[pvrusb2] B&W video problem...

[Mike Isely]

I have been spending too much time chasing a problem here, and I'm rapidly 
running out ideas.  This whole episode was the trigger that caused me 
update the driver to fix tda9887 issues, but I think now that this might 
have been a coincidence.  Maybe someone here can read this and see an 
idea...

I have two PVR USB2 devices here.  The newer one has been having a problem 
where the color just "stops working".  I don't have to do anything to the 
device to cause this - just sitting there the color can spontaneously 
cease and I'm left looking at B&W video.  It can take anywhere from 1 
minute to maybe 8 minutes of streaming before the color drops out.

I have been dealing with this problem for a while.  In fact, this symptom 
is what caused me to tumble into the whole tda9887 mess a month ago. 
However with all that solved now I still see the same problem.  Every 
combination of support modules and kernel versions I've tried (including 
the new 2.6.15 stable release) eventually does the same thing.

Other interesting bits:

When it is showing B&W, I am seeing a tiny checkerboard pattern on some of 
the gray levels - which I believe is the colorburst frequency bleeding 
through unfiltered.

Once this happens, the ONLY way to clear the problem is to power cycle the 
device.  Re-plugging (while powered on - which clears all the software 
state) doesn't help; it's still B&W right away immediately afterwards. 
Reloading the driver modules (even all of them) makes no difference 
either.  It's almost as if there is some piece of "state" in the device 
that nobody is controlling, and once this tips a certain way the only way 
to influence it again is to power cycle.

Meanwhile my older PVR USB2 device continues to work perfectly.  They're 
both NTSC devices.  The older stable device has "29022 Rev C523 
NTSC/NTSC-J" on the white tag underneath.  The newer unstable device has 
"29032 Rev D2A3 NTSC/NTSC-J" on its white tag.  The software sees the two 
devices being nearly identical, except the newer (unstable) device reports 
a Hauppauge type 103 tuner, mapped to type 43 in V4L - and it has a 
tda9887.  The older (stable) device reports a Hauppauge type 23 tuner, 
mapped to type 2 in V4L, and no tda9887.  Other than that the two devices 
are identical.

I had been thinking all along that this problem is tda9887 related since 
B&W video (in NTSC at least) on old TV sets is usually suggestive of 
out-of-whack tuning.  But once it starts happening, it happens SOLIDLY 
from that point forward, no matter what channel I select.  (Fine tuning 
the frequency doesn't help.)  I've also since re-snooped what the Windows 
driver is doing and have compared that to what tuner.ko and tda9887.ko are 
doing and they almost exactly match (even the frequency divider value 
matches, but the charge pump flag for the tuner is different and fm auto 
mute for the tda9887 is different - but I don't think either difference 
can explain this).

Has anyone here seen this problem?  And (perhaps more significantly) does 
anyone here have a device whose model number on the white tag underneath 
the device matches the info on my newer (unstable) device - and if you do 
have you seen any funny behavior like this?

   -Mike

-- 
                         |         Mike Isely          |     PGP fingerprint
      Spammers Die!!     |                             | 03 54 43 4D 75 E5 CC 92
                         |   isely @ pobox (dot) com   | 71 16 01 E2 B5 F5 C1 E8
                         |                             |