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 was thinking it might help you make that sausage faster without having to patch in that EQ and compressor to fix what the EE took away.
It would basically have a single knob like an EE: Out, 75, 150, 300 etc.
For in phase it would act like an EE.
For out of phase more like the original LF was mono.

But if you don't see it as a legitimate tool I'll forget about it.
Gold
Posts: 678
Joined: Sat Dec 01, 2007 5:20 pm

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

Post by Gold »

mediatechnology wrote:I was thinking it might help you make that sausage faster without having to patch in that EQ and compressor to fix what the EE took away.
I do two kinds of cutting. Stuff I master and stuff I don't. If I haven't mastered it and I think there might be tracking problems I'll switch in an EE at 150. Usually that with an HPF at 40Hz is enough to get things to cut. It might not be the perfect solution in every case but it will rarely cause undue harm. Cutting happens live with little time for audition. If I am being payed for a "flat cut" then the person who prepared it has to do a good job. If they don't, I take modest steps to keep the record from skipping.
It would basically have a single knob like an EE: Out, 75, 150, 300 etc.
For in phase it would act like an EE.
The EE doesn't have an effect on the mono signal as far as I know. Did I miss something?
For out of phase more like the original LF was mono.
Whether adding LF back into the center channel or just removing the side information sounds better is subjective. I believe adding low frequency to the center channel was the approach Haeco took. I don't have those schematics though. Some liked it. I'm not as big a fan. I wouldn't call it illegitimate. It's just not an approach I am interested in. Others might be.

If I hadn't heard things like this I might be interested. I believe the "stereo enhancer" in Samplitude/Sequoia adds LF back to center when you turn the knob towards Mono. I'm not a big fan of it there either.

Another practical problem when cutting is that adding level is bound to cause more problems than subtracting level.
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 »

Paul when I ran tests on this using program material I didn't experience level build-up from the additive process.
I think that's because the level of side relative to mid is usually quite small.

The EE will have no effect on mono sources that are mono by virtue of being panned center.

When they become off-center due to being panned or develop phase relationships due to chorus or whatever the EE will lower the level.
I suppose that's why other folks add low end back.
The VAB84 technique doesn't seem to do that.

I did some tests panning LF sources to check levels and listening to program and even with fairly high cutoff frequencies couldn't hear large level or balance changes. I was quite surprised by that.

There's a curious property about MS that blows your mind the first time you experience it and the VAB84 topology takes advantage of that.
If you encode Left and Right into Mid and Side and then polarity invert Side and reconstruct Left and Right a strange thing happens: Left and Right swap.
The VAB84 exploits that property by dividing the spectrum.
Above the LP filter cutoff Left feeds Left and Right feeds Right.
The filter output is inverted and fed back into Left and Right.
Below the cutoff Left becomes Right and Right becomes Left to provide true crossfeed.

I need to look at the Haeco and see if that's what they did.
Haeco et al may just boost LF to compensate for the typical EE LF losses.
Other than Neumann's original explanation in 1984 I've not seen anything about Side to Mono transformation published.
Gold
Posts: 678
Joined: Sat Dec 01, 2007 5:20 pm

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

Post by Gold »

mediatechnology wrote:Paul when I ran tests on this using program material I didn't experience level build-up from the additive process.
I think that's because the level of side relative to mid is usually quite small.
In the case where the low frequency uncorrelated information is small there is no reason to switch it in. The lathe automation can take care of it. You need to use it when there is a lot.
There's a curious property about MS that blows your mind the first time you experience it and the VAB84 topology takes advantage of that.
If you encode Left and Right into Mid and Side and then polarity invert Side and reconstruct Left and Right a strange thing happens: Left and Right swap.
The VAB84 exploits that property by dividing the spectrum.
Above the LP filter cutoff Left feeds Left and Right feeds Right.
The filter output is inverted and fed back into Left and Right.
Below the cutoff Left becomes Right and Right becomes Left to provide true crossfeed.
That is very interesting. I'm not sure I understand. I'll read through the thread again and look at the VAB docs again. The sliding filter is the part of the VAB I don't like. I remember skimming a lot of discussion about the resultant filter slope earlier in the thread. How does the result of this method differ from a 6dB/oct slope on the side channel?
Gold
Posts: 678
Joined: Sat Dec 01, 2007 5:20 pm

