|
From: Stefan R. <st...@s5...> - 2010-11-16 13:07:58
|
On Nov 16 Joan Pau wrote: > 2010/11/15 Stefan Richter <st...@s5...> > > > There is a TSB81BA3 physical layer chip on that card. This is an early > > 1394b device with a long list of serious hardware errata (TI literature > > SLLZ015C). [...] > Oh, we didn't know that. This card had worked without problems for quite a > lot of time with a 1394a camera in another underwater robot (bigger than the > one where it is being installed now, cooling matters!) . Since we need to > buy another PCI firewirecard (there are two robots), is there any hardware > compatibility list up-to-date? There is no catch-all 1394 compatibility list maintained anymore. We have compatibility with virtually all OHCI-1394 controllers. Some very rare ones pose difficulties in practice (ALi, NVidia, one on Audigy sound cards, some Ricoh ones) depending on the usage pattern. OHCI-1394 controllers all adhere to the same open link layer controller specification (well, to one out of two-and-a-half marginally different versions of the spec). They only have different bugs. > Because the snapshot linked in the wiki seems > a bit old (the TSB82AA2 is said to "work great" but it seems that it is not > the case now). AFAICT TSB82AA2 works very well overall. But the TSB81BA3 which is used on most TSB82AA2 based cards is a failure. Newer versions of it, TSB81BA3D and E, are much preferable but rarely or never found on TSB82AA2 cards. But as I said, the majority of TSB81BA3's problems is with 1394b hardware only, not with 1394a hardware. > We would not expect to find a PC104 firewire card in there, > since they are not very common in use, but at least we could compare the > chipsets that come into the cards from several vendors and choose one > already tested and working. Mostly. But there is some variability in case of link layer controllers with external physical layer chip (lspci does not show the latter, hence this information is often unknown). And there is some variability WRT board layout, components for bus power supply etc. A poorly laid out board might "work great" in PC A with cable B and device C but might fail miserably in PC X with cable Y and device Z. However, such issues do not occur very often, I think. They should not happen at all if designers and manufacturers followed the specs and the recommendations and testing procedures from 1394 TA, but sometimes corners are cut. Anyway. Let's wait for your kernel logs. -- Stefan Richter -=====-==-=- =-== =---- http://arcgraph.de/sr/ |