On Sat, 2010-03-13 at 12:46 -0500, Dale Pontius wrote: > Andy Walls wrote: > > The PCI bus uses "reflected wave swithcing", which means that one > > bad/dusty/oxidized PCI board connection can affect the other boards. > > Yikes - I never realized that, but then I never really thought about it, > either. Wouldn't that also mean that a lot of issues become dependent > on how many slots there are and how many slots are populated, etc? In > other words, plugging more cards in might make a system flakier - or > more stable, same for moving a card from one slot to another. (And > that's not just moving it to another IRQ.)
Yup. But in practice an arrangement of physical slots for a single PCI bus segment can only have 5 physcial slots. Southbridge chips can have a large number of devices integrated in silicon though. The idea was to not deal with termination of the PCI bus "transmission lines" and rely on the reflected waves coming back to give you the proper signal levels. Fun, huh? The PCI local-bus specification is a very interesting read. At the PCI SIG (.org) website, the freely available developers' material and briefs from years back may provide a nice tutorial with pictures and rationale. Copies of the spec's themselves are costly. > Dale > (In a past life, I was a Rambus licensee, speaking of signal > propagation. I specifically did voltage reference and regulation, > temperature sensor, and CAD tools. But I was aware of the rest of the > design, and particularly marveled at the I/O and DLL designs. Now that > I write it, the I/O and DLL simply ARE the crux of Rambus, along with a > protocol to make it all work together, and a few bits and pieces. The > array was our design that we were bolting on.) Neat. Didn't the owners of the Rambus patents recently try to sue every memory manufacturer on Earth? BTW, if you do a little digging on patents by Rockwell-Collins and Conexant you can find patents and papers that somewhat explain DLLs and SRCs that are most likely used in the CX25843 and CX23418. For example, I suspect Fritz Rothacher's thesis work directly impacted the CX25843 design: http://www.guest.iis.ee.ethz.ch/~rota/ since he is later listed on some Conexant patents. Regards, Andy _______________________________________________ ivtv-users mailing list [email protected] http://ivtvdriver.org/mailman/listinfo/ivtv-users
