MS Decoding/Encoding transformers

Hi,


 Today I was talking to the wonderful Brian Sowter today about MS transformers. Her was talking about doing a special based on their 3603. Now I am confused. He was talking about 1:1+1, Basically. I was was trawling through the NET, and I found this Neumann brochure for a Z 240 Matrix transformer. This is 1:0.7+0.7.
 Why should you want 1:0.7+0.7, not 1:1+1? Is Brian wrong(heaven forfend and absolutely NO disrespect intended!), or is this Neumann transformer not intended for passive use perhaps? Confused, me . . .

  Any input gratefully received, as always.


    Kindest regards,


   ANdyP

HI again,


  I also found this. I know it is for microphones, but could I do something similar with line level? Should I substitute the 250 ohm resistors for, day, 600 ohms? I guess this would need buffering in and out(?)

 Surely it can't be THIS easy . . . just 4 resistors? I am guessing it would both encode and decode, yes?


   Kindest regards,



         ANdyP

How about this?
Bruno2000


http://www.picocompressorforum.com/forum/php/viewtopic.php?f=6&t=112

Hi Bruno,


  thanks for the link, but they are dealing with ACTIVE MS en/decoding. I am interested in this too, but my questions refer to PASSIVE decoding, firstly with transformers, and secondly with a resistor "bridge" thingy . . .


  kindest regards,


      ANdyP

Ratio isn't some terribly crucial thing.  You could go crazy and do it with a pair of UTC A-10's, if you really wanted.  Every set I've built has been with a pair of 1:1 transformers, with one side of the 600 split into two sections.  1:0.5+0.5, in other words.  My wild guess with the Neumann transformers might be that they are attempting to step up slightly to account for losses in the cores.  That doesn't really add up either.  No idea. 

Hi Doug,


  thanks! Glad I'm not the only confused one . . .


  Kindest regards,


  ANdyP

HI,



  anyone know anything about these Z 240's? I found 'em in a sales catalogue from 10 years ago. They were priced at $870 for the pair . . . must be fashioned from pure un-obtanium! - and I need FOUR such similar beasties . . .


  Perhaps CJ has "Tested" 'em sometime?


    . . . . CJ? . . . .



    ~Kindest regards,



        ANdyP

can i use lundahl 1540 in 1+1:1+1 wiring ?
or do i need to use it 1:1+1 wiring

strangeandbouncy said:

HI again,
  I also found this. I know it is for microphones, but could I do something similar with line level?

Click to expand...

You could, but remember it works only for floating outputs.

Should I substitute the 250 ohm resistors for, day, 600 ohms?  I guess this would need buffering in and out(?)

Click to expand...

You could not load it with 600 ohms directly. The open-circuit equations are:
L= M/2 +S/2 R= M/2 - S/2 

 Surely it can't be THIS easy . . . just 4 resistors?

Click to expand...

In essence, it's not much simpler than two transformers. It fulfills the same equations.

I am guessing it would both encode and decode, yes?

Click to expand...

Yes, it would

strangeandbouncy said:

Hi,
Today I was talking to the wonderful Brian Sowter today about MS transformers. Her was talking about doing a special based on their 3603. Now I am confused. He was talking about 1:1+1, Basically. I was was trawling through the NET, and I found this Neumann brochure for a Z 240 Matrix transformer. This is 1:0.7+0.7.
 Why should you want 1:0.7+0.7, not 1:1+1? Is Brian wrong(heaven forfend and absolutely NO disrespect intended!), or is this Neumann transformer not intended for passive use perhaps? Confused, me . . .
  Any input gratefully received, as always.
    Kindest regards,

   ANdyP

Click to expand...

Actually, one could debate about the fundamental equations of LR-to-MS conversion. If you consider using 1:1+1 xfmrs, the results for heavily correlated program will give M=>2 and S=>0, which is not the proper way to handle headroom.
A 1:0.5 + 0.5 xfmr will give in the same conditions M= 1, which is the key to use the same operating level on MS components as LR, and ensures perfect capability of handling mono signals.
Now, for moderately correlated program, using cos(alpha) = 0.707 (which is optimistic IMO - most modern recordings are somewhat more correlated) the 1:0.7 +0.7 xfmr will give M=1 and S=1. This will give M= 1.4 for mono input (3dB overload).
It is clear that there is always a risk, depending on the stereo image, of overloading or undermodulating, which makes it necessary to have some degree of level adjustment.

