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: 5444
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 had a chance to revisit the Neumann VAB-84's technique to transform LF Side information to mono to provide an improved Elliptical Equalizer.

This is the circuit built using the Precision MS Matrix. The board really is a Swiss Army Knife of Mid/Side Processing.
In this implementation the Mid channel isn't even used.

Image
Emulating Neumann VAB-84 Side to Mono Transformation Using the Precision MS Matrix

Here's a quick breakdown of the signal flow.

Note that the points shown on the board correspond to these drawings:
http://www.ka-electronics.com/Images/jp ... iagram.jpg
http://www.ka-electronics.com/Images/jp ... _Large.JPG


Left and Right enter the Decoder inputs as balanced signals at points EM and ES.
The Right output appears on FS Out and is carried into the FS Input.
The Left output appears on FM Out and is carried into the FM Input.
The FM and FS inputs are differenced by U11 (internal to the board) to provide a Side Output.
Side is at the GR Output.
Both Left and Right feed the summing nodes at GL In and GR In through 10K resistors.
(Internal resistors R4 and R5 on the MS PCB are made 0 Ohm.)

The "Encode" section of the matrix board provides balanced I/O for the external EQ and to provide polarity inversion.

The Side input at CS In is single-ended and converted to balanced to drive the DS Output.
The balanced output from the EQ is received at Input AR.
The filtered Side output is at BR Out.
BR Out feeds a summing node for the right channel through a 20K resistor.
The BR Output also feeds BR In to provide polarity inversion.
The CS Output is polarity inverted with respect to BR In.
The CS Out is summed into the Left channel summing node through a 20K resistor.

The overall Left and Right Outputs appear at HM and HS.

The Mid channels are not processed: IC1, IC3, IC5 and IC10 are not used on the PC board.

I've been playing around with various filter 3 dB points and have found that an aggressive 300Hz (-3dB) 12dB/octave response isn't that audible in terms of mid-band separation. I also played around with a 600 Hz filter and the effect on imaging was minimal. Neither the EE70, EE77 or VAB84 are capable of a LF crossover greater than 6 dB per octave. Having a steeper filter option allows a larger chunk of LF to be mono'd (when required) without as great an effect on imaging.

Steering Math

Let's take a look at the math the VAB-84 uses to convert Side to Mono.

Image
Nuemann VAB84 Block Diagram Showing Low Pass-Filtered "Side" Converted to "Mid" By Side to Mono Conversion.

Lout = L -1/2(L-R)m
Rout = R -1/2(R-L)m

Where "m" is the filter transfer function.

Assume for sake of discussion that the filter is infinitely steep.
If the Side information is within the bandwidth of the filter then m = 1.
If the Side information is outside the filter pass band then m = 0.

Outside the pass band of the filter (where m = 0) L=L and R=R because the difference terms disappear.

Inside the low pass filter pass band (m=1) the following occurs:

L = L -(L/2) + (R/2) = L/2 + R/2
R = R -(R/2) + (L/2) = R/2 + L/2

Thus, below the LP filter cutoff, L and R are summed.

I hope to provide some sound files to provide comparisons.
I'm not having a hard time finding things mastered with a wide low end.
The next step is to build a conventional EE70/EE77 subtractive EE and provide comparisons to no EE and the VAB84 additive type.

What I've found is that aggressive amounts of EE (a high crossover) using the VAB-84 method isn't that audible when the Side info is folded into mono.
Some of this may be due to the fact that no information is actually lost, the other may be due to a higher filter order.
Another interesting attribute is that as tones are swept through the passband and filter skirt of the VAB84 EE the mono sum remains constant.
There is no perceptual reduction in low end as the VAB84 EE is switched in or out or if the result is mono'd.


I should also point out that if one wanted to get really creative, a bandpass filter could be used to reduce separation of a particular frequency range for more difficult repairs and/or cleanup.
User avatar
mediatechnology
Posts: 5444
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 been running real listening tests comparing an EE77 topology vs. the VAB84 topology.

Both have a 300 Hz -3db crossover to mono. The EE77 is 6 dB/octave the VAB84 emulator 12 dB/octave.

When there is a lot of LF difference information, particularly in the sub-bass region, having some material mono'd in the low end seems to provide some increased tightness or punch on 'phones or small speakers with either the "77" or the "84" circuit.

Though there might be a very slight decrease in perceived low end using the "77," what I hear it do more is smear in the midrange and reduced midrange imaging. Midrange width collapses and on stereo mic'd or chorused instruments there is some phase cancellation.

I got really curious as to how the "77" and "84" compare with completely uncorrelated material.
In theory the "77" should cancel uncorrelated material the "84" should fold L into R and R into L.

I figured the best way to create uncorrelated material was to use noise.
I generated noise in the Left channel, copied Left into Right, and then reversed Right in time to create ultra-wide noise.
The resulting oscilloscope vector looked like a round ball of steel wool.

I then processed this source using both the "77" and "84" techniques of LF blending.
The levels are matched within about 0.15 dB on tone.
The peak levels of the two files confirm this.

Here is the result. The "84" sample is first, then the "77" sample.
You may want to loop this offline.

http://www.waynekirkwood.com/content/VA ... Hz_6dB.wav (1.6MB .wav file)

The "84" sample seems to sound slightly spectrally fuller and wider than the "77" sample.
The "84" sample also has about 0.5dB greater power despite only having a 0.15dB peak level advantage.
The "84's" vectorscope pattern is also wider in the midrange.

When the results are mono'd both sound exactly the same.

