Topic: Oddball microphone technique (OMT) - part 2

[Gutbucket]

It's fun for me to analyze this stuff, so please don't sweat my critiques.  Your recording stands on its own.

As far as redundancy goes, it's hard to go wrong with overly sufficient clear directness.  A lot of folks are happiest with a straight SBD afterall, even though that's not sufficient by itself for many of us spatial teleportation junkies.  My argument is more about increased flexibility when a recording calls for it, yet this recording doesn't really need that other than the level of the guitar perhaps (I didn't specifically listen for guitar level, and am just going off your comment about it). 

I also make the argument above because its far harder to convince tapers to point extra microphones in directions other than toward the stage and PA! And in doing so they are right in the pickup of clear direct sound being the most important component.. it's just not the only one to the exclusion of all else, and most times the other stuff comes along for the ride in any case, just in a less controlled way.  Lots of folks run extra microphone pairs, lots of folks mix those pairs together, but few give much consideration to how those pairs work together as a single array, and that's what I see myself arguing for most of the time here at TS.


And oh setup pressures and how they can modify the best laid plans at best, if not outright take the helm and steer the shit ship-o-fools!  I lost the first two sets of four last Friday night partly due to that and running a new recorder (Zoom F8). The taper's blues!

I suspect that if the SBD feed is good the best one can usually do with two additional microphone channels is run an overly wide pair of spaced omnis to mix with it.  Wide omnis provide essentially everything a SBD feed does not without being so extreme as to try and eliminate as much of the direct-arriving SBD content as possible like a rear-facing pair does.  The omnis provide a more well-balanced take on the entire room sound and perhaps more importantly on the audience reaction, including the enthusiasm forward of the recording position rather than isolating and focusing on distracted folks behind.  I should rephrase that concluding statement in my previous post- I'm mostly curious how well a rear-facing pair will substitute for the wide omni pair mixed with a straight SBD, not because I think it would be a better option, but specifically because the rear-facing pair pushes the "not picking up direct sound from in front" thing to an extreme.

[heathen]

I took an OCT2 array out for a spin this weekend.  Here's the result: http://taperssection.com/index.php?topic=189468.0

So far I really like this setup.  The stereo separation of the supercards is complimented nicely by the strong center of the middle mic.  I could've dialed in more center to give it more direct sound, but I kept the mix nearly equal between sides and center (I think I put the center only 2 dB higher than the sides).  I'm definitely going to run this sort of configuration more often.

[Gutbucket]

Heathen has kindly granted his permission for me to post here a our PM discussion about OCT (Optimum Cardioid Triangle) made prior to the recording he linked above yesterday.  It mostly concerns the spacing of the center channel forward of the L/R pair(s). I figure it may be of interest to others following the thread.  Although much of what I say below I've posted about previously in the OMT part 1 thread, here it is pulled together in the context of OCT.

I hope you don't mind me bugging you with a question out of the blue.  Having recently acquired my first pair of supercards, I want to give some sort of OCT setup a try.  While researching OCT, I came across discussion that when the ultimate goal is a stereo mixdown, the center mic should be placed further forward than the 8 cm for OCT, resulting in OCT2.  I think I saw mention of 40 cm for OCT2, but I'm not sure.

What I haven't found the answer to, though, is why the center mic should be farther forward when the ultimate goal is a stereo mixdown?  Is it to avoid some sort of phasing issue?  Is it purely to introduce greater decorellation between the center and each of the sides?

With my current setup I can achieve up to about 70 cm of spacing between the sides (that's counting the reach from the length of each mic body).  Achieving 40 cm of forward spacing for the center is not currently feasible for me, though.  Part of my desire to understand the reasoning behind the additional forward spacing in OCT2 is to figure out if I need to find a way to rig up that additional forward spacing, if OCT turns out to be something I want to use more.

I should be able to try some of this out tomorrow, and with three bands in the lineup I want to try three variations on the OCT theme.  Hopefully that will be somewhat enlightening.  As always, though, I'd appreciate any insight you can give into the nuts and bolts of why a certain array does what it does.  If you don't have time for this, no worries!

