M/S-stereo setup

Discussion in 'rec.audio.pro' started by Ben Krämer, Aug 28, 2003.

  1. > This is embarrassing because my point was to insist that
    > the _other_ person be as clear as possible. No doubt this
    > illustrates some important spiritual principle about the risk
    > of going around challenging other folks...


    No, it just indicates that in the rush of posting a reply, a "slip of the brain"
    occurs.
  2. Mike Rivers

    Mike Rivers Guest

    In article <birov8$k84$1@news1.chem.utoronto.ca> reedijk@hera.med.utoronto.ca writes:

    > To have such microphones requires that
    > they (figure 8 and cardioid) have identical capsules in an identical
    > position with regard to the mic bodies. This is probably only possible
    > with multi-pattern mics. Multi-pattern mics tend to be large diaphragm
    > things with uneven polar frequency responses. So unless B&K makes
    > a small diaphragm side-address multi-pattern mic, I don't think this
    > concept can be be pratically implemented.


    So you're saying that M-S stereo can't work in practice? I don't think
    that's what you really mean.

    I suppose that in theory you might be correct (Bob Olhsson please
    insert the correct quote here) but somehow people still seem to be
    making acceptable recordings by using combinations of front and
    side-facing microphones. It's certainly possible to construct a poor
    implementation, but it's also possible to make something fairly simple
    work. "Work" here meaning that you can get a good recording, not that
    you can exactly duplicate another microphone setup (which may or may
    not be capable of making a good recording in that circumstance).

    > Alternatively, if you worked backwards, and took a pair of KM184s,
    > set them in X/Y, matrixed them to MS and generated a "figure 8" response
    > 90 degrees to the intersecting angle of the mics, just how useful
    > would it be? I have never tried this, but I imagine it would
    > be a fairly weird response.


    The mid microphone of an M-S setup has a useful output by itself -
    it's a well balanced mono recording of whatever's in front of it. The
    side microphone represents everything BUT what's in front of it. So
    while you can listen to the mid mic by itself (and if it doesn't sound
    darn good, it's in the wrong place) but there's little practical use
    for LISTENING to the side mic by itself.

    However, it's definitely possible to convert a Left-right stereo
    recording (whether it was done with a single X-Y pair or individual
    mics, pan pots, and boxes of reverb) to sum-and-difference components
    which represent what would come out of an M-S microphone. You can then
    manipulate these components as you would the outputs of a mid and side
    microphone and re-mix them to left-right in whatever proportion sounds
    right. This is a pretty common technique, and there are even mastering
    tools and techniques that employ this principle to reduce the effect
    of leakage between mics in a live multi-mic recording.

    Bottom line is that microphone selection and placement isn't automatic
    or something that you can do by formula or textbook. It's an art and a
    skill. Deciding what setup will work in a particular situation is
    something that's gained from working experience, not theory or
    reading. For making casual or documentary recordings, a single point
    stereo setup "where it looks right" will get you something that you
    can use. But if you want the best quality recording, you need
    experience. The more experience you have that you can apply - like
    working frequently in that hall with that kind of music - the quicker
    you can set up. But a first-time job with two mics can take at least
    as much time to get right as any mulit-mic setup.


    --
    I'm really Mike Rivers - (mrivers@d-and-d.com)
  3. David Satz

    David Satz Guest

    Rob Reedijk wrote:

    > The concept of adding and subtracting signals from capsules in one
    > geometric orientation in order to arrive at the equivalent siganal
    > of capsules in another orientation relies on your mics having
    > even polar responses in volume and frequency response. (Not really
    > volume, but you get my drift).


    Thanks for clarifying your earlier remarks. Now it's evident that you have
    a very specific objection that has a rational basis and can be discussed.
    It concerns the performance limitations of achievable M/S microphone setups
    and the question of what constitutes "acceptable" or "good" results, if I
    understand you correctly--and I think that by now, I do.

