In 1979 I was debugging some Z80 assembler by putting a radio next to the processsor and picking up the RF noise. I could tell by the quality of the sound which loop it was in, and trace the execution as it happened. This was on a device that only had a serial line as its interface, and the problem was in the processing of the data, so it would occasionally just hang, making "standard" debugging impossible.

Later (1980) when I wrote a BASIC to Z80 compiler I used the same technique to profile the code and achieved a 5x speed-up by knowing which loops ran slowly.

Several years ago I was stuck in a data centre over night upgrading our systems with other team members who were working remotely. I was desperately tired and done my part, but had to wait around to monitor things while the rest finished up. Instead of staring at the screen, trying in vain to keep my eyes open, I put /dev/audio to good use by writing a simple script to wake me when tcpdump picked up the traffic I was looking for. I was able to lay my head down, guilt free, and steal 20 or 30 minutes of precious sleep between testing sessions.

Could you tell us more? This sounds like a near-epic hack. I have an idea of how it would work. Whenever I tried dumping to /dev/audio it always came out too fast.

could have been as easy as 'tcpdump port XXYY' and nothing was being output until the event he was looking for... would have made a fast screech, i bet.

That was basically it. I knew I would start to see traffic on a certain port when the service was restored.

>a good place to start, would be yes > /dev/audio, which outputs a high whining as the value 'y' is sent as audio over and over.

The key point of this combination is that yes outputs "y\ny\ny\n...". If you just send one character over and over you don't hear anything - I think the difference between adjacent character values determines the frequency of the output.

>Plus if there are any sound files on your drive saved in a lossless format they just play at random and then disappear back into the soup.

They need to be uncompressed (ie. contain raw audio data), not just lossless.

One of the better stories I've read was the guy who piped the output of ping to his speakers, turned the volume up, and went around the office wiggling the token ring cable listening out for the place with the unreliable connection.

  cat /dev/urandom > /dev/audio

I use it as a white noise generator, to cover ambient noise at night. Anyone know how to make a low pass filter (to make pink/red/brown noise)? I tried repeating the random source, to increase the wavelength, but it gets little R2D2 chirps and clicks in it (because of the step-like waveform):

  cat /dev/urandom | sed 's/./&&&&&&/g' > /dev/audio

I know it's probably not what you're asking for, but if you just want a white/pink/brown noise generator this is an excellent little flash app: http://www.simplynoise.com/

Thanks, that's what I'm using at the moment. I'd prefer a local solution rather than a webapp (but perhaps it doesn't make any difference?).

I tried their download brown noise loop, but it doesn't (always) loop seamlessly, even using:

  mplayer /home/user/8usb/SimplyNoise.com\ -\ Brown\ Noise.m4a -loop 0

I also tried downloading their flash swf itself, but it doesn't run locally (I expect it needs to phone home for the sound files).

The only local one I've tried is boodler: http://boodler.org/ (it's a "soundscape generator" written in python, which from memory has at least white and pink noise reasonably out-of-the-box).

You could try loading the swf in a browser, letting everything load up, and then checking /tmp.

An intriguing idea, but nothing there unfortunately (guess they keep it in ram.)

Weird. I would guess that would be browser/configuration dependent? I use firefox for this, and whenever I load up a flash movie (to be honest, I haven't tried it with anything else), it sits in ram until it's loaded, and then it gets cached in /tmp as a file named something like Fla[0-9a-z]+ .

In about:config there's the options:

browser.cache.memory.enable

browser.cache.disk.enable

and related settings. Disabling the memory cache might let you grab it.

If I get the spare time today, I'll see if I can't get the swf to pull out.

Sox (http://sox.sourceforge.net/) is your friend. In particular, check out the "synth" and "filter" command line options.

philh's post about (about yes's "y\n") gave me a tip. Instead of repeating the random character, do this:

  cat /dev/urandom | sed 's/./   &/g' > /dev/audio

Many years ago this facet of Linux was the thing that really made me realize just how astonishingly wonderful the central idea of UNIX (everything is a file) really is. It also puts in stark contrast those systems that haven't grokked this beautifully simple idea.

The irony is that the only reason Linus Torvalds spent so much time on Linux in the early days is because he accidentally made his dialer program use his minix partition to dial up his university. The dialer was fine with it... it even 'talked' back, destroying the data. He couldn't re-install Minix because he didn't have a copy of it. So he was stuck with Linux, and he was forced to improve it if he wanted to use his expensive computer.

Interesting. Sources?

It sounded like a dial up modem when I did this:

  cat /dev/mem > /dev/audio

Also, for some reason, this command makes it "beep" on every eleventh character you type. Probably something to do with buffers.

  cat /dev/input/by-path/platform-i8042-serio-0-event-kbd > /dev/audio

Or try it with core dumps. http://everything2.com/title/How%2520to%2520get%2520people%2...

even more fun

  wget http://www.wikipedia.com --output-document=- > /dev/audio

Download a webpage, output to soundcard. Works pretty well, I need to find a bigger page to work with though.

In a way it sounds a bit like modems of yore connecting :)

Edit: For a cool audio representation of the file size, try comparing Wikipedia with the Google homepage :)

Is this possible on a Mac?

Yes, but you'll need to do the mknod yourself, and have a program running in userspace reading from the block special file (or fifo).

I've used portaudio from Python on OS X before for doing this kind of audio manipulation.

/dev/audio doesn't exist. i think sounds go through core audio or something. i'm not really sure.

the command say is pretty interesting though

As far as I can tell, just by loading up Linux in a VM, but of course, that's cheating :)

Find all the .wav files and play them: find . -name '*.wav' -exec cat {} > /dev/audio \;

I guess I don't have any at the right frequency, though :(

If you have a remote machine that has a microphone plugged into it, you can use this to listen to what's going on in the room remotely.

From the original document:

> In order for noise to happen there must be at least some variation.

It's pretty neat. The data pretty much represents the position of the speaker.

Audio stuff is pretty fascinating; I read a great book on explaining DSPs for audio processing. I'll have to dig it out of my closet.

Reminds me old days I tried to listen to the linux kernel :D. A small typo I did filled up the terminal with lots of garbage. Later I learned I can rcover such a situation with..

$reset

If your terminal gets messed up from catting random characters to stdout, type 'reset' and press enter.

If you have Windows, you can play along to a degree using Cygwin and /dev/dsp instead.