Good question.  In practice I don't have a definitive answer.  I've varied how far forward my center channel, but for the most part its as much a result of practical mounting issues than either a rigorous investigation or adherence to the recommended dimensions.  More on that later..

On the theory side- OCT2 pushes the center microphone farther forward and implements a time-delay on that channel to compensate for that forward shift.  I think that modification was developed because it retains good correlation across the 3 channels for front-arriving sound to a degree similar to standard OCT, yet increases decorrelation of the the center signal verses the sides for sound arriving from outside the SRA window (sides & back).  That's likely advantageous because OCT was conceptualized for "pair-wise" 3-channel L/C/R playback with minimal crosstalk from the channel on the opposite side for off-center positions, and the 3 signals remain fully discrete until they are reproduced by the speakers and only then do they "sum in the air" between the speakers and listener.  In contrast, when mixing the 3 discrete channels to 2, phase-correlation interactions between all three channels summing in the electrical ream are far more distinct than the signals summing in air, so increasing the forward spacing of the center channel effectively pushes the microphones out of close proximity for all directions except the forward direction.

That's part of what informed a number of lengthy posts of mine in the OMT part1 thread concerning the importance of having sufficiently differentiated signals across the multiple microphone pairs one intends to mix together.  Granted in that case, the main point I was trying to get across was an argument against the somewhat typical taper approach of mixing two or three near-spaced pairs all mounted in close-proximity on the same stand and pointing the same direction, but its really addressing the same basic issue.

In a practical sense, I'm not sure how much it really matters.  I've had little problem mixing down my OMT stuff to 2 channel regardless of the forward spacing and have never gotten around to playing with time-shifting the center channel in post. 

My OMT setup generally has the center channel pushed forward about 10" or so (say 25cm) simply due to the mounting arrangement I use.  But sometimes the outrigger arms are not at +/- 90 degrees but angled forward or backward somewhat due to any number of reasons, including mounting against a column or on stage avoiding and keeping out of walking paths of the musicians, which either increases or decreases the spacing somewhat.

Coincidentally I'll be re-rigging it this weekend for a bunch of recording next weekend, and am changing the mounting for other reasons which will shift the center microphone backwards, closer to the OMT(1) standard.  That will make for an opportunity to see if I hear a difference in the 2ch mix.

Way back when I was running 4 omnis mounted in spheres in a sort of mini-Decca tree diamond-shaped arrangement, the front and back pair were spaced more than the supercards I'm using now for OMT.  All four mics were on telescopic antennas and the spacing varied with the situation, but was typically 100cm or so for the Left/Right pair and usually something like half-that front/back pair.  I'd guess the center-front mic was usually between 30 & 50 cm forward.  Those also mixed well and I never did any time shifting.

continued below..

[Gutbucket]

More thoughts-

The three things which primarily informed the evolution of directional portions of my OMT setups were:
Decca tree
OCT
Michael William's MMAD (multichannel microphone array design)

My OMT stuff is sort of a mix and match of all those, and with that in mind you might want to sort of mentally asses the similarities and differences between them in thinking about OCT vs OCT2 and other variations.

Decca tree uses sphere-mounted omnis and increased spacing and is familiar to most folks. Its intended for 2 channel mixing.

OCT uses 180 angle opposed supercards as a way of limiting acoustic cross-talk from the opposite side, and the use of supercardiods at such wide angles in combination with a narrower base-spacing than Decca tree setups, makes necessary shifting the center channel rearward (in position or time) to achieve stereo image "linking" between the L-C and C-R pair segments.  OCT is intended for 3 channel playback (but works fine for a 2 channel mix IME) and the latter OCT2 modification was specifically developed for 2ch mixes.

That kind of image-linking between pairs is what Williams focuses on primarily (there is obvious cross over between each of these approaches, if emphasis on different aspects). Williams is less concerned about acoustic crosstalk from microphones outside each individual pair segment, and is more concerned with smooth imaging "linking" without gaps or overlap between multiple segments around the array, sometimes all the way around 360 degrees.  His arrays are intended for discrete 3, 5, 6 or 7 channel playback and always use microphones of the same polar pattern all the way around the array.  I've used them on their own on stage for some jazz trio stuff and like them.  I did not like OCT as much in those initial on-stage tests because the sideways facing supers were not as on-axis to the musicians and picked up more sidewall reflections.  I now use the OMT setup everywhere though including on stage and I think the omnis help with that, and sometimes I angle the side-facing supercards more forwards when I think that will help, partly informed by those tests.