    I think that your objection can be narrowed down somewhat, but a certain
    core of it can't and shouldn't be dismissed because the imperfections of
    the microphones are still an audible factor. In other words, I not only
    respect what you're saying but even agree to some extent. (I have my own
    critique of M/S, but it's along different lines.)

    In this age of digital recording and DSP, the recording and matrixing
    processes are no longer the hazard to signal integrity that they once were.
    Back in the days of matrixing transformers and typically 15 ips analog tape
    (with its wobbly inter-channel phase at high frequencies), that was not yet
    the case. But nowadays the recording and matrixing processes are
    considerably more tractable than the physical/acoustical side of things.

    Maybe you've overstated things somewhat in your first sentence. In order
    for an M/S equivalent to exist for a given X/Y microphone setup, the
    microphones used for X and Y only need to have the same directional
    pattern as each other, whatever the practical quirks and limits of that
    pattern may be. The only point at which a strict demand is placed on
    the "perfection" of any microphone's pattern in this entire picture is
    the figure-8 microphone in the M/S pair--it must be truly symmetrical
    at all frequencies of interest.

    In practice the deviations from ideal directional patterns work in such
    a way that they generally tend to bring X/Y and M/S into somewhat greater
    convergence with one another. It's not a random situation; most real-world
    microphones deviate from their nominal pattern, but those deviations follow
    certain typical "patterns" of their own. For example, cardioids lose their
    directivity at both ends of the frequency spectrum; pressure transducers
    ("pure" omnis) become directional at high frequencies; pure pressure
    gradient transducers roll off at both low and high frequencies. Details
    vary, but these general themes are always present.

    Because all first-order directional patterns are some mixture of the two
    basic types of response (pressure and pressure-gradient), then these main,
    typical imperfections of the X/Y microphones' pattern will be duplicated
    in the corresponding "M" microphone's response once you conjure up the
    equivalent M/S pair, and vice versa.

    For example, the M/S equivalent of a pair of X/Y cardioids would generally
    be a forward-facing omni plus, of course, the side-facing figure-8. But a
    cardioid results from a 50/50 mixture of pressure response and pressure
    gradient response. Thus the typical low-frequency loss of directionality
    in the X/Y cardioids is matched by the fact that the "S" microphone will
    have a low-frequency roll-off while the "M" (omni) will not. So after the
    matrix, the M/S pickup will yield a decreasing separation between channels
    as the frequency goes lower--which is just what occurs with X/Y cardioids.

    This is not an isolated example, but rather the norm. No microphone is
    perfect, but the primary and (currently) unavoidable imperfections of
    real-world microphones work such that real-world M/S and X/Y setups tend
    to produce more similar results to each other, not less similar. And
    that was one of the two main Q.E.D.s of this debate, as I understand it.

    The only assumption of "pattern perfection" that I'm making here is that
    the figure-8 in the M/S pair should have directional response that is as
    purely symmetrical as possible. It simplifies things if its frequency
    response is also flat, but that isn't necessary to the "QED".


    > To have such microphones requires that they (figure 8 and cardioid)
    > have identical capsules in an identical position with regard to the
    > mic bodies. This is probably only possible with multi-pattern mics.
    > Multi-pattern mics tend to be large diaphragm things with uneven polar
    > frequency responses. So unless B&K makes a small diaphragm side-address
    > multi-pattern mic, I don't think this concept can be be pratically
    > implemented.


    Again I think that you've overstated things somewhat; you've taken what is
    basically a high-frequency problem and generalized it to the entire range.
    Those high frequency issues are unsolved problems, but they occur on both
    the X/Y and the M/S sides of the universe.

    Remember that this is (or at least was) about the ability to transform M/S
    to X/Y or vice versa in the real world. Again I'm saying that the real-world
    limit of our ability to bring good microphone capsules into true coincidence
    tends to bring the two approaches into greater, not lesser, convergence on
    a practical level.

