. He says as much in one of his articles. Now I do respect Ethan, but he does have a vested interest in promoting acoustic products (he makes and sells them) and he is directly contradicting Toole, who straight out says that room comb filtering is not as audible as the measurements would indicate (again chapter 7). I am prepared to entertain Ethan's proposal, but it is always unwise to dismiss Toole.

Unfortunately, Toole does not provide a good explanation as to why room comb filtering is less audible. I have my own theory of room comb filtering. Be warned, this is purely my conjecture and it is unsupported by any science I am aware of! I suspect that it is because comb filtering is position dependent. A microphone measures comb filtering at one point in space, but our heads move around in a soundfield. Thus we are exposed to hundreds of little comb filters in the upper frequencies which continuously change as we move our heads - in other words, it is uncorrelated. On the other hand, comb filtering which is embedded in the signal remains correlated at every position in the room. I think this is why our brains latch on to it.

I don't know. I need to spend more time thinking about it. @DonH56 would love to hear your thoughts.

I appreciate your taking the time to respond in detail. I read the quote a couple of times and then gave up and pasted it into perplexity.ai and asked it to explain it to me like I'm 5. It appears that perplexity didn't understand the quote either. Glad to hear that that there may not be as much "there" there as the measurements would suggest. Either way, I'm still a happy camper.

 

[DonH56]

Thank you for that insight. I presume that green = 0.1ms, blue = 1ms, and purple = 10ms?

Not that I don't believe you, but I will have to simulate that myself to see.



Not really, Toole said the opposite. He said that adding a diffuser might improve the appearance of the ETC, but the audible differences are slight. If I have misinterpreted him I am sure someone on ASR would take great delight in straightening me out

If you were to deliberately introduce comb filtering into your music signal (via a VST) and listen with headphones, you will easily hear the consequences. It sounds oddly hollow and has a weird metallic sheen. Yet, comb filtering is normal in listening rooms. I don't think my listening room sounds hollow or metallic. There is a dramatic difference between comb filtering embedded in the signal, and acoustic comb filtering which reaches your ears! Why is there such a difference?

If you were to ask Ethan Winer, he would say that our brain plasticity has adapted to the sound of the comb filtering of listening rooms, and very few of us have experienced lack of comb filtering in a proper acoustically treated control room. He says as much in one of his articles. Now I do respect Ethan, but he does have a vested interest in promoting acoustic products (he makes and sells them) and he is directly contradicting Toole, who straight out says that room comb filtering is not as audible as the measurements would indicate (again chapter 7). I am prepared to entertain Ethan's proposal, but it is always unwise to dismiss Toole.

Unfortunately, Toole does not provide a good explanation as to why room comb filtering is less audible. I have my own theory of room comb filtering. Be warned, this is purely my conjecture and it is unsupported by any science I am aware of! I suspect that it is because comb filtering is position dependent. A microphone measures comb filtering at one point in space, but our heads move around in a soundfield. Thus we are exposed to hundreds of little comb filters in the upper frequencies which continuously change as we move our heads - in other words, it is uncorrelated. On the other hand, comb filtering which is embedded in the signal remains correlated at every position in the room. I think this is why our brains latch on to it.

I don't know. I need to spend more time thinking about it. @DonH56 would love to hear your thoughts.

I did a quick search but could not find Floyd's site; my old bookmark is broken, and the Harman site has been substantially reworked. Take the following comments as my speculation (though backed by some research and experimental data I have no intention of trying to find again).

My thoughts on comb filtering are that it is far less important as frequency increases and the accoutrements of a normal room like furniture, pictures on the walls, lights, etc. are enough to largely mitigate it. A wavelength at 3 kHz is only about 4.5" so there are plenty of surfaces in a typical room to diffuse (and/or absorb) those (and higher-frequency) reflections, and speakers are usually farther than that from adjacent walls, so SBIR is less an issue. That makes comb filter effects (amplitude variations) much smaller (nulls much shallower) than at lower frequencies. The ear/brain tends to be an integrating device so a lot of closely-spaced peaks and dips that are not too large are probably more readily ignored -- masking? Psychoacoustics and such is not my area of expertise by any means!

At 1 kHz a wavelength is about 13.5", doubles to 27" at 500 Hz, so the lower midrange and upper bass is more affected since there are fewer surfaces large enough to provide adequate absorption or diffusion, and that is where the fundamentals of many instruments (and voices) lie. That is the range (say 100~1000 Hz) that exhibited the greatest comb filter effects to my ears and drove me to damp the back wave of my Maggies. Sonically I could hear the amplitude changing with frequency, which could mess up the image as I moved around, but also caused strange amplitude modulation up and down the scale even if I sat still. If everything was not perfectly symmetric in the environment and my listening position, the image would wander side-to-side as a scale was played through the speakers, so a centered solo player/signer would suddenly dart to the side then back passing through notes. I really hated that.

I do not want to speak for Ethan but my memory is that he was of generally the same opinion. IIRC he sold (sells) panels with or without membranes to increase HF reflections (?) I may be thinking of another company... I think I spoke with Floyd about it way back when but do not remember any details; I was focused on optimizing my system at the time, and my room was pretty dead (many absorbers -- Ethan would've loved it!) so reflections were not a real concern.

FWIWFM - Don

 

[Tim Link]

There's no doubt the room qualities have a profound effect on our perception of the sound. But that doesn't mean it confuses our ear about the timbre of the instrument or reproduced instrument or voice we are listening to. We hear through the room, sort of, to properly interpret the timbre of the source, but we still hear the room, and it can be detrimental to our overall experience. If the room's reflections are diffused it changes the room's character in a noticeable way, making it less distinctly colored while still nearly as lively as before. The room effect is profoundly more noticeable when a sound in the room is recorded through a typical microphone than when we're listening there in person. We have two ears, and they have directional detection capabilites that allow us to better separate the direct sound from the reflections. A dummy head stereo mic. can do that too, but not a regular mic.

As to the effectiveness of diffusion, a story I've related before is of a recording studio's tracking room where they were recording a vocalist with two mics. One was up close to the singer, and the other was far away in the corner of the room. The far mic. is to pick up a stronger amount of the room's sound so they can mix that in to taste if desired. As they were tracking a singer, guys in the monitoring room were baffled that they couldn't really notice much of a difference between the close mic. and the room corner mic. Our client who owned the studio thought this was hilarious and gave us a call. We had treated the room with a combination of mid-bass absorption and high frequency diffusion. The result was that the room was largely devoid of its own sonic character.

 

[bmc0]

I presume that green = 0.1ms, blue = 1ms, and purple = 10ms?

Yes.

Not that I don't believe you, but I will have to simulate that myself to see.

I'm not sure there's anything to "believe" as it follows directly from the definition of ERB.
I generated a simple test file (three channels, each with impulses at t=0 and t=<delay>, where <delay> is 0.1ms, 1ms, or 10ms) and then loaded it into RoomEQWizard to plot the ERB-smoothed frequency response.