My current OMT setup is pretty much OCT for L/C/R (but using a center supercard) and gets more Willams when I angle the L/R supers forward, and with respect to how I setup the rear-facing supercard pair so as to have a reasonable pair-wise relationship on each side to the L/R pair.

Interestingly Williams writes about what his "Magic Array" which is his "universal" multichannel recording arrangement intended for producing 2, 3, 4, 5, 6, 7, or 8 channel output by choosing specific output channels (no mixing).  There is some parallel to OCT2 there in that he does this by starting with a typical near-spaced stereo config (think DIN, cardioids 12"/90 degrees), extends that to 3 channels by placing a center microphone well-forward of that pair and delaying it (further forward and more delayed than OCT2 due to the L/R channels being angled less and closer spaced), extends to 4ch quad by muting that center channel and adding a rear facing DIN pair (creating 4 DIN segments all the way around), extends to 5 channels by unmuting the center channel, extends that to 6 channels by adding a rear-facing cardioid as far behind the array as the front one is forward, etc..  The full 8 channel array looks like a sea urchin, with a an IRT cross arrangement in the center (4 ch DIN) with four additional cardioids extending outward from the center of each pair, something like 2 meters.

Its crazy looking and not very practical, but conceptually informative.  I've posted about and linked a photo in the OCT part 1 thread, but I'd have to go look for it.


I suppose all of this is a long way of saying, "fun to think about, but don't worry about it too much!"

[heathen]

I keep going back to the question of forward spacing with the center mic in a OCT/OCT2 array.  For the typical taper scenario (recording a PA, with maybe some sound coming from the stage, maybe not), how much of a difference does the center spacing make?  And I should qualify this by saying how much of an audible difference, because there's obviously a measurable difference in the time of arrival.  I'm only concerned about what is audible though.  Part of my wondering about this is driven by practical concerns.  It's far easier to set up a center mic that's only slightly forward of the side mics than it is to set up a center mic that is 25-30 cm (not to mention the 40 cm I've seen discussed in the context of OCT2).  If the difference between a few cm forward and 25-30 cm forward isn't audible, then all the better. 

[Gutbucket]

That's my current conclusion as well-  Center forward spacing doesn't seem to matter much in comparison to the practical constraints of trying to push the center mic farther than 12" or so forward.  What I'm still curious to determine is how much difference there is between no forward spacing at all, the OCT suggested spacing of 8cm (just 3"), and the approximate 25-30cm (10-12") spacing you used above and which I've been using.  I don't think its a major difference, but I've now nailed the other variables down enough that I'm coming back to this detail.

The reworking I'm currently doing will pull the center microphone back in, because adding the fig-8 to it to form a M/S pair made it heavier and harder to support ~12" out.  So I plan to do some listening to determine if I hear any difference.

My general thoughts on this are that the microphones intended to pickup direct sound from in front should be relatively time-aligned with each other to keep transients aligned and not smeared.  And any rear-facing ambient microphones shouldn't be spaced overly far back due to inevitable leakage of direct sound into them, although a bit more spacing there is likely desirable.  That translates generally as having the microphones arranged more or less in a horizontal line, using spacing more in the Left/Right dimension than the Fore/Aft dimension.  And it works out nicely that such an arrangement is more practical to setup as well!


More from the discussion between Heathen and myself concerning center mic spacing and the OCT2 specification about delaying the center microphone to compensate for is farther forward positioning-

For the other bands I angled the supercards towards the front, so I'll be posting those and that may make for an interesting comparison.

Right on.  Glad it worked out well for you. 
Did you delay the center channel a or just mix it in straight?
What are your impressions on angling the supercards forward vs not?

I tried delaying the center channel and didn't hear any difference, so I just kept it undelayed.  When I calculated how much time the delay needed to be, it was so small that I'd be shocked if it made an audible difference.

