Simple mid/side matrix w/ opamps

I've got a version of this breadboarded already, and it works. I've since been through the idea with a bit more thought, and this schematic is the result.

What do you think? It's meant to be used as an insert for a pair of mixer channels, or similarly "safe" line-level applications. I figure that I need to buffer the inputs a bit before mixing, though I could probably get away without doing that. The gain of the summing amps is meant to add 6dB (2x voltage) so I can give the volume pots a bit of "gain".

(23-9-2005 schematic has been updated to reflect discussion in this thread)


I'm actually going to have two of these in the box so I can convert left/right inputs to mid/side, screw with the levels, and then back to left/right. Meant for tweaking the width of my XY stereo mics, as well as a generic box for goofing around with M/S techniques. There will be normalled outs/ins between the sections.

Oops - I can see that I got my summing amps mixed up. The top one should be the right output, the bottom one the left output.

Edit: OK, it's fixed.

[quote author="Scodiddly"]Oops - I can see that I got my summing amps mixed up. The top one should be the right output, the bottom one the left output.

Edit: OK, it's fixed.[/quote]

As long as you are not adjusting the controls in mid-recording it looks fine. Otherwise I would do some more a.c. coupling around the potentiometers so that op amp bias currents and offset voltages don't give you scratchy wiper noise.

Much as I like to avoid unnecessary coupling caps in the signal path, I've been won over to a.c. coupling with potentiometers after realizing just how senstitive the ear is to such noises. You could use FET op amps with super low offset voltage (or trimmable offset voltages) and negligible bias currents in lieu of the extra caps.

I assume you are processing nice big heallthy line-level signals here so that noise is not a major concern. If it were there would be other somewhat lower noise configurations.

When I had a somewhat similar matrix synthesis design chore once, I used discrete devices at the input in a phase splitter arrangement, generating inverting and non-inverting signals at the outset, and then combined them thereafter.

This is kind of an early prototype, "goof around" sort of project. I've cobbled together a m/s matrix before with mixer inputs and tried plugins, but really a dedicated box is what I've been wanting. I'll probably do a bit nicer version inside a stereo mic preamp in the future, sort of an "everything stereo mic" box.

It's pretty fun to play with just the breadboard version. The very first thing I put in to test it was Tom Waits' "Rainbirds", since I'm learning the bass part on that. With upright bass dead center and the piano in stereo, I found that my output had the bass in the Mid output and the piano in the Side output and with a surprisingly good separation.

Update: I built the box version, and it works! Nothing that fancy to look at, since it's crammed into a 4x3x2 aluminum BUD box. Now I just need to glue a couple things down and let it dry overnight.

One interesting thing about the M/S matrix is that it'll convert L/R to M/S as well. So my box has two of the above circuit, with a normalled conection between them. I've now got a box that'll do L/R input through M/S processing and back to L/R out, or vice-versa, or will do two simultaneous M/S->L/R or L/R->M/S conversions.

Hi Scodiddly,
I would make one suggestion.... the way the pots are arranged, you can't retain accurate sums on the L and R when you alter the pots.
Suggest that you use 10K pots in the feedback of the M and S buffers, then the 'sum and difference' ICs will be fed from zero impedances at all times... and the thing will work better!
Ted Fletcher

Hmmm... I don't see how I've got a problem with the current circuit. The whole virtual ground thing should be working in my favor, keeping the impedance at the volume pots from making any difference to the sum & difference amps.

The gain structure is only correct if the pots are fully up.
With the pots halfway, the gain of the signals into the 'left' amp are zero, but the gain of the signal into the pos input of the 'right' IC is very roughly -7dB (without going into all the sums.)
The fix is to make the feedback resistors variable on the 'M and S' amps and drive the 'L and R' amps direct from their outputs, that way the gain stays predictable because the impedances driving both the VE amps and the non-inverting amp stay the same, so their relative gain stays the same.

I love M and S! :grin:

The gain structure is only correct if the pots are fully up.
With the pots halfway, the gain of the signals into the 'left' amp are zero, but the gain of the signal into the pos input of the 'right' IC is very roughly -7dB (without going into all the sums.)
The fix is to make the feedback resistors variable on the 'M and S' amps and drive the 'L and R' amps direct from their outputs, that way the gain stays predictable because the impedances driving both the VE amps and the non-inverting amp stay the same, so their relative gain stays the same.

I love M and S! :grin:

Well, you're going to have to show your work. :wink:

I just did some tests, and at the worst I'm only a couple of dB off.

OK, I was too hasty.... it's not as bad as that: in fact the gains match perfectly when the faders are set at electrical halfway.... but because the loading on the wipers of the faders is not the same, there's likely to be some deviation at intermediate points.
The arithmetic started to get tedious.... I must put the values into my analyser.... :sad:

One point to remember.... when going from L/R to M/S and back again, the gains of the amplifiers have to be very precise; any error gets translated into crosstalk between the channels.

Ted Fletcher

Well, you're right anyway. I switched over to linear-taper pots in the feedback loops, and it lines up much closer now - more like 0.1dB.

I was using voltage divider pots because I hate what happens when feedback-loop pots get noisy - they get REALLY NOISY!!!!

Amen to that!
But it's worth it with good pots.

Glad it's working well.

Ted

This project justified itself last night, in an unexpected way. I'm participating in an online recording project, playing bass. So far we've got a rough mp3 with sequenced drums & bass, plus some scratch guitar and vocals. I wanted to record a different bass part, but the sequenced (synth) bass was pretty distracting. But since bass is almost always dead center with no effects, I just ran the mp3 through my mid/side box and dialed the center down a bit. Presto, much less bass!