Analog Audio Mux/Demux over copper wire???

June 21, 2020 12:20 PM

I have an idea for an audio-muxing system to transmit multi-track audio over a standard analog transmission. I am wondering what would be required to build this...

What I'd like to do is be able to create a muxer/demuxer system for audio signals.

The muxer would be a small "mixer" style console with like 8 inputs.

I would like to mux the inputs into a single signal to send through a single audio wire. Ideally an 1/8" jack/wire (a la headphones etc...)

At the receiving end (say, a computer) I would like to receive the signal and demux the single unified signal back into original 8 separate tracks of audio.

I know this is possible in theory, but I have a few questions:

1) What would be the best form of multiplexing along these lines?
2) Would QAM/FM somehow be related to this process?
3) Would it required ADC/DAC conversion? Or can you do this as a pure analog signal?
4) What would the required bandwidth be of such a wire?
5) What sort of bitrate would be needed and is it possible, again, to do it over such a small wire?
6) If not possible over 1/8" (not sure what gauge that's considered) what gauge would be required for muxing 8 separate 22,050 hz audio streams?

I imagine each addition of a stream would require amplitude reduction to not oversaturate when adding the frequencies? (this is why I ask about FM).

TDM seems to be almost "packet" based? Since I want a more "realtime" approach it seems I want a less serial approach and more "parallel" approach (that is each second of time contains the second of all 8 channels at once, not 8 seconds sequentially carrying them, and divided into a single wave).

---
*If you are curious why I ponder this - I have a groovebox, and each channel is combined into a single stereo signal to be sent, there are no separate jacks. I realize I can't make the device itself change, but I would like to see a proof of concept mixer that can do this, and wonder what it would take to build it... What technologies, what limits, and requirements, and such.

Would the mux/demux process be relatively instant (for our perception?) My concern is - looking at telephony, it seems the wiring would probably need to be a lot more than typical whatever gauge a standard 1/8 inch wire is, and require coax or some heavier gauge wire which might make it moot.

Any ideas or tips of where to look, what components would be needed? Would this be capable in analog? Would it require conversion? Is it easier in digital? Etc...

posted by symbioid to Technology (11 answers total)

I have no idea if this is possible over a single analog wire. It seems unlikely, because A. at this point someone would have already done it, and B. digital technology already exists that does this, so there's zero advantage to an alas system. The ADAT lightpipe interface (which is standard on many audio interfaces and mixers now) gives eight channels over an optical connection. I use it to give me eight additional inputs on my audio interface.

So, that doesn't answer your question, but I can't tell if this is just a theoretical question or if you're hoping to build this. If you want to build this, the technology already exists digitally/optically and it works great.
posted by jonathanhughes at 12:28 PM on June 21, 2020 [2 favorites]

Best answer: This is absolutely possible over a single wire using analog techniques alone.

In its native form, an audio signal covering the range of human hearing occupies around 20kHz of bandwidth.
When you put such an audio signal on a single wire in the usual fashion, it occupies the 0-20kHz frequency range.

One way to multiplex more signals onto the same wire would be to simply shift the others into different frequency ranges. So channel 1 would be 0-20kHz, channel 2 could be shifted up to 20-40kHz, 3 would be 40-60kHz, etc. In practice, you would want to leave some gaps in between though, to make it easier to separate them cleanly. You can keep cramming in channels until you start running into the bandwidth limits of your piece of wire. And you don't have to start at the bottom of the frequency range either - you could shift everything up to frequencies in the MHz range, and there are a number of reasons it might make sense to do so.

This approach is called frequency division multiplexing. You're probably more familiar with it in the context of radio, but exactly the same thing can be done over a wire, and indeed this is exactly how telephone companies used to cram multiple calls into the same trunk line.

Everything needed can be done with analog circuitry if you want to. To implement this, you would need circuit blocks that can shift the frequency of a signal up and down, which are called mixers (not what you're used to that word meaning from audio), and other blocks that select or remove desired frequencies, which are filters - which are exactly as you're used to.
posted by automatronic at 12:52 PM on June 21, 2020 [10 favorites]

Best answer: You could do this with an FM radio transceiver IC for each audio channel. Have each audio channel mapped to a different FM channel, and connect all of the transmitter outputs to the same coax cable.
posted by monotreme at 1:13 PM on June 21, 2020 [1 favorite]

Best answer: This is typically done digitally these days. For example HDMI is capable of 8 channels of audio, 24 bits per sample, up to 192 KHz. And carry all of the video in the same stream.

Let's assume you want uncompressed CD audio quality. That's 44.1 KHz at 16 bits per sample.

So your required serial bit rate is 44.1K times 8 channels times 16 bits. That is 5.65 megabits per second, which is a fairly low and feasible bitrate for any sort of tiny wire. Or you could use those cheap TOSLINK LED optical cables used to connect your DVD player to your audio receiver.