OK, so I know that using fully uncorrelated noise is pure Geekdom.
But, it proves that there is a difference in how uncorrelated material is processed by the EE77 and VAB84.
That was the point. They sound and work different.

So far the main takeaways are:

(1) Some mono'ing of low end is a good thing (depending on the source) even if it's not going to vinyl. Too much LF stereo is not a good thing.
(2) If you gotta do it, the one that stays out of the midrange is the better option.
(3) Depending on the source material there may be less LF loss by using the "84" and there will certainly be a wider midrange.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

VAB-84 at 12 dB/octave vs EE-77 at 6 dB/octave

Post by mediatechnology »

I did a crosstalk curve comparison for the Neumann VAB-84 topology modified for 12 dB/octave versus the 6 dB/octave response of the EE-77.
The EE-70, EE-77, and MS techniques for elliptical equalization seem to be topologically limited to 6 dB/octave.
The VAB-84 could have been 12, 18 or 24 dB per octave but wasn't.


Image
Crosstalk Curves of Neumann VAB-84 Emulator Modified for 12 dB/octave vs EE-77 Emulator at 6 dB/octave

Compare sound files processed with the VAB-84 emulator and the EE-77:

VAB-84: https://proaudiodesignforum.com/content ... -84_EE.mp3
EE-77: https://proaudiodesignforum.com/content ... -77_EE.mp3
Nescafe
Posts: 22
Joined: Sat Aug 03, 2013 10:39 am
Location: Jakarta-INDONESIA

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

Post by Nescafe »

Ha!!!!! This is very interesting, I have check both audio files and looks like the VAB84 is what I'm after, like bx_digital mono maker plugin.
In the schematic shown above, the inputs, outputs and inserts inside the block correspond to the PC board's silkscreen designators.
This a bit confuse me, is that mean that "GR O" is U11 out and "GR I" is U9B inverting input?

And if I want to try a switchable R for various frequency, do I need to worry about DC?

Thank You.
User avatar
mediatechnology
Posts: 5444
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 »

This a bit confuse me, is that mean that "GR O" is U11 out and "GR I" is U9B inverting input?
That is correct.
Do note that the U9B stage is modified to make the on-board resistors 0 Ohms (they are usually 20K) so that summation can be done off-board.

You could switch resistors in the filter to vary the cutoff.
For a second order filter that's a 2 pole switch, for third order a three pole.
If it were me I'd pick 2-3 filter frequencies and build three filters switching outputs to reduce the switch to a single pole.
Op amps, resistors and caps are cheap compared to a three pole switch and all that wiring.
To be honest with a third-order filter I don't think you'll hear much difference for a cutoff below 200-300 Hz.
Even at 300 Hz -18dB/octave it gets out of the way of the midrange fast.
Nescafe
Posts: 22
Joined: Sat Aug 03, 2013 10:39 am
Location: Jakarta-INDONESIA

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

Post by Nescafe »

Hi Wayne,

Thank You for the confirmation, I just finish read and learn Your explanation slowly and carefully, so the difference between 2 version that You create: the last one is using balance filter and the the first one is unbalanced filter right?
I will go with the first one (unbalanced) so I can use the other half as monitor encoder, the funny things is the elliptical filter will be in the monitor control box :mrgreen:
Will starting next week, need to etch + buy some resistor and cap for filter, do You think fet input opamp are appropriate for the filter? I have some left from previous project, and also need to decide 23 freq to use :evil:
20140410_225649SS.jpg
20140410_225649SS.jpg (115.25 KiB) Viewed 22455 times
User avatar
mediatechnology
Posts: 5444
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 »

Thank You for the confirmation, I just finish read and learn Your explanation slowly and carefully, so the difference between 2 version that You create: the last one is using balance filter and the the first one is unbalanced filter right?
The second one can be used with either a balanced or unbalanced filter. It's a little more refined using balanced I/O and the precision inverter but does indeed use both halves.

If you switch resistors in the first example you will want to always have the largest value in circuit and then use a switch to parallel the values to raise the frequency. The reason for that is that they provide the bias current for the op amp. In addition the overall filter input needs to see a DC path to ground: The previous stage provides that.

I do understand why you would want an EE in the monitor path.

The choice of op amps is pretty flexible.
Nescafe
Posts: 22
Joined: Sat Aug 03, 2013 10:39 am
Location: Jakarta-INDONESIA

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

Post by Nescafe »

Hi Wayne,

Thank's again for the resistor explanation, got it. And what I mean with EE in the monitor box is not because I use it for monitoring, the EE is for processing, but since it only need a half of MS card, I use the other half as M/S encoder in my monitoring control box :mrgreen: that's because I cannot cut the card into 2 pieces :mrgreen:
User avatar
mediatechnology
Posts: 5444
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 »

And what I mean with EE in the monitor box is not because I use it for monitoring, the EE is for processing, but since it only need a half of MS card, I use the other half as M/S encoder in my monitoring control box :mrgreen: that's because I cannot cut the card into 2 pieces
Got it.

My working theory is that a good EE for mono LF crossover/LF blending might benefit smaller speaker systems particularly those using a sub-woofer. As I understand it most mono subs use LF addition (like the EE70/77) which effectively subtracts difference information rather than folding L>R and R>L at LF like the VAB84. With really small speakers it forces both pistons to work together.
Nescafe
Posts: 22
Joined: Sat Aug 03, 2013 10:39 am
Location: Jakarta-INDONESIA

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

Post by Nescafe »

Ah....I just realize that, Thank's Wayne.
Post Reply