I haven't listened to the other sets with the angled supers as much yet, but I did a quick comparison between just the 180* supers and the ~90* supers and, when compared side-by-side, there seemed to be a very big difference, with the ~90* supers sounding much more direct (whether that's a good thing or bad thing in the final mix remains to be seen).  I was surprised how the 180* supers sounded on their own...not what I'd prefer to listen to, but also not totally unusable.

[heathen]

Some helpful info from Schoeps about OCT/OCT2:

When  one system  must  be used  for both  5.1 surround and 2.0  stereo  recording, downmix  compatibility becomes  a  prime  consideration.  OCT 2 offers  special  advantages in such  situations. OCT in general  has good  downmix  proper-ties  already,  since  a sound  from any given  direction  cannot  be  picked  up strongly  by all three  microphones  at the same time. Thus  level  differences,  rather  than  arrival-time  differences,  provide the primary  directional  cues,  and this minimizes  harmful  comb-filter  effects  in the direct  sound  when  mixing down  to two  channels.  The OCT 2  approach  improves  downmix  quality further with  its decreased  inter-channel  correlation of diffuse-field  pickup.

https://schoeps.de/fileadmin/user_upload/user_upload/Downloads/Kataloge_und_Broschueren/Anwenderbroschueren/SCHOEPS_surround-brochure.pdf

[heathen]

Sorry to keep hammering this thread, but I keep thinking about this stuff.  Based on my gorilla math, with the center mic 28 cm forward in the OCT2 array, that would put the center mic .8 milliseconds "ahead of" the axis of the side mics.  I didn't think this would be enough time difference to have an audible effect, but now I'm not so sure.  The delay between our ears is about .5 milliseconds, and that certainly has an audible effect in the sense that it impacts how we locate a sound.  So it would stand to reason that an even greater delay, i.e. .8 milliseconds, should have an audible effect, right?

[EmRR]

I'd say you can hear it if you were to A/B apples to apples, but whether it's good or bad is another question.  Since they are calling for 40 cm for OCT2, it would seem a matter of taste about the amount of correlation/de-correlation, with the hard and fast versions there as starter guides.   It likely makes a bigger difference if you are mixing for surround. 

FWIW, Decca Tree center mic is about 5 feet in front of the L and R mics.   Omni's with beamy directional pattern in the treble. 

[Gutbucket]

The delay between our ears is about .5 milliseconds, and that certainly has an audible effect in the sense that it impacts how we locate a sound.  So it would stand to reason that an even greater delay, i.e. .8 milliseconds, should have an audible effect, right?

Well yes, maybe.  Consider that this delay is not between Left and Right, but between center and sides.  So it's sort of a Mid/Side thing rather than a Left/Right thing.  Also there are three channel pair relationships to consider which are all superimposed upon one anther rather than just a single one.  We don't have just the left+right relationship, but the left+right, left+center, and center+right relationships all interacting.  That interaction is complex.  It is interacting in the electronic realm (mixing) in the air (speakers to ears) and perceptually in our ear/brain system. Taking just the perceptual part of that, the Left/Right microphone channel relationship remains fixed while the left/center and center/right relationships change on the move from OCT to OCT2.  Perhaps the left/right part perceptually "anchors things" leaving more freedom for the mid/side part to not be perfectly aligned. Or perhaps the Mid/Side difference is perceived as depth, or as first arrival presence emphasising "centeredness".

The OCT 2  approach  improves  downmix  quality further with  its decreased  inter-channel  correlation of diffuse-field  pickup.

^
Good to have my memory/presumption of what that is intended to achieve confirmed, thanks.

To step back for a moment for folks following along, correlation between adjacent channel pairs is desirable for sharp imaging, while low correlation ("increased decorrelation") is desirable across all channels for the ambient, diffuse-field sound arriving from all directions.  How to achieve both of those things at the same time, and to what extent one should go in order to achieve it is what is at issue. 