Hi,


  Abbey Road D'Enfer, many thanks for your reply. It makes more sense now! Why not go for a smaller output ratio, say 1: 0.3+0.3? Surely it is easier to make up gain than deal with the consequences of overload? or does that take you near the saturation of the output windings?


  Hmmmm...


  ANdyP

All in all, you could use almost any good performance xfmr with a dual secondary, within the limits of noise build-up and headroom preservation.
A 1:0.1+0.1 xfmr will not saturate later than a 1:1+1, it's the primary that counts. (For the purists, this is a first-order approximation, since secondary load affects saturation - in fact more load on the sec decreases flux)
But IMO, active electronics is the ideal. MS matrixing is such a trivial operation for electronic circuits that I reckon nothing can justify the use of an expensive xfmr in comparison to its (relatively) poor performance, especially considering it needs to be surrounded by some complementary active elements.
The common motivations for the use of xfmr's in MS matrices are:
- adding "character": I would rather use a xfmr somewhere in the chain just for that and use a clean dependable matrix.
- minimizing parts count in the chain: this is an anthropological (think audiophool) point of view, that nothing justifies technically. A single xfmr does much more deterioration to a signal than a whole rackful of properly designed active circuits.

Isn't the original reason for 1:0.7+0.7 coding also the fact that usually (always) one decodes somewhere too, and with this ratio one can use the same trannie, or coder/decoder box on both ends and retain level through the chain. It makes a lot of sense.

The math is very simple. If you use 1:1+1, you get:

Coding:

M=L+R
S=L-R

Decoding from above:

L=M+S=(L+R)+(L-R)=2L
R=M+S=(L+R)-(L-R)=2R

So, you get 6dB boost. Not nice. You have to have "divide by two" somewhere, or use "divide by SQRT(1/2)" in both transformations. There we get that "obscure" 0.707.
I would love to have those trannies available more widely but the heyday for passive MS-coding is gone, and so are most of the manufacturers.
Some plugins and mastering software use 0.707, but because you obviously can't go over zero, for example Weiss uses the divide by to in coding. (If I remember)

-Jonte

Sorry, divide by SQRT(2) of course. 1/SQRT(2) = 0.707.

Good clarification, Jonte. It makes sense on a manufacturer POV.

I also think that the impedances get "reflected" better that way. Used to be important when gear had 600 Ohms inputs. With 1:1+1 or 2:1+1 you will cause some mismatch.

Hi Jonte,


  many thanks indeed for that. Makes perfect sense. Question is, where can one buy such a transformer? is anything available now?


  Kindest regards,


    ANdyP

Thank you for thanking

I'm certainly took much stuck to using Lundahl stuff and don't know some of the smaller manufacturers, shame on me. How about the Japanese? Tamura, Tango? One customer of mine actually knew some 1:0.7+0.7 trannies, but I wasn't familiar with the brand and have forgotten it totally. I only remember that they were very expensive. Probably some high end stuff.

How about asking Sowter to make them? Lundahl is probably not willing to design anything for such a small market, unfortunately.

I feel that the impedance thing might have been one design argument, really. And it is not essential anymore, so manufacturers suppose we are happy with general purpose trannies for the job.

Stupid question: Are people generally aware of this level and math issue? I'm only asking because I didn't pay any attention to it before I actually had to design MS-things for mastering use.

Best Regards,

Jonte

strangeandbouncy said:

Question is, where can one buy such a transformer? is anything available now?

Click to expand...

Hopefully it does not look like spamming, but as I said in another MS thread I could make either toroidal, or UI any ratio you need, no problem.

Best, Mark

Andy, Jonte, you may be interested to know that in stereo FM transmission, the equation is different. The ratios are 1:0.45 + 0.45. That leaves 10% for the 19kHz subcarrier. This is very important because FM transmission operates as close as possible to the max.
Why do I mention that?
Because it's not very dissimilar to what we see today in mastering - to my knowledge one of the areas of sound production the most concerned by MS. I know that some mastering engineers, when they receive a file that's already normalised at 98 or 99%, attenuate the signal digitally to operate their analog equipment at +4-ish, in that case, whatever the matrix coefficients, doesn't really matter. But some leave the signal as close as possible to the level it was delivered to them, and then they have to run some of their analog stuff pretty close to clipping. In that case, a coefficient of 0.5 is safe, 0.7 hazardous, and 1:1 a disaster.