Topic: 4-mic Phased Array Recording

[voltronic]

FLACs now added to the Dropbox folder, as I said a few dB higher and with some RX noise reduction to reduce the HVAC drone.

[2manyrocks]

PM boojum to see what suggestions he may have.  Nice guy.

[Gutbucket]

As TF says they are starting points although he mentions the 67 cm omni spacing quite often.  Clearly he has found that consistently reliable for what he does, but we don't have his equipment nor are we recording at Abbey Road...

Perhaps similarly (only in the sense of being consistently reliable, in all other aspects I'd never presume to compare anything I say or do with TF! I'm just a amateur hack in that sense), the reason I often recommend ~3' or one meter as a good starting point for a pair of spaced omnis is from doing the 'spacing adjustment while listening' thing in a number of rather typical taper situations FOB outdoor in amphitheaters recording bands playing through a PA.  I find it's simply a good, safe general starting point and spacings of less than that are usually just not as good.  More is sometimes better, but riskier without another mic in the middle, in which case I double it to 6'.  And that's primarily based on how the bass sounds and how open and spacious the recording sounds overall, rather than SRA imaging aspects. 

I later found that 3-4' also worked well on stage or at the stage-lip, where the situation is different in many ways, but perhaps not so different in others (like the quality of the bass, based on that spacing-distance/frequency/phase relationship)

[2manyrocks]

I would keep it private particularly since we want his personal input. 

TV rabbit antennas sure make it easy to play with the spacing of little Omni's. 

[DSatz]

Sorry to be late arriving. I sense that someone's little experiment for pleasure is getting blown up into something much more important than what it really was/is.

For one thing, the theoretical speculation can be set aside: Tony Faulkner calls these setups "phased arrays" only because years ago, he wrote up one experiment that he'd done with a pair of side-by-side figure-8s facing forward. He knew that he needed some kind of catchy name for the article, and someone suggested to him that they looked somewhat like a simple RF antenna. While that's a doubtful comparison even on a visual level (unless you use progressively shorter microphones evenly spaced in front of a taller one, or the horizontal equivalent), that's the name that he and Studio Sound magazine eventually went with.

The thing is, with RF antennas, you know what frequency or small band of frequencies you're designing the antenna for. Thus you know the wavelength, and you dimension and space the array elements accordingly. Then the antenna is usable only at that frequency or in that narrow band; in any other range of radio frequencies it is worse than useless. But in audio, there can be no phase alignment between any two or more spaced transducers because wavelength is inversely proportional to frequency, and audio has 1000:1 range of frequencies and therefore a 1000:1 range of wavelengths. Whatever setup you have that's aligned for one group of (harmonically related) frequencies, it will always be 180 degrees out of alignment for another, equally large group of frequencies. That's why the 3:1 rule-of-thumb exists. I'm not aware of any special exemption that's been granted on the basis that he's Tony Faulkner. Any positive results that one might get with such microphone arrangements are unlikely to be reproducible in many other recording situations.

And no surprise, Mr. Faulkner doesn't generally use these methods for his professional recordings. They're just things he finds it interesting to play with and occasionally write an article about. He's become known for them, people request this kind of recipe from him and he obliges them. Good for him, but please let's not misconstrue what it's about.

--best regards

[Gutbucket]

Are we making this out to be overly important by analyzing underlying principles?  I don't think so.  I admit that I enjoy examining the underlying basis of things such as this more than most.  For me that's one of the best ways to gain a deeper understanding of how things work, and I usually find I can then apply that understanding to other areas.  With regards to recording and microphone arrays at least, I'm thankful this forum is a place to discuss the theoretical underpinnings as much as the practical day to day live music taping stuff.  If it was only day-to-day practical taper stuff I'd have lost interest long ago and wouldn't be hanging around here as much.

I don't really care so much if TF uses this regularly or not, but I am interested in how it works, what I can learn from that, and if I can apply that knowledge to what I'm doing.

IMHO, rather than looking for parallels with RF antenna theory, we can find clearer parallels with other applications in audio which are more directly applicable to understanding underlying aspects of microphone arrays such as this.  Multiple microphones on stage intended for separate sources which are all summed together is defacto and the 3:1 rule helps avoid interaction complications.  Multiple microphones arranged in a line and summed together, intended for pickup of a single source or all of them in combination, form a simplified line-array and is far less common.  There is plenty of information available on line-array speaker systems on the reinforcement and reproduction side.  The 'sensing' side of things using microphones in a line-array configuration is closely analogous.

Google the phrase: subwoofer array steering and take a look at most any of the resulting links.  The interaction of multiple  microphones arranged in a line perpendicular to the source and mixed together with each other are roughly analogous to the behavior of a "broadside array" of subwoofers.  A lot can be learned from reading some of those articles and considering the basic acoustic principles which apply to both sensor arrays and driver arrays.  The polar diagrams showing the resulting lobing and how it varies with frequency for all angles off the median plane visually illustrates much of what we are describing here in words.

