mid-side conversion with transformers, will this work?

Hello.

I was trying to convert L+R stereo audio into M+S stereo using transformers, but screwed up bad with wiring last weekend. I think I see the error in my ways, but I need help verifying this.

What I have here is a pretty generic line-level application. A standard 1+1:2+2 transformer input, or alternatively 2+2:1+1 output. What's between input and output does not interest us in the conversion. It could be some 1:4 stepped up thing, or maybe 1:2+2 class AB amp. In the example below, it's wired 1:4.

I can see it clear as as day when written down or programmed in C.

Left + Right = Mid
Left - Right = Side

But I'm confused with this simple math on the transformer windings.

Is the following correct?



1. Left signal goes to both transformers top windings in phase.
2. Right signal goes to transformer 1 bottom winding in phase, creating a mid sum.
3. Right signal goes to transformer 2 bottom winding out of phase (reversed), and this is the side sum.

Thanks!

Haven't really checked it in detail nor build it, but here's the Jensen MS transformer matrix wiring for comparison...

http://www.jensen-transformers.com/as/as055.pdf

Seems ok
left and right sine waves in phase = full signal on traffo 1. no signal on traffo 2.
left and right sine waves out of phase = no signal on traffo 1. full signal on traffo 2.
I'd say wire it up and see how it performs

Kingston said:

Hello.

I was trying to convert L+R stereo audio into M+S stereo using transformers, but screwed up bad with wiring last weekend. I think I see the error in my ways, but I need help verifying this.

What I have here is a pretty generic line-level application. A standard 1+1:2+2 transformer input, or alternatively 2+2:1+1 output. What's between input and output does not interest us in the conversion. It could be some 1:4 stepped up thing, or maybe 1:2+2 class AB amp. In the example below, it's wired 1:4.

I can see it clear as as day when written down or programmed in C.

Left + Right = Mid
Left - Right = Side

But I'm confused with this simple math on the transformer windings.

Is the following correct?

1. Left signal goes to both transformers top windings in phase.
2. Right signal goes to transformer 1 bottom winding in phase, creating a mid sum.
3. Right signal goes to transformer 2 bottom winding out of phase (reversed), and this is the side sum.

Thanks!

Click to expand...

It just can't work! You've just forgotten that transformers are not active summing units. When you connect two sources on two windings of a transformer, the almost 100% coupling sends one into the other, that means that both your outputs are passively (and badly) mixed. When in addition you couple them out-of-phase in the second transformer, you mix them out-of-phse. The overall result is low distorted signals.
In order to use transformers as summers, you must use two separate transformers 1:1+1 (or 1:0.7 + 0.7 or 1:0.5 + 0.5, depending on what theory is considered), one receiving the L channel, the other the R channel. You put in series one secondary of the L xfmr with one secondary of the R xfmr to obtain M. You then put in series one secondary of the L xfmr with one secondary of the R xfmr out-of-phase to obtain S.
This subject has already been discussed in this forum, with explanations why the transformers should not be 1:1:1 (basically, if L and R are at reference level, M would be 6dB above)

But it's done directly on transformer windings on Fairchild 670, isn't it? certainly works well there.

I'm trying to figure out the formula for the windings and ratios pictured above.

Also, I don't care about volume differences resulted from the ratios in the conversion, like the 6dB louder mid channel.

Just how it can be done. I did not quite understand the above explanation.

Kingston said:

But it's done directly on transformer windings on Fairchild 670, isn't it? certainly works well there.

Click to expand...

The signals coming from the first stage are not applied directly to the 3rd xfmr. The 150R resistors provide necessary load isolation.

I'm trying to figure out the formula for the windings and ratios pictured above.

Click to expand...

I don't understand what formula. What are the variables, what is the data, what is the unknown?

Also, I don't care about volume differences resulted from the ratios in the conversion, like the 6dB louder mid channel.

Click to expand...

OK, it's your privilege, so that means you have a wider choice.

Just how it can be done.

Click to expand...

Doing what?

I did not quite understand the above explanation.

Click to expand...

A transformer is a passive unit. The energy coming in reflects coming out. Let's assume all windings are identical. If you feed one winding with 1V, you will get 1V on all windings. The distinction between primary and secondary is arbitrary. So if you feed two windings with separate signals, they will reflect into each other and be mixed. Now if you put isolation resistors (or better, an active buffer) the mixing effect will be attenuated (or even totally avoided).

abbey road d enfer said:

Kingston said:

But it's done directly on transformer windings on Fairchild 670, isn't it? certainly works well there.

