Hector Martin

"As the PC audio signal is a square wave" What? > final int SAMPLE_RATE = 40000; A 20kHz signal at 40kHz sample rate is a sine wave, not a square wave, because Nyquist. Either this guy is using a horrible sound card with aliasing, or there's harmonic distortion somewhere.

It is if you go through a complete sample and reconstruction process intended to reproduce a bandlimited input signal. But if you're bitbanging a speaker you get a lot of the squarewave harmonics. The natural rolloff won't be the same as an antialiasing filter.

Yeah, but this is a Java applet running on a PC, so I would *hope* the signal chain is bandlimited and the resampling in use, if any, is mostly free of harmonic distortion.

Ah ha, a piece I didn't know. I would hope the same, yes. I saw 'PC audio' and read it as 'PC speaker'. If it really is the 8253-ish-driven speaker, it's likely a bitbanging interface. If it's a soundcard and you're getting 40/60kHz out... yikes.

Yeah, I'd expect some funky harmonics out of PC speaker tones. Would be a fun thing to record and see. Legacy ones are definitely driven via a single digital output; I wonder what happens with modern HDA setups though.

I think for HDA there's two ways, either you feed in the i8253 signal into a pin on the codec, or the BIOS emulates the whole thing using an HDA widget to internally generate the tones inside the codec. I wonder if either or both of those approaches have band-limited output.

Incidentally, that's 12kHz instead of 10kHz because the uber-crappy HDA internal generator works using an 8-bit divider from 12kHz (48kHz/4). Which makes PC Speaker music completely out of tune if played through it.