Comb Filtering : Please engineer / smarty-pants ... learn me sumtin

Really good around 9m45s? Dude, it was ROCKING!
Gosh, this morning I didn't even know what comb filtering was, and now I'm all ready to follow the, er, band.

This is my recording of the same, ^, from around that same 9m45s mark.
sample of mine: https://soundcloud.com/user-271082364/live-phase-cancellation-comb-filtering
Lots of peripheral crap in this one - sorry.
Scene: Art Museum, abstract art exhibit, and this performance. The audience was quiet and respectful; but the kids in the day care next door, and the street noise,.... [rolls eyes smilie here][slaps forehead smilie here]
Hey,.. At least it gets you an idea of what went down.
 
This is my recording of the same, ^, from around that same 9m45s mark.
sample of mine: https://soundcloud.com/user-271082364/live-phase-cancellation-comb-filtering
Lots of peripheral crap in this one - sorry.
Scene: Art Museum, abstract art exhibit, and this performance. The audience was quiet and respectful; but the kids in the day care next door, and the street noise,.... [rolls eyes smilie here][slaps forehead smilie here]
Hey,.. At least it gets you an idea of what went down.

Only got 3:10 of it...then went to your organ recording.
 
That is all of the sample @ 3:10
I broke the entire 16 minutes (or whatever length it was) into separate tracks whenever they changed the oscillator tone.
It was such a noisy location that listening to the whole thing is a bit tedious; city buses, Harleys, kids screaming, traffic signal start up and stops, etc. But, in that 3:10, things get pertty crazy in places.
 
Comb filtering is when two signals arrive at different times causing one signal to partially or totally cancel the other signal. Now if the arrival times are very close together there is little comb filtering. If the arrival times are fairly large the ear will select the first signal and mentally suppress the second, known as the Haas effect. Which case applies to you, I have no clue. You may be somewhere in-between which would be a worst case. Wiring speakers in series is absolutely the worst thing you can do. The damping factor of the amp is totally cancelled leaving your drivers to not only interfere with each other acoustically but electronically, too. Now if speakers are self controlled and don't depend on the amp to keep the driver cones in check you are OK. If not where the speakers depends largely on the amp to control the speaker cones, then wiring your speakers in series totally negates that valuable part of the systems necessary qualities.
 
Comb filtering is when two signals arrive at different times causing one signal to partially or totally cancel the other signal. Now if the arrival times are very close together there is little comb filtering. If the arrival times are fairly large the ear will select the first signal and mentally suppress the second, known as the Haas effect. Which case applies to you, I have no clue. You may be somewhere in-between which would be a worst case. Wiring speakers in series is absolutely the worst thing you can do. The damping factor of the amp is totally cancelled leaving your drivers to not only interfere with each other acoustically but electronically, too. Now if speakers are self controlled and don't depend on the amp to keep the driver cones in check you are OK. If not where the speakers depends largely on the amp to control the speaker cones, then wiring your speakers in series totally negates that valuable part of the systems necessary qualities.

I've read more on the series effect and I understand more now. Thanks. This entire exercise can be summed up with this equation : 2+2 = 0 .
 
In general the Haas effect happens with time differentials greater than about 10 milliseconds with a window of about 30 milliseconds. It can result in a differential of around 10 dB.

Because hearing is subjective these numbers are not cut in stone for everyone.

The period of a 5000 Hz tone is 0.2 milliseconds. For phase delay of 180 degrees needed for full cancellation (this would create the notch that leads to comb filtering) the period would half or 0.1 milliseconds.

This would be a locational differential of about 1.35 inches for the sound sources. For a 10 KHz tone the numbers are smaller by half or about 0.67 inches. Remember that these numbers are for maximum cancellation Shorter distances and smaller time differentials will still result some level of cancellation (signal reduction at certain frequencies).

Of course my examples are single frequency, but when you look a the full spectrum, these cancellations based on the differential phase (time of arrival) of the individual frequencies of the spectrum generate the typical comb filter response as previously mentioned.

