Taperssection.com
Gear / Technical Help => Microphones & Setup => Topic started by: Gutbucket on August 20, 2020, 03:17:17 PM
-
This thread is dedicated to the discussion of ambisonics in general, as related to recording and the manipulation of recordings using ambisonic techniques. It's intended to cover B-format and its manipulations, the general use and operation of dedicated ambisonic microphones, "native" ambisonic microphone arrays made using standard microphones, and how ambisonics relates to Dual Mid/Side 3-channel and traditional 2-channel Mid/Side techniques.
Although we might touch on it, this thread will not have much if anything to do with dedicated ambisonic playback systems or ambisonic techniques used to manipulate standard 2ch stereo and multichannel surround playback systems on the reproduction side of things. Interest in ambisonics at TS is almost completely dedicated to 2 channel stereo output (and in rare cases "traditional" multichannel surround playback).
This thread was started because a bunch of general ambisonics discussion started happening in both of Len's recent CoreSound OctoMic threads, off topic to his intent for them.
Links to those threads:
Comparing OctoMic's Polar Patterns to the World's Best Mono Mics (https://taperssection.com/index.php?topic=194944)
The World's Finest Blumlein Array (https://taperssection.com/index.php?topic=194934)
I assumed we had a thread dedicated to the general discussion of ambisonics (I thought I started one years ago), but a search determined we do not. Here are links to a few related threads that came up during a brief search:
2008 general ambisonic microphone thread- Ambisonic mics --- has anyone here recorded any shows? (https://taperssection.com/index.php?topic=109916)
2010 to current general ambisonic microphone thread- Team Ambisonic (Soundfield, Core Tetra/OctoMic, Oomagamma Brahma, Senn Ambeo VR) (https://taperssection.com/index.php?topic=135045)
Also 2010 Soundfield and Coresound Tetramic (and ambisonic mics generally) (https://taperssection.com/index.php?topic=141195.0)
M/S vs Hypers/Wides/etc for Live Recording? (https://taperssection.com/index.php?topic=115814.0)
[that's just a start. I'll edit this post to add additional links to other relevant threads]
-
[reserved for offsite links to explanations of ambisonics in general]
-
[reserved]
-
Here's a link to an old thread: https://taperssection.com/index.php?topic=141195.0
-
Marking to gain knowledge. :D
-
Here's a link to an old thread: https://taperssection.com/index.php?topic=141195.0
Thanks! Link now added above. This was the original thread I was searching for. Still good to have them all consolidated with links here now that the discussion has come around again with new interest in ambisonics over the past several years.
-
Rather than starting here by explaining the basics of ambisonics (which is what the 2nd and 3rd post slots above are intended for, and the linked threads touch on in part), I'm thinking we might just jump right into it and continue discussion of a couple interesting topics brought up by questions kuba e was asking in the recently locked threads. Specifically these two separate discussions: proximity effect or the absence of it in ambisonic microphones, and ways of altering pattern with frequency in 1st order ambisonic systems.
I'll start with the second topic-
kuba e asked [my paraphrase here] Can we modify pattern such that it varies by frequency by EQ'ing the (omni) W-channel? In other words might we be able to specify, say, an omni at low frequecies transitioning to a supercard at high frequencies? Or say, a subcardioid at very low frequencies smoothly morphing to a hyper-cardioid through the midrange at some transitional frequency and then to a cardioid at some high frequency, or whatever.
The easiest way of altering pattern by frequency will be with the use of more advanced ambisonic plugins. This is essentially what the Schoeps PolarFlex system does (originally done using hardware now through software) except applied to ambisonics. But this being a thread about how ambisonics works in general, lets discuss what's going on "under the hood" and kuba e's specific question. The part below in italics is mostly from a PM I sent to kuba e earlier today, copied here with his permission and mine (to honor, in spirit, the rule regarding no public posting of PM discussions).