Your notion of packets and realtime isn't quite right. For CD quality, you only need one sample for every channel 44.1K times a second. The 5.65 megabit rate from above will do that. You packetize so that every 44.1K of a second, you gather 16 bits of data for each of 8 channels, put them all together and you have one sample for every channel simultaneously.
posted by JackFlash at 2:07 PM on June 21, 2020 [3 favorites]

I don’t see how it’s possible to do this in the analog domain while maintaining the same fidelity.
posted by STFUDonnie at 2:55 PM on June 21, 2020

I mean, it can be done in analog like automatronic says, but it's a lot cheaper and easier to have an 8 channel copper snake or to digitize it and send it over a Cat 5 cable, which is what the concert industry has moved to. The theory of muxing and demuxing isn't all that hard but getting something that can do it with acceptable audio quality and cost is where it gets tricky.
posted by Candleman at 2:56 PM on June 21, 2020 [1 favorite]

Best answer: Yeah, nthing others curious about whether you want to try to build this just to build it, or as a bit of thought experiment, or if you're actually looking for a functional solution.

If you want to build it for the heck of it, automatronic's explanation makes sense to me.

But if you're looking for a practical solution, I think you have to consider something digital, which is a solved problem with multiple systems already in wide use:

DANTE audio networking

MADI

AES50 (pdf link)

EtherSound
posted by soundguy99 at 3:19 PM on June 21, 2020 [2 favorites]

Best answer: send it over a Cat 5 cable

If, for whatever reason, you did want to do a solder-it-yourself approach, you could get 4 pairs or 7 signals + 1 ground out of a Cat 5 cable, the latter of which almost meets your spec, if you're mainly interested in minimising the number of cables.

I can imagine there's a lot of incorrigible diyers who could do that today from parts they have lying around.
posted by ambrosen at 5:12 PM on June 21, 2020

Nthing automatronic's answer above. This kind of problem was solved a century ago with AM and then FM. That sounds like radio not wireline, but the same principles are at work, where an audio channel is modulated up to some higher frequency.

I don’t see how it’s possible to do this in the analog domain while maintaining the same fidelity.

You have an insufficient respect for how much bandwidth is available on an analog wire and how awesomely high fidelity analog can be. As long as you control the parasitic capacitance on the cable, you have a HUGE amount of bandwidth available on that wire. See also "DSL".
posted by intermod at 8:18 PM on June 21, 2020 [6 favorites]

As I understand it the OP wants to split a signal already mixed inside the device into seperate outputs. This can't be done.
posted by Kosmob0t at 3:20 PM on June 22, 2020

Response by poster: Yeah, nthing others curious about whether you want to try to build this just to build it, or as a bit of thought experiment, or if you're actually looking for a functional solution.
posted by soundguy99 at 5:19 PM on June 21

As I understand it the OP wants to split a signal already mixed inside the device into seperate outputs. This can't be done.
posted by Kosmob0t at 5:20 PM
------
It's a weird question I know. My curiosity is based on Kosmob0t's phrasing. I'm obviously aware it can't be done as a hack to the device. And any solution I build will never actually be what I'd want for anyone else.

I would have no problem with the "just get a cat5 or TOSLink or S/PDIF or whatever" if that shit was already built into the equipment. I hazard the cost of doing that (as well as routing audio over cat five) and then needing an extra bit of hardware to support that which would defeat the purpose.

I have a soundcard that takes an 1/8 copper wire. I have a groovebox that outputs the data on a copper wire (or RCA jacks). There are a lot of such boxes that don't have a single audio signal that I can't just throw in a mixer and be done with it. Since they're internally mixed by the time it gets to the output, it's pointless in that regard with pretty much any current box.

But I want to know if it would be possible, and what the technical hurdles would be.

Would the tech to mux the audio into and out of the signal be more expensive than just adding 4-8 more connections for each individual part? It seems to me it could be cheaper to get a chip to mux it and some software to demux.

So in this case, the ultimate question is fidelity of signal. And WHY isn't this a standard if signal fidelity is possible. The only thing I can think of if fidelity isn't an issue would be cost. Same reason there's not multiple ports.

I guess I'm mostly curious what proof of concept could be built. Not to apply to existing software but to show it could be done and how it could be useful. I have a feeling I'm not explaining why this would be useful (*IF* the quality was there, obviously if you're just gonna get telephone quality signal well, fuck that).

I realize there's a lot of other engineering questions involved (especially for things that do integrate a mix on the output - like post processing (for example, on the Novation Circuit - there's a global combofilter that's applied to the signal after all channels are mixed; when do you mux it then? Do you filter each channel separately and then mux it? (this is just an example of things I can see causing problems)).

I think DSL is a good example of where my line of thinking is (I work for a fiber service provider which is where I learned about muxing signals as a concept (TDM in particular w/r/t Voice Signals, as well as how signals are muxed in the optical spectrum from the office to the hub)... I don't work in the super technical end of things - so this is why I ask here - I'm obviously underinformed)).

Thanks for more of the links, soundguy, not sure it's what I'd need, but I'll check it out when I'm not heading to bed!
posted by symbioid at 11:02 PM on June 22, 2020