Click to expand...

The signals coming from the first stage are not applied directly to the 3rd xfmr. The 150R resistors provide necessary load isolation.

Click to expand...

oh.  


I did not realise this. I thought the transformers alone did the trick.

How can I work out what resistors to put there, looking at transformer datasheets, and knowing the ballpark for expected input and output impedance of the transformer?

http://i47.tinypic.com/2hx6qrn.jpg

Not sure why this wouldn't work?

Yes, this is a much better configuration (there's a different one, which consists in reversing the inputs and outputs - that's the beauty of L/R to M/S conversion, it is symetrical so the same configuration works for M/S to L/R). No stress on the outputs, better frequency response, no attenuation...

I don't see any resistors.  ???

L and R are still being mixed in the windings, "almost 100% coupling" as you said. wouldn't that cause the exact same thing as you mentioned on the first post?

Like I said in the beginning of this thread, I'm confused. Now even more than in the beginning.

abbey road d enfer said:

In order to use transformers as summers, you must use two separate transformers...

Click to expand...

I did draw two separate transformers in the first post, just not in the same configuration as onlymee.

Kingston said:

I don't see any resistors.  ???

Click to expand...

No need for them, because the transformers are not parallel driven; they are driven in series.

L and R are still being mixed in the windings, "almost 100% coupling" as you said. wouldn't that cause the exact same thing as you mentioned on the first post?

Click to expand...

Look at the way the windings are connected, see how the voltages are shared.

abbey road d enfer said:

No need for them, because the transformers are not parallel driven; they are driven in series.

Click to expand...

Yes, they have to be wired in series. In your pic in 1st post they were wired in parallel.

[OEP]
http://i47.tinypic.com/2hx6qrn.jpg

[Jensen]
http://i48.tinypic.com/xqbcl5.png
http://www.jensen-transformers.com/as/as055.pdf

[BBC encoder & decoder]

I see it now. thanks all of you.

I suppose in the case of series connections done in my picture, it would become exactly like the OEP wiring above.

But with a catch. It's no longer 1:4, but in fact 2:4 ratio. That somewhat limits my options. I suppose another way would be to wire it like the above BBC example, with encoding on secondary, but then I could not have a center tap anymore.

I wonder how much attenuation those resistors cause in the parallel version.

Kingston said:

I see it now. thanks all of you.

I suppose in the case of series connections done in my picture, it would become exactly like the OEP wiring above.

But with a catch. It's no longer 1:4, but in fact 2:4 ratio. That somewhat limits my options. I suppose another way would be to wire it like the above BBC example, with encoding on secondary, but then I could not have a center tap anymore.

Click to expand...

Earlier on you said you didn't care about level!?...

I wonder how much attenuation those resistors cause in the parallel version.

Click to expand...

The attenuation is directly related to the crosstalk protection you want to achieve, and depends very much on the impedance of the source.
BUT WAIT!
I've checked the complete cct dgm of the 670. The M/S encoding is done at the input by putting half primaries in series; same for the outputs. I'd been fooled by the complex output switching.

abbey road d enfer said:

Kingston said:

I see it now. thanks all of you.

I suppose in the case of series connections done in my picture, it would become exactly like the OEP wiring above.

But with a catch. It's no longer 1:4, but in fact 2:4 ratio. That somewhat limits my options. I suppose another way would be to wire it like the above BBC example, with encoding on secondary, but then I could not have a center tap anymore.

Click to expand...

Earlier on you said you didn't care about level!?...

Click to expand...

with options I meant my choices for transformer, so I don't have to go shopping. But I figured it out already.

I will use 1+1:2+2 input transformers, mid-side conversion done on secondaries, thus keeping the step up at 1:4. But I will use 4+4:1+1 on output where I need a center tap, with conversion back to left-right done again on the secondaries. with series secondary wiring the output becomes 2ct2:1

I don't understand why you need a center tap...

That has nothing to do with the mid-side conversion, so don't worry about that little detail.

I'm just musing around, trying to keep things generic.

Not really needed, just for the sake of having it complete. Here's the encoder/decoder wiring of the 670. When comparing to the BBC typology, the transformers are the other way round or, rather, the wirings for adding/substracting (L+R=M & L-R=S) are on the "outside". And there seems to be a centre tap on the input ("inside") -- haven't checked what it's used for though.

INPUT [L/R --> M/S]



OUTPUT [M/S --> L/R]

Update only.

(Had been PM'ed about it, cos links above are gone.)

Attached is the Jensen encoder. For decoder just read from right to left.