Topic: Side mic from two Core Sound Cardioid Stealthy mics?

[TheJez]

I don't have much choice anyway 

Actually you do.. or might, depending on the gear you have at hand.

My gearbox is rather limited. No pre-amp, no multitrack recorder. Just a very limited set of mics, a stereo recorder (FR-AV2), a soldering iron, near zero budget and some curiosity. I feel a bit like MacGyver here. 
It's been very educational so far already, even without having to heat-up the soldering iron yet! When I get near the electronics shop, I'll get the XLR-connector to start wiring things up!

[TheJez]

if you look at this as an exercise in understanding what is going on with mid-side processing., it's a great idea.  If you look at this as a way to make pleasing recordings, it's a lot of pain for a worse result.

I certainly don't plan to use this as a replacement for my normal concert taping! Unless I'm blown away with the results of this experiment, which is rather unlikely.

[Gutbucket]

Even if no one actually ends up doing it I think this discussion about how one might produce a DIY fig-8 pattern, along with exploration of the challenges involved and how they might be addressed, makes for a good TS thread.  It makes for a good exploration of the fundamentals involved.  Actually doing it with the explicit goal of useful recording will probably only suit someone who finds this stuff fun to mess around with and is sufficiently motivated, willing to do the initial setup calibration tests and the needed post recording work that's required to get it to work well.. which ideally means having an extra recording channel available.  Despite lacking that extra channel, TheJez seems sufficiently intrigued to give it a go, and may be fortunate if the mics he's using work well enough without additional corrections.  So hope remains in addition to the theoretical discussion.

If the primary goal were a high-quality Mid/Side recording rig at reasonable cost, I'd take a serious look at the Rycote BD-10 (fig-8) which I think go for around $800.  It seems a proper fig 8.  Or to play around with the idea at significantly less cost, the X8 remains available from Naiant at a small fraction of that price.  Anther path of reason is a dedicated Mid/Side microphone.  I was about to try the BD-10 myself a few years ago when ready to move on from the X8, but went with an integrated Mid/Side mic instead (AT BP4029, about the same cost new as BD-10), which fit my open recording requirements and was available a significantly less cost second-hand from another TS member. I've yet to actually incorporate any Mid/Side stuff into my stealth setups, but thinking about ways I might do that has been a driver for some of my more imaginative fig-8 riggings.

[Gutbucket]

Okay.. back to producing a fig-8 using two back to back cardioids.  Regardless of actually doing it for real or not.  Conceptually it works.  Getting it to work well practically and minimizing the hassle in doing so is the challenge.

We've talked about how a fig-8 pattern needs symmetry and a sufficiently deep center plane null, and how achieving that requires a good enough match between the two microphone outputs prior to summing them to a single channel. We've also experimentally determined that the mics TheJez is using might match close enough, which is fortunate since he's going to need to sum them into a single recorded channel prior to recording due to the restrictions on channel count.  If he were instead recording them to two separate channels, additional things could be done to get the level, frequency and phase responses to match as closely as possible.  A "matching EQ" type plugin might work really well for that.  Place the two mics immediately next to each together as in Method A. Record something, then use one channel as the target and have the matching EQ adjust the other channel to match.  Save those settings.  Then put the mics in the intended back-to-back configuration and make the desired recording.  Make sure not to accidentally swap mics or mic channels, and apply the matching EQ using that saved setting each time prior to the polarity flip and summing.  Use the resulting Side channel in the conversion from Mid/Side to L/R.

^ The same method can be used to improve the match of any pair where the response of the two mics differ somewhat.  It can improve the stereo recording quality that is achieved from a lesser quality pair of mics.  If the two mics sound audibly different, best to match the less good sounding mic to the better sounding one.

A close match is always good for any stereo technique, but is more important in some cases than others.  In this case its particularly important, because otherwise the shape and symmetry of the virtual fig-8 pattern that results from the sum will suffer spatial distortion.  This is very similar to optimizing an ambisonic mic, where all four of capsules must match very closely in order for the virtual polar patterns that result from conversion of the raw A-format output from the microphone to ambisonic B-format WXYZ or directly to L/R stereo, to accurately reflect the desired polar patterns.  Using that analogy, the raw output from the two cardioids is analogous to A-format from the ambisonic mic, and the Side channel which results from the post summing is analogous to the Y channel of B-format.  Mid/Side is analogous to B format, which is then converted to L/R stereo.

[Gutbucket]

A close match between channels isn't the only requirement.  Another is the physical geometry of the arrangement of the two mics. Here's a question I don't yet have a good answer for:  How close together do the two back to back cardioids really need to be?

Conceptually-
The closer together, the more consistent the resulting pattern at all frequencies.  Spacing between them will cause the fig-8 pattern to begin to break down above some frequency. The center null plane remains but the smooth fig-8 pattern pattern begins to go multi-lobe and spiky instead of forming a smooth two-lobe 8 shaped pattern.  Does that matter?  Maybe not so much at the highest frequencies where our directional hearing albitites aren't particularly astute.  The polarity inversion and Mid/Side matrixing will make whatever complex pattern that occurs up there behave symmetrically at least.  It may end up being more pseudo-stereo up there than a reflection of accurate image positioning, but that probably doesn't really matter as long as it sounds good.  There are different kinds of stereo as discussed earlier in the thread, perhaps that kind of more "PA like stereo" up top is good enough.  We can't achieve perfect coincidence, but only need to get it close enough.  How close does it need to be?