    On a side note, B & K of course has no monopoly on any principles of
    physics, and there have been small-diaphragm multi-pattern microphones
    with good polar patterns for many years. These tend to be single-diaphragm
    designs with mechanical pattern switching rather than the Neumann and AKG
    approach with twin diaphragms. Unfortunately the Schoeps approach (that's
    whose multi-pattern microphones I'm saying have the best polar patterns)
    doesn't support continuously variable directivity--the capsules have two
    or three definite pattern settings that they "click" into.


    > Alternatively, if you worked backwards, and took a pair of KM184s,
    > set them in X/Y, matrixed them to MS and generated a "figure 8" response
    > 90 degrees to the intersecting angle of the mics, just how useful
    > would it be? I have never tried this, but I imagine it would
    > be a fairly weird response.


    It would depend on the distance between the diaphragms and yes, at high
    frequencies things would definitely take on some aspects of strangeness.

    You might be interested to know that for many years Neumann offered a
    microphone very much like what you're describing. The KM 86 contained
    two complete capsules of the same type as the KM 84 (actually it was first
    introduced in the tube era as the KM 66, so it was the capsules from the
    KM 64 or U 64); these were not just the membranes and backplates, but two
    of the entire capsule head assembly with just a shorter contact pin coming
    out of the back of each.

    There was an appreciable spacing between the diaphragms in this arrangement.
    That didn't matter in the cardioid setting, in which only the front capsule
    was polarized. Regarding its figure-8 performance, the lack of coincidence
    disturbed the high-frequency polar pattern (it's peaky off axis) but helped
    significantly with the low-frequency response. Neumann's Web site has some
    product documentation for the KM 86, including response curves, on:

    http://www.neumann.com/infopool/history/produkte.php?ProdID=km86

    (choose "Documents" from the tiny horizontal menu near the top of the page)
    and there's a detailed and interesting page about the KM 66, with more
    information on its design, on:

    http://www.neumann.com/infopool/history/produkte.php?ProdID=km66

    (choose "Description" from the menu).


    --Best regards
  4. ScotFraser

    ScotFraser Guest

    << Multi-pattern mics tend to be large diaphragm
    things with uneven polar frequency responses. So unless B&K makes
    a small diaphragm side-address multi-pattern mic, I don't think this
    concept can be be pratically implemented. >>

    Neumann, Schoeps & AKG have all made small diaphragm multipattern mics.


    Scott Fraser
  5. Sugarite

    Sugarite Guest

    > This is the way I have always conceptualised MS. Perhaps I am incorrect.

    I've read ahead, and you are correct, but you try to justify conceptual
    issues rather than just listen and appreciate what's going on.

    > The concept of adding and subtracting signals from capsules in one
    > geometric orientation in order to arrive at the equivalent siganal
    > of capsules in another orientation relies on your mics having
    > even polar responses in volume and frequency response. (Not really
    > volume, but you get my drift).


    All coincident stereo mic formats represent the sound pressure levels
    transpiring at one location broken down into two component vectors,
    generally left and right (XY) or mono and side (MS). Whether the capsules
    are matched, mis-matched, half-shot, or siamese twins, the effect is to be
    observed and appreciated, not contemplated and discarded.

    > To have such microphones requires that
    > they (figure 8 and cardioid) have identical capsules in an identical
    > position with regard to the mic bodies.


    To me that's like saying all guitars must have 6 strings, tuned E-A-D-G-B-E.

    > This is probably only possible
    > with multi-pattern mics. Multi-pattern mics tend to be large diaphragm
    > things with uneven polar frequency responses. So unless B&K makes
    > a small diaphragm side-address multi-pattern mic, I don't think this
    > concept can be be pratically implemented.


    Pity that seems to prevent you from trying it with the available mics and
    hearing what happens.

    > Alternatively, if you worked backwards, and took a pair of KM184s,
    > set them in X/Y, matrixed them to MS and generated a "figure 8" response
    > 90 degrees to the intersecting angle of the mics, just how useful
    > would it be? I have never tried this, but I imagine it would
    > be a fairly weird response.


