ANDing Audio amp

[Ian_K]

Is there an easy way to AND two audio signals together so the output will just be the similarities but not the differences. I have a feeling their is an easy way but I can't think for the life of it what it is or find any info. I have visions of getting bogged down with A to D and D to A converters which I don't perticually want to do.

Regards

[ppppenguin]

I think you may be trying to open a large can of worms here. Audio signals are analogue so the the logical AND operation is meaningless. The nearest analogue equivalent to logical AND is to instantaneously select the larger of the 2 signals. I have used this in video processing where it's called a non-additive mix.

"Similarities" and "differences" are very vague terms. Picking out what's subjectively similar in audio signals can be a very processing intensive task.

[Paul Stenning]

If you wanted the differences and not the similarities, and didn't care about the phasing, inverting one then summing would do the job. That is how the rear channel of basic analogue surround-sound systems was derived, usually followed by a delay of around 20mS.

I can't think of an easy way of doing what you are trying to achieve though. You need something that compares the two and only keeps things that are common to both, and rejects everything else. I think it would need to be done digitally.

[llama]

To some extent it depends what you want to do with it. Suppose you feed the two signals into left and right speakers and have an excellent stereo set up. The sound-stage should give you a rock-solid centre image of the common stuff. But I can't think of a way to extract it electrically, unless some surround sound processor can generate a centre-speaker signal that's better than just adding left and right.
Graham

[ProdMan]

Feed both channels into a diff op-amp and a summing node on another op-amp.

The output of the diff amp then connects in to the non-inv input of the other op-amp.

Buffering etc may be required on the A and B inputs if fed from a hi-Z source.

Something like the sketch below - if my frayed mind is working...

The non-inv input on the second op-amp will need an R to 0V

Or use a DSP! Not for the faint-hearted.

[Ian_K]

Thanks PodMan for the circuit and everyone else for their comments.

[Paul Stenning]

All that circuit does is produce 2B, which is just the same as one channel amplified. I can't see how that solves the original problem.

[Skywave]

Originally Posted by Ian_K

Is there an easy way to AND two audio signals together so the output will just be the similarities but not the differences.
Regards

Having thought at length about this, I've concluded that the answer is "No": 'no', that is, in the sense that it isn't easy.

Conceptually, my take on this is:

Define your term "similarities". Let's say it's this: that at any instant in time, the instantaneous frequency and amplitude of each of the two signals is each within a pre-defined "window" which we will call "equal". (Let's leave phase out of this for the time being, to try to keep it simple).
Using a time-sampling method and comparators, when the freqs. are 'equal' in the time-sample and ditto the amplitudes, then arrange for the two comparator outputs to produce the required resultant signal. You'll need some sort of filtering on the O/P to remove the switching noise arising from the sampling. If the highest freq. in the input signal is F, then Shannon's Theorem tells us that you'll need to switch at 2*F as a minimum sample rate.
By definition, this processing will produce a signal which will sound somewhat distorted compared to each of the pre-processed signals - but it will meet the requirement of "AND" - ing two "similar" analogue signals.

Just how you actually achieve this, circuitry-wise, I'm not sure. I'll leave that to the analogue Gurus!

Al / Skywave.

[arjoll]

I've thought about this a couple of times. Everything I could think of using op amps etc looked like it would just give me one of the source signals, so that was obviously the wrong track.

Getting differences is easy - an op-amp wired as a differential amplifier. Subtracting this from the input isn't going to give anything useful though - you get up with something like (L+R)-(L-R) = 2R. Adding it to the output - (L+R)+(L-R) = 2L. Looking suspiciously like we're decoding FM stereo now

If by similarities you're wanting to not have any cancellation of one the signal by the other you could rectify them before summing (1N914 in the feedback loop of an opamp). Instead of cancellation instantaneous inputs of say +5 and -5 would turn into +5 and 0 so output (unity gain) of +5 instead of the 0 you'd get without rectification. You'd basically have a steadily varying DC, but I'm guessing that putting a blocking cap on the output would give you a (kind of) usable signal again.

I do think that all you'd end up getting is a nice distorted sound that would be great for a guitar pedal

[Kat Manton]

Hi,

It sounds like what you're essentially trying to do is to synthesise a centre channel from a stereo source.

I've found a discussion about this here - it seems like there's no easy way to do it with analogue electronics. Even with DSP, it seems like you'll end up with some distortion or other artifacts.

Cheers, Kat

[arjoll]

That's an interesting discussion. I would have thought that synthesizing a centre channel would be a matter of (L+R)/2 but I understand the issues with narrowing the stereo image.

Depending on the content I wonder if putting a bandpass filter in the 'fake centre' channel for (say) speech frequencies or even a lowpass filter with cuttoff at (say) 3-4k would produce the desired result. It would help knowing what the desired result was but this would potentially steer speech frequencies to the centre channel. Maybe with subtracting a portion of this from each other channel.

Something in DSP would be a bit tidier though, being able to have a look at what content was in each channel so that (for example) only speech that was the same level in each channel was steered toward the centre. Don't ask me how on earth you'd do that

(these last two paragraphs assume that the desired result is moving speech to a centre channel)

[ProdMan]

Sorry, I was trying to show the principle rather than design an entire system...

The idea was to do the same thing to derive 2A and then sum 2A and 2B together.

If someone could express what is required in math terms, it may be easy to use analogue computing methods to achieve...

[boiss]

Some ten years ago I made an amplifier with active crossovers. ie 2 x Treble amps 2x30W, 2 x Mid amps 2x30W and a single Bass Amp 50W. Each amp had it's own speaker. The Bass amp used some Op amps to sum the base after the active crossover network. It appeared to sound OK, but I guess that the Bass part is somewhat simpler.
Cheers,
Boiss

[Ian_K]

Hello All

I'm sorry I should of filled you all in as to what I am up to.

If you play an old mono record 78, LP etc on a stereo turntable the mono sound as expected will hang in the middle of the sound field. But the hiss pops and crackle will often to be more on the left or right. What I am trying to attempt is to create a box that will separate only what is in the centre of the soundfield and output it. I don't know wether anyone has attempted this before but I haven't been able to find any info anywhere.

I have been trying this out using software filters with mixed results until I found an addon 'Centrecut' for 'Winamp' in the link that Kat provided. http://www.moitah.net/download/latest/dsp_centercut.zip
Thanks Kat.

This can be set to output just the centre channel. I have fed several Stereo recordings of 78's through this with startling results. It can reduce the background noise by about 75% across the whole frequency range with no real distortion of the sound.

I have now had a go at building ProdMan's circuit to see how well it performed. While the first stage differential amp outputs the differences (only the noise from a record) as it should when it comes to using that to cancel out the noise using the second op amp it just turns into sound soup instead. I am unfortunately very rusty on op-amps, Is their something additional I need to add to the circuit to make it work?

[Kat Manton]

Hi,

What I've found when playing worn mono vinyl and 78s with a stereo cartridge etc. is that summing the left and right channels improves the signal/noise ratio considerably anyway - it'd be worth comparing "playing them in mono" with more esoteric techniques.

Regards, Kat

[Ian_K]

Very true Kat. As can inverting one channel.

I have put up a before example http://www.iankarley.plus.com/in.mp3 and an after at http://www.iankarley.plus.com/out.mp3 if anyone is interested.
The before is straight out of the phono amp with just a bit of equalisation applied. the after was produced just by feeding it through the filter I mentioned above. The recording was also compressed before being processed which I think is generating some artefacts as well. This is only intended as a rough and ready demo.

Hopefully I will be able to turn this technique into a practical circuit without too much trouble.

Regards