Stereo Width Experiments
I am getting interested in building a hardware stereo image widener circuit. I have seen a few ideas floating around based on the Middle(M) +Side(S) technique. The basic premise can be summarized into the following equations:
Encoding M and S
Middle = L+R
Side = L-R
Translating that into basic op-amp circuits, the Middle signal could be created with a summing amplifier.
I am not sure if a non inverting summing amplifier will be as stable as inverting amplifier, but its a simple to test to find out. If Rf is a short and Ri is open then the Middle = Left + Right.
For the side signal a difference amplifier can be used.
If all the Resistors are the same value then Side output = Left-Right
Decoding M and S
The basic equations are as follows:
L = M+S
R = M-S
Decoding can be done using the same summing and difference circuits, just with different inputs.
You need a bit more circuitry if you want to handle balanced inputs and outputs. You also have to cater for the overall 6dB gain that you get if you go LR -> MS -> LR. You can do this by giving each route a gain of -3dB, or just arrange the first unit to have -6dB and the second 0dB. This avoids possible internal overload.
I've built lots of different types of M-S en/decoders, but the one that proved most robust in practice was a passive box using 1:1+1 transformers. It was quite critical on the type of construction that the transformer used, as the splitting/recombination tended to break down at low frequencies.
You might also look at using THAT1240 and 1246 devices, since they have internal resistors of the necessary precision.
Thanks Boswell, I am a fan of the THAT IC's as well. Doing it will transformers would be interesting and challenging!
Surely while these circuits are interesting they do not increase width because that is predetermined by the actual audio. Maximum width is simply what the side microphone recorded. Minimum width is what the mid mic recorded. Your options are between the two. What are we trying to do here? The circuits here don’t seem to do anything that the old split and invert one and have two channel faders on the mixer didn’t? Have I missed the point?
This has little to do with width, at least in the first instance - it's all to do with the precision of the decoding. Once you have an accurate L-R decode, this can be processed and mixed normally, although the decoded L-R as a stereo pair has a sound that could not have been recorded using X-Y.
When I'm doing both PA and recording for a live event, I sometimes use M-S for miking a single wide instrument such as a hammered dulcimer. On these occasions, I send the M-S as two tracks to the recorder but just send the M channel to the PA if it's mono, and decode to L-R on the fly if it's stereo. To give me both width and pan capability, I use an insert lead on the (balanced) S pre-amp output to send phase and anti-phase to consecutive tracks (at 6dB lower) on the PA mixer to blend with the M channel, giving me the width and pan for the occasion. A separate M-S decode and mix is always necessary for the CD of the event.
Hi Paul, your correct. I wanted to start blog about the thought process behind a design. Starting with M+S was my first stop. I want to build that first, measure the results and then potentially move on to other methods. Possibly explore short delays (sub 10ms) and variable phase (ALL pass filters), etc...
I just havent got to it yet :)