    I have tried it, and believe it or not, it sounds *exactly* like a pair of
    KM184's in XY format matrixed to a "figure 8" response. Weird, eh?

    Fortunately, we're recording for an *audience*, not a team of scientists.
    "Usefulness" is not determined by the recording technique, but by the
    impression that can be made through recording technique.
  6. > The concept of adding and subtracting signals from capsules in one
    > geometric orientation in order to arrive at the equivalent signal
    > of capsules in another orientation relies on your mics having
    > even polar responses in amplitude and frequency response.


    You're conveniently overlooking the fact that discrete cardioid mics don't have
    even polar or amplitude responses either. It makes no sense to criticize M-S
    miking for the same problems that exist with other forms of miking.
  7. James Boyk

    James Boyk Guest

    MS has the obvious disadvantage of requiring more processing. So what's it's advantage? That one of the mikes is facing front center, where in many cases one or more important sources will be. Since mikes with diaphragm dimensions of appreciable size--that
    is, all studio condenser mikes--will have off-axis anomalies at high frequencies, the point of MS is that at least the M mike will be handling those center sources pretty well since it's facing them directly.


    If one uses mikes without such anomalies, which in practice means ribbons, one can still do MS; but its "advantage" isn't needed. This leaves you with only the disadvantage of extra processing. So a possible conclusion is:


    MS goes with condensers. XY goes with ribbons (and preferably 90-coinc ribbons; that is, "Blumlein pair" miking.)


    One reason, I would guess, for much negative opinion about coincident miking among engineers is that they've tried condensers in XY and the result was lousy. Well, it is, most of the time, especially if the mikes were large-diaphragm. But even with 1/2" di
    aphragms and even with Blumlein miking (90-coincident 8's), one still misses having a "palpable" quality at front center. With ribbons, suddenly all is well.


    There's one situation when it would make sense to use MS even with mikes not having off-axis anomalies, but I don't think anyone actually does it. This is in the FM broadcast chain, where M and S would *not* be matrixed to XY but passed directly through to
    the broadcast sum (M) and difference (S) signals. This would skip the matrixing normally present in the broadcast chain. The only matrixing would be what's always done anyway, automatically, by the listener's FM receiver.


    Finally, if one doesn't hear the advantages of good coincident miking---not only in imaging, but in integrity of attacks and absence of comb-filter alterations to frequency-response---I can't imagine how to convince one of it.



    James Boyk
  8. James Boyk wrote...

    > Finally, if one doesn't hear the advantages of good
    > coincident miking---not only in imaging, but in integrity
    > of attacks and absence of comb-filter alterations to
    > frequency-response---I can't imagine how to convince
    > one of it.


    It's odd you mention comb-filtering effects, because one of the problems with
    coincident miking is that, as it does not generate the "correct" psychoacoustic
    cues, in playback the ears hear comb filtering that make the sound seem more
    distant and reverberant than it does "live" at the mic position.

    These effects are one of the reasons (another of which is the ability to
    arbitrarily adjust the balances) multi-miking is common, despite the fact it
    ruins the overall integrity of the sound.
  9. David Satz

    David Satz Guest

    James Boyk wrote (in long lines of text that scrolled off the edge of the
    display page--please check your software settings!):

    > [T]he point of MS is that at least the M mike will be handling those
    > center sources pretty well since it's facing them directly.


    This is a common misunderstanding. True, the mono "M" pickup is from a
    front-facing microphone, so its off-axis anomalies will be inflicted
    mainly on the sound sources that are away from center--depending on the
    degree of direct vs. reverberant sound at the microphone's location.

    But stereo L/R signals result from the matrixed sum and difference of this
    "M" pickup with the "S" microphone--and front/center is exactly where the
    "S" mike has its null. The frequency response at and around the null of
    any directional microphone (including your beloved ribbons--tweak, tweak) is
    normally the most ragged response that the microphone has. So at the very
    angles of incidence where the "M" microphone has its smoothest response,
    the "S" microphone has its least smooth response and vice versa.