OCT2 goes to more a involved extreme to achieve this than OCT(1) does.  That's because OCT(1) was originally designed for 3 speaker Left/Center/Right playback where the center signal remains discrete and is not electronically mixed with the Left and Right channel signals until the signals pass through the speakers and "mix in the air of the room".  Phase interactions between correlated signals manifest more strongly when signals are mixed electronically, than when they mix together in air after being reproduced by speakers in a typical listening setup of speakers set some distance apart from each other in a non-anechoic room. 

To make this real-world, consider polarity-inversion.  Take a signal and duplicate it to two channels.  You have a 100% correlation between those two channels.  Invert polarity of one channel, mix them together without changing gain and you get total, or near total cancellation.  But play both signals thorough a stereo system and invert polarity ahead of one of the two speakers and you don't get total cancellation in the room.  You may notice attenuation of low frequency energy, because that remains more correlated in the room than higher frequencies.  This difference between mixing in the electronic realm verses in air is what OCT2 is intended to improve for mixing the 3-channel microphone feed down to a 2-channel stereo signal. To further extend this real-world example- If the signals in both channels are not correlated, they won't cancel when summed electronically, regardless of the polarity inversion of one channel. Think uncorrelated pink noise in two channels. It sounds the same when one channel is polarity inverted. If the two channels are correlated at low frequencies but not at high frequencies then only the low frequency part will be attenuated.

We can decorrelate the diffuse pickup by using a microphone setup geometry that introduces differences of pickup pattern + angle and/or by differences in time-of arrival (spacing).  Using microphones with a tighter directional pickup pattern combined with increasing the angle between the microphones will decrease inter-channel correlation of diffuse-field  pickup. But with 180 degree oriented L/R supercardioids, pattern and angle have already been maximised as much as possible, so increasing the spacing between the L/R pair and the center microphone is the only remaining way to do that using setup geometry.  OCT2 pushes the center microphone forward then introduces a delay to compensate.  The compensating delay realigns wavefront arrival for direct sound coming from the forward direction (it keeps the signal correlated for that angle of arrival).  For sound arriving from other directions it effectively increases time of arrival non-alignment and therefor decreases correlation. 

Can we do the same by changing the microphone setup geometry in some other way?  There are a couple options- we can push the L/R supercardioid pair wider, yet that will narrow the SRA.  We can move the center microphone up or down out of the horizontal plane, keeping time of arrival more or less the same for horizontal arriving sound, but increasing the distance between the center mic and all others for sound arriving from all directions other than the horizontal plane.  Neither of those are very practical, although I have seen some odd 3-channel arrangements using a standard near-spaced-stereo center pair plus a center microphone positioned a couple meters higher on the same stand.  The 3 positions remain more or less closely aligned for horizontal traveling plane waves, but not for sounds arriving from above or below the horizontal plane.

[Gutbucket]

I'd say you can hear it if you were to A/B apples to apples, but whether it's good or bad is another question.  Since they are calling for 40 cm for OCT2, it would seem a matter of taste about the amount of correlation/de-correlation, with the hard and fast versions there as starter guides.   It likely makes a bigger difference if you are mixing for surround. 

FWIW, Decca Tree center mic is about 5 feet in front of the L and R mics.   Omni's with beamy directional pattern in the treble.

Yes on mixing for surround, but that's the inverse of the problem.  I've never seen any mention of whether OCT2 is actually considered better for surround (or rather, 3 front channel playback) or not due to the increased decorrelation of diffuse pickup.  The OCT2 modification seems to be targeted fully at better optimizing the setup for mixing to 2 channel, without compromising discrete 3 channel L/C/R output. 

Probably not surprising, but for playback using 3 front speakers I do have a pretty strong preference for microphone setups which use three near-spaced microphone channels over those using three coincident channels or those those deriving a center.  For mixdown to 2-channel stereo I find I have a preference for a coincident center pair versus a discrete Left/Center/Right near-spaced pair (although that preference is far less strong and not a deal-breaker), mostly prefer some combination of both, and wouldn't want to give up the sideways facing pair.  That's why I've now implemented a M/S center full-time along with the near-spaced side-facing L/R supercards.  So ignoring the rear-facing channels, I'm doing a variation on OCT using a M/S center position and omnis spaced about twice as wide as the side-facing OCT supercards.


