Hi,

believe it or not I recently tried m/s eqing for the first time and found a pretty strange behaviour. No matter if I use mid or side, low or high shelf or any bell, no matter if I rise or lower a band the stereo image gets narrowed according to the amount of dB I dial in.

I made this for demonstration purposes: in the channel, a sine wave at 1kHz panned completely to the right. In the master, an eq set to Sum (aka mid) low shelf at 80Hz set to -12dB. Result: the the pan moves to the left.
I can set the sine to 10kHz or use a high shelf, it's still the same. And when I use a bell at 1kHz the wave is almost centered...
How can I use m/s eqing if it ruins the pan?
And just for the books, this happenes of course with other m/s eqs too and of course with all kinds of sources.

 

Maybe this will help you :

https://theproaudiofiles.com/mid-side-processing/

https://unlockyoursound.com/stereo/

 

The phase thing again, really? Seems I didn't check it.

 

Last edited: Feb 27, 2019

As i think LP just minimizes this problem, more accurate, not completely) need to test more.

 

Ha ha. You funny man! What you got here is a well set up trolling arrangement. Cuz the sinewave is panned hard left, it is not present in the right channel at all. To be in the mid, a signal needs to be in both L and R with +1 phase, our sine is only "half in the mid" or not in it at all depending on how you wanna look at it, just tipping a toe in. The EQ plug first sums L+R to get this "mid", then divides by 2:

Sine signal from Right + 0 signal from Left, divided by 2 = mid

Now the sine, at half amplitude, is processed by filter. Then "mid" is decoded back to stereo by adding it to both L and R. And, the wtf moment is here - you hear a distinct pan change, since some sine has just "appeared" in the opposite channel.

It's a degenerate stereo case because of the hard panned signal. Similar to how you can troll mathematicians by tricking them into dividing by zero.

To get the stereo to un-degenerate, put the same sine in the opposite channel with inverted phase. Now you got a basic ultrawide non-monocompatible sine, which will be not be present in the mid channel. Adding some center panned sine to this arrangement will create a basic reproduction of your typical stereo field and you can process both mid and side in this "sandbox" to heart's content with less "wtf" results.

 

Last edited: Feb 27, 2019

Err, sorry man, but you're completely wrong, no trolling at all here. I made this for visible demonstration purposes because I could hear that effect before when I tried it on drums, guitars and a whole mix.

And btw, the sine wave is panned hard right, not left, as you wrote in the first line.

But the explanation you gave could be the reason why this occurs with music too. What do you think?

And why should a hard panned signal be degenerated, please?

 

Last edited: Feb 28, 2019

Yep, I think that's the reason. I meanwhile found a m/s eq with linear phase and it didn't influence the pan at all.

Sorry, at least here with the non linear phase eq, it didn't (depends also on the slope a bit).

 

Last edited: Feb 27, 2019

It made me smile anyhow!

Hard panned signals are a degenerate case from the point of view of stereophonic sound. The whole point of stereo is that we aim to reproduce two correlated signals and based on the amplitude and phase differences between the two, the ears/brain get tricked into thinking there's a whole "sound field" in front of them. Actually there isn't and the illusion is not very natural either, it's more like a cheap parlour trick but over time we have all become very used to hearing it and know that's what it sounds like.

So a situation where a signal is hard panned, is a degenerate out-of-bounds situation from the point of view of stereo (and hence mid/side processing), but at the same time stereo itself is a degenerate situation from the point of view of a system that would represent sounds 360 degrees in a sphere around the listener, so including up-down. There is such a system, called Ambisonics.

Ambisonics is essentially mid-side technique but generalized mathematically to the full sphere, not just left-right difference and sum. Signals in Ambisonics get "encoded" by a kind of mid-side process which separates out a Mid (called W in Ambisonics) and 360 degree "side signals" for left-right, up-down, and front-back. Then you can easily decode this representation onto any number of playback speakers in realtime, one of the huge advantages of Ambisonics.... But the exciting part for us, I think, is that you would get even more leeway doing mid-side type processing on such signals than with stereo. It would "degenerate" far less. To my knowledge those processors don't yet exist but as Ambisonics rides the wave of VR and binaural repro for gaming, those tools will all be created for us to play with.

 

As MPM said, you want FIR processing. One implementation is izotope Ozone's LP M/S EQ.

ambisonics uses the M/S principle to store in-phase sounds in a sphere.

Yes, but if you actually USED ambisonics you would find that it's not superior in all ways to stereo or surround. In stereo, you can store WHATEVER you want in ANY channel or position, anywhere, including out of phase things. In ambisonics, in-phase sounds are preferred because you have trouble storing out of phase things, and what you can store highly depends on how many channels you have. You'd need maybe 16-25 channels ambisonics for what we take for granted in stereo or 5.1 surround. ofc, academia knows this and spatial PCM is preferred over ambisonics for HQ representation. Also,

A number of ambisonics deformers / manipulators exist. See https://docs.google.com/spreadsheet...bttqMLClikcLoE7pAkf4jmYn8jhGfqJx1TirK/pubhtml

probably The ONLY reason ambisonics is used for VR, is that rotation / head tracking is easy and has no loss in quality. (unlike surround). As said above, ambisonics has already degraded the sound to be in-phase (within limits of the spherical harmonics) . This is why cinema since the 60s doesn't use ambisonics since it's "LQ" . It uses surround.