    For the stereo signal, an M/S pickup recreates the polar response defects
    of the corresponding X/Y pair by indirect means. When you think about it,
    it could hardly be otherwise--the "Law of Conservation of Goodness" applies
    here as it does everywhere else. Otherwise known as, "there's no such
    thing as a free lunch--but if there were, it would be something that you
    don't want to eat."


    > One reason, I would guess, for much negative opinion about coincident
    > miking among engineers is that they've tried condensers in XY and the
    > result was lousy.


    Seemingly a valid reason for a negative opinion, I would think <g> ...

    Let's not forget also that many engineers (and non-engineer listeners)
    prefer a stronger sense of spatiality and envelopment than coincident
    stereo pickup methods (X/Y and M/S) generally provide. Sometimes this
    situation can be helped by boosting the low-frequency signals from the
    "S" microphone of an M/S pair, or its virtual equivalent; sometimes not.

    Really good-sounding coincident stereo pickups require really good hall
    acoustics, and there just seems to be no way around that.

    --I'm arranging to have measurements made with some ribbon microphones,
    since I frankly don't believe the hype about their supposed total lack of
    off-axis response peaks. You might think that the people making the claims
    in such extreme terms would be the ones offering the evidence, but ... no.

    In the meantime, I can say for sure that among condenser microphones there
    is fairly wide variation in off-axis response, with capsule diameter being
    a large factor but certainly not the only important one. Plenty of small-
    diaphragm mikes have what's technically known as "crummy" off-axis response,
    for example.

    --best regards
  10. Scott Dorsey

    Scott Dorsey Guest

    Rob Reedijk <reedijk@hera.med.utoronto.ca> wrote:
    >Alternatively, if you worked backwards, and took a pair of KM184s,
    >set them in X/Y, matrixed them to MS and generated a "figure 8" response
    >90 degrees to the intersecting angle of the mics, just how useful
    >would it be? I have never tried this, but I imagine it would
    >be a fairly weird response.


    No, it works very well and it is just as effective at pattern control
    as an M-S mike. Try it!
    --scott

    --
    "C'est un Nagra. C'est suisse, et tres, tres precis."
  11. James Boyk

    James Boyk Guest

    David Satz wrote:

    > ...True, the mono "M" pickup is from a front-facing microphone,

    so its off-axis anomalies will be inflicted mainly on the sound
    sources that are away from center.... But ...front/center is
    exactly where the "S" mike has its null. ...the null of any
    directional microphone... is normally the most ragged response....
    So at the very angles of incidence where the "M" microphone has its
    smoothest response, the "S" microphone has its least smooth
    response and vice versa.


    Yes; but the *level* of the S signal is way way down there; so the M signal dominates and its smoothness (for front center sources) should prevail.


    (Actually, your "vice versa" does not apply if M is a cardioid, as in the classic MS setup.)




    > Really good-sounding coincident stereo pickups require really good hall acoustics, and there just seems to be no way around that.



    Absolutely. But good acoustics are also a requirement for fine playing by the musicians; so we need them anyway. (It's just too bad that L.A. doesn't have any, except for one small hall and--I hope--the new Disney Hall.)



    > --I'm arranging to have measurements made with some ribbon microphones,
    > since I frankly don't believe the hype about their supposed total lack of
    > off-axis response peaks. You might think that the people making the claims
    > in such extreme terms would be the ones offering the evidence, but ... no.


    Of course it depends on the mike. The original BBC research report on the design of what became the Coles 4038 gives the curves for that mike. I personally haven't seen any 'hype' about lack of off-axis peaks in ribbons... There was an article in 1957 or
    so by B&O's engineer who designed the B&O ribbons, about off-axis *phase anomalies* in condenser mikes. You can hear these in the thin sound of applause on condenser rec'gs.


