4 ms·

Author here, thanks for posting this :) Would be happy to answer any question.

Dowwie 8y ago
Very cool concept! Radio stations used to share software over the airwaves by using something like this. How long would it take to transmit a 3MB jpeg and then encode it,?
- roman_zeyde 8y ago
  Thanks! Using an audio cable (for better SNR), I have managed to achieve and sustain 60kbps for long periods time - so 3MB file should take ~7 minutes to transmit.
jacob019 8y ago
Can work above the audible spectrum?
- roman_zeyde 8y ago
  Currently, it does not - mainly because I've tried to use all available bandwidth (to increase the bit-rate). Please take a look at https://github.com/quiet/quiet-js https://github.com/quiet/quiet-js for an ultrasonic data transmission.
ggerganov 8y ago
In your experience, how reliable is the communication between 2 devices when using speakers and microphone (no cable) at distance ~3 meters? What bandwidth do you typically achieve in such conditions?
- roman_zeyde 8y ago
  ~3 meters is indeed quite a challenging scenario. I've tested it with the lowest bit-rate supported (1 kbps), but it doesn't work reliably. You can try my Android app [1] to reproduce the experiments for short transmissions. [1] https://play.google.com/store/apps/details?id=bit.zeyde.audiomodem https://play.google.com/store/apps/details?id=bit.zeyde.audi...
  - ChuckMcM 8y ago
    Nicely done, I'll note for the readership here that a similar technique works fine at 300 baud across the room using FSK. Something you might consider is to write a web camera backend for it and use the screen/webcam. That has a lot more bandwidth and has higher SNR as the selector effect of the web camera is so much better than that of a microphone.
    - brian-armstrong 8y ago
      FSK across the room? Does that have noise restrictions? I’ve found all the phase modulation schemes work poorly across the room. DSSS works ok but is very slow. I would be curious how reliable FSK is in that situation.
      - ChuckMcM 8y ago
        With 1270Hz/1070Hz you've got over 3 complete cycles to identify the tone at 300 baud, at higher frequency pairs it gets better. The person who wrote it was Tom Lyon at Sun but I bet I could reproduce it with a gnuradio-companion flow graph. That would make it pretty easy for anyone to play with it.
        brian-armstrong 8y ago
        Interesting, I will have to give it a try then. Thanks!
      - darkmighty 8y ago
        I don't have first hand experience but I believe a big problem with sound channels is multipath propagation/challenging frequency response. FSK naturally deals quite well with those; it's similar to OFDM but more robust due to single carried symbols. You can extend the symbol period as long as you like while using an adequate number. Finally in challenging environments it'd a good idea to use error correction, not just error detection as done here -- you can (in theory) achieve reliable transmission over arbitrarily noisy channels by adding redundancy.
jedberg 8y ago
I’m wondering, did you build this just to see if you could, or did you have a use case in mind?
- roman_zeyde 8y ago
  Both :) The initial use-case was performing air-gapped communication for signing Bitcoin transactions [1] on an offline machine. [1] https://github.com/spesmilo/electrum/pull/964 https://github.com/spesmilo/electrum/pull/964
codezero 8y ago
Was this inspired by any of the dialup modem protocols?
- roman_zeyde 8y ago
  Indeed, I am using OFDM [1], which uses several carriers, each modulated via QAM [2]. [1] https://en.wikipedia.org/wiki/Orthogonal_frequency-division_multiplexing https://en.wikipedia.org/wiki/Orthogonal_frequency-division_... [2] https://en.wikipedia.org/wiki/Quadrature_amplitude_modulation https://en.wikipedia.org/wiki/Quadrature_amplitude_modulatio...
  - viraptor 8y ago
    I'm a bit confused. How does QAM work with audio if you have only one channel rather than I/Q? Are you using multiple frequencies for that?
    - roman_zeyde 8y ago
      QAM is using both carrier's amplitude and phase to encode information, so I/Q can be thought as [1]: I = amplitude*cos(phase) Q = amplitude*sin(phase) [1] http://whiteboard.ping.se/SDR/IQ http://whiteboard.ping.se/SDR/IQ
      - viraptor 8y ago
        I'm just sitting here and staring at this sentence: "The modulated signal rides on a carrier of a given frequency, but the base band signal got no fixed frequency at all. Because of this, we have the possibility to encode the two-dimensional I/Q signal onto the one-dimensional RF signal without losing anything." I may need to sleep on it...
someguydave 8y ago
Have you tested this software on macos?
- roman_zeyde 8y ago
  No, I've used it only on my Linux machine. I guess it should also work on MacOS, since it depends on `numpy` [1] and `portaudio19-dev` [2] Debian packages (which should be installable via Pip/Homebrew). [1] http://www.numpy.org/ http://www.numpy.org/ [2] https://people.csail.mit.edu/hubert/pyaudio/ https://people.csail.mit.edu/hubert/pyaudio/
brian-armstrong 8y ago
Do you have ideas or plans on how to increase the transmit speed?
- roman_zeyde 8y ago
  The original goal was transmitting ~10kB in a few seconds over an audio cable, so I didn't try to improve the transmission speed much further. I guess that adding error correction coding will allow increasing the highest possible bit-rate (currently I'm using only CRC-32 for error detection).
zaarn 8y ago
How reliably does this work over FM transmission or VoIP? (If it doesn't I would guess and say that you'd have to add some mechanism to detect reliable frequency bands like in existing modems)
- roman_zeyde 8y ago
  Unfortunately, VoIP doesn't allow reliable OFDM communication... It's preferable to transmit and receive the audio as-is (without any compression).