The Microtech Gefell  KEM 970 and 975 line-array microphones are examples of applying the same underlying principles to sound pickup side of things.  Other more arcane and complex audio applications are 2 and 3-dimensional beamforming sensor arrays consisting of tens or hundreds of microphones used in applications such as sonar arrays.

The practical take away, applicable to the design of simpler microphone arrays used around here, is that the interactions are incredibly complex and unless very carefully designed and implemented, simpler is often better!

There are a number of discussion threads here at TS concerning the use of two pairs of microphones atop a single microphone stand out in the audience, which is not an uncommon audience taper setup.  As four channel recorders became common, a lot of members here started doing this and many since decided that it's not a predictable or worthwhile improvement over a single stereo pair.  In some of those other threads I've analyzed why that may be the case, and suggested some alternate setups which should be more predicable and potentially more useful.  In practical terms that usually boils down to using less microphones and/or more spacing between them, and considering the microphone array as something new, rather than the combination of two closely placed stereo configurations with which we are all familiar.

It's the practical applications and the quality of the resulting recording which are most important of course.  I'm just suggesting that we can arrive at good practical solutions more quickly and consistently by understanding and considering underlying principles, rather than trying things randomly, throwing together combinations of two standard stereo setups like a shot in the dark. 

When I see an opportunity to discus these things around here I jump in with both feet!  Apologies for the overly big cannon-ball splash soaking the poolside sunbathers.

[DSatz]

> There are a number of discussion threads here at TS concerning the use of two pairs of microphones atop a single microphone stand out in the audience, which is not an uncommon audience taper setup.  As four channel recorders became common, a lot of members here started doing this and many since decided that it's not a predictable or worthwhile improvement over a single stereo pair.

This, to me, is the key information. People are trying these four-microphone > two-playback-channel approaches largely because they have the equipment available. And I'll add: because it's fun to try things out, and "hey, you never know."

OK. But arrival-time issues don't go away just because we would rather not think about them. If two or more microphones are spaced apart and their outputs are summed (whether before or after recording), the same direct sounds will have arrived at the microphones at different moments in time, and will therefore have been picked up in varying phase relationships depending on the specific frequencies occurring at that moment. The result is a fairly strong frequency- and direction-of-origin-dependent cancellation of some frequencies, along with some selective reinforcement of others. In other words the frequency response of your recording will be screwed up by a considerable amount and in a spatially complicated way.

By varying the parameters (microphone orientation and spacing, relative mix levels) you might, in a given case, be able to tweak things until the result sounds interesting or even good. But then you will have gained nothing that you can reliably apply to the next recording that you make, in some other space or with a different arrangement of forces on stage. If (say) a stereo amplifier did to your signals what this type of setup does to the incoming sound, you would never regard it as more than a toy. A basic knowledge of acoustics will tell you that there can be no generally-applicable stereo recording technique with two or more microphones per playback channel, unless the microphones for each channel are damn near coincident (e.g. Straus-Paket, Polarflex).

This is not new territory; it was explored before stereo recording was even in the picture.

--best regards

[2manyrocks]

...so what are these 4 channel recorders good for other than to put a main pair and take a soundboard feed? or main pair plus a couple of spot mics? 

[DSatz]

Those are two good uses, and you can also put up two pairs of mikes in arrangements that make sense, record both pairs, and pick your favorite afterward.

Or you use three channels for double M/S, or four channels for Ambisonic B-format.

You can split the outputs of a stereo preamp and record the signals at full gain in two channels and at -6 dB in the other two, in case the peaks go higher than you expected.

Over in another thread I talked about recording M/S with one kind of "M" microphone above the figure-8 and another kind beneath it, so that I can choose whichever sounds better for each piece on the concert.

But never let the availability of further channels steer you away from using the simplest mike setup that gets the job done. Add microphones only when that is the only way to get the balance you need, and do it with your purpose clearly in mind, and an awareness of what the additional microphones will do to the overall pickup. Don't just put up extra microphones because you have extra recorder channels, then hope that your good intentions will magically exempt you from the physics of arrival-time differences.

Nearly all of the first 2,000+ live recordings that I made were single-pair. When I started using spot mikes occasionally, the results were only rarely a real improvement. There's a skill to choosing and placing spot mikes and mixing them in, which it took me quite some time to learn to a tolerable degree. And that's where I really do rely on having a four-channel recorder, since I would never in 100 years be able to get the relative levels between the overall microphones and the spot mikes right in a live concert situation, especially given that I only monitor with headphones. (BTW I also tend to use microphones that have deliberately rolled-off low-frequency response as spot mikes, e.g. speech cardioids with flat high-frequency response.)