    James Boyk
  12. Mike Rivers

    Mike Rivers Guest

    In article <bj0rhd$3p0$1@naig.caltech.edu> boyk@caltech.edu writes:

    > There's one situation when it would make sense to use MS even with mikes not
    > having off-axis anomalies, but I don't think anyone actually does it. This is
    > in the FM broadcast chain, where M and S would *not* be matrixed to XY but
    > passed directly through
    > to the broadcast sum (M) and difference (S) signals. This would skip the
    > matrixing normally present in the broadcast chain. The only matrixing would be
    > what's always done anyway, automatically, by the listener's FM receiver.


    That's an interesting thought. I wonder if it works? Bob Orban, got
    your FM receiver tuned in?

    M-S is really good at capturing ambience, often too good. One solution
    to this that was used by a few classical engineers is what's been
    dubbed "Double M-S" by its innovators, Neil Muncy and Curt Wittig.
    They conceived of the idea and experimented enough to figure out how
    to make it work.

    The technique involves mixing two M-S setups. With conventional M-S
    recording, the mic pair is positioned to get the best compromise
    between image of what you're recording and room ambience. In double
    M-S, one pair is placed close enough to get the desired spread and
    imaging without regard to the sound of the room (which rarely will be
    as you want). A second pair is placed out in the room, beyond the
    critical distance (where the reverberant sound dominates the direct
    sound) and is used to add the desired amount of ambience.

    Neil and Curt's development of the technique was during the late
    1980's when we didn't have phase-accurate multitrack recorders that we
    have today, so everything they did with this technique was mixed live
    to two tracks, much of it recorded on a VCR using a PCM adapter. Being
    able to record the output of each mic individually and mix in a better
    listening environment could bring this technique back to life among
    new users.


    --
    I'm really Mike Rivers - (mrivers@d-and-d.com)
  13. James Boyk

    James Boyk Guest

    Mike Rivers wrote:
    > That's an interesting thought. I wonder if it works?



    I sure don't see any reason for it not to work. It's just doing exactly what's done already, but skipping some processing. How can this be anything but good for the sound. (I only wish FM sound was good enough that we could hope to appreciate such subtle i
    mprovements. Here in LA, we don't seem to have any stations that understand what musical sound is about. They're only about "advertising" themselves to the largest audience. Even at evening prime listening hours, everything is conditioned on the idea that
    you're listening in your car or other high-noise environment. You'd think they could have 2 hours per day of WIDE dynamic range. This would mean running things about 20 dB lower than they usually do. (I've watched the limiters in the control room while pla
    ying my recordings; and at 15 dB lower than usual, they were still kicking down all the time.) What a pleasure it would be, though. And if this were combined with LIVE broadcasting...Heaven! But do we get this? No. We get
    full orchestra blowing its guts out, followed by announcer 10 dB LOUDER!



    > M-S is really good at capturing ambience, often too good. One solution
    > to this that was used by a few classical engineers is what's been
    > dubbed "Double M-S".... The technique involves mixing two M-S setups...

    one pair ...placed close enough to get the desired spread and imaging
    without regard to the sound of the room... [and the] second pair ...placed
    out in the room....


    To whatever degree the 2nd pair is mixed in, it will compromise the virtues of coincident miking. It might be a worthwhile compromise sometimes; but still....



    James Boyk
  14. Bob Cain

    Bob Cain Guest

    Scott Dorsey wrote:
    >
    > Rob Reedijk <reedijk@hera.med.utoronto.ca> wrote:
    > >Alternatively, if you worked backwards, and took a pair of KM184s,
    > >set them in X/Y, matrixed them to MS and generated a "figure 8" response
    > >90 degrees to the intersecting angle of the mics, just how useful
    > >would it be? I have never tried this, but I imagine it would
    > >be a fairly weird response.

    >
    > No, it works very well and it is just as effective at pattern control
    > as an M-S mike. Try it!


    Absolutely agreed. A case where theory and practice nicely
    coincide. A quick way of checking it out is with the
    Stereoimager plugin from Waves should anyone have that. It
    applies this transform as well as others including
    shuffling.


    Bob
    --

    "Things should be described as simply as possible, but no
    simpler."

    A. Einstein

Share This Page