Re: [Alsa-user] Alsa Devices Impossible to Address
Brought to you by:
perex
From: Petr A. <pet...@mi...> - 2006-08-12 22:10:11
|
Lee Revell <rlrevell <at> joe-job.com> writes: > How many USB controllers do you have? There's no way you can run more > than one or two USB audio devices per controller - USB does not have > enough bandwidth. > > Lee Sure, in fact I had similar idea to use more usb-audio adapters with HUBs and as I researched about bandwith issues, there is possibility to have USB 2.0 HUB which does "Transaction Translation" for every slave port. This feature calls exatly that "Transaction Translator", or shortened "TT". But almost none of HUB vendors notice about its possibilities (sure, its defined by HUB chipset, but almost none have it in specs too :-). I have found only one which is assured by specs to have TT on each port - so I bought 4 of them to test even daisy chaining, adapter unique identification (very strange, need to re-connect to the same ports everytime and before boot-up...), adapter initialization order on tree of HUBs and so on.... If usb device contains unique serial ID, then "MAY BE" to have ii identified in system permanently no matter of port where is inserted, but it is very rare case... "TT" feature on USB20 hubs does "injection" of slower 1.1 devices (base or full speed) to HIGH speed streams of 2.0. Most USB hubs specs does not care about that, but if you have multi TT HUB, then this hopefully works simillary as network SWITCH (which also allows speed translation, as opposite to network HUBs, which are designed only for eighter 10 or 100 Mb...), ...HOPEFULLY - I think, that I have not reached any limits there yet... In fact, I want to test realtime simultaneous recording from all usbaudio devices then routing them into mixing matrix with effect plugins (jack, jamin, ardour) and then back to all usb outputs ... Sure, card sync issue is trouble, but I am curious WHY jack does not have builtin internal MASTER clock to which all unsynced inputs can be resampled ?!?! - OK, MOST Windows multitrack recorders does not solve this too, expecting single multitrack adapter or more (profi class) synced by SPDIF or so ... - BUT, at least Cakewalk Sonar NEVER had this problem even with 2 DIFFERENT internal PCI cards while recording (Cakewalk uses lowlevel WDM exclusivelly, its THE company that KNOWS HOW TO do low latency audio on windows without ALSA :-)) - BUT even Cakewalk does not solved CM108 based LIVE mode trouble described below... HUBs I have in testing are this: http://www.shopping.hp.com/webapp/shopping/product_detail.do? product_code=DM866A%23ABA&aoid=32861&jumpid=reg_R1002_USEN I tested to record at 48kHz/16bit/moono from CM-108 based usb phones to Adobe Audition and some more multitrack recorders successfully; phones was connected to 2 HP QuadTT HUBs and each one was conneted to third one and this one to single USB20 port on PC (everything was tested on Windows, for now) Most of my testing was done in nTrack multitrack recorder (ntrack.com), latest versions, but there is one more BIG issue with CM108 based adapters - they have REALLY only MONO (single channel) input a STEREO (two channels) output which causes difficulties for ntrack in LIVE mode (routing inputs through effect plugins to outputs in realtime). Creative Live! External works, cards with SONIX SN11116F (as Hercules Muse Pocket LT) or Phillips based Griffin iMic (which has dual channle +20dB preamp, even not highednd) too. !!! ANY CM108 based adapter causes start of periodic clicking (by frequency given by buffering setup) after activating LIVE mode. Still no progress here.¨ I think that usbaudio.sys or nTrack even SONAR does not know how to deal with 1- in/2-out adapter. Everyt my test on Windows was using usbaudio.sys class driver only. I dont know where is really unsolvable trouble, but I want to test this in Linux ALSA too (unfortunatelly, I still have no knowledge about it .... grrrr - but currently I am trying to play with ubuntu and it look fine for first attempts, ten I want to try gentoo or some realtime multimedia distro, may be PlanetCCRMA for Fedora, or Gentoo) Another possible issue of usb audio streaming through HUBs may be added latencies with each HUB on way or possible problems during VERY intensive ISOCHRONOUS transfers through them. May be that this is not so much tested scenario and many HUB chips cannot solve some uncommon situatuion well. BUT, it seems that IT MUST BE POSSIBLE TO HAVE MULTITRACK DONE THIS WAY, somehow :-)) Petr |