Yes, in 1st order ambisonics, EQing the W channel will alter the pickup pattern by frequency (without altering the axis). Your understanding of the fundamentals of 1st order Ambisonics as an extension of Mid/Side that you posted the other day in the OctMic Blumlein thread is correct. Altering pattern by frequency is essentially the same as with Mid/Side, where EQing either Mid or Side channel differently than the other will alter pattern in a way that corresponds to the difference in EQ. Same goes for altering axis. But a significant difference from Mid/Side is that those two aspects are no longer inseparably linked when deriving about multiple outputs, such as a stereo pair. Pattern and angle can be adjusted separately and need not affect each other. With basic sum/difference combination you can vary pattern with frequency without change of angle if both microphones are oriented in the same direction, or change angle without change of pattern using two crossed figure 8's. If you want a stereo output from a Mid/Side sum/difference arrangement change of pattern and change of angle always remain linked.
Consider this- in addition to being able to change pattern with frequency, one could also change angle with frequency. Imagine a pair of cardioids (or any 1st order pattern) that remain cardioid across all frequencies but the angle between them changes by frequency. That might be useful. With ambisonics, one can vary both aspects (pattern and angle) by frequency, separately from each other. Mid/Side varies both at the same time, but only on its own terms, not separately, and in a way that is inverse to what is typically useful for a stereo pair (pattern getting wider as angle gets narrower, instead of the opposite).
My advice is to think of the fundamental workings of ambisonics separately from the practical implementations of ambisonic micrphones. The fundamental workings are B-format manipulations as extension of Mid/Side, or rather sum/difference, which is the basis upon which both Mid/Side and Ambisonics work. "Native" ambisonic micrphone arrays using 2 or 3 figure-8 microphones and an omni (all as close as possible to the same point in space) seek to derive that directly from the outputs of the microphones, along with level and frequency matching. In that case the microphones used correspond directly to the W,X,Y,Z b-format channels. Think about the basics of ambisonics in that way, and not as mics on a sphere and the implementation specifics of how that gets converted to W,Z,Y,Z b-format components, which is more complicated.
Dual Mid/Side is a good conceptual bridge between 2ch Mid/Side and ambisonics. A horizontal only "native ambisonic" microphone array such as the setup Nimbus uses (you've probably seen photos of it), or what TS member EMRR did a few times using Sennheisers (pictured in the OMT thread) is fundamentally the same as a dual mid/side setup. The more typical and practical Dual M/S implementation is to use a single figure 8 Side and two cardioid Mids. It's done that way for a few practical reasons: Its easier for people who already understand regular Mid/Side (just adds a rear facing cardioid Mid); more folks have 2 cardioids and a single fig-8 than an two fig-8's and an omni; and reportedly Schoeps tried it both ways with their capsules and doing it with two cards and a singe Fig-8 worked somewhat better for them. Regardless, all of these physical arrangements can be more or less easily understood as using the same fundamental sum/difference basis as Mid/Side.
Where things get more complicated are ambisonic microphones with capsules arrayed over the surface of a sphere. That's done with 1st order ambisonic microphones for a few practical reasons: one can't really get all 3 or 4 microphones in the same spot but this arrangement gets closer to doing so, and an ambisonic microphone using four cardioid elements is easier to implement and less costly for a manufacturer than using figure-8s. Its done that way for higher orders because its the only way, one cannot make a "native 2nd order ambisonic array" because it would require a the addition of native 2nd order microphones (which do not physically exist) to an existing native 1st order arrangement, which already has a hard time achieving close coincidence.
Where I think your summary in the other thread goes somewhat astray is in considering 2nd order entirely different from 1st order. On a fundamental level it is not. It adds complexity by introducing additional spherical harmonics which do not correspond to native 1st order patterns, but the basis upon which it works in B-format is the same, only further extended. Where it is very different is in practical implementation (mics arrayed across the surface of a sphere), yet in the same way that a 1st order ambisonic microphone made using 4 cardioids on the faces of a tetrahedron (4 mics on a sphere) differs from a "native array" ambisonic arrangement of 3 figure-8's + omni.
Ambisonics in terms of basic B-format is complicated enough just in understanding the theory. The engineering of practical A-format implementations used to derive b-format is a much greater complication and best left out of this discussion. The two locked threads which inspired this one are intended to be about how well the engineering of that particular microphone has been implemented. The OT discussion over there was more concerned with concerned ambisonic theory in general (universal b-format rather than specific a-format).
