Understanding Comb Filtering: Causes and Audio Examples

[00:00:00]Comb filtering is the result of two or more  identical signals interacting at a slight delay. Each signal contains many frequencies,  each frequency with a different wavelength.

[00:00:11]When the signals are aligned in time every  frequency lines up perfectly and the two sounds add together to create a louder sound. When  the signals are slightly shifted in time some frequencies add together and some cancel out. The  resulting graph resembles a comb. Hence the name, comb filter. In this video you'll hear for  yourself what a comb filter sounds like, you'll learn the common causes of comb  filtering, and how to eliminate comb filtering. But if you're new to this channel,  my name is Kyle. Welcome to Audio University.

[00:00:44]You've certainly heard the effects  of comb filtering in the past.

[00:00:47]Maybe you didn't notice it or maybe you didn't  know what to call it, but severe comb filtering can ruin the quality of a recording so it's really  important to learn to identify it by listening.

[00:00:59]To learn what comb filtering sounds like, let's  start with a very extreme example. In addition to the fundamental frequency of my voice, my  voice is also made up of many other frequencies.

[00:01:10]So what does comb filtering sound like? If I  copy this recording of my voice to another track, the two copies will work together and the  overall level will increase by six decibels.

[00:01:21]So what does comb filtering sound like? However,  listen to what happens when I delay one copy.

[00:01:31]So what does comb filtering sound like? Due  to the fact that the two copies of the signal are equal in level, this example is very extreme.  The closer in level the two signals are, the more extreme the interference will be. If I turn down  the delayed copy of the signal by six decibels, the comb filter effect becomes much more  subtle. So what does comb filtering sound like?

[00:01:57]As I said before, comb filtering occurs when two  or more identical signals interact at a slight delay. In this section, I want to go through the  four most common scenarios in which comb filtering occurs in the real world. The first cause of comb  filtering is latency or delay caused by digital signal processing. All digital systems add some  bit of latency to the signal and this could be a big problem. Imagine a situation where you want  to add parallel compression to the snare drum.

[00:02:28]In a parallel compression setup, there  are two copies of the original signal.

[00:02:33]The dry snare signal passes through from  input to output with no additional processing.

[00:02:38]The wet snare signal passes from the input to a  compressor and then mixes with the dry signal at the output. Because the wet signal is processed  by the compressor and the dry signal isn't the dry signal could reach the output before  the wet signal, which would result in comb filtering. Fortunately, most modern digital audio  systems have delay compensation which ensures a fixed latency across all audio channels.  In an effort to keep all signals aligned, the system will intentionally delay each signal  so that the latency is constant across the board.

[00:03:16]Another cause of comb filtering is reflections.

[00:03:19]Right now you're mostly hearing the direct  sound of my voice to the microphone.

[00:03:24]However, when I bring a barrier closer there  are now two pathways that my voice can take one directly to the microphone and another from my  voice, to the barrier, to the microphone. Now of course the distance that the reflected path takes  is longer than the distance that the direct sound takes, meaning that there are two copies of my  voice reaching the microphone, one at a slight delay. If you listen closely you can hear that  the frequencies affected most by the comb filter shift depending on the distance away from the  barrier. This problem can occur any time that a sound takes multiple pathways from the source  to our ear or from the source to a microphone.

[00:04:07]For example, the sound from a guitar  amp cabinet takes a direct path from the speaker to the microphone and also takes a  reflected path from the speaker, to the floor, to the microphone. The sound from your speakers  might be perfectly accurate at the source but by the time it reaches your ear, it could be wildly  inaccurate. Not only do you need to consider the sound reflecting off of the walls, but also the  desk, the ceiling, and any other nearby sources.

[00:04:35]To avoid comb filtering from reflections it's  important to remember what I said earlier.

[00:04:40]The closer in level the two copies of the signal  are, the more extreme the interference between those signals will be. One way we can widen  the difference in level between the two signals is to add distance to the reflected path. The  further the reflected sound has to travel, the quieter it will be by the time it reaches  the listener. One of the most effective ways to improve the sound quality of your speaker setup  is to use absorption panels to break up the early reflections from the speakers. To do this,  you'll determine the shortest indirect pathways from each speaker to the listening position  and place absorption panels in those spots.

[00:05:19]Comb filtering can also occur when placing  multiple microphones on a single sound source.

[00:05:25]Right now both of these microphones are  about the same distance away from me, so it takes the sound of my voice about  the same time to reach each microphone.

[00:05:33]However, if I start to move one mic further away,  now the sound of my voice has to travel further to reach that microphone than it does to the closer  mic and so, when these two signals are mixed together in your headphones, there will be some  interference which will result in comb filtering.

[00:05:49]The same thing can happen when you place  multiple microphones on a single instrument.

[00:05:53]Let's take a guitar amp for example.  I'll record a short bit of playing.