 

Last edited: Feb 28, 2019

With a m/s EQ, it doesn't matter what filter shape you use.. the principal is the same.

If you use a m/s eq to boost the mids.. it will narrow the stereo image. If you use it to boost the sides.. it will broaden the stereo image, or even clip the stereo field if you push it too much.

Same goes with cutting. Cutting mids broadens the stereo image. Cutting sides narrows the stereo image.

Most who implement m/s EQ in mixing should already know this. And if not, use iZotope insight, or any analyzer that shows a visual of the stereo image.. and observe what goes on with the stereo image when doing each of the things I mentioned above.

Did anyone notice the image you posted shows a mono channel panned hard right, an isolated tone at 1k which the low shelf cut does not have any effect on at all? You're cutting nothing. Any L/R swap pertains to the difference in routing between the channel, plugin, and master.

 

Well, that's what I thought too, but that's not what's happening, as you can see in my pic. The sum/mid LS is affecting the pan of the extremely panned sine wave. Again, I made this on purpose to show its weirdness.
I have no idea how the routing could affect the panning in this case. You may like to have a look at this

Spoiler: low shelf

The lesser the LS is turned down, the lesser the impact on the pan position.

And furthermore, even a rise with a high shelf in Diff/side mode narrows the pan, see

Spoiler: high shelf

But as mild pump milk already said this has to do with the non phase linearity of this plugin.

 

it has also to do with the way the signal is decoded by plugin as well ,
MS EQ it is usually Input LR Signal Split Into 2-4 Channels (Plugins Variant) that would equal to Mid And Side
Then Process The Signal (EQ) Mid And Side Individually, Ultimately Sum Back Into L&R Output

Signal > Mid > EQ Processing
> Mid Side Sum = Output LR
Signal > Side > EQ Processing

my point is
The chain where EQ Processing happens , Depending on the technique (Non-Linear , Linear , Dynamic EQ ,Etc)
they each introduce Sample Delay to the processed Signal if its more calculation the cpu has to do more then likely it would need more buffering time , a good plugin company compensate those delays using delay compensate.
but even at 1 sample rate of delay can cause phasing issue especially when working with the width of a signal.

I've came across EQ Plugins where the developer were trying to save CPU so they'd only "process what you dial in "
for say you EQ out the low end on Side but left the Mid untouched, even tho it sounded wide but the sides were delayed (By Samples).
and the mid timing were on point, because the mid weren't processed. but you wouldn't really know that unless you spent hours on different material with the same EQ and settings and Bounce and matching the waveform.

The M/S encoding back to Stereo L&R suming as well , every company would have their way of doing this. just certain things i noticed through out the years maybe this helped ?

 

Look at my first and only comment before this. I said this:

If you use a m/s eq to boost the mids.. it will narrow the stereo image. If you use it to boost the sides.. it will broaden the stereo image, or even clip the stereo field if you push it too much.

Same goes with cutting. Cutting mids broadens the stereo image. Cutting sides narrows the stereo image.

These images you shared demonstrate exactly what I said. You are cutting mids(sum) in the first image. Notice the stereo broaden? You are administering it back in. Notice in the image when you don't cut the mids(sum) causes no stereo spread?

In the second set of images where you boost the sides(diff). Do you notice how the stereo broadens a little in the image where you boost the sides a little? And do you notice how the stereo broadens alot more in the image where you boosted the sides alot more? It's doing exactly what it's supposed to do. Your adding "sides".. stereo into the signal you panned hard right.

So even though you for some reason chose to demonstrate the application of a m/s EQ in some awkward scenario where it's on a mono channel panned hard right (In that scenario, little to no one would say "I know what this track needs.. some m/s EQ!!).. it's still doing precisely what it's supposed to as I stated earlier.

 

Last edited: Feb 28, 2019

If a hard panned signal is turned towards the center you call that 'broaden the stereo image'???
I tried it with two signals at 100R & 100L and 70R and 70L and in both cases the signals are moved towards the center with this Sum/mid LS. Means, the image is narrowed, that's clearly both, audible and visible. And this does not happen with a linear phase eq.

 

Man.. the signal is not "turned" towards the center. When you cut the mids or boost the sides.. panning is not what's occurring. What's happening is you are adding sides back to the signal. Sides means L/R, not just mono R. It's a stereo function. And yes, you are broadening the stereo phase of the signal when you do it. The L/R (stereoized) signal is obviously wider than the mono R signal. I have no clue why you'd think when you boost sides (L/R), and so you see some signal get added back to the left.. you get the notion what you're doing with the EQ is "panning"? Hope you see what I'm trying to explain now bro. If not, just use a visual analyzer that has a stereo image display on the master and you'll see exactly what goes on when you boost or cut with a m/s EQ. EQ filters are not pan knobs.