Decca Tree is not time compensated for alignment and uses larger spacing forward to the center position.  But it's also placed farther "into the orchestra" than most other main microphone arrangements, such that the orchestra wraps around the array, necessitating an Orchestra Angle of 180 degrees or more.  It's sort of an oddball in itself really.

[kindms]

RE array.

kind of interesting this discussing is happening now. i just added some clamps to our 15mm rail setup and a 90 degree. If i pick up a few more 12" sections we could probably find a place to run some funky forward experiments

basically a huge + right now would be 3ft split 1 foot section in the middle but that could easily be increased

[heathen]

I just did a quick mixdown of a song to compare no delay on the center mic with .8 milliseconds delay.  All I did was get the sides to equal level as the center, adjust the delay on the center (or not), and mix them together.  I'm curious if anyone hears differences between these.  Here's a download link: https://we.tl/t-G5iWimxWaY

I'll hold off on my own comments for the time being...

[EmRR]

They definitely sound different.  Think I prefer the non-delayed as being more forward-present in the image.  Delayed center sounds set back, cloudier by a hair.

It strikes me a delay is not the same as actual mic position, though it does give some idea about the spacing. 

[Gutbucket]

Thanks for making both samples available.  I'm listening at work through cheap earphones and find a slight preference for the delayed version, but really need to listen at home. [edit, upon further listening on the cheap earphones I find I prefer the non-delayed version due to the center imaging position being stronger, which better anchors the snare and high-hat in the center]. I may not be able to do so for a couple weeks though as I'm crushed at work in the short-term and will be gone the next two weekends, but I will definitely come back to this as it's been something I've wanted to get a better real-world grasp on on for a while now.

Last night, while thinking about re-arranging my OMT setup for recording this weekend, I decided to plug a few variables into the on-line Image Assistant app to help tweak my spacings, with the app set for 3 ch OCT - http://www.hauptmikrofon.de/stereo-surround/image-assistant.  It's been a while since I messed with that app and it's interesting playing with the out-front center microphone spacing.  I'm thinking you might want to mess with it a bit as well, Heathen.   

With the 25cm forward center supercard mic placement I've been using, the ~63cm L/R spacing, and no center delay, I actually cannot get the Left-Center and Center-Right segments to link smoothly unless I angle the L/R supercards forward +/- 45 degrees.  That's interesting.  Especially because I've been recording indoors for the most part recently and angling the L/R mics forward by that amount, and have been getting really good results that way. But the SRA is then something like 130 degrees total which seems over-wide.

I'm planning on setting the center M/S pair 5"/13cm forward this time while keeping the ~63cm L/R spacing the same.  With L/R supers angled +/- 80 degrees, which is about as much as I can angle forward when using the big Shure windscreens on them for outdoor recording (without also angling the entire telescopic arms forward with the omnis on the far ends), segment linking falls in line producing a 120-ish degree total SRA.  That's wider than I'd like, but to get a 90-ish degree SRA I'd need to push the L/R supercards out to 1m total.  I don't wish to do that this time, but might play with that if I can make it up to Suwannee Spring Reunion in March.  The amphitheater there has been my experimentation lab for this stuff - good and familiar.

If I then angle the L/R supercards +/-45 degrees, Image Assistant wants a 2m spacing between the L/R supers.  That seems extreme, is outside the suggested OCT range and likely outside the practical "good" envelope. However, if I push the center mic back out to 25cm again, L/R spacing drops back to about the 60-ish cm total spacing I've been using.  Hmmmm.

Tentative conclusion- When angling the the side-facing directional L/R pair forward, it's more reasonable to push the center microphone further forward by a small amount than the L/R pair wider by a large amount.  So I'm going to look into making the center microphone position adjustable using a short telescopic antenna the next time I re-rig things, so that when choose to I angle the L/R pair forward, I'll also push the center microphone farther forward at the same time.  I'm thinking I may end up with two configurations I can run for same pre-rigged setup and easily switch between without having to measure each time: 180-degree opposed L/R pair with less forward center extension (OCT), and +/-45 degree forward angled L/R with more forward center extension.  We'll see. 

I tend to like the omnis wider, regardless.