The one good thing I have to say about the kinds of techniques discussed in this thread is that you're only recording one microphone's signal per recorder channel, so you always have an unmixed, one-mike-per-channel stereo recording in the can. Don't be afraid to use it. Most of the time, I think you'll find it more natural sounding, and less tiring to listen to, than any mix you can make using more microphones in more places, unless you have really planned and tested your proposed mix in terms of the specific sound sources that you're working with.

--best regards

[voltronic]

But never let the availability of further channels steer you away from using the simplest mike setup that gets the job done. Add microphones only when that is the only way to get the balance you need, and do it with your purpose clearly in mind, and an awareness of what the additional microphones will do to the overall pickup. Don't just put up extra microphones because you have extra recorder channels, then hope that your good intentions will magically exempt you from the physics of arrival-time differences.

snip
...I would never in 100 years be able to get the relative levels between the overall microphones and the spot mikes right in a live concert situation, especially given that I only monitor with headphones.
snip

I can't speak for others, but it was the other way around for me.  I purchased a 4-channel recorder specifically to try techniques such as this because I was impressed with the sound of recordings made this way, and I already owned the right type of mics to try it.  Your technical explanation of the frequency / phase interaction problems is over my head, I will admit so I'll let the grown ups here discuss such matters .  But again, I wanted to try this because I enjoy the sound and I wanted at least the option of a second pair of mics to choose from in post after recording a concert.  I never expected it to be some kind of magical thing, just another option to try.

If you have not done so, I urge you to listen to the samples in the links I posted, especially post 52 (the DPA array), 56, and 72 in this thread: https://www.gearslutz.com/board/remote-possibilities-acoustic-music-location-recording/730017-faulkner-arrays-2.html  Maybe I'm hearing things, but I think they're quite convincing. 

Regarding your other comment I quoted in part, of course you wouldn't be able to balance the main pair and spots in that situation - that's why it's nice to have the channels to manipulate in post or discard if they aren't useful.  Are you saying that makes it not worth trying?

[voltronic]

Another idea is to use the extra channels to record a spaced pair of omnis, but be prepared to use a low-pass filter on those channels before mixing.  That will eliminate the frequencies where phase cancellation will cause a problem, while giving you the benefit of omni LF response.

I did read a comment on GS from someone who did exactly that with one of these setups.

[DSatz]

voltronic, the frequency/wavelength/phase interaction is a crucial perspective. I have a day job that I have to go to now, and I suspect that by the time I log in tonight, someone will have nicely filled in the basics. It's a perspective that everyone needs to consider when thinking about microphone placement, and specifically about mixing the signals from more than one microphone together under any circumstances.

Just to get the ball rolling, though: With some audio software or test equipment, you can record an audio frequency "sweep tone"--a simple tone that starts at one end of the audio frequency spectrum (say, 20 Hz) and slowly rises in pitch/frequency until it gets to the other end of things (say, 20 kHz), with a constant amplitude. This is a very useful signal. If you record it as a two-channel wave file and play it back, the tones that you hear should sound as if they're coming from a point between your loudspeakers. You could then take a copy of that two-channel signal, mix the two channels 50/50 into a mono signal, play it back again through the same two loudspeakers, and the effect would be indistinguishable (no surprise since each speaker would be playing back exactly what it had played back before).

Now if you go back to the two-channel sweep tone and take one of its channels and shift it in time by, say, 2 milliseconds in either direction, you can play that signal back and the mono effect will remain at low frequencies. But something weird will start to happen in the midrange and above. I want to keep this completely non-mathematical for now, so let me just say that at some point the clear center placement will dissipate and dissolve, then return as the tone goes higher still--and then it will dissolve again and return again, more and more rapidly as the frequency continues to rise.

Mix _that_ two-channel signal into mono (again 50/50) and instead of the weird phasiness, you'll get a centered-seeming tone with an amplitude that rides up and down quite severely. Wherever it was "phasey" sounding in the two-channel version, it will be lower in amplitude in this mixed version (even disappearing completely for a moment), and where it was well centered in the two-channel version, it will temporarily be several dB louder in the mixed version--so-called "destructive" and "constructive" forms of interference respectively. As before, the alternation between rising and falling signal strength will become more and more rapid as the frequency rises.

That (for reasons we can get into) is called a "comb filter" effect. And it is basically what you're doing to your signal any time you mix two inputs together that contain substantially the same material, where one is time-delayed relative to the other--including when the source is speech or music rather than a test tone. The sweep tone simply makes the point more obvious, since it isolates the effect on one frequency at a time. The problem with many multi-mike setups is that the differing distances from the sound sources to the various microphones cause time delays of this kind.