-
I’ll toss out the practical idea that, purely because I can envision it more clearly, to mix patterns with respect to frequency ranges, I’d probably generate multiple patterns from an array and then mix them together through a crossover network. I have mixed Blumlein with AB omni’s in that manner, which is even more offroad than mixing multiple coincident virtuals.
-
Thank you Gutbucket for this thread. I will join after the weekend.
Emrr, how does your crossover network work? I remember that I was using linear eq when I edited Mid/Side. I don't remember if I did eq bass or treble. But there was an audible difference between linear eq (no phase shift) and classical eq (with phase shift).
-
I rarely find cases where linear EQ is artifact free or sounds better to my ears. It has it’s place. Things like vocal plosives, anything that is a spot fix versus overall, it can be more invisible.
I make a copy of the track, then apply opposite hpf and low pass filter to each with matched frequency, mix back together. Really a ‘reverse crossover’ approach.
-
I went to do mic placement research for a music livestream/podcast/radio show from a theater this week. The first two seasons had an audience, not possible now. They have a pair of Line Audio CM3's at the center lip of the balcony, essentially rear of the room. Now that there's no audience, ambience can be grabbed from anywhere, and it's probable we find something we like eventually and fly something permanent.
I took a native B format array with MKH30's and an MKH20, and had an AB pair of MKH20's with the treble boost on at about 27 inches on the same stand. I moved that around during soundcheck and show to get samples for review. I captured that on the F8n, along with a mono copy of the broadcast mix.
I also put a pair of DPA 4060's in boundary mounts about halfway back in the room, mounted 8 feet up on the side walls. Those went to the house PA/record mixer, so I'm waiting to get those tracks and the CM3 tracks sent for comparison.
This show gets a live mix, but is also multi-tracked for safety. Usually the live stage mix is combined with the CM3's, and that's it unless there's a mix error to fix for the podcast version.
The context and eventual mix goal plays a big part, and is a big strength with ambisonics. In reviewing the tracks, mixing with the broadcast mix in place, I actually like the native array set cardioid XY90 and pointed to the BACK of the room, because there's a lot of direct PA in anything forward facing. Opposite of our normal taper problem. I also like the AB pair for spaciousness, but only when it's way back in the room to diminish the direct PA contribution.
It's a converted movie theater, so standard sloped floor. The PA is mounted really high, and pointing down at a sharp angle, so optimal mic height varies a lot with depth of placement. A proper ambisonic mic with up/down panning would be really useful in a place like this, at least for assessment purposes.
Here's a pic from a previous show there.
(https://live.staticflickr.com/65535/48756380576_c0c9152e18_z.jpg)
-
Received the other tracks, pretty interesting to compare. The distant spaced CM3 (or 4) pair is WAAAAY back in the diffuse field...only another 20 feet maybe from where I was. You get reverb, but no definition. The wall mounted 4060's are immediately exciting sounding and the width adds to that, but after awhile I realize there's distracting timing differences on percussive sounds.
Using the same mics in the native array pair, I went to mid-side. You can use the figure 8, or the omni, or both to get cardioid. You can flip polarity on the omni to make cardioid MS turn around 180º and point to the back of the room. All achievable with ambisonics, but the comparison process is different and thus gives different perspectives on the results.
By itself, no board feed, Something between cardioid and figure 8 sounded best. f8 has the most pinpoint center clarity, but it's all in the middle at the exclusion of some other stuff. Cardioid was good, but a little too ambient still.
Putting the board feed back in and going back to the ambisonic plugin path, Blumlein started to stand out, but eventually I moved the pattern slightly toward supercardioid to lose a bit of the rear sound. From there you can start to envision how placing different sets of mics with different patterns more forward or further back might achieve similar results to the varying ambisonic patterns, ignoring the time delay aspect.
It's worth taking B format (WXYZ) patterns and setting up bussing paths to make the direct comparisons between MS and ambisonic encoding techniques, it really helps envision the possibilities and the finer points of the virtual patterns achieved.
