On Fri, Oct 12, 2001 at 07:43:04PM +0200, CeDeROM wrote: > I made such circuit on AT89C52 microcontroller, but chip is too slow for > that job.
Odd .. I'd expect you could get away with using a timer or something, but those timers may not be very good at PWM generation .. > So I've diecided to use PC as a "fast microcontroller". I'd really advise you against this. PCs are not good microcontrollers. Trying to do a class D in software is .. well .. I'd be interested to see if it could be done, but I doubt very much it would work very well, if at all; the OS just won't be able to give you processor time when you need it. And you'll need a lot of it. > But why not to write kernel module - the PC would stand on Linux. > > So I have to write special driver for this amplifier. > I think that writting it in assembly language would be the fastest > choice. > The speed is critical - I have to generate a signal of frequency more > than 1MHz > (or more - faster output frequency = better sound). > I also, with sampling frequency of 11kHz/22kHz/44kHz, have to take > samples form > /dev/dsp (or some other sound output) and give them to output (serial > port). What you could probably get away with is doing the necessary EQing and signal conditioning in software, and sending samples to a PWM generation circuit. But you'd need a fast interface. One thing you could try is using a S/PDIF output, connecting it to a S/PDIF receiver, and converting that into PWM using a micro or even a CPLD. There is absolutely _no way_ a regular RS-232 serial port (is that what you meant?) will give you the bandwidth you need for this. Even better: with this setup, you wouldn't need a kernel module at all; you'd simply use existing ALSA drivers to generate a S/PDIF stream. Of course, your stream would not be "true" digital audio - it would instead be specially made to generate PWM for your H-bridge - but that's fine; the soundcard wouldn't care. > I'd like not to invoke any libraries to program (totally platform > independecy). Hate to say it, but _total_ platform independence is an ideal we all (except Microsoft) strive for, but can never really achieve. Use libraries - save yourself some time. Plus, by running on Linux, you're already very platform independent: and if you don't write a kernel module, you can run on other _processors_ simply by doing a recompile, if you don't use assembly in your user-space code. > As for now I have one question - are there some timer functions in > kernel? Oh yes, but they are 100% unsuitable for doing real-time PWM, even at baseband. They're just event timers. > If I finish it, lets make it part of ALSA project...? Now _that_ would be interesting: class D support as part of ALSA! You could build a super high-powered amp for, say, US$100, including case. :) But still, I don't think this sort of thing really belongs in ALSA, necessarily - I think it would be better done as a user-space program, using the PC for the DSP stuff and an external S/PDIF circuit (or something) for PWM. (USB is another possibility.) cheers -- Michael Ashton <[EMAIL PROTECTED]> ---- Pray for: - the souls of those who have lost, and will lose, their lives, as a result of the US terrorist attacks; - the souls of those responsible, that they may see the destruction they have wrought, and repent, before it is too late; - the United States, that its people will return to God for strength, and forgive; for the attackers "know not what they do"; - the agents of justice on Earth, that they may speedily do God's will, whatever that may be. _______________________________________________ Alsa-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/alsa-devel
