In my understanding, most of the usb audio problems come from the fact that usb is, by default, asynchronous. Latency becomes a huge issue, then, since there's no way to ensure data is getting to its destination on time, especially if there are other devices sharing the buss. USB does, however, have an isosynchronous mode, which is, I think, used with the "audio class" devices you mentioned. This mode is designed to solve major latency issues, but has not been used by many manufacturers and is, therefore, very expensive. That's my two cents.
Tschuss, Greg
"Joseph A. Sarlo" wrote:
I've been using the Swissonic Studio D USB audio interface under Linux with my laptop and after some initial difficulties it seems to work fine. The latency is nominal. I haven't done any hard-core testing, but the total latency (including audio buffer) is definately under 50 ms. It seems all the horror stories about usb are not true (well, most of them anyway). I can even use a usb joystick on the same port (with a hub) at the same time, which makes sense since full duplex 44.1k 16 bit stereo should only require about 350KByte/s of bandwidth and USB (ideally) offers 1.5MByte/s.
The trick with linux is that the drivers only work with devices that are 100% usb audio class compliant. I've only found two devices that can claim that, the Swissonic Studio D and Sound Technology's SoundBuster. The prolem is they're both kind of expensive and the Soundbuster is more of a data aquisition device for acoustic measurements. You also are going to want to get the latest usb kernel modules. I had some problems using a 2.2 backport that all went away when I moved to 2.4.1. Also, I have an OHCI controller which everybody seems to say doesn't wor well, but it's been working fine for me. Some things are a little weird but nothing unworkable.
-- __________________________ | | Joseph A. Sarlo | Computer Music Dept. | Peabody Conservatory | Johns Hopkins University | | jsarlo@peabody.jhu.edu |__________________________
On Fri, 30 Mar 2001 pix@test.at wrote:
I'm really interested in the 'remote gui' option rather than the netsend/receive option, as it doesnt really require any functional changes to existing patches. but at the moment im not in a oaition to test it, only having the one pc. but real soon now i'll be getting a laptop (curently in the 10-14 business days tease-zone), so i'll probably give it a whirl then.
another thing i hadn't thought about until today was looking into the possibility of using multiple gfx cards under x4. it might not be possible, but if it were possible to do multi-head using two different kinds of cards (i know its possible with expensive cards that support multihead) than you could just have the gem window appearing on whichever one has video out. maybe i'm ascribing mythical superpowers to x4 here :)
also, a friend forwarded me some mailing list exceprts about huge latencies using usb-audio solutions on laptops under windows (using audiomulch). has anyone on this list played with usb-audio (especially under linux), and what kind of latencies have they got? i'm hoping the latency isnt inherent in usbaudio, but i wouldnt be surprised if it was...
(then again, my current patch has so much inherent latency, that this wouldnt really be a problem for me anyhow, but i'm just curious).
On Wed, 28 Mar 2001, Timmy B wrote:
HI Pix, Miller and others
the way I have this running is to seperate the control to a laptop and the rendering to a linux box with a good graphics card. i use netsend/netreceive to send info from the laptop to a patch i have written on the box which opens the gem window at maximum size. i use pd -nogui on the box to get rid of the gui. i use twm as the window manager as it should be the simplest and fastest. i have added my own .twmrc to remove title bars and boundaries from the window, but i cannot get the positioning requests to be honoured, so i have to drag the top left corner of the gem window to the top left corner of the screen manually. suggestions welcome
i chose netsend as the control mech because i believe it would be the simplest and fastest rather than using some remote control / X-window confusion.
this is planned also to be very portable - for a workshop i need only take the video card and linux install CDs, all the extra code is on the laptop, so i can de-Bill a PC at the venue and make it do clever (well, half-clever) things [hopefully] quite quickly....
haha
tm
Miller Puckette wrote:
This should work... you can feed Pd a command line option "-guicmd" to set how Pd will start the GUI up, which could use rsh to start up on another machine, or even set the DISPLAY variable. In either case Gem's window should start up on the machine Pd is running on.
I got this working once but I have another problem, which is that I don;t have a good way of sizing and positioning the Gem window to cover all of the display. Should be easy, but it didn't just work instantly when I tried it.
cheers Miller
my current thought is to have the pd interface on another machine using the new gui options... i dont know if this is a possibility tho (i havent tried it yet)... also if it did work, would the gem window open up on the machine running pd or on the machine running pd-gui (i'm hoping it would open on the machine running pd).
pix.
--
\ / Tim Boykett mailto:tim@timesup.org \ / TIME'S UP / Industriezeile 33 B /\ A-4020 Linz /xx\ ph/fax:+43/732-787804 /xxxx\ http://www.timesup.org