-
It's worth taking B format (WXYZ) patterns and setting up bussing paths to make the direct comparisons between MS and ambisonic encoding techniques, it really helps envision the possibilities and the finer points of the virtual patterns achieved.
Now perhaps stretch what you're doing to second-order and see what you could do.
It'll be nothing like what you're used to, unless you just use the second-order mic to improve the first-order B-format patterns. That's too easy.
-
It's worth taking B format (WXYZ) patterns and setting up bussing paths to make the direct comparisons between MS and ambisonic encoding techniques, it really helps envision the possibilities and the finer points of the virtual patterns achieved.
Now perhaps stretch what you're doing to second-order and see what you could do.
It'll be nothing like what you're used to, unless you just use the second-order mic to improve the first-order B-format patterns. That's too easy.
Yes! I am listening to that stuff too, 2nd order gives very different options. In this context though, as an advisor on a show with limited budget, the goal is to spec a path for a permanent flown (affordable) pair, so focusing on first order patterns. Right now I have opinions about where to try the existing Line Audio mics, so they don't have to buy anything.
-
If coincident arrays are a possibility, you could do it with one second-order mic.
You would't have to worry about cable twist, as you can rotate B-format in post - the same as with the directivity pattern and pointing angle selection.
-
I rarely find cases where linear EQ is artifact free or sounds better to my ears. It has it’s place. Things like vocal plosives, anything that is a spot fix versus overall, it can be more invisible.
I make a copy of the track, then apply opposite hpf and low pass filter to each with matched frequency, mix back together. Really a ‘reverse crossover’ approach.
Thank you for the correction. I think I tried the linear eq on Mid/Side of close spaced pair and it occurred to me that the linear eq sounds cleaner. But I'm not good for post processing, I can easily overhear some artifacts.
-
Where I think your summary in the other thread goes somewhat astray is in considering 2nd order entirely different from 1st order. On a fundamental level it is not. It adds complexity by introducing additional spherical harmonics which do not correspond to native 1st order patterns, but the basis upon which it works in B-format is the same, only further extended. Where it is very different is in practical implementation (mics arrayed across the surface of a sphere), yet in the same way that a 1st order ambisonic microphone made using 4 cardioids on the faces of a tetrahedron (4 mics on a sphere) differs from a "native array" ambisonic arrangement of 3 figure-8's + omni.[/i]
Thank you Gutbucket. You are right, I didn't write it well. The basis of all ambisonic is the same - everything is based on measuring acoustic pressure and pressure gradient.
Maybe it's better to say this:
The 1st order ambisonic (3 fig.8+omni, tetrahedron ..) assumes a coincident capsules.
On contrast, the higher order ambisonic (HOA) assumes noncoincident capsules. Therefore, it is possible to obtain higher orders. If we consider the capsules as a coincident, the HOA would not work.
I'll be happy if someone correct me. The following will probably be familiar to most, but I'd like to make sure I'm looking at it correctly.
The principle of the 1st order ambisonic is simple. The B-format can be obtained directly from 3 fig.8+omni or it is obtained by the summing and subtraction the signals of the capsules e.g. at a tetrahedron. Here is an example of the four coincident cardiods at a tetrahedron (for better understanding, take a look at a picture of e.g. TetraMic):
W (omni) = FrontLeftUp + FrontRightDown + BackLeftDown + BackRightUp
X (fig.8. in X direction) = (FrontLeftUp + FrontRightDown) - (BackLeftDown + BackRightUp)
Y (fig.8. in Y direction) = (FrontLeftUp + BackLeftDown) - (FrontRightDown + BackRightUp)
Z (fig.8. in Z direction) = (FrontLeftUp + BackRightUp) - (FrontRightDown + BackLeftDown)
But HOA is more complicated. Here is B-format matrix calculation. The positions of the capsules are hidden in the matrix:
B = W ^ -1. (Y ^ t.Y) ^ - 1.Y ^ t.pa
If we want to derive the 1st order pattern of HOA at least four signals are needed as in the 1st order ambisonic. Probably the signals from all capsules are used in the 1st order pattern for error correction (capsules position, signal). Therefore, the 1st order pattern of HOA should be more accurate than the 1st order ambisonic.
