Using the Precision MS Matrix for Mono Crossover LF Blending

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Using the Precision MS Matrix for Mono Crossover LF Blending

Post by mediatechnology »

I receive a lot of questions about using the MS Matrix in disc mastering transfer consoles to provide mono crossover.

I ran a quick test with a 12 dB octave Butterworth unity gain Sallen-Key HP filter in the Side channel.
The -3dB cutoff was approximately 20 Hz.
The "send" was point C, the "return" into point F.
This bypasses the 1646 and 1246 insert line drivers and receivers.

Image
Precision MS Matrix with 12 dB per octave 20 Hz high pass filter in the Side channel to Provide Mono Blending vs. Bypass

Note however that the derived response is 6 dB per octave.

I often see and hear about references to the Neumann "elliptical" Filter shown here:

Image
Neumann EE70 "Elliptical" Filter Schematic

There is an inductor that bridges L and R to blend low frequencies to mono.
But the EE70 does not appear to be an elliptical filter topology...

This is an elliptical topology.
The second-order elliptical part begins at R5.
It's a state variable filter with the low and high pass outputs summed.

Image
http://sound.whsites.net/articles/active-filters.htm#s7

I'm curious to see how a true elliptical filter in the Side channel would steepen the response.
I don't think it will.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by JR. »

I may not be following you accurately.

-12 dB, 2 pole crossovers are generally disliked because two symmetrical 2 pole filters are 180' out of phase at crossover so there is a suck-out at the crossover point.

AFAIK all derived crossovers where a real filter is subtracted from dry signal result in -6dB one pole response for the derived section so asymmetrical response. Actually not bad for some applications.

Not sure where you are heading with this.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by mediatechnology »

John -

The application is for a low frequency "crossover" to mono for disc cutting.

(I think this is sometimes called "vertical crossover" due to the sum/difference mechanical vectors.)
So not a crossover in the speaker sense, but mono blending to prevent large difference information at low frequencies that are difficult to cut.
Large low frequency phase/polarity differences cause the groove to narrow causing skipping.

However, the derived subtractive speaker crossover seems to provide an analogy.

In a derived speaker crossover the derived output will be 6 dB per octave regardless of the steepness of the "un-derived" output.
If the LPF is 12, 18 or 24 dB dB per octave, the derived HP (or BP )response will be 6 dB octave regardless.
(See: http://sound.westhost.com/articles/derived-xovers.htm)

It appears - and this was my hunch - that when Side is filtered and then summed/subtracted to reconstruct L and R, its the same process as a derived speaker crossover.

Thus, to provide vertical crossover for disc cutting, the filter order in the Side channel only has to be 6 dB/octave because any higher-order filter won't provide an advantage.
The result is going to be a crosstalk blending curve that has a 6 dB/octave slope regardless.
That's the main take-away...

So what's with the "elliptical filter" in the Neumann EE70?

A true elliptical filter in the side channel will have a steep slope, but the derived result, mono blending, will still have a gentle 6 dB per octave slope.
So why is the Neumann filter, which blends L and R at a 6 dB per octave rate with an inductor, called "elliptical?"
It's not an elliptical filter topology.
The need for a "mono crossover," by definition, obviously didn't exist until stereo recordings.

Could it be a filter that blends mono to benefit "elliptical" styli and thus be called an "elliptical" filter?
What is it about the EE70 that's elliptical?


EDIT: Found the origin of the term "elliptical equalizer" as it relates to Neumann:
From abbey road d enfer
This is what was called an "Elliptic equaliser" by Neumann, because it modified the width of the ellipse on the vectorscope on vinyl lathes.

It's called an elliptical equalizer due to it changing the vectorscope ellipsoidal pattern not because it's an elliptical aka Cauer filter.
Solved that mystery.


There's also this neat thing called the "Vertikal Amplituden Begrenzer" aka VAB-84 that dynamically controls width:


Image
Neumann VAB-84 Block Diagram Showing Dynamic Control of Side Information

https://proaudiodesignforum.com/images/ ... nn-VAB.pdf

It appears to be a feedback topology dynamically-controlled derived high pass filter to limit the vertical (Side) component.
It's also 6 dB per octave and inside de-emphasis/pre-emphasis filters that more heavily weight low frequency information corresponding to the "RIAA" curve.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by JR. »

OK gotcha..

I suspect the classic inductor shunt across the cutting head is low order, so making this high order does not appear to be important.