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

Post by Gold »

Okay, now I think I understand it. It's pretty cool. I would like to hear it. If you eliminate the inverter do you get the subtraction of the side channel? If so you could put a switch in.
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 »

In the case where the low frequency uncorrelated information is small there is no reason to switch it in. The lathe automation can take care of it. You need to use it when there is a lot.
True.

What I also found is that the LF is highly uncorrelated it should be mono'd regardless of whether it's going to wind up on vinyl.
Having something fully-centered at LF when played back on speakers effectively doubles the LF driver area vs. something panned hard to one side.
On 'phones it takes LF that would sound slightly outside the head and centers them inside the head.
The LF is more "focused" and more suited for playback on smaller speakers.
How does the result of this method differ from a 6dB/oct slope on the side channel?
A 6 dB per octave slope on the Side channel using an MS process produces the exact same result as an EE70/77.
A steeper slope inserted into the Side channel still gives you 6 dB per octave because the filtering is subtractively derived.
I don't see how the EE-series could have been made higher order.
In fact everything I see commercially produced in the modern era is 6 dB per octave. (ADT, Maselec MTC-2)

The VAB84 was 6 dB per octave but can be a higher order with the crosstalk/desired LF blending results also being that order.
I think Neumann may have kept the VAB84 to 6 dB per octave since the filter slid and was constructed with a VCA.
They also may have just not tried anything steeper.
So in addition to having the property of "Flank signal components are simultaneously transferred to the other channel" it can be made with a steeper slope and thus provide a lot of LF blending to mono without affecting the mid-band.

ADT mention that with their box set to 300 Hz (the -15dB separation point) you can begin to hear it.
That's because with a 6 dB per octave slope the separation at 1 kHz is only about 24-25 dB.

This is VAB84 emulation with an 18 dB per octave 50 Hz corner frequency.

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

You can't hear the steepness of the higher-order filter but it sure gets out of the midrange fast and can be made much higher in frequency without affecting the midrange if the source material requires it.
Okay, now I think I understand it. It's pretty cool. I would like to hear it. If you eliminate the inverter do you get the subtraction of the side channel? If so you could put a switch in.
Here's another weird aspect. If you take the inverter out of the Left channel and move it to the right - effectively inverting Side - then you get an odd LF comb filtering phase cancellation effect that is not useful.

I haven't figured out a way to switch it easily (there may be one) since one topology (MS Side filtering) injects Side into Mid and the other (VAB84) injects filtered Side into Left and Right. It might be a complex switch.

Using the MS matrix board to do the VAB84 emulation uses the "Decode" half of the board leaving the "Encode" available for a conventional Side HP filter.
It might be easier to switch outputs to change methods.

Thanks for looking at this Paul.
I'll try to figure out a way for you and all of us to listen to it
Gold
Posts: 678
Joined: Sat Dec 01, 2007 5:20 pm

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

Post by Gold »

Using the MS matrix board to do the VAB84 emulation uses the "Decode" half of the board leaving the "Encode" available for a conventional Side HP filter.

Since you only have the encode side would you do the same kind of thing where you subtract the side channel HPF from the main signal by summing it with the main signal reverse polarity?
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 »

Since you only have the encode side would you do the same kind of thing where you subtract the side channel HPF from the main signal by summing it with the main signal reverse polarity?
You mean only the encode side available when the decoder is used for the "VAB84" configuration?

There may be an answer to your question since there's a lot of ways to accomplish the same thing.
I'm going to have to think about it and the answer will probably come when I'm not thinking about it. :lol:
Sort answer is I don't know.