-
^Yes. A tetrahedron is the simplest non-coincident arrangement of evenly spaced sensors intersecting the surface of a sphere. Four points are the minimum required to define a 3-dimensional spherical volume.
-
They have a pair of Line Audio CM3's at the center lip of the balcony, essentially rear of the room.
[snip]
I took a native B format array with MKH30's and an MKH20, and had an AB pair of MKH20's with the treble boost on at about 27 inches on the same stand. I moved that around during soundcheck and show to get samples for review.
I also put a pair of DPA 4060's in boundary mounts about halfway back in the room, mounted 8 feet up on the side walls.
[snip]
In reviewing the tracks, mixing with the broadcast mix in place, I actually like the native array set cardioid XY90 and pointed to the BACK of the room, because there's a lot of direct PA in anything forward facing. Opposite of our normal taper problem. I also like the AB pair for spaciousness, but only when it's way back in the room to diminish the direct PA contribution.
Received the other tracks, pretty interesting to compare. The distant spaced CM3 (or 4) pair is WAAAAY back in the diffuse field...only another 20 feet maybe from where I was. You get reverb, but no definition. The wall mounted 4060's are immediately exciting sounding and the width adds to that, but after awhile I realize there's distracting timing differences on percussive sounds.
Using the same mics in the native array pair, I went to mid-side. You can use the figure 8, or the omni, or both to get cardioid. You can flip polarity on the omni to make cardioid MS turn around 180º and point to the back of the room. All achievable with ambisonics, but the comparison process is different and thus gives different perspectives on the results.
By itself, no board feed, Something between cardioid and figure 8 sounded best. f8 has the most pinpoint center clarity, but it's all in the middle at the exclusion of some other stuff. Cardioid was good, but a little too ambient still.
Putting the board feed back in and going back to the ambisonic plugin path, Blumlein started to stand out, but eventually I moved the pattern slightly toward supercardioid to lose a bit of the rear sound. From there you can start to envision how placing different sets of mics with different patterns more forward or further back might achieve similar results to the varying ambisonic patterns, ignoring the time delay aspect.
This all rings true with my experience leading up to mucking around with what we've been calling OMT arrays.
Especially with regards to surround recording and playback, I found ambience channels most useful when the microphones are placed around the same distance from the stage as the forward-facing array due to time of arrival reasons. With true surround playback as the goal, time-of arrival becomes a multi-dimensional problem which places increased constraints on the use of delay for time correction. One one cannot simply delay the main pair to achieve sync with more distantly placed ambient microphones without compromising the imaging of reflections and direct arrival sounds that come from directions other than the front. That means using level differences generated through directional sensitivity rather than distance to achieve needed reduction of front main sound pickup in comparison to the pickup of ambience and non-frontal direct arriving sounds.
After some additional experimentation I realized that regarless of specific application, I wanted near-spaced or closer time coincidence of all microphones in the front/back orientation and greater time-of-arrival differences in the Left-Right orientation, and all my later OMT setups evolved with this basic geometry in mind. Wider to either side while relatively compact front/rear. [edit- and this is why I plan to revisit the use of TetraMic or some other ambisonic-microphone between a pair of wide omnis.. or subcardioids.. as the central component of an easily setup basic OMT array].
At one point when I was using just a pair of wide-spaced omnis for ambience, routing the omnis to the surround channels for playback, rather than also including rear-facing directional mics, I empirically determined that in order to achieve optimal surround channel playback levels without pulling too much front image into the rear I needed about 6dB less sensitivity to front-arriving sound in the omni channels (this with outdoor ampitheater recordings that didn't otherwise have any room reverberation issues). That 6dB reduction figure corresponds with the reduction in sensitivity to the rear of a subcardioid and if I had a pair I would have used them pointed away from the stage to achieve that, hopefully in a sufficiently transparent way and without overly compromising the low frequency extension I was getting from the omnis. I even started a TS thread around it at the time to discuss which cardioids and subcardioids sound most natural 180-degrees off-axis, through their region of minimal sensitivity. Since I have no subcards I instead I went back to using sphere attachments on the omnis and pointing those rearward, which is less effective but is at least doing something in that direction. Granted, all of this was about attempting to get just the right amount of front bleed into the rear channels simply by arrangement of the microphones and adjustment of level alone, without any post-production cross-mixing. Achieving enough bleed for good seamless imaging and blend all the way around, but not so much as to begin to noticeably pull the front around to the rear. Essentially extending the way that a single stereo pair overlap each other just enough, yet not too much, around the rest of the surround playback array. In mixing things for 2-channel output we have more control, and greater channel isolation is useful when it can be achieved, but the timing still has to work right.