If we review where stereo separation becomes important I suspect it is not until somewhere in the midrange, and if a complex low bass sound has higher frequency components spread to stereo and the low bass mono, we will probably still localize it properly.

So I'd KISS with low order filters all around.

Derived will also sum perfectly to mono should that happen.

Note: there may actually be some modern tracks with hard panned low bass. Old school recording knew not to do that (or they should have known).

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by mediatechnology »

...There may actually be some modern tracks with hard panned low bass. Old school recording knew not to do that (or they should have known).
I cited an article in another thread (somewhere around here) about the modern requirement to have center-panned low frequencies for proper ear bud translation. "Out of the head" low frequencies do sound weird and lack punch. I think many MEs check for that and fix it when it happens.

I question the need for Side corner frequencies (-3dB point) much higher than the first two octaves since the mid-range separation loss, given the 6 dB slope in separation vs. frequency is pretty high.

This 500-series EQ with an "elliptical filter" (in quotes because there's nothing elliptical about it) has switch frequencies indicate the point where there is 15 dB of separation. http://www.adt-audio.com/Audio_Modules/ ... r_EEQ.html
Elliptic EQ
The elliptic equalizer reduces the stereo base (sic) width below an adjustable edge frequency and can reduce mono compatibility problems caused by phase reversed signals in both stereo channels or time shift effects at low frequencies. Since human ear is not able to locate anything below 300 Hz, reducing the base (sic) width below that range causes no audible effects but 'centers' the bass range to improve the mono compatibility. The section increases the crosstalk with a steepness of 6 dB/octave. The frequency control is calibrated in that frequency that causes 15 dB crosstalk. The control range starts at 25 Hz and reaches up to 2 kHz. A setting above 300 Hz will be audible; however, may result in a better compromize (sic) of the particular mix..
I suspect that the Neumann EE70 has similar curves to the one I posted at the beginning of the thread showing the MS Matrix used as an "elliptical" filter. I'm not sure what the 150 and 300 Hz switch calibrations mean on the EE70. If the filter was set to 300 Hz and it was the 3 dB point then there is virtually zero separation mid-band.

I think I need to find a set of filter values for the MS Matrix and then re-label the cut-off frequencies to some separation value at a particular frequency.
For the ~ 20 Hz Fc shown in the first post, the -15 dB separation frequency is 60 Hz.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by mediatechnology »

I don't think the 150 and 300 Hz labels were the -3dB frequencies...
I'm thinking those panel markings may have been "x" amount of separation at the marked frequency.
That high a frequency (for a -3 dB point) and at 6 dB octave there would be no mid-band separation.
I did see one post where people were discussing how ME's typically modified the frequencies to be much lower.

I should point out that even a 20 Hz (-3dB) filter has enough loss in the pass-band to reduce separation to around -40 dB at 1 kHz.
The filter gain never approaches unity in the pass-band: It's an asymptote that never actually gets there.
You can see in the curve how in the last upper octave - a 1000x higher point than the cutoff - it gets close.

I think there is a need for a dynamic filter Like the VAB-84 that would have a cutoff that could be driven into the subsonic to "get it out of the way" when not needed and open up when the LF L-R gets too high. Sort of a modern VAB-84 but improved and feed-forward. I doubt the LF cutoff had much frequency control range in the VAB-84 since it was likely FET-based.

With a dynamic filter reaching into the subsonic the ME would have the best of both: Reduced LF separation when the vertical component gets too high but no mid-band penalty when LF L-R is not too high.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by JR. »

While this has probably been mentioned in other threads the old school phono carts had 20-30 dB L-R separation. I do not know what our ear to ear separation is while listening in an real room.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by mediatechnology »

the old school phono carts had 20-30 dB L-R separation.
I think those are reasonable numbers. Lets use -30 dB to keep numbers round.
I think I may have seen a 35 dB figure quoted once in a review and was astounded.

The 20 Hz -3dB filter in the OP provides ~ -42 dB of separation at 1 kHz.
An 80 Hz filter, two octaves higher, would reduce the encode separation at 1 kHz to -30 dB.
Encode separation combined with the -30 dB separation of the cartridge (assuming the crosstalk is incoherent) degrades crosstalk by 3 dB to -27dB.
An 80 Hz filter is enough to begin to degrade overall crosstalk performance.

