On Mon, 2010-01-04 at 22:13 -0500, Devin Heitmueller wrote: > Hey Andy, > > On Mon, Jan 4, 2010 at 9:27 PM, Andy Walls <awa...@radix.net> wrote: > > The changes in question (mostly authored by me) are based on > > documentation on what offsets are to be used with the firmware for > > various DVB bandwidths and demodulators. The change was tested by Terry > > on a Leadtek DVR 3100 H Analog/DVB-T card (CX23418, ZL10353, XC3028) and > > some other cards I can't remember, using a DVB-T pattern generator for 7 > > and 8 MHz in VHF and UHF, and live DVB-T broadcasts in UHF for 6 MHz. > > > > (Devin, > > Maybe you can double check on the offsets in tuner-xc2028.c with any > > documentation you have available to you?) > > At this point the extent to which I've looked in to the issue was > validating that, for a given frequency, the change resulted in a > crappy SNR with lots of BER/UNC errors, and after reverting the change > the signal looked really good with zero BER/UNC. I haven't dug into > *why* it is an issue, but I examined the traces and looked at the > testing methodology and can confirm that there was definitely a > regression and Robert narrowed it down to the patch in question. > > I was kind of hoping that one of the people that helped introduce the > regression would take on some of responsibility to help with the > debugging. ;-)
I take responsiblity for the change. However, if fixing a known problem unmasks another problem, then is that a regression? I puzzled over the docs for a while until I had the "Aha!" moment and understood what they were saying. I'm really confident about the freq offset changes - especially since using the wrong center freq in channels.conf is an easy way to mask incorrect freq offsets in the driver module. I'm less confident about the xc3028 firmware segments as extracted and repackaged for linux. I was not involved in that development and I conveniently (for me) assume it is correct -- although that may be an assumption worth challenging. I also do not know the source of the commanded DTV freq's that are in use in the reported problem case. Using the wrong DTV center freq can cause the same problem symptoms as moving the offset used in the tuner-xc2028 module (two wrongs making a right). I just found a nice authoritative Australian source on DTV freq licensees (see my other foloow-up email), so hopefully Robert can double check that. Of course testing with a DVB-T generator instead of a broadcaster's signal would eliminate any doubt about the center freq in use. > I think I have one of the boards that will demonstrate the issue (a > Terratec board with xc3028/zl10353), and will try to find some time > with the generator once I wrap up the xc4000 work for the PCTV 340e. OK, thanks. I have no hardware with which to test. Regards, Andy > Devin -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