Besides the outdoor amphitheater recordings (discussed in past OMT threads), much of this was determined over the course of a few years at a monthly jazz trio gig with a very quiet audience, perhaps not dissimilar to the theater music livestream/podcast/radio show EmRR describes above, but in a somewhat smaller room (PA not used). There my primary focus was a 2-channel mix for the artist and for CD's handed out to patrons at the next month's show, the secondary focus being exploration of ambient mic'ing, 3-channel stereo and surround recording techniques. I ended up preferring two methods for room/ambience channels which might be worthwhile trying at the theater above. My main array usually consisted of a near-spaced L/C/R supercardioid triplet on stage near the stage-lip. I found the best arrangement for the ambience channels was either a wide spaced pair of DPA 4060 boundary mounted to the front of the short front wall under the stage-lip facing out into the audience, or a pair of cardioids clamped to the lip so as to hang just beneath it and face out into the audience, sometimes wide sometimes near-spaced. Either way, the placement achieved sufficient reduction of sensitivity to on-stage sound by using the stage lip as a baffle, even when using cardioids facing the audience (their rear-rejection allowing them to be mounted somewhat higher than the omnis, yet still needing some baffling from on-stage sound diffracting around the stage lip). At the same time the placement produced a time alignment that worked "natively" as those microphones were only a few feet more distant in the front/rear dimension than the main pair. Boundary mounting omnis on the side or rear walls required time-alignment and picked up too much direct sound from on-stage. Rear facing cardioids placed further back in the room still required time-alignment in 2-channel mixes and didn't work as well for surround channels for surround playback.
Fortunately this general arrangement of relatively narrow front/back spacing relying on pattern or baffling for front/back differentiation, combined with more side-to-side array spacing working in combination with pattern for left/right differentiation is something I realized I could also manage from a single stand when recording from the audience as well as on-stage.
There are photos and discussion of that in the primary tread about on stage taping. You've likely come across it. I will find it and link here. Here's a link to start of that particular discussion in the on-stage taping thread. (https://taperssection.com/index.php?topic=1625.msg1888271#msg1888271) Discussion of room ambience pair options begins on the next page following that link with a crappy photo of a wide spaced DPA 4060 pair taped to the back wall.
Cool to read the quoted portion of the posts above in which I find confirmation of these observations.. including achieving sufficient reduction of direct PA sound in rear-facing channels, and how that can be what is most needed to support an otherwise well-balanced if dry direct-board recording.
-
After many years, I understood how ambisonic with omni capsules works. Everything is very nicely explained in this discussion:
https://taperssection.com/index.php?topic=207169.0
And the summary is in this post:
https://taperssection.com/index.php?topic=207169.msg2433233#msg2433233
-
Glad to help, although I really only understand it myself in a somewhat cursory way.
How coincident ambisonic mic arrays work is more conceptually straight forward to me than arrays of omnis. "Native" ambisonic arrays are simple enough, producing WXYZ channels directly. The "B to A format" manipulation that need necessarily occur with a tetrahedral array of cardioids is somewhat more mysterious, particularly when considering the compensations that might be made for non-perfect coincidence. And the "B to A format" manipulation necessarily for an array of omnis to be converted to WXYZ channels, be they embeded in a sphere or in free-space (requiring different manipulations) is well beyond me.