In professional recording situations, where dozens of microphones may be mixed down to two channels (or even just one channel for P.A. at live events), the effect gets kind of swamped by the sheer number of different inputs--and of course they aren't all picking up the same material for the most part, especially if the 3-to-1 rule is being followed. In a pragmatic sense you can record with a pair of microphones, or carefully blend in a few spot mikes, OR go all the way to the other side and use 50 microphones with or without a main pair--but where the comb filtering problem really gets you the worst is when you add the signals from just one or two additional microphones to the signals from a main pair, unless you're very careful about what those additional microphones are picking up. If they're coincident with the main pair, there are no arrival-time differences so this problem doesn't occur. If they're far from the main pair, the signals are so dissimilar that there's a random relationship between them and again, the comb filtering won't generally be obvious although technically it will exist to some small extent.

Gotta go. See y'all later.

--best regards

[Gutbucket]

I'll only add that some complexity in the phase relationship between channels is often desirable can be good and of course.  Else we would all gravitate to only using coincident setups where both channels have identical phase at all frequencies.  The important thing is getting the right amount with the right qualities- enough good sounding phase complexity without bad sounding problems.

Its 'good' if you like how it sounds. It's 'bad' if you don't.  In the end, that's all that matters.

The technical take on all this is interesting (or boring, depending), but is not the point of making recordings.  Its useful in figuring out why something may sound bad or sound good, and in helping to figure out how to make it sound the way you'd like more consistently, in various situations.  It's a tool for understanding what's going on and figuring out how to manipulate things, rather than an end in itself, helping to illuminate which path may be most productive.

My approach to running multiple mic arrays for stereo is this- As long as each channel is being recorded individually, and the main pair is set and ready, go ahead and run the extra mics if you have them and its not too much complication.  You then have two separate recordings which you can compare at least, and redundancy if you have a problem with one.  Go ahead and play around with mixing them, it can be enlightening and might work out well.   If you like it better than either on their own, then great.  If not, you might think about why, and how some of what is being discussed in this thread may apply.   It's more about the intent of the recordist, rather than refraining from running the extra gear.  There is no problem with the extra stuff as long as you consider a single stereo pair primary and can keep it separate.   It's when your approach switches from "a single stereo pair.. plus some other mics that may be interesting in comparison or maybe mixed together" to "I'm setting this up with the intention of mixing all the microphones together" that a more considered approach becomes important in getting consistent results that are an improvement on the single pair on it's own.

If someone wants to try adding a second pair with the intention of mixing the two pair together, the fist thing I'd suggest is putting up a wide-enough spaced pair of omnis.  But place them twice as far apart as you would if they were to be used on their own, or more.  That will make it much more likely that the combination with the main pair will work well.  The problem is that's usually not practical from a single stand.   It's interesting to play with low-pass filtering the omnis like Jon mentioned, and that may be more important in getting everything behaving well together if the omnis aren't spaced far enough apart from the primary pair.  If spaced enough, I find I usually like the addition of the random phase complexity at higher frequencies as well though, and don't end up using the low-pass.  A more nuanced EQ curve is often what works better for me in that case.

BTW, I listened briefly to the mp3s and prefer the four-mic mixes to either pair alone.  They seemed to have more dimension and held my interest more.  I don't care if that's not accurate or synthetic or whatever, I just prefer how they sound.  They present the better illusion.  But I'm probably more tolerant of a bit of phasyness than others.  I also tend to like near-spaced configs like ORTF more than X/Y most of the time despite of or perhaps because of the phase interactions.

Here's a thought-
People are familiar with the term "mono-compatibility" in regards to issues in mixing two channel microphone configurations down to single-channel mono.  Similarly, a four microphone configuration mixed down to two-channel stereo raises "stereo-compatibility" issues.

[Gutbucket]

Over in another thread I talked about recording M/S with one kind of "M" microphone above the figure-8 and another kind beneath it, so that I can choose whichever sounds better for each piece on the concert.

Great idea, quoted so it doesn't fly under the radar.

[voltronic]

DSatz, thank you for the clarification.  I understand comb filtering and also the 3-1 rule, but now I can better see where you concern is for the mic placements we're discussing here. 

I would suspect Gutbucket may also be correct that in certain circumstances these interactions may produce good-sounding results.  That may or may not be what I'm hearing in these recordings that I enjoy - I've definitely heard recordings that have phase issues before, but I don't here those characteristics in these recordings.  Admittedly, my experience is limited, but I enjoy the sounds I'm hearing.

Regarding the time delay causing said phase interference, I think it's worth noting that in the mic placements I'm describing the flanking omni mics are obviously not coincident but are still somewhat close -  in the experimental one I've posted here and some of the other spacings linked, each omni mic is 10cm from the inner subcardiod.  I know that is close to the wavelength of 3400 Hz, but I'm not sure if that means that's the frequency above which there would be phase interference between each "side" pair.