MidSide Enc Dec question

[richiyobs]

Hi folks
trying to have a mid side encoder decoder with insert points to use with external devices....

is this a possible way?

thanx a lot for your attention
best

 

[gyraf]

Does not compute. Try posting a schematic?

/Jakob E.

 

[ccaudle]

What software is that schematic file from? Not very convenient to post a raw schematic file with no information on which schematic editor will open it properly. Export as PDF is the usual way to post a schematic.

 

[richiyobs]

sorry .... here is a pdf
software is LTspice

just breadboarded and seems to work.... but i can only use a pot to control Sum going to Decoding

attenuating sum, can listen to the side part of the material
so side one must be always present to get a rebuilded image

the pot after the Difference amp make the decode works strange, how i can say... if totally attenuated signal is bad rebuilded.... polarity reversed maybe ... i got only one channel on the stereo mix, and it s strange cause with no side should get only MID....instead i get MID but only on one side of the meters

so no pot for Side level, cause to make it works side signal must be always present at R17 and R18

insert point should be not affected, i hope..


thanx for your attention

 

[abbey road d enfer]

The way the summer and substractor are done is not very elegant. It could be better. Replace the 1k's with 10k's and short R9, 10, 13 & 14.
Now the difference output has a different impedance between the legs, which is not very good for balancing impedance. But the main issue is that crosstalk is not constant, which expalis the strange behaviour regarding stereo image.
The passive matrix you use at the output works only if the signals are properly buffered.

 

[richiyobs]

ok i see.....Thanx for your suggestions
i ll check better designs....

best

 

[ccaudle]

My first sketch would be differential to single ended buffers for each input, then diff-amps to do the differencing. Probably unity buffers, then a single diff-amp for L, and dual diff-amps for the R input stages, so you get CM rejection on each input, and end up with L+, R+, and R-. Or could just use the combo buffered diff input stages like INA1650 and an inverter on the R+ signal to generate R-. With two diff-amps on the output stage you can get L-R, and L-(-R), i.e. L+R to get S and M signals. I would have to think about gains, I don't remember if there is a convention of having attentuation on the sum (M) signal so you don't get overload problems or not. Could wire it all up with a handful of TI INA parts. I haven't checked availability lately, I know TI's inability to make small switching regulator parts has been causing me big headaches for over a year, but I haven't looked to see whether their analog amplifier parts are similarly constrained.

 

[abbey road d enfer]

I would have to think about gains, I don't remember if there is a convention of having attentuation on the sum (M) signal so you don't get overload problems or not.

There is still a debate about it. Some sources suggest direct summing, such as M=L+R, without any consideration for the level increase.
For two L and R perfectly correlated signals, which is the case wih a single mic picking up a source placed on-axis, the M signal is 6dB louder, which may result in overload. In order to restitute the original levels when decoding, a 0.5 factor must be applied, which gives L=1/2(M+S) and R=1/2(M-S)

In order to compensate that, many recommend, aptly IMO, a 0.5 factor at coding, which prevents overload, resulting in M=1/2 (L+R) and S=1/2 (L-R)

The disadvantage of both methods is that the coding and decoding matrixes are different.

Now, there are some who advocate 3dB attenuation when encoding, accompanied with same 3dB attenuation at decoding, which gives
Left = (M+S)-3dB Right = (M-S)-3dB at coding, followed by decoding Mid = (L+R)-3dB Sides = (L-R)-3dB
The only advantage is making the MS to LR decoder identical to the LR to MS coder, but it still presents a risk of overload due the potential 3dB increase in level.
I think it's not of much practical use, since one of the major considerations for using MS is the ability to change relative levels.
Wayne Kirkwood, in the Pro Audio Design site, advocates a hybrid solution, where the level of the S signal can be increased by 6 or even 12dB, with corresponding attenuation at decoding.

The possibility of having out-of-phase L and R signals, which may result in overloading the resulting S component, must not be discounted, however most musical recordings show a large correlation between L and R, which tends to ascertain the near 6dB increase in the M signal and a significantly lesser S signal.
However, for nature or sound effects recordings, the possibility of largely decorrelated L and R is real.

 

[richiyobs]

following Kirkwood example
attached 2 attempts on the same design
builded a proto based on -same polarity- schem.... but strange behavior decoding out even if simulation is fine... so reversed the difference on LTspice to check if it s important in that stage of the signal.... but same result, the out is still well decoded ..... so at this point i m a bit confused

what s wrong? what i m missing?

thanx in advance for your support

best

 

[richiyobs]

this is the proto on test....

thanx again for your attention

 

[richiyobs]

Checked my proto... I did some shorts soldering!
Also added pot after insert point.... And is working now
I can finally listen to Encoded signals
But not sure about difference amp polarity

 

[abbey road d enfer]

following Kirkwood example
attached 2 attempts on the same design
builded a proto based on -same polarity- schem.... but strange behavior decoding out even if simulation is fine... so reversed the difference on LTspice to check if it s important in that stage of the signal.... but same result, the out is still well decoded ..... so at this point i m a bit confused

what s wrong? what i m missing?

I believe you need to check circuit analysis. It's a bit difficult to figure out the circuit cause of poor resolution.

 

[abbey road d enfer]

Checked my proto... I did some shorts soldering!
Also added pot after insert point.... And is working now
I can finally listen to Encoded signals
But not sure about difference amp polarity

I see you have opted for 12dB attenuation, which IMO is not standard, unless you did it to accomodate reduced headroom somewhere in the chain...?

 

[richiyobs]

I believe you need to check circuit analysis. It's a bit difficult to figure out the circuit cause of poor resolution.

Fine thanx for checking it!
And thanx to your notes I ll fix it...
What do you mean for poor resolution in analysis? Any parameters into Ltspice? Or graphics...

 

[richiyobs]

I see you have opted for 12dB attenuation, which IMO is not standard, unless you did it to accomodate reduced headroom somewhere in the chain...?

I Did wrong calculations... I would realize a 6dB pad at input buffer to not overload the sum Amp... So fixing the gain at first buffer I will get it working properly... Right?

 

[richiyobs]

I believe you need to check circuit analysis. It's a bit difficult to figure out the circuit cause of poor resolution.

here at the out... it means i get only half portion of the coded original signal , or half level?
sorry for troubling

 

[abbey road d enfer]

Or graphics...

Graphics. i had trouble finding the orientation of opamps.

 

[richiyobs]

attached a better res schem

and input vs encoding vs out

 

[richiyobs]

Running Rew for real measurement
Attenuating side made out wave as the sum one... With all components the original signal, 2 sines 1k@L 500Hz@R, is complete

 

[abbey road d enfer]

here at the out... it means i get only half portion of the coded original signal , or half level?
sorry for troubling

Half ot the mathematical level. But it doesn't matter as much as making the level compatible with the eaquipment it is connected to.