On Fri, 19 Nov 2004 02:24:59 +0100, Florian Schmidt <mista.tapas@xxxxxxx> wrote: > On Thu, 18 Nov 2004 16:31:03 -0800 > Mark Knecht <markknecht@xxxxxxxxx> wrote: > > [snip - thank you very much for clarifying that] > > > Anyway, before I depress you too much, I'll stop going there. > > Nah, i'm not depressed. I expected something like that though not quite as > high. > > > > > > > From my POV a more interesting idea would be to do an external sound > > device, probably 1394 based since that will work for more people. > > Please remember that a PCI card is almost useless for laptop users > > unless we're trying to put this into a cardbus formal also. That adds > > money. > > > > If it was 1394 based then you can put a 1394 adapter in your PC for > > $20 and then everyone uses the same audio unit. We control how it > > works, so we can follow specs or do it in our own standard. You get > > the advantage of probably more channels and better SNR, but you do > > have to package the unit in a box or some type to be of general use. > > > > Anyway, those are some ideas for you to chew on. Hope I haven't poked > > a balloon with a pin here... > > > No, i find the alternative of a 1394 device completely acceptable, too. But > how much cheaper would it be? I suppose the logic for talking to the 1394 bus > [?? i don't know anything about firewire, except for 1. serial, 2. faster > than USB] can be put into a FPGA again, right? > > For the sake of an example, let's choose a stereo full duplex device with > 48khz only samplerate, and with medium to good quality AD/DA's. So we got > like 15-30 bucks for the DA/AD's. What's the rest? How expensive is the FPGA > to control all this? What else is needed? Ok, a case mustn't be pretty, so i > suppose anything will do -> 2-3$ :) > > Flo > Well, I guess my thought is that with a 1394 device we can just buy standard links and Phys and sort of wire things together without going to a full-fledged bus structure like PCI. I may be wrong about that. The most common Link devices are 1394 OCHI and they are generally PCI based so that puts us back into all of this stuff. Think about this from another direction though. The 2-channel cards are always going to be cheap since that's what most of the world wants and that's where the big companies provide value. I don't think we can do anything at that number of channels that makes a lot of financial sense. However, what about a much bigger devices, competing more or less with an RME HDSP 9652 or a DigiFace? If we went that direction then there isn't all that much incremental cost to add the channels (we can talk about this later) but we get another degree of flexibility. Think of this like a Digiface, or a MultiFace, or a Delta-1010 class device, but most importantly with built-in reprogrammable DSP capability. wouldn't that be cool? How about the ability to run LADSPA plugins in hardware. To me that sound exciting, and would open up a really interesting new way to use all the programming talent we have around these forums. If we could come up with a process to take a ladspa plugin and map it to gates in an FPGA then Linux audio would suddenly jump up to the level of product like the higher end Creemware machines and start approaching some of the lower-end pro versions of Pro Tools. Instead of a $400 2 channel PCI card we might end up with a $600 16-in/16-out device with hardware signal processing on board. To me this is probably a better place to go. If we do all this work ten we want to start working towards an architecture that will last. - Mark