Here's the long answer:

The encode side could also be used but it's less convenient to implement for either "EE or VAB mode" it since it would require some additional outboard circuitry and lacks a key insert and summing node the decoder does. (Insert point G.)
I don't see how the encoder could be easily configured to provide an identical preview path.
Its hard to wrap the head around...

(The real breakthrough occurred once I remembered out of the blue the discovery that reversing Side polarity swapped Left and Right.
That's when I realized (when I wasn't thinking about it) that the Neumann did the same thing but only at LF.
Glad I found that English translation to confirm it.)


I originally tried a slightly different VAB configuration using both halves but it was less economical circuit-wise.
The encoder could be used at the input and output of the processor however to encode Left and Right into M and S for level metering.
One could switch the encoder back and forth between un-processed and processed to check overall side "elliptical" width by monitoring side level.

Doing EE-70/77 style requires an almost full end-to-end encode/decode.
Insert point C is used for the send.
The high pass filter is fed from CS out and returned into FS in.
(Or the HPF can be inserted into D and E for fully-balanced I/O and M carried through the insert.)

A single mode system - either EE70/77 or VAB84 - would require two boards: One for Preview the other for Modulation.
A switchable dual mode EE/VAB 84 might require four boards.

Also the EE uses a HP filter, the VAB a LP filter.
One could use a shared Side filter and it be either a state-variable or a unity gain Sallen-Key LP filter.
If a unity gain Sallen-Key is used, the high order output would be used for the VAB LP filter and the HPF for EE emulation could be derived.
The EE mode can be a derived first-order since the EE is always going to be 6 dB/octave regardless.
Gold
Posts: 678
Joined: Sat Dec 01, 2007 5:20 pm

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

Post by Gold »

I was proposing doing the the normal EE response in the same style as the VAB. Maybe I'm not thinking clearly but wouldn't it be possible to do the same kind of external circuitry you did except instead of the complicated derivations just mix back in a reverse polarity LPF'd side channel signal derived from the S output to the main signal that passe straight through?
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 »

Okay, now I think I understand it. It's pretty cool. I would like to hear it.
Thanks for that BTW! I'm pretty sure the guy a Nuemann that thought it up was a major MS geek.
Takes one to know one I guess. :oops: :geek:
I was proposing doing the the normal EE response in the same style as the VAB. Maybe I'm not thinking clearly but wouldn't it be possible to do the same kind of external circuitry you did except instead of the complicated derivations just mix back in a reverse polarity LPF'd side channel signal derived from the S output to the main signal that passe straight through?
I think you nailed it.

Like this:

Image
Neumann VAB84 Matrix. Vertical is Difference or Side.

You need both S and -S.
(Not shown but Left is the top input so "-S" sums into Left.)
In a balanced environment getting -S is easy.
You just have to make sure that the correct polarity of S gets fed back in to the respective channel and that the filter is non-inverting or you've otherwise built a comb filter.

You also need to make sure that Side is 1/2 Side.
I'm not sure I grok the math of that yet but Neumann was correct about the -6 dB loss.
I tried 0 dB and it was whacked.

I could do a simpler circuit than using the MS board but it has all the required "bits" other than the filter.
I really think you could almost bootleg the thing with an outboard EQ and some creative patching.

I think the IC count, including balanced IO would be:

2X 1246 for Input
1X 1240 for S
1X LME49860 for the LPF and -6dB stage (12 dB or 18 dB per octave)
2X 1240 to sum -1/2S into L and 1/2S into R
2X 1646 for Output

The single-ended I/O VAB84 EE "core" is 3X 1240 and a dual op amp.
Fortunately to switch the filter frequency (second order) all you need are 2 poles.

This block diagram show the Side scaled by 1/2.
"m" is the filter transfer function and within the passband of the filter the value of m is "1."
Image
Nuemann VAB84 Block Diagram Showing Low Pass-Filtered "Side" Converted to "Mid" By Steering Into Left and Right Using Summation.
Post Reply