Why isn't phase cancellation a problem in M/S?

[ScienceOne]

Why doesn't the polarity-flipped mult track for the figure 8 mic cancel out some of the omni's signal? Or does it?

[midiot]

Why doesn't the polarity-flipped mult track for the figure 8 mic cancel out some of the omni's signal? Or does it?

Isn't M/S a fig-8 and a cardioid?

[gyraf]

It is.

And the phase cancellation is what actually makes the stereo difference.

Simple as that.

[Punkity]

Actually, M/S recording can be done with anything for a mid mic. With an omni there the system acts sort of as if there were two cardioids. A cardioid there is equivalent to two hypercardioids, a wide cardioid=two cardioids (not in the same way an omni yields two cardioids), another figure-8 gives you a blumlein, etc..

The logic of what happens with a M/S set-up takes a little while to get your head around, but if you read enough about it and look at enough diagrams the fog should clear.

[ScienceOne]

Well I guess my questions deals more with levels. Say your polarity-flipped channel is the one on the far right. Wouldn't that side be quieter than the left because everywhere that the left mic is summing with the center mic the right mic would be negating?

[KennyLusk]

Here's a decent article on MS mic'ing. The article is further down the page. I just bought on eof these MS mic kits from PAiA because for $54 you can't beat having the matrixing done for you in hardware.

Further down the page is an article with some details about making MS work.

http://www.paia.com/msmic.htm

[Professor]

It seems the confusion you're having is around the Mid+Side vs. the Mid-Side that are used to create the resulting left and right signals. Remember that when you see a graph of a fig-8 pick-up pattern that the 0? side is considered positive while the 180? side is considered negative. So you take the cardioid center and add one lobe of the fig-8 while subtracting the other which yields another cardioid turned 45? to one side. When you invert the fig-8 on the other side, the 0? lobe becomes negative while the other is positive. When that is added to the cardioid mid, again one lobe is added while the other is subtracted but the lobes are reversed and so the resulting cardioid is turned 45? the other way.
So a cardioid mid plus & minus the fig-8 side yields a 90? XY pair, but with the added bonus of being able to vary the gain of the mid vs. side mics and so adjust the angle after the recording. So, in a slight advantage over a simple XY pair, the MS pair can swing from a narrower, say 60? image to a wider, perhaps 120? image. If the signal is all mid, it is mono at 0? while all side yields, well just the side mic. If the mixed down signal is summed to mono, in an AM radio broadcast for example, the positive and negative fig-8 mic is cancelled out and the result is simply the mono center.

As for the choice of center, I'll have to double check the math, but I'm fairly certain that a cardioid center yields two cardioid resultant channels. Omni in the middle yields two wide cardioids. Hyper yields two hypers. And the one I find most intriguing is that a fig-8 mid mic yields two fig-8s, basically taking the mid side array and rotating it 45?. But since both mics are fig 8, either one can be considered the 'mid' and therefore the pattern can be rotated 45? in either direction.
Where this can be very useful is in a live recording of a stage that will be setup differently for several groups, but there is only one engineer recording. Simply setup a Blumein pair at the front-center of the stage. If a particular setup is placed too far to the left, run the pair through an MS decoding matrix back in the studio with the left mic as the center, and the stereo image rotates to capture the group. If another ensemble is setup too far right, just reverse the MS decode so the right mic is the 'MID' and the image rotates the other way.

There is also a surround technique that someone developed that utilizes three capsules in an MS array. The typical fig-8 side mic, but with a forward-facing cardioid mid and a rear-facing cardioid mid. Take the two mids, sum them to mono and they yield an omni that can then be decoded to a wide-cardioid stereo. Take the two mids and sum them with the rear-facing mic phase-reversed and they yield a fig-8 mid that can then be decoded. For surround, they take the front-mid as the center, the front-mid decoded with the side as a L&R Front pair, and the rear-mid as decoded with the side as a L&R rear pair. The whole system can be summed all the way down to mono without much cancellation. I believe that Schoeps markets the whole setup as a single package.

If you're still having trouble wrapping your noodle around the phasing, I suggest printing out a couple of polar pattern charts (just visit a mic maker website like www.audiotechnica.com and download the .jpg files of their mic patterns) and actually do the math on the additions. Remember that the polar pattern graph is simply a graph that shows you the relative strength at different angles. Read the value for a cardioid at 0? and fig-8 at 90? (the 'side' mic), add them together and divide by 2 (for the sake of size on the graph) and plot it on a clean polar graph (just draw one). Now work your way around the plot at 30, 60, 90, 120, etc. and you will see the resulting pattern emerge. Remember that the back side of the fig-8 is negative and it will all work out. Then you can 'reverse the phase' of the fig-8 so the front side is negative and draw the result again. This is the fastest way to understand what is happening when you combine coincident mics.

-Jeremy

[Punkity]

Well I guess my questions deals more with levels. Say your polarity-flipped channel is the one on the far right. Wouldn't that side be quieter than the left because everywhere that the left mic is summing with the center mic the right mic would be negating?

