On Fri, 2009-09-11 at 17:04 -0400, Devin Heitmueller wrote: > On Fri, Sep 11, 2009 at 4:56 PM, Andy Walls <awalls@xxxxxxxxx> wrote: > > + /* FIXME: hard-coded for NTSC support */ > > + format->fmt.vbi.sampling_rate = 6750000 * 4 / 2; /* FIXME: ??? */ > > > > The BT.656 Pixel clock for both 60 Hz and 50 Hz systems is 13.5 > > Mpixels/sec. Unless you switch to PAL or NTSC square pixel timing for > > some reason, I would expect this particular number to stay the same for > > both PAL and NTSC. > > > > A pixel in BT.656 consists of 1 Y sample and 1 U or V sample. So, in > > the VBI active line regions, I would have expected 2 * 13.5 Msamples/sec > > for VBI data. I don't know if any app in userland actually pays > > attention to this number though.... > > > > Regards, > > Andy > > Hey Andy, > > Well, when I said "hard-coded for NTSC support", I was referring to > the entire block of code that followed, not just the one on the > sampling rate. I expect the line count and start offsets to be > different for PAL as well. Ah, OK. Yes they are. > > In reality, the "???" was because I actually didn't know what the > proper value needs to be. I copied it from another driver without > fully understanding its role, and changing it caused userland to break > (zvbi-ntsc-cc in this case). > Worth noting that we are actually delivering 8-bit VBI info, as > opposed to 16-bit yuyv (there are some USB bandwidth issues I'm > actively working and doing 16-bit put me over the limit), so I'm not > confident that the generalization for ITU656 applies. OK. You can always work backwards to get the number: For example, if your hardware uses the BT.656 pixel clock of 13.5 Mpixels/sec and BT.656 pixel count of 720 pixels per active region of a horizontal line, then: 720 pixels/line active region / 13.5 Mpixels/sec = 53.3333 usecs / line active region And so your VBI sample rate is: n VBI samples/line active region / 53.3333 usecs/line active region with n being whatever the driver provides per VBI line. Ah and now I see you're only providing 720 samples per line. So yes, 13.5 Msamples/sec is the right number. Mea culpa. Regards, Andy > As I get the PAL support working, that area will certainly get some cleanup... > Cheers, > > Devin -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html