Few years ago, I thought the same, that "old PC motherboards are under-appreciated", but trying to use them is just not very practical.
The whole thing is so big -- you need the PC itself (with huge PSU), plus a monitor, plus a keyboard. It takes a lot of power, too, and any sort of battery-powered device is out of the question. And the inputs are not even that well protected -- while the audio i/o is pretty robust, internal headers are not.
Compare them to chinese Arduino clones or esp8266 board -- for $3, the price of a sandwich, you get a tiny board with excellent (or at least decent) documentation and lots of protected inputs and outputs. You can do hard real time too, with a bit of effort. And since all data/code is stored on your main laptop and PC you don't have to worry about backups or data loss from destroyed devices.
(Note: if you want to support reusing old items, then raw atmega chips + serial-usb cables from old cell phones work as well, but require a bit more soldering)
(Note: and if you really want to use old motherboards, don't forget about serial and parallel ports! Lots of pins, pretty protected, and very easy to programs)
5 years ago when VR was the hype we measured latency using them. By syncing the audio stream to the high precision clock and basically putting in a photoresistor into the microphone port we could measure the latency of actually submitting a frame into the moment it physically appeared in the VR glasses.
25 years ago I treasured my paper copy of an article about how to do general purpose I/O on a printer port, and always kept an old PC around for experiments for that. But with Raspberry Pi computers, interfacing is so much easier. And no need to worry about blowing up a precious motherboard, because they don't cost much. And the Pi4 is as fast as oldest PCs I still have, core 2 duos.
The A2D of soundcards is pretty usefuyif you can get your signals into the audible range. The ZetaSDR (second schematic at http://www.qrz.lt/ly1gp/SDR/ ) is probably the most rewarding project for the effort I put in. It plugs into a soundcard and I was able (with the help of an antenna tuner) pick up signals from around the world, including some incredibly weak ones.
Nice idea! First I was thinking he will use NE555 to convert the sensor reading to variable frequency and reading that. Generating a signal then reading it back with the same soundcard is neat.
I wonder if a frequency approach might be more accurate than relying on the left/right channel gain accuracy, providing the signal changes on timescales much longer than the wave period. Just generating pulses at some frequency might be the easiest method, which wouldn't rely on frequency measurement or FFTs.
I wonder if temperatures can be read simply by reading electrical noise and training some sort of regression model on it, instead of using an actual sensor.
I wonder, there must be even some sort of butterfly-effect correlation between the room temperature and e.g. the number of errors in non-ECC RAM or some foo.
> I wonder if temperatures can be read simply by reading electrical noise
Yes and no. On one hand, semiconductor properties are temperature dependent, therefore the "noise" from amplifier is temperature dependent.
On the other hand, temperatures of the sound card circuitry depend mostly on computer case temperature, not ambient temperature, but that is somewhat combatable with external cards. Furthermore, amplifiers have multiple stages with various feedbacks and filters designed to among other things reduce (i.e. distort) noise, which is bound to reduce effectiveness.
Even if one succeeded in training an ML model on sound card noise data, noise profile will most likely be too different on different sound card designs for the model to work.
All in all, I suspect that there is enough data in the noise to measure temperatures with one or so degree accuracy, but such a solution would not scale to different hardware.
Semiconductors are sensitive in non-linear manner to both temperature and light. I was experimenting with latter, it was large germanium(!) transistor w/o cap and joystick port for analog input. What I really learned from that was sound card input is very easy to fry.
That's probably just a convenient, cheap, and robust connector to source instead of something more obscure. Grill probes are more likely to be a thermocouple rather than a thermistor as they can usually handle higher temp ranges.
How cheaply can this be made? it would be great to create a line of dishware for elderly, blind, or disabled people to let them know when there's enough
1. water in their cup
2. food on a plate
etc...
The whole thing is so big -- you need the PC itself (with huge PSU), plus a monitor, plus a keyboard. It takes a lot of power, too, and any sort of battery-powered device is out of the question. And the inputs are not even that well protected -- while the audio i/o is pretty robust, internal headers are not.
Compare them to chinese Arduino clones or esp8266 board -- for $3, the price of a sandwich, you get a tiny board with excellent (or at least decent) documentation and lots of protected inputs and outputs. You can do hard real time too, with a bit of effort. And since all data/code is stored on your main laptop and PC you don't have to worry about backups or data loss from destroyed devices.
(Note: if you want to support reusing old items, then raw atmega chips + serial-usb cables from old cell phones work as well, but require a bit more soldering)
(Note: and if you really want to use old motherboards, don't forget about serial and parallel ports! Lots of pins, pretty protected, and very easy to programs)