Ahhh yes.

I know that you know alot of what is comming up, but I don't know what you know, and I think reviewing the big picture many times works for understanding M/S micing.


A figure-8 has two lobes. One has the opposite polarity of the other. This is hardwired into every figure-8 mic. If the two lobes had the same polarity, the mic wouldn't have a figure-8 but an omnidirectional pattern instead.


With a M/S matrix, the signal from the fig-8 mic is split and one of the signals has its polarity reversed. Now we have two signals: one where the right hand lobe is nominally positive and the left hand lobe is nominally negative, just like a normal signal from a fig-8, but the other signal reverses this situation and the right hand lobe is now nominally negative and the left hand lobe is nominally positive.

The mid signal entering the M/S matrix uneffected and is always positive (except if the mid mic is hyper/super cardioid or fig-8, but don't worry about those special cases right now).

When we mix a signal from the mid microphone with these signals from the fig-8 mic, the mid signal reacts differently with each:

--With the non-reversed polarity signal, all the sound picked up from both the mid mic and the right hand lobe of the fig-8 mic is reinforced (remember the right hand in this case is nominally positive and the mid is also positive), while all the sound picked up from the mid mic and the left hand lobe is attenuated (the left hand lobe is nomally negative while the mid is positive). All this stuff comes out the right speaker.

--With the reversed polarity signal the situation is the reversed. All the sound picked up from both the mid mic and the right hand lobe of the fig-8 mic is attenuated (now the right hand in is nominally negative and the mid is still positive), while all the sound picked up from the mid mic and the left hand lobe is reinforced (the left hand lobe is now positive while the mid is positive). All this stuff comes out the left speaker.

Clear as mud? Sorry if I get a little pedantic but a degree in philosophy will do that to a person.

One of the things I think is getting you hung up is that there is a big difference between polarity and phase.

[Punkity]

Dang, the Professor beat me to it.

[Professor]

It's all good - obviously we were typing at the same time. And it never hurst to read the explanations given by several folks, expecially when they're saying the same thing.

-J

[Punkity]

Professor, I'm pretty sure an omni will yield the same as back to back cardioids. My thinking is since the ideal omni pattern=1 and the ideal fig-8 pattern = cos A, then with and omni mid, you are going to be having a situation where the patterns = 1-cos A, which is the definition of the ideal cardioid pattern. (Boy I hope my memory is working today). This is just quick and sloppy thinking; however, if this is the case, there is no way a cardioid mid can yield a virtual pair of cardioid mics.

Anyway, I'm signing off for the day.

Cheers.

[ScienceOne]

Woah. You guys are awesome. I guess I really get it now. What book or class did you pick up to learn how trig applies to mic placement? If it's a book, where can I get it???

[Professor]

The only textbook I can think of is Ron Streicher's Stereo Handbook, but I don't have a copy handy to verify.
The sum & difference question has been bugging me though, so I went to the most authoritative source I could think of - Wes Dooley & Ron Streicher's 1982 AES article on M-S Stereo (which ScienceOne should find in his Hotmail inbox).
According to the experts:
When an omni is your Mid microphone, you will always decode a pair spread 180? wide or back-to-back. If you decode with an equal gain on mid/side or 50/50, you will get two, back-to-back cardioids. If the signal is 70/30 mid/side then you will get back-to-back wide-cardioids. If the signal is 30/70 mid/side, then you get back-to-back hypercardioids. But as noted, the angle is always 180? with an omni center.

With a cardioid middle mic, it seems we were both right. At a 50/50 mid to side ratio with a cardioid mid, you will get a pair of hypercardioids spread 126.9? wide. But if you swing that ratio to 70/30 mid/side you get a narrower 81.2? angle of nearly perfect cardioids. And if you swing the ratio to 30/70 mid/side you get a 155.8? angle on mics that are partway between hyper & fig-8.

The only arrangement that yields a consistent polar pattern is the fig-8 mid plus fig-8 side. When this is decode 50/50 mid/side, you get a pair of fig-8s spread 90?, and as you increase the mid mic, the image narrows while if you decrease the mid, the image widens.

And for what it's worth, the polar patterns are:
omni = 1
fig8 = cosA
cardioid = 0.5 (1 + cosA)
hyper-card = 0.25 (1 + 3cosA)
super-card = 0.37 + 0.63cosA


-Jeremy

[stuntbutt]

Blumlein and MS with a figure 8 as the Mid mic are different deals. Actually, I'm not sure that MS with a figure 8 as the Mid is a good idea. Wouldn't that make any sound coming from the rear appear on the wrong side in the stereo field?

[Scodiddly]

A really good book on microphones (and which happened to be owned by my local library! ) is John Eargle's "The Microphone Book". Great discussion (with some math) of various pickup patterns and a lot on stereo techniques, as well as some really intriguing chapters on things like linear arrays of omnis.