On Sun, Jan 19, 2014 at 12:32 PM, Mauro Carvalho Chehab <m.che...@samsung.com> wrote: > Em Sun, 19 Jan 2014 11:50:55 -0500 > Devin Heitmueller <dheitmuel...@kernellabs.com> escreveu: > >> On Sun, Jan 19, 2014 at 8:39 AM, Mauro Carvalho Chehab >> <m.che...@samsung.com> wrote: >> > It seems that subsequent tuning makes the device worse, reducing the >> > maximum effective packet bandwidth. Btw, this happens with both xHCI >> > and EHCI drivers, so, it is not related to any USB 3.0 issue. >> >> I'm pretty sure I saw this and had a patch. I don't recall the exact >> circumstances under which it happened, but I believe it had to do with >> stopping and then restarting the streaming on the em28xx too quickly. >> The state machine inside the em28xx gets confused and you end up >> getting a misaligned stream (which is why you see hundreds of >> different PIDs in output from tools such as dvbtraffic). > > Do you still has your old tree? I'm not seeing it anymore at kernellabs.
Yeah, I still have the tree. It's not the HG tree that you saw a couple of years ago - it's on one of my private git trees because it was part of a commercial project. I'll have to dig around and see if I can find it. >> >> > Enabling some demux logs, it is possible to see that there are too many >> > FEC errors: >> > >> > [73514.186880] dvb_dmx_swfilter_packet: 4546 callbacks suppressed >> > [73514.186933] TEI detected. PID=0x17f data1=0xc1 >> > [73514.186965] TEI detected. PID=0x1c68 data1=0xbc >> > [73514.186993] TEI detected. PID=0x17f data1=0xc1 >> > [73514.187022] TEI detected. PID=0x1c68 data1=0xbc >> > [73514.187049] TEI detected. PID=0x17f data1=0xc1 >> > [73514.187080] TEI detected. PID=0x1c68 data1=0xbc >> > [73514.187105] TEI detected. PID=0x17f data1=0xc1 >> > [73514.194878] TEI detected. PID=0x1c68 data1=0xbc >> > [73514.194906] TEI detected. PID=0x17f data1=0xc1 >> > [73514.194927] TEI detected. PID=0x1c68 data1=0xbc >> > [73521.569205] TS speed 402 Kbits/sec >> >> Are these actually valid PIDs you're expecting data on? If not, then >> it could just be the issue I described above. Does the TEI check >> occur after it has found the SYNC byte? > > Yes. it keeps repeating several times, even after finding the SYNC. > > This patch makes the behavior stable: > > http://git.linuxtv.org/mchehab/experimental.git/commitdiff/88c9318cbea60d189a9b10cbc4c5a73f8b002336 > > Even so, the bitstream of my test signal is 19Mbps, while the measured > speed with the valid packets is being about 3Mbps. Something is seriously wrong then - I had it delivering all 19 Mbps. >> Probably worth mentioning that while I got signal lock on ATSC, I >> didn't any significant analysis into the quality of the SNR. It's >> possible that additional optimization of the frontend is required in >> order to achieve optimal performance. > > It is unlikely to be due to some optimization, as I'm not injecting > any errors via the RF generator. Sorry, I wasn't clear. I didn't intend to suggest any RF optimization is causing the issue you're seeing. I was just saying that I didn't do any optimization of the RF path so while it works under ideal signal conditions it may not work as well under more adverse signal conditions. In other words, I did what most of the other LinuxTV developers do - I got a successful signal lock and said "good enough". ;-) On that note, it's probably worth mentioning that particular design has an LNA controlled by one of the GPIOs on the DRX-J. So if you're consistently having poor RF performance (especially with a generator), try sticking an attenuator in-line, and/or make sure that the LNA is actually being properly disabled. > Btw, I noticed that there are two "extra" firmwares, that aren't currently > used, defined, on your driver, at drxj_mc_vsb.h and drxj_mc_vsbqam.h. > > Do you remember when those should be used? Or are them just two variants > of the main firmware, with support for just VSB and just ClearQAM? I would have to look at the source again to be sure, but if I recall it was just so you could reduce the size of the firmware if you didn't care about the other modulation scheme. Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- 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