[LAU] multi-channel recording

[Patrick Shirkey]

On 12/28/2009 08:18 AM, Ken Restivo wrote:
> On Sat, Dec 26, 2009 at 11:07:05PM +0000, Edward Barrow wrote:
>    
>> I've been getting by so far with an amd64 box running debian, jack, ardour
>> etc through Terratec DMX-6fire (Envy24 based) card, but now find that I need
>> to record more channels simultaneously.
>>
>>
>> I would like (and can just about afford) the M-Audio FastTrack Ultra, but it
>> doesn't look as though it is going to be easy to get it to work under Linux.
>> It's a USB 2.0 device - it needs those 480Mb/s to record 8 channels
>> simultaneously - so it isn't class-compliant.
>>
>> Has anyone succeeded in getting this interface to work properly in Linux?
>>
>> If not, can anyone suggest alternatives in the same sort of price range?
>>
>>      
> This is probably the most FAQ on LAU these days (well, maybe thirdmost, after questions about PulseAudio and Xruns, and slightly ahead of questions about RT kernels).
>
> Being shut out of the USB 2.0 market is an upcoming and serious threat to the continued viability of Linux audio, IMHO.
>
> USB 2.0 devices are affordable and great-- for everyone except Linux users. We have to pay more and get Firewire-- iff we have, or can get, a laptop that has Firewire, which seems to be on the way out.
>
> I sure hope someone can either get a manufacturer of USB 2.0 cards to open up the specs enough to write a driver for it, or, alternately, that the Ethernet-based Open Source multichannel audio interface project bears fruit soon enough. I'm personally more excited about the Ethernet interface because it'll be open and hackable, but for cheapness and convenience, the USB2.0's are going to be hard to beat.
>
> Someone posted a link to a USB 2.0 maufacturer that, IIRC, said they'd be willing to help someone write a Linux driver. Did any USB developers take them up on that?
>
>    



The other way is for people to buy a device and work with the various 
developers interested in usb-audio to get it to figure out the quirks.

This is how most of the usb-1.0 devices were made to work anyway. If you 
look at the quirks list for the snd-usb-audio driver a considerable 
number of usb-1.0 devices are on that list. The thing is that many of 
those devices have been bought, tested and triaged years back so for 
anyone buying a device recently it would seem like there were no hassles 
associated with the usb-1.0 driver.

In reality it took nearly 5 years for it to get to a point where most 
popular cards were supported ootb or with minimal driver tweaking. It 
took almost two years to get my device working properly and it had 
plenty of non standard quirks that had to be understood without any 
assistance from maudio in the slightest. We did it in the end though. 
The final nail was figuring out that the inputs are little endian and 
the outputs are big endian. That one took a while as it mostly just 
sounded like the driver was broken when it was actually just a case of 
specifying the correct endianess when testing with arecord.

It turns out that there are a couple more manufacturers that chose that 
approach too. But at the time of testing it was a very strange "quirk" 
that got a few laughs round here when it was realised what they had done 
and reinforced the idea in many peoples minds that usb-audio is a 
crapfest of interstella proportions.




Patrick Shirkey
Boost Hardware Ltd