Here is a generic example.

upload_2017-11-10_19-47-53.png


BTW the waw, waw, waw that is heard in some of the videos in not necessarily due comb filtering. It is what is called beat note, and comes from a difference in frequency of two tones. When the two tones are the exact same frequency it is called "zero beat", that is the waw, waw, waw goes away.

On the other hand comb filtering is the result of differences in the the phase/time of arrival of two instances of the same frequency and results in differences in amplitude not beat notes.
 
This is a screen shot of the .wav form for the Alvin Lucier - James Tenney Homage:
Screen Shot 2017-11-10 at 6.53.19 PM.png

You can see the constancy of the oscillator from the left side moving to right just before 2:45. At just before 2:45, the bass note comes into play, and the constancy of the oscillator shape goes all warbly.
 
Here is a great example of comb filtering. He he uses software to apply the phase shift (relative timing).

The phase shift (timing differential) could come from speaker placement, the use of multiple speakers or technical issues with the amplification chain and at higher frequencies it can come from changes in the position of one's head.

He starts at 1 millisecond. The impact on the spectrum of the drum is shown. You can see the dips in amplitude the result in the comb like appearance of the spectral display.

Note there are no beat notes, the waw, waw, waw sounds.


Here is a still shot showing the comb filter effect.

This example shows the deep nulls that happen to the audio spectrum when the phase shift is close to what is needed for maximum attenuation of the frequencies involved.

Note that this is for 1 millisecond. 1 millisecond equals 1000 Hz. Note the major deep null at 1000 Hz.

upload_2017-11-10_23-35-51.png

Smaller amounts of phase shift, timing differential (less than 180 degrees) will still cause frequency specific reduction, but the amount of reduction will be less. This can still have an overall noticeable impact on sound quality.
 
Here is a good example of the addition of two sine waves (tones) of different frequencies, resulting in beat notes or the waw, waw waw, sound.


This is not the same as comb filtering. In comb filtering the signals are the same frequency, where as these are not.
 
Perfect example of a beat frequency is to play a 32 Hz pedal C on an organ and 34 Hz C#, etc. simutaneously. Sounds like an airplane...., rrr rrr rrr..
 
Is it just sometimes LUCK that determines whether two speakers are going to compliment or comb?

It is likely that there will always be some comb filtering.

With the variables involved, it is difficult to predict the impact an sound quality.

Other things come into play, differences in frequency response, dispersion characteristics, sensitivity and so on. The set up of speakers and the listening environment will play a part.

And in terms of sound quality, let us not forget the listener.
 
It is likely that there will always be some comb filtering.

With the variables involved, it is difficult to predict the impact an sound quality.

Other things come into play, differences in frequency response, dispersion characteristics, sensitivity and so on. The set up of speakers and the listening environment will play a part.

And in terms of sound quality, let us not forget the listener.

Yes, each variable is a worm-hole that can take me down a new path of dissecting causes and effects ... so before I end up putting tape on the nose-piece of my glasses... let me thank you and the others for putting me a step closer to understanding.
However, I have just un-wired my 4-speaker set up, and I am now back to two-speaker stereo ignorant bliss. I am worried that if I continue down this minutiae vortex my currently unbearable (for family) audio hobby will further spin out of control. Sometimes a person must knowingly step into the harness and reigns ... simply out of self awareness.
Thanks!
 
Last edited:
This one screwed with my head for a long time.

I'll start with what I see when I look to the heavens,...
IMG_2751_zpsp7eqrvxx.jpg

That mic array is two mics short of the entire stage front package at this concert, below.
DPA-omni ......................................................... EW-omni -- *array* -- EW-omni .......................................................... DPA-omni
IMG_2721_zpsugv2nkxo.jpg
*Center Array* = Royer Stereo Ribbon and DPA 4028 sub-cardioid stereo pair as ORTF stereo
Wide L/R Flanks are DPA compact series omnis

This concert hall has just been made "famous" to all TV watchers as the concert hall with the insane triangle solosit in the chamber orchestra (Geico ad?). If you watch the ad, you can see the same hoops as the first image. /rambling.