[00:06:03]Now let's take a look at the waveforms  from the two mics inside the DAW.

[00:06:09]You can see that the sound from one  mic is a little behind the other mic.

[00:06:16]To fix this, I'll adjust the distance of  the mics until the waveforms are aligned.

[00:06:29]Now you can see that the waveforms from  both mics are more closely aligned.

[00:06:35]Another scenario where this might be a problem  is when you're recording two nearby sound sources with separate microphones. The sound from the  trombone reaches the trombone microphone but it also bleeds into the trumpet microphone, and  vice versa. Of course, it takes longer for the sound from the trombone to reach the trumpet mic  than it takes for it to reach the trombone mic.

[00:06:57]When both of these mics are mixed together in  your final mix, you'll hear some bit of comb filtering. You may have heard of the 3:1 rule  in audio. In this situation, the 3:1 rule would tell us that the distance from the trombone to  the nearest mic (in this case the trumpet mic) should be at least three times the distance  of the trombone to the trombone mic. The idea is that by the time the sound from the trombone  reaches the trumpet mic, it will be so quiet that it won't cause too much interference. Remember -  the closer in level the two signals are, the more extreme the interference will be and therefore  the more extreme the comb filtering will be.

[00:07:38]The last cause of comb filtering I want to  show you is multiple speakers. This one is also extremely common because we almost always  use two or more speakers in a sound system.

[00:07:48]In a mixing studio environment, as long  as the listener is in the right position the sound from the left and right speaker will  reach the listener at the same time. However, what happens when the mixing engineer steps over  to the equipment rack to make some adjustments?

[00:08:02]Now the sound from the left  speaker takes a longer path than the sound from the right speaker. I  want to offer you a speaker placement guide to make sure that you get this right in your  listening space. You can download that for free at audiouniversityonline.com/speaker-placement-guide/.  This is even a bigger problem in live sound reinforcement. You've got a left speaker, a right  speaker, and maybe some front fill speakers at the front of the stage. You can do your best to  align all of these speakers, but you can really only align them at one point in space. Taking  one step to the side will change the distance between the listener and each speaker. The  experience can't be the same for every listener because they're each in a different position.  Unfortunately, the best you can do is try to direct the speakers so that their dispersion  patterns overlap as little as necessary.

[00:08:57]If you're enjoying this video so far,  please do me a favor by hitting the "Like" button to help the video reach more  viewers. I really appreciate your support.

[00:09:06]Before you go, I want to help you better  understand what exactly happens within a comb filter. To demonstrate this for you, I'm going to  use some test signals. The first signal I'll use is called white noise, which is made up of equal  energy per frequency. When you play white noise, the frequency graph shows a  more or less straight line from the lowest frequencies we can hear to  the highest frequencies we can hear. I'm also going to add another copy of the white  noise, but I'll delay it by 5 milliseconds.

[00:09:40]When I play those two copies of the white  noise together, we'll get a comb filter.

[00:09:49]That's because white noise is made up of all  frequencies and, based on the amount of delay between the two copies, certain frequencies will  be cancelled and certain frequencies will be summed together. The second test signal I want to  use is called a sine sweep. It sounds like this.

[00:10:09]This sine sweep test signal sweeps from 20 hertz to 20 kilohertz. Again, I'm going to copy  this sine sweep and place it on the next track. I'm also going to delay this by  5 milliseconds. Let's hear what happens.

[00:10:29]That was a little fast. Let's  slow it down a little bit.

[00:10:34]These two signals are the same sine sweep, but  they're over 30 seconds rather than just 1 second.

[00:11:10]Let's take a closer look at the sign sweep  signal to get a better understanding of why comb filtering is happening. You can  see that at the beginning, the peaks and troughs of both copies of the signal line up.

[00:11:35]When we get to our first dip at 1 kilohertz you can see that the peak and the trough are  lined up, which makes a complete cancellation.

[00:12:03]Again, the peaks and troughs  are more or less lined up, and then again we get to the point where a peak  and a trough are lined up. This pattern continues, notching a frequency at each octave. You'll  notice that because the delay between the two signals is still five milliseconds,  the frequencies that are notched by the comb filter are the same as we saw with the  white noise demonstration. Check this out. So as we can see when there's 5 milliseconds  of delay between the two signals, the first frequency that is notched is just  over 1 kilohertz. Let's see what happens if we increase the amount  of delay to 10 milliseconds.

[00:13:30]With 10 milliseconds of delay rather than 5  milliseconds of delay between the two signals, we can see that the first frequency to be notched  by the comb filter has gone down a full octave, now to just over 500 hertz. With that logic, if  we move the delay to about 2.5 milliseconds we should see that the first notched frequency will  be an octave higher than 1 kilohertz. Let's see.

[00:14:02]If you enjoyed this video, make sure to check out the video that's  on your screen now. I'll see you there.