From a recording blog:
M-S is great for stereo imaging, especially when most of the sound is coming from the center of the ensemble. Because of this, it’s less effective on large groups, favoring the middle voices that the mics are closest to.
Its not true stereo. If you record something hard left it will replay at the same level in both speakers , but one will be out of phase, this will not sound hard left. What MS excells at is capturing the room, and allowing it to be adjusted in the mix. From this I would think that using MS technique for playback will give you similar results, anything hard panned will just sound phasey.
M-S is great for stereo imaging, especially when most of the sound is coming from the center of the ensemble. Because of this, it’s less effective on large groups, favoring the middle voices that the mics are closest to.
Its not true stereo. If you record something hard left it will replay at the same level in both speakers , but one will be out of phase, this will not sound hard left. What MS excells at is capturing the room, and allowing it to be adjusted in the mix. From this I would think that using MS technique for playback will give you similar results, anything hard panned will just sound phasey.
Going from normal wide angle stereo with say 2m speaker separation to a narrow angle setup with say 20cm speaker separation (if practically feasible) will require up to 20dB LF boost to get back to the S-channel output that the wide angle setup is capable of.
Further to that, the restricted output at LF in most recordings is simply because of dipole attenuation and its compensation can provide for reassessment of what stereo is capable of at such frequencies (if the dynamic range is available).
This is just not true. Please see my earlier post on MS encoding and what the so-called "MS speaker" turns out to be.
No. This confuses LR and MS signals. S is not equal L, just as -S is not equal to R.
ANY properly applied coincident microphone technique can produce "great" stereo imaging. There are a number of means to produce MS encoded signals - or LR which are a simple, linear transformation thereof. Sometimes engineers opt for different microphone arrangements because they exploit certain microphone characteristics such as reducing HF directivity, sometimes they find arrangements that are not strict MS-encoding but which better fit the situation (such as attenuating sounds at the rear). But MS and LR are simply different representations of the same signals.
I did wonder if this would come up.
What you've omitted is the "mid" microphone.
Here goes:
We'll take the example with cardioid mid and dipole side mics.
We'll set those mics to have a flat response and equal sensitivity, and we'll listen to them at equal gain.
When a source at 45 degrees to the left of the array makes a sound, here's what happens...
Left channel: both sources combine equally. Due to the mic polar patterns, each one will be about 3dB down at that angle. Still, they're coincident mics and will combine well, meaning voltage doubles. Two sources combined well add 6dB, so the left channel will have that source at +3dB.
Right channel: The "side" signal is inverted, and cancels perfectly with the "mid" signal. The right speaker remains silent.
If you follow this with the angle set to 90 degrees, the mid mic will be at about -6dB, and the side mic will be at 0dB. The right side will have a small amount of inverted signal.
Now, you can alter the "side" gain so that you can choose the angle at which the speakers are silent - it's a consequence of the maths behind all this.
Having listened to plenty of mid-side recordings (and made a few), I can say for a fact that it is true stereo.
Chris
I agree that M/S can differentiate L/R, and width, because of the anchorpoint M.
M/S:
..M
S|-S
'True' stereo
L..|R
-R.|-L
Correct me if I'm wrong but I think stereo adds another dimension, because the L/R anchorpoint is not fixed in the depth dimension, like M/S.
Explaining 'stereo depth of field' is no magic if M/S can already explain width and L/R location cues IMO.
M/S:
..M
S|-S
'True' stereo
L..|R
-R.|-L
Correct me if I'm wrong but I think stereo adds another dimension, because the L/R anchorpoint is not fixed in the depth dimension, like M/S.
Explaining 'stereo depth of field' is no magic if M/S can already explain width and L/R location cues IMO.
Last edited:
Given two microphones to record, and two speakers for playback, how do you think "depth" could be transmitted to the listener?
Serious question.
Is depth a function of phase & level differences between the speakers?
ie, any coincident mic setup is faulty because the stereo image is only level based.
Chris
Serious question.
Is depth a function of phase & level differences between the speakers?
ie, any coincident mic setup is faulty because the stereo image is only level based.
Chris
I am correcting you
There is no basis for any such statement whatsoever. LR and MS are linear transformations of one another. MS is merely more amenable to useful processing. But the two representations are entirely equivalent.No, you have this the wrong way round. Coincident mics always give better depth because they better record the patterns in time from early reflections that portray depth. This information is smeared with non-coincident techniques.
Microphone spacing produces instead the lateral "phasiness" akin to a more reverberant field. It is an artefact of a lack of fidelity in a recording, but favoured by some.
Just a quick note to say that I got the speaker up-and-running, with 2x Tang Band mini-subwoofers and 2x Bose 2.5" drivers. 400Hz 4th order crossover and a lot of LF boost. Got the figure-of-8 speaker sounding decent, and then fired up the "mid" speaker.
The result wasn't great, but I think I know why: microphones are small, so they can be physically very close together. That means they're coincident over most of the frequency range, and the M/S <=> L/R stereo transform can work.
Speakers, however, are not small enough to pull off that particular trick.