If the "150" and "300" Hz filters represent the -3 dB points, then they are enough to degrade the encode process far beyond what the playback cartridge is capable of. I would be interested in seeing those EE77 schematics. Based on what I've seen so far every "elliptical EQ" is first-order.
emrr
Posts: 570
Joined: Sun Mar 02, 2008 10:27 pm
Location: NC, USA
Contact:

Re: Using the Precision MS Matrix for Mono Crossover

Post by emrr »

I didn't see this till now, having posted the other derived crossover observation a few hours later in the other thread. Makes sense to add and document this function.

It's always fun to have vinyl come back when the band never mentioned vinyl as an end product. Most noticeable have been wide stereo synth parts when the entire mix has been reduced in width, sounding like it had mild phase shift on the synth.
mediatechnology wrote: This 500-series EQ with an "elliptical filter" (in quotes because there's nothing elliptical about it) has switch frequencies indicate the point where there is 15 dB of separation. http://www.adt-audio.com/Audio_Modules/ ... r_EEQ.html
The ADT line has some very interesting pieces in it. I've heard good things about their work. Those are actually 5RU modules, I believe all of their pieces mount in racks of their unique spec.
Best,

Doug Williams
Electromagnetic Radiation Recorders
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Using the Precision MS Matrix for Mono Crossover LF Blen

Post by mediatechnology »

I've looked at the VAB84 again and realized that it is a completely different animal than the LF blending of earlier Neumann "elliptical equalizers."

Phase and polarity differences (Left relative to Right) are what make stereo.
Pan something hard Left and the Side information will be the opposite polarity in Right relative to Mid to produce full cancellation.
By definition the "Side" component contains out-of-phase material.
Another curious property of the MS Matrix is that when the polarity of Side is reversed, Left and Right swap.

When simple LF blending occurs in an "elliptical equalizer," Side information that gets blended will partially (or fully) self-cancel.
If a LF element is Side-heavy (wide) and gets progressively blended to mono it will lower relative to the overall mix.
It's not fully mono compatible.

I don't know what the filter order is in the VAB84 and my analysis is early and may be fluid.
It looks like the VAB84 converts Side information to Mid below a certain frequency and then adds it to Left and Right.

The VAB84 low pass filters out everything in the side channel above "x" Hz.
It then recombines the low frequency Side information with reversed polarity into Left and Right so that it reinforces Mono, not cancels it.
That's how it converts Side to Mid.

An MS Matrix subtracts Side to increase width or separation.
The VAB84 does the opposite: It sums Side into Left and Right to reduce separation.


Image
Neumann VAB84 Matrix. Vertical is Difference or Side.

In a conventional MS Matrix Mid would be extending into the summation stages, not Left and Right, and the polarity of Side feeding Left and Right would be reversed from the above figure.
In the VAB84 Left and Right are fed into the summation stages and the polarites of Side feeding Left and Right, which has been filtered, is the opposite of what it is in an MS Matrix.

The VAB84's filter is dynamic. Assume for the sake of discussion that the filter is static.
The broadband Left and Right inputs pass through to their respective outputs. (Ignoring the "RIAA" weighting filters.)
The filtered side information in the VAB84 then is summed into Left and Right but in an opposite polarity of what it would be in an MS matrix.
Rather than subtract Side from Mid to reconstruct Left and Right Side is added to Left and Right at low frequencies to reinforce mono, and decrease LF separation.
It appears that LF elements hard-panned no longer drop out of the mix due to phase cancellation but fold down to mono.
The VAB84 is "mono compatible."

I haven't fully proven this but what I did do is a quick setup with the MS board similar to the topology of the VAB84.
I used a second order filter (50 Hz -3 dB) and was able to obtain a 12 dB/octave crosstalk curve.


Look at the second-order 12 dB/octave crosstalk/blending curves:

Image
VAB84 Emulation Crosstalk/Blending Curve Using the Precision MS Matrix

Image
VAB84 Emulation Frequency Response Curve Using the Precision MS Matrix

Update:

I've found that I can also do third-order low frequency cross-talk curves.
This is with a 50 Hz corner frequency.


Image
VAB84 Emulation Crosstalk/Blending Curve Using the Precision MS Matrix With An 18 dB Per Octave Slope

The cutoff frequency was set at 50 Hz.
The slope is steep enough that it can be set much higher without affecting mid-band separation.

So the question becomes "how do I make the MS board do this?"
It did after all generate those curves.
Turns out it may be pretty easy but I think there may be more than one way to do it.

Update 2:

This is the crux of the Neumann EE77's LF blend.
It's 6 dB/octave.
150 and 300 Hz appear to be the 3 dB points.

Image
Neumann E77 Low Frequency Blend
Post Reply