On Tue, Dec 1, 2009 at 4:18 AM, Vincent McIntyre <vincent.mcint...@gmail.com> wrote: > Hi Rob > > I missed your followup and tested the 'revert.diff' patch, attached > for reference. > I have been slow replying because I've been scratching my head over the > results. > > I used 'signaltest.pl' to test[1], which uses tzap under the hood. > Perhaps this is not the best choice, but I wanted something that I thought > would > allow objective comparisons. That's trickier than I thought... > Unfortunately I only discovered last night how easy 'vlc > ./channels.conf' makes doing quick visual checks. I didn't have enough > time to re-patch and test again. > > My test procedure was: > - get a baseline with tzap and signaltest.pl > - patch, make, install. cold boot. > - test with tzap and signaltest.pl > - revert patch, for the moment. > > I tested two kernels, and both cards. I tested all the tuners but I'll > spare you that for now. > > * 2.6.24-23-rt + v4l (c57f47cfb0e8+ tip) > > I got rather different baseline results. All channels had > significantly higher BER > than I'd noticed before. After patching, snr on some channels > seemed a little higher > and BER was lower. On ch9, I think snr was up and BER improved a little. > > here are the signaltest summary tables: > without patch. usb device (dvb0) usbid db78:0fe9 > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 76.0 % 322.6 672.4 Seven > 191625000 76.0 % 320.2 1783.3 Nine > 219500000 76.8 % 329.8 2948.2 Ten > 226500000 76.9 % 296.6 4885.0 ABC > 571500000 77.0 % 542.0 7529.4 SBS > 578500000 77.1 % 539.5 10669.7 D44 > > with patch. usb device (dvb0) usbid db78:0fe9 > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 76.6 % 2.3 0.0 > 191625000 77.0 % 235.5 83.3 > 219500000 76.9 % 288.0 501.8 > 226500000 76.9 % 295.1 1416.4 > 571500000 77.0 % 523.4 3980.0 > 578500000 77.1 % 549.9 7409.4 > > without patch. pcie device (dvb1) pciid db78:18ac > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 71.2 % 3.1 0.0 > 191625000 21.7 % 645.4 246.4 > 219500000 73.6 % 1.9 1632.0 > 226500000 73.5 % 2.8 1632.0 > 571500000 73.9 % 13.6 2134.6 > 578500000 72.7 % 58.2 6393.4 > > with patch. pcie device (dvb1) pciid db78:18ac > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 73.2 % 4.0 0.0 > 191625000 74.0 % 37.0 0.0 > 219500000 73.9 % 0.0 0.0 > 226500000 73.0 % 4.6 0.0 > 571500000 74.2 % 76.7 193.6 > 578500000 72.8 % 213.8 4480.3 > > > * 2.6.31-14-generic + v4l (19c0469c02c3+ tip) > Hard to say if I'm seeing an improvement. > > before patching - adapter0 usbid db78:0fe9 > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 75.5 % 293.7 1926.4 > 191625000 75.9 % 363.2 2993.3 > 219500000 76.7 % 304.5 4225.8 > 226500000 76.9 % 223.8 6153.3 > 571500000 77.0 % 491.7 8726.0 > 578500000 77.1 % 558.9 11787.1 > > adapter0 repeat usbid db78:0fe9 (not sure what happened to UNC here..) > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 75.9 % 327.9 13893.6 > 191625000 76.0 % 392.8 14939.0 > 219500000 76.7 % 252.0 16052.0 > 226500000 76.8 % 254.0 18063.1 > 571500000 76.9 % 533.2 20644.1 > 578500000 76.9 % 464.1 23836.8 > > after patching - adapter0 usbid db78:0fe9 > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 76.3 % 2.5 0.0 > 191625000 76.8 % 227.6 119.0 > 219500000 76.8 % 262.6 604.5 > 226500000 76.8 % 282.7 1545.4 > 571500000 77.0 % 486.8 3541.7 > 578500000 77.1 % 521.5 6537.7 > > > before patching - adapter1 pciid db78:18ac > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 70.9 % 0.0 0.0 > 191625000 69.8 % 2.7 0.0 > 219500000 73.2 % 4.1 0.0 > 226500000 73.4 % 4.5 0.0 > 571500000 74.0 % 0.0 0.0 > 578500000 72.3 % 125.7 3589.3 > > after patching - adapter2 pciid db78:18ac (enumeration order changed) > Frequency Signal Ber Unc > ========= ======== ======== ======== > 177500000 73.6 % 0.6 0.0 > 191625000 74.2 % 0.0 0.0 > 219500000 74.0 % 4.9 0.0 > 226500000 73.3 % 163.2 349.7 > 571500000 74.4 % 267.0 1014.6 > 578500000 72.7 % 70.7 4906.0
The stats above suggest that reverting the patch in question significantly reduces the instances of uncorrectable errors. It looks like indeed that a regression was introduced. Mauro, any thoughts on this? I had heard things from a couple of other zl10353/xc3028 users that the tuning performance was pretty crappy compared to the mrec xc3028 driver. This could definitely be the cause. Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com