Very cool concept! Radio stations used to share software over the airwaves by using something like this. How long would it take to transmit a 3MB jpeg and then encode it,?

roman_zeyde 8y ago
Thanks! Using an audio cable (for better SNR), I have managed to achieve and sustain 60kbps for long periods time - so 3MB file should take ~7 minutes to transmit.

jacob019 8y ago

Can work above the audible spectrum?

roman_zeyde 8y ago
Currently, it does not - mainly because I've tried to use all available bandwidth (to increase the bit-rate). Please take a look at https://github.com/quiet/quiet-js https://github.com/quiet/quiet-js for an ultrasonic data transmission.

ggerganov 8y ago

In your experience, how reliable is the communication between 2 devices when using speakers and microphone (no cable) at distance ~3 meters? What bandwidth do you typically achieve in such conditions?

roman_zeyde 8y ago
~3 meters is indeed quite a challenging scenario. I've tested it with the lowest bit-rate supported (1 kbps), but it doesn't work reliably. You can try my Android app [1] to reproduce the experiments for short transmissions. [1] https://play.google.com/store/apps/details?id=bit.zeyde.audiomodem https://play.google.com/store/apps/details?id=bit.zeyde.audi...
- ChuckMcM 8y ago
  Nicely done, I'll note for the readership here that a similar technique works fine at 300 baud across the room using FSK. Something you might consider is to write a web camera backend for it and use the screen/webcam. That has a lot more bandwidth and has higher SNR as the selector effect of the web camera is so much better than that of a microphone.
  - brian-armstrong 8y ago
    FSK across the room? Does that have noise restrictions? I’ve found all the phase modulation schemes work poorly across the room. DSSS works ok but is very slow. I would be curious how reliable FSK is in that situation.
    - ChuckMcM 8y ago
      With 1270Hz/1070Hz you've got over 3 complete cycles to identify the tone at 300 baud, at higher frequency pairs it gets better. The person who wrote it was Tom Lyon at Sun but I bet I could reproduce it with a gnuradio-companion flow graph. That would make it pretty easy for anyone to play with it.
      - brian-armstrong 8y ago
        Interesting, I will have to give it a try then. Thanks!
    - darkmighty 8y ago
      I don't have first hand experience but I believe a big problem with sound channels is multipath propagation/challenging frequency response. FSK naturally deals quite well with those; it's similar to OFDM but more robust due to single carried symbols. You can extend the symbol period as long as you like while using an adequate number. Finally in challenging environments it'd a good idea to use error correction, not just error detection as done here -- you can (in theory) achieve reliable transmission over arbitrarily noisy channels by adding redundancy.

jedberg 8y ago

I’m wondering, did you build this just to see if you could, or did you have a use case in mind?

roman_zeyde 8y ago
Both :) The initial use-case was performing air-gapped communication for signing Bitcoin transactions [1] on an offline machine. [1] https://github.com/spesmilo/electrum/pull/964 https://github.com/spesmilo/electrum/pull/964

codezero 8y ago

Was this inspired by any of the dialup modem protocols?

roman_zeyde 8y ago
Indeed, I am using OFDM [1], which uses several carriers, each modulated via QAM [2]. [1] https://en.wikipedia.org/wiki/Orthogonal_frequency-division_multiplexing https://en.wikipedia.org/wiki/Orthogonal_frequency-division_... [2] https://en.wikipedia.org/wiki/Quadrature_amplitude_modulation https://en.wikipedia.org/wiki/Quadrature_amplitude_modulatio...
- viraptor 8y ago
  I'm a bit confused. How does QAM work with audio if you have only one channel rather than I/Q? Are you using multiple frequencies for that?
  - roman_zeyde 8y ago
    QAM is using both carrier's amplitude and phase to encode information, so I/Q can be thought as [1]: I = amplitude*cos(phase) Q = amplitude*sin(phase) [1] http://whiteboard.ping.se/SDR/IQ http://whiteboard.ping.se/SDR/IQ
    - viraptor 8y ago
      I'm just sitting here and staring at this sentence: "The modulated signal rides on a carrier of a given frequency, but the base band signal got no fixed frequency at all. Because of this, we have the possibility to encode the two-dimensional I/Q signal onto the one-dimensional RF signal without losing anything." I may need to sleep on it...

someguydave 8y ago

Have you tested this software on macos?

roman_zeyde 8y ago
No, I've used it only on my Linux machine. I guess it should also work on MacOS, since it depends on `numpy` [1] and `portaudio19-dev` [2] Debian packages (which should be installable via Pip/Homebrew). [1] http://www.numpy.org/ http://www.numpy.org/ [2] https://people.csail.mit.edu/hubert/pyaudio/ https://people.csail.mit.edu/hubert/pyaudio/

brian-armstrong 8y ago

Do you have ideas or plans on how to increase the transmit speed?

roman_zeyde 8y ago
The original goal was transmitting ~10kB in a few seconds over an audio cable, so I didn't try to improve the transmission speed much further. I guess that adding error correction coding will allow increasing the highest possible bit-rate (currently I'm using only CRC-32 for error detection).

zaarn 8y ago

How reliably does this work over FM transmission or VoIP? (If it doesn't I would guess and say that you'd have to add some mechanism to detect reliable frequency bands like in existing modems)

roman_zeyde 8y ago
Unfortunately, VoIP doesn't allow reliable OFDM communication... It's preferable to transmit and receive the audio as-is (without any compression).