I'm surprised that you could get a modem pair to connect with a plain wire. An old dial telephone (e.g. a 500 set) was an entirely passive device, relying on power from the central office end of the wire. The carbon mic didn't need an amplifier - it's just a resistor, with resistance varying according to sound pressure; a transformer subtracted out the mic from the combined signal and sent the difference to the speaker.<p>I did a quick search on 1670 schematics and came up with very little. Note however that several sources identify the USR101 chip as a ROM - that's clearly not the case, as a schematic shows no address or data bus, and you can see a line from the coupling transformer to an RC network and several pins on the device. It's almost certainly the analog front end and the modulator/demodulator.<p>If you could find a datasheet for that device you'd probably be able to figure out how the whole thing works, but I'm guessing those were never available to the public and have been lost to the mists of time.
It might be trying to detect dial tone, then hanging up and trying again after a second or so - the click would be the mechanical relay you see in the schematic. (note that standard phone lines don't send dial tone until they detect an off-hook condition, i.e. current flowing down the line. Even then, older exchanges had a limited number of "dialers", and wouldn't send dial tone until one had been connected to the line)
Sounds a bit like constructive interference between two sound sources that are almost the same frequency but not.<p>Or RC “ringing”. That click stopping the sound makes me curious to see a plot of bytes per second sent and received. Is the modem resetting itself every time it clicks and introducing a small pause or data loss? Or is it completely divorced from the sound.<p>These modems don’t have fans built in right? If you isolate the speaker does it still make the noise?