Comb filtering from original source compared to reproduction from speakers

[hex168]

Any thoughts (or references) regarding comb filtering from original sources compared to its reproduction through the chain from microphones to speakers? For example, a cymbal produces high frequencies from a fairly large object and comb filtering would occur based on its dimensions and the wavelengths being emitted. A microphone would pick up the signal from a point in space (close enough) and the speakers would produce an analog of the original emitted from a smaller object than the original.

If a listener was in a room with the cymbal, comb filtering might be perceived not only in frequency response but also in its variation with small head movements. This would be different from the sound field reproduced by speakers. Is this an audible effect, or can it be completely ignored? If the difference is audible, could the comparison be less perceptively different if the speaker had multiple tweeters, so that the sound field from the speaker contained comb filtering more similarly distributed in space compared to the original source (for some range of source sizes, at least)?

(As far as I know, the only member here who has mentioned deliberately using widely separated tweeters is @dualazmak, but I was not able to find his explanation for doing that.)

 

[STC]

a listener was in a room with the cymbal, comb filtering might be perceived not only in frequency response but also in its variation with small head movements. This would be different from the sound field reproduced by speakers. Is this an audible effect, or can it be completely ignored?

A speaker can only do so much. After all, it’s a man-made lyrebird — remarkably good at mimicking sounds, but never quite the same as the real thing.

Stereo has its own built-in compromises (including comb filtering), yet we’ve adapted to it over a lifetime and usually find it natural enough. Live instruments like cymbals sound different depending on distance and small head movements, and no recording/playback chain can fully capture every perspective of the original event. That’s the fundamental limit.

Multiple tweeters might help a bit by making the high frequencies feel more distributed, but they bring their own trade-offs too.

ST

 

[voodooless]

Audio recordings are totally artificial. They will never represent what you would have heard in the room. Comb filtering is only one of so many things.

Particularly for drums, they will sound nothing like the real thing anyway, because of the way they are recorded, but also because of the processing that is added on top. An attempt at an authentic drum sound is rather rare nowadays.

Eventually, the recording process is just as much a part of the creative expression as the performance of the artists. The only thing we can do as listeners is accept that whatever made it the final product is as it was meant to be.

 

[Keith_W]

Toole (3rd edition), Chapter 7.1.1: "Comb filters are measurable in any room when a direct sound and a reflected version combine at the microphone. However, there is a great difference in what we measure and what we hear depending on where the two sounds come from."

He then describes an experiment by Clark (1983) which compared subjective impressions of:

1. Two speakers radiating the same sound,
2. A single speaker with a reflective panel deliberately placed to produce a time-delayed reflection that replicates the same pattern of comb filtering,
3. A single speaker with an electronically delayed signal that was added back to itself to create the same pattern of comb filtering

Actual measurements were shown to prove that the pattern of comb filtering was the same. The subjective results were:

1: "moderate to pleasing effect"
2: "very small effect"
3: "greatly degrading effect".

Toole does not provide an explanation as to why the experiment produced different results from the same comb filtered pattern, except to say that listeners "adapted" to the room. This is my explanation (not his): a static microphone can measure comb filtering in a single point in space, but the head is not a single point in space, and it moves around. So the pattern of comb filtering changes depending on head movement, and our brain filters it out. On the other hand, comb filtering that is mixed into the signal is the same no matter where your head is. So it is not filtered out. I have to add that this is speculation on my part - it may be somewhere else in the book (which is why I thought of it) but I can't be bothered looking for the reference right now

 

[Keith_W]

I will also say:

Single instrument --> single microphone: no comb filtering, assuming no reflections
Single instrument --> two microphones: comb filtering (law of reciprocity)
Single instrument --> one microphone, but electronically mixed with other microphones from other instruments in the same recording studio --> comb filtering

 

[STC]

Toole does not provide an explanation as to why the experiment produced different results from the same comb filtered pattern, except to say that listeners "adapted" to the room.

In Chapter 7, Toole explains the Clark experiment by showing that while a single mic measures the same comb-filter pattern in all three cases (two speakers/phantom, reflective panel, electronic delay), our two ears + brain experience it very differently. Room reflections fill in notches, patterns differ between ears (binaural/central spectrum), and we adapt/spectrally compensate — making natural acoustic combs mostly benign or even pleasing, while the pure electronic one is degrading.

Keith, I know you’re the expert in room measurements and REW — this is just my early-days experience…..the more I tried to ‘fix’ everything by chasing perfect graphs and clap tests, the more sterile and lifeless the sound became. Now I trust my ears more in the actual listening position (though I still value measurements as a starting point).

ST

 

[voodooless]

Obviously, speakers will generate some kind of comb filtering in some cases. They do, however, have nothing to do with the comb filtering present at the recording location.

As for drum recordings:

This is a nice anecdote: Peter Gabriel, Phil Collins and producer Hugh Padgham stumbled upon a drum sound, completely by accident, that would go on to define the 80s era of music

 

[Keith_W]

Keith, I know you’re the expert in room measurements and REW — this is just my early-days experience…..the more I tried to ‘fix’ everything by chasing perfect graphs and clap tests, the more sterile and lifeless the sound became.

I am far from an expert in either of those. I only have that appearance because i'm the only one willing (or stupid enough) to write a guide about interpreting those measurements. Nobody else was going to do it, I got tired of repeating myself over and over again, so I just put it in a document and said "look there".

Re: trying to fix something and getting bad results. This is the rub: you have to take appropriate measurements of what you want to fix. If the measurement you take isn't representative of the reality of your hearing, and you try to correct that, bad results will surely follow. You could start by asking yourself "what do the measurements of a perfect loudspeaker look like?". Create some simulations so that you know what to look for, and then look up how the professionals take those measurements. Repeat them a few times to make sure they are robust, repeatable, and consistent. Examine every aspect of the measurement for artefacts before you try to correct it. Once you have that measurement, you can apply extremely fine DSP and correct every last detail and you'll have your loudspeaker as perfect as its physical limitations will allow. Then put that speaker in your room, correct the bass, and you're done.

So I wonder how automated DSP systems like Dirac, where all the measurements are taken from the listening position, can possibly do a good job. Answer: they can't. If they are smart, they will leave high frequency problems alone. You simply can't rely on those measurements taken at MLP. This isn't a question of whether they have smarter DSP brains than I do (they are most certainly smarter than me on that) - it is a fundamental limitation of the measurement technique.

 

[Chris A]

You simply can't rely on those measurements taken at MLP. This isn't a question of whether they have smarter DSP brains than I do (they are most certainly smarter than me on that) - it is a fundamental limitation of the measurement technique.

A pro-loudspeaker engineer (who tutored me along with a few other guys) consistently made the case that the anechoic chamber EQ dial-ins of loudspeakers were the best to use for in-room listening, (Of course, there are suitable adjustments for quarter wavelengths about the size of the smallest dimension of the listening room--i.e., the Schroeder transition region).

It took me a while to figure out why this is true (i.e., anechoic measurements are best). After taking a lot of measurements over many years at different distances--with/without temporary floor absorption to capture floor bounce during the measurements--it's now obvious why.

Chris