Practically-
Is the two inch spacing as achieved by TheJez's Lego mount (+T) close enough? I don't know. I do know that the outside dimension of the Naiant X8 which uses two back-to-back cardioid elements to form a fig-8 measures 34mm (1.33"), so the center-to-center spacing between it's diaphragms is slightly less than that.  I used Naiant X8 as a Side channel mic for years and was quite happy with it, although I always used that Mid/Side pair in arrays that included additional mics - a less demanding application than a single Mid/Side pair intended to be used on its own would be.  Two inches looks to be about the size of the Oktava MK-012 fig-8 adapter which combines two cardioid capsules to produce a fig-8.  I've not used that, but I think EmRR has and might be able to comment on it when used as Side channel.

[Gutbucket]

Let's back up a bit.  This is can be simpler than it seems.  Think of TheFez's back-to-back cardioid arrangement as an X/Y pair.. because that's essentially what it is..

[snip..] why not just record x/y @ 90* with the cards you have, decode from stereo to raw mid-side, then back to stereo with the desired m/s ratio.  I believe there are dsp plug-ins that will allow you to process stereo> m/s> stereo in one step.

Sure you can do that.  If you don't change the ratio of Mid to Side, what comes in goes out, and is essentially how vinyl record recording and playback and FM stereo transmission along with other things work.  We can take L/R stereo and convert it to sum/difference, then convert it back again.  The difference signal is equivalent to the Side channel of a Mid/Side pair. 

Things may get more complicated if we alter the ratio (level balance) between the Sum and Difference (Mid and Side) channels before converting back to L/R again.  It's then no longer a fully lossless conversion, and errors begin to accumulate.. from non-perfect coincidence, from a more extreme change of ratio.. and need to be listened for.  This is where the cautions morst and grawk mention about converting L/R to M/S and back come into play.

For any X/Y pair, the level of the difference signal varies with choice of polar pattern and X/Y angle.  Opening up the X/Y angle to 180 degrees, along with using a more directional pickup pattern, maximizes the difference signal as much as possible.   TheFez's back-to-back cardioid arrangement can be thought of as an X/Y pair which has been optimized for this particular role, from which the Mid content will be discarded, leaving only the Side channel information.

[kuba e]

Thank you for a very nice discussion. Also thanks to Gutbucket for the contributions—we completely forgot that it’s necessary to make sure the capsules are matched. If they have different sensitivities, it might be possible to solve it by adding a resistor to the more sensitive output. But that would be tailored to a specific preamp. Because the required resistor value is depended on the microphone’s output impedance and the preamp’s input impedance.

[TheJez]

A "matching EQ" type plugin might work really well for that.

Thanks for the tip! I could use this to verify how much the mics match (or mismatch) (or even to correct a mismatch...if I would have a recorder with more than two tracks )
Unfortunately, long working days prevent me from making speedy progress with this experiment.
So far for all the (very interesting and educational) theory; I finally got the the XLR connector and soldered the two signal lines (left and right) to the appropriate pins of the balanced input of the recorder! (and ground of course to the remaining pin ).  I walked around my living with my poor-man's mid/side mic and a few sound sources here and there (tv, stereo set) and a good isolating headphone connected to the recorder to monitor the result. It didn't sound bad at all and I had the impression that playing with the m/s balance on the recorder made a noticable difference in the spaciousness of the sound. So far so good. I do think (again) that the noise of the CoreSound mics might be the biggest spoilsport for this experiment...
Next step will be finding some nice spots to put the setup in practise. Hopefully the upcoming vacation period will provide some oppertunities and time. I'm thinking about the local ice rink full of screaming and shouting kids, the railway with passing trains, things like that. We'll see, I'll keep you posted! Thanks again to all for your support, especially Gutbucket! This is greatly appreciated.
Merry Christmas to all!

[Gutbucket]

Right on.  Let me know how the test recordings work out once you can make a few more.

A "matching EQ" type plugin might work really well for that.

Thanks for the tip! I could use this to verify how much the mics match (or mismatch) or even to correct a mismatch...

Yes, use of a good matching EQ routine to more closely align the response of two mis-matched mics that will be employed as a stereo pair is likely to improve the quality of a lot of recordings made with cheap mics!  I've long intended to start a thread here at TS about doing just that, but never got around to it.. so here it is!

Many tapers already EQ each channel to best effect by ear as needed.  This matching EQ technique would not eliminate the usefulness of doing that in more subtle ways for specific recordings as required.  Rather, it would be a base-line correction that would be applied to all recordings made with that particular pair of mics, prior to other post work.  In addition to matching frequency response it would also correct for differences in sensitivity.. so no need to run slightly different gains on the recorder to compensate for one mic being slightly hotter.  The initial response measurement of pink noise with the mics placed parallel and immediately next to each other need only be done once.  Likewise the matching routine only need be run once on those noise files and the resulting compensating EQ curve saved. Then just apply that curve each time, and be careful not to mix up the two mics or which channel each is assigned to.

[edit- To be clear, this is in no way intended to substitute for using properly matched high-quality stereo microphone pairs, yet should make for better-quality stereo recordings made using not-so-high-quality mis-matched mics.]