From: Martin P. <mar...@sy...> - 2011-04-16 15:37:47
|
On 2011-04-16 11:22, Stefan Richter wrote: > On Apr 16 Jonathan Woithe wrote: >>> Host controllers: >>> 03:0a.0 FireWire (IEEE 1394) [0c00]: VIA Technologies, Inc. VT6306 Fire II >>> IEEE 1394 OHCI Link Layer Controller [1106:3044] (rev 46) (prog-if 10 >>> [OHCI]) >> >> Here we could have a problem. This particular VIA chipset has given trouble >> for at least one other user in the past (this was confirmed when that user >> changed to a different firewire card - they had no further problems). > > The one card with this chip that I currently have in use (a no-name > CardBus card) works fine for me --- just as fine as Texas Instruments or > Agere based cards --- with a BeBoB based or with a DICE based audio device. > > I did a single test of this card with both of these audio devices > together, i.e. aggregated. After half an hour or longer use, FFADO went > into a loop of timestamp discrepancy reports. But I haven't tested device > aggregation systematically yet to tell whether such an event would be less > likely to occur with other FireWire controllers. > > There are few things to keep in mind though: > > - It seems that different hardware revisions of VT6306 exist (all of them > listed as "rev 46" by lspci though). > > - Not all FireWire controller cards are created equal. Cards with the > same chip may differ WRT quality of board layoutand components or > whether their EEPROM content makes any sense. > > - VT6306 is one of the many FireWire controller chips with unreliable > Cycle Timer readout. With ohci1394 + ieee1394 + raw1394, FFADO works > around this issue. I wonder if this makes things more fragile than > with chips that don't have this bug. With firewire-ohci and > firewire-core, the kernel already works around this chip bug so that > FFADO and other applications that need it get correct Cycle Timer data. I don't know why firewire cards are sold that don't work, usually capitalism brings a rapid end to such practices. MOTU themselves recommend using only TI cards for their product. In the past month I have tested a couple of MkIs with MOTU's latest Windows driver and they don't work properly, using various firewire cards. (I think MOTU doesn't support MkIs any more.) Usually the MOTU driver fails to recognize the interface, or it hangs duting configuration. I tried the same interfaces with FFADO which does recognize the interfaces and can set the baud rates, but that's all. So right now I can only use MOTU MkIs on MacOS. Martin |