Example using omnis embedded in a rigid sphere-
(https://media.springernature.com/lw685/springer-static/image/chp%3A10.1007%2F978-3-030-17207-7_6/MediaObjects/472601_1_En_6_Fig3_HTML.png)
Example using omnis arrayed in free-space-
(https://www.odysseyprosound.com/images/products/secondary/trinnov_audio_3d_measurement_microphone-1.jpg)
-
These are interesting pictures, specially Trinnov. I looked for an ambisonic microphone from Trinnov. I found one picture. I don't know what capsules they used. It is true, it looks like science fiction.
I have only an amateur idea of how to cope with the limitations for low and high frequencies in high order ambisonic of omnis. Fig 8 (and other hoa directional patterns) can be composed of different omni pairs and sum together. Then the resulting signal would be stronger. For low frequencies, they can use pairs that are furthest from each other. And for high frequencies, they can use pairs that have the shortest distance. I believe that what I can imagine for fig 8 also could be applied to other hoa patterns, which are more difficult to imagine.
I'm just thinking out loudly. Hoa with cards or hypercards could have less difficulty with low and high frequencies because they can place the capsules closer together and use the directivity of the capsules. And hoa with omnis could solve this problem by using more capsules.
-
Array geometry is a trade off. Closer coincidence of directional capsules allows for operation to a higher frequency before spatial aliasing occurs and the polar pattern goes all wonky and spiky. Greater spacing in a spaced omni array allows for operation to a lower frequency before incurring signal to noise problems from the weak difference signal. Higher order require a rapidly increasing channel count either way. Most spaced arrays use omnis, but I believe there area a few papers on the use of bi-directional figure 8s in place of the omnis. That may allow for a physically smaller array, or perhaps improved S/N ratio for a given spacing.
As far as spaced beamforming microphone arrays go, as far as I'm aware they fall into two basic categories: Those placed on a rigid surface such as a closed sphere, and those placed in free space, such as those arranged to form an open sphere. Doesn't have to be a sphere if not interested in fully-periphony. The high-end Tinnov HOA recording system I was unable to find a photo of was horizontal only, using 9 DPA 4006 omnis as I recall, all positioned at at the same height carefully arrayed in the horizontal plane. I've seen spiral planar arrays used in visual audio imaging systems where the output is a visual camera image overlayed with colors mapping the sound emitting areas and level of emission. I think those systems target things like noise assessment of cars for NVH reduction and other applications like that and are "perpendicularly end addressed" rather than "side addressed" with regard to the orientation of the array.
FYI- the small Trinnov omni array pictured in my previous post above (that one does come up in a search) is a 3d calibration mic and is not intended for general recording. Note the differing heights of its elements in addition to the horizontal spacing of them. The photo kuba posted shows a significantly higher channel count planar array oriented vertically rather than of horizontally [edit- returning to view the zoomed in image it appears not all of the mic elements inhabit the same plane].
-
Once my cinella DMS mount arrives (probably a couple of months), I'll do a comparison between my sennheiser ambeo vr and mk41v/mk4v/mk8 DMS. With the investment in the cinella mount ($$$$), I'm sure my sunk cost bias will have me favoring it, though.
-
Cool. Will be interested in seeing and hearing that.
I imagine that in addition to any improvement related to using top quality Schoeps mics, the geometry of the Cinella DMS mount will be superior, as long as it features all three mics arranged such that their diaphragms are positioned one directly above the other, thereby achieving something close to true coincidence for a wavefront arrival from any horizontal angle. It's limited to horizontal plane use only though. No way to include an additional vertically-oriented fig-8 or pair of back to back directional capsules to include height with the same tight coincidence.
The Ambio is fully periphonic, but is geometry is less fully coincident in the horizontal plane. Although it's tetrahedral capsule arrangement is not as tightly coincident in the horizontal plane, the degree of coincidence it does achieve is on average more uniform for all approach angles including those outside of the horizontal plane. Not sure how high the frequency where the response is no longer coherent and the pattern goes wonky (that's determined by the diameter of the sphere on which all four capsules lie). If the capsules are positioned close enough to each other, that threshold may be sufficiently close to the upper threshold of hearing that it won't matter much (perceptually at least) when the directional pattern breaks down and goes wonky at frequencies higher than that.