I suspect that a large investment in very thin planar speakers might produce interesting results (I'm thinking of the B&G Neo series). I think having them mounted in a T-shape (ie, a dipole source just in front of a monopole) might work.
So, I'd consider this experiment to be a failure this time, but at least now there's direction for further study.
Chris
The result wasn't great, but I think I know why: microphones are small, so they can be physically very close together. That means they're coincident over most of the frequency range, and the M/S <=> L/R stereo transform can work.
Speakers, however, are not small enough to pull off that particular trick.
I suspect that a large investment in very thin planar speakers might produce interesting results (I'm thinking of the B&G Neo series). I think having them mounted in a T-shape (ie, a dipole source just in front of a monopole) might work.
So, I'd consider this experiment to be a failure this time, but at least now there's direction for further study.
Chris
I will say it again for good measure... The loudspeakers being discussed in this thread have nothing in particular to do with "MS" encoding. Rather they exploit two distinct techniques:
(1) "Narrow angle stereo";
(2) (Optimal) three speaker stereo.
With appropriate care, the results are stunning - at least for one person in the right place anyway.
(1) "Narrow angle stereo";
(2) (Optimal) three speaker stereo.
With appropriate care, the results are stunning - at least for one person in the right place anyway.
It is not the separation that is an issue and drivers can even be mounted sufficiently close to circumvent any coincidence dips (that incidentally beset normal wide angle stereo). As I indicated previously, S-channel compensation is highly desirable.
Any talk of LR/MS transforms is not relevant.
Nice of you to inform us what the thread I started is about.
Hats off if you're getting drivers with less than 1" centre-to-centre (as my mics are). I couldn't do it with what I have here - as I said, I suspect thin planar speakers will be required.
LR/MS transforms are perfectly relevant - this is an unusual setup, so I wanted to make sure everyone understood where the signals were coming from.
Chris
Hats off if you're getting drivers with less than 1" centre-to-centre (as my mics are). I couldn't do it with what I have here - as I said, I suspect thin planar speakers will be required.
LR/MS transforms are perfectly relevant - this is an unusual setup, so I wanted to make sure everyone understood where the signals were coming from.
Chris
The thread was about a so-called "MS speaker" that transpired not be be so, but a narrow angle, three-speaker stereo arrangement as I have described above.
Your remarks regarding microphone spacing are also not of relevance to this speaker arrangement (barring the boost that is required). The critical distance for the speaker is the difference in path length between the two "S,-S" drivers and the listeners ears. It is not the reciprocal of an MS microphone.
I have tried and succeeded with this arrangement several times, including with three Manger MSWs: It is not a practical problem. Above all, I believe the approach can deliver outstanding results and does not deserve to be overlooked because of misunderstandings that are commonplace.
So MS transforms and MS microphone techniques are not relevant either, other than MS representations providing a convenient way to process the S-channel.
To reiterate (again), the setup featured exactly two speakers. One monopole, one dipole.
I know this, because I set it up today.
You're welcome to make suggestions, but so far you've quoted other methods without providing much/any background to them.
What I think you're suggesting is to increase the width of the dipole speaker, so that it ends up as two out-of-phase monopoles spaced apart. I can see that there's potential there, but it seems to me like it's heading back towards conventional L/R stereo with spaced speakers.
You claim excellent results, so I'll give it a try. If I'm way off on my interpretation, I'd appreciate some guidance - a sketch would clear up a lot of mystery.
Chris
A dipole is essentially two acoustic sources - hence the three speaker description. Three acoustic sources might be a better way to state it - my apologies for the confusion.
I have outlined what is needed to get such a three speaker system to work well. There are no 'other' methods - we are talking about the same type of loudspeaker system, just confusing the underlying principles.
No. The narrower angle the better as you can achieve perfect HF compensation of the S/-S channel outputs - and as I have said previously - if you provide the compensation in the digital domain, you can even compensate for the linear phase term that is normally ignored in such schemes.
Adding extra spacing for LF drivers does reduce the LF S-channel boost required. I used 15" drivers below 400Hz, hence greater separation was inevitable. But this did not compromise the HF narrow angle that remains desirable; Longer wavelengths at LF relax the constraints somewhat.
The results are excellent, at least at the optimal position. I really hope you do give it a try. I have never been able to listen to conventional stereo set-ups since without hearing their inherent limitations. But you need to acknowledge that you are describing a speaker with three acoustic sources.
MS inputs are just one possible arrangement, although IMHO, for the narrow angle case it is the best. Implementing HF S-channel compensation even avoids the need for the two or three band shelving filters in the MS matrix decoders invented by Michael Gerzon for wide angle/multiple listener set-ups.
If I get chance I will add more details. For now I would recommend reading up on the "stereo dipole" investigated by the ISVR and Gerzon's hierarchical matrix stereo decoders. Both are published in AES Journal papers but I think both are also described in plenty of freely available web publications too. Edeko et al's equally seminal AES paper on conventional stereo is also recommended if you can get hold of a copy.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.