The bassist. That is the biggest bass violin that you'll likely ever see, with a massive and deep voice.

The tracks were all discretly recorded as mono tracks, intended to be mixed later.
But, the arrays are intended as multiple discrete stereo recordings:
DPA-omnis as wide Spaced Omnis
EW-Omnis as 3' Spaced Omnis
Royer Ribbon as Mid-Side stereo
DPA sub-cardioids as ORTF stereo
And, a stereo piar of DPA4028 sub-cards at the rear of the audience, facing the stage.
In total, 5 stereo recordings were done.

As I stated earlier, I am a two channel stereo specialist for 35 years. I'm just recently, in the last couple of years, trying to learn mixing.
Now, none of the stereo patterns mentioned above would ever have issues with overt comb filtering issues, when properly placed, as they are minimalist arrays, with long proven track records. The one exception might be the widest spread omnis, which will often end up with the "hole in the middle", which the center array fills.

In my mixing effort, I tried mixing the widest DPA omni tracks with the center ORTF tracks, which should have been a safe mix. NOT.
The right center mic was focused on the big bass face plate. The left wide flanking omni was focused on the big bass rear plate.
When mixed, that giant voice of a bass, and I'm talking wall shaking voice,... it flat out disappeared.
Listen to the discrete tracks, massive voice. Mixed, bass disappears. This is some strng phasing shit, and its mind boggling.

OK,... lets try flipping polarity of mics.
Yea! the bass is back. But, it is now cleanly placed in the fourth violin spot from the left, right between the last standing violin, and the first seated violin.
WTF?
I tried various polarity flips between the four tracks, and, could never land him in his position in the mix.
The problem, I surmise, arose with the two left mics relating to each other in a + polarity for the face-plate of the bass, and a + polarity for the back-plate of the bass, which caused phase cancellations.

Eventually, I found a combination of things, along with a bit of HPF edit sweeps to the bass tracks to reduce the commonalities that were causing the phase cancellations.
The stereo tracks,.. tracked, split, enjoyed.
 
Last edited:
A number of posts show the combing pattern in the frequency response, but Imaybe this graphic helps show how two sources can interact to produce lobing. I believe is the cause of the comb pattern in the frequency response (relative to this discussion, anyway).

LNFigure2.gif
 
A number of posts show the combing pattern in the frequency response, but Imaybe this graphic helps show how two sources can interact to produce lobing. I believe is the cause of the comb pattern in the frequency response (relative to this discussion, anyway).

View attachment 1043398

Your image reminds me of what isperhaps the polar opposite of what your image is eluding to.
In stereo recording we use two mics to create a ghosting center image; an image that might make a person think that there is a third mic involved, a center mic.
We work with phasing in a positive sense, in that we can utilize the shared commonalities (random phase cancellations) that cancel to center, to create that center image with just the pair of mics. With that shared commonality canceling to center, we capture the center player in their proper position in the playback imagery.
xy.png
 
Your image reminds me of what isperhaps the polar opposite of what your image is eluding to.
In stereo recording we use two mics to create a ghosting center image; an image that might make a person think that there is a third mic involved, a center mic.
We work with phasing in a positive sense, in that we can utilize the shared commonalities (random phase cancellations) that cancel to center, to create that center image with just the pair of mics. With that shared commonality canceling to center, we capture the center player in their proper position in the playback imagery.
View attachment 1043406

So sometimes frequency cancellations are engineered to aid in imaging... and yet in other instances frequency cancellations destroy imaging... Hmm... I'm trying to wrap my tiny brain around this... I need time.
 
A number of posts show the combing pattern in the frequency response, but Imaybe this graphic helps show how two sources can interact to produce lobing. I believe is the cause of the comb pattern in the frequency response (relative to this discussion, anyway).

View attachment 1043398

Isn't it interesting, the two perspectives between recording and playback...? I'm learning a lot.
 
Back
Top Bottom