[pvrusb2] HVR-1900 tuning drift problem?

[Gary Francis]

Hi,
     I have an HVR-1900 with a problem that behaves like tuner frequency 
drift when viewing or recording from DVB. I appreciate any help you can 
offer to diagnose and resolve this problem. I am using the latest 
mercurial download of v4l-dvb with a 2.6.28 kernel.

I first noticed this problem as a periodic degradation of the decoded 
image, consistent with errors in the digital stream. Approximately every 
10 minutes I would see picture disturbances that gradually ramp up in 
severity until the video was no longer watchable, with severe audio 
breakup. The video would then suddenly correct itself and play normally.

To investigate further I have spent some time recording signal levels 
under various conditions with tzap. I have found that bit error rates 
behave consistent with the picture errors I observe. The bit error rate 
will idle around a fixed level (say 0x600), then begin to ramp up. As 
bit error rate increases, I also see a reduction in both signal to noise 
ratio and received signal level. As bit error rate exceeds approximately 
0x4000 uncorrectable errors begin to increase. As bit error rate exceeds 
approximately 0x5000 it will suddenly drop back to the initial quiescent 
level with corresponding increases in signal to noise ratio and received 
signal.

I say approximately every 10 minutes as this is typical, although I have 
observed cycle times as low as every 90s, and as long as several hours.

I have observed the fastest cycle times after a cold start. This showed 
an initial cycle of about 10 minutes, but had decreased to 90 seconds 
after 2 hours. This leads me to believe that the root cause is something 
in the tuner drifting with temperature. After the 2 hours I tried tuning 
to another channel with a different frequency, then back to the original 
channel. The problem persisted with 90 second period. I then removed the 
pvrusb2 and tda10048 modules and inserted pvrusb2 again. After this the 
no further problems were observed for the following 2 hours. Maybe the 
HVR-1900 samples some temperature dependent parameter at boot, and with 
the equipment at a stable temperature after the initial 2 hour soak, the 
drift is much less evident.

I have not observed the problem using Windows. Does the windows driver 
cause some sort of periodic reset to avoid the problem? I have not used 
the hardware with Windows very much, so it is possible that I have just 
not noticed it. I plan to repeat the cold start test using Windows soon.

In case the problem was resolved in the tda10048 firmware that windows 
used, I spent some time tracking down the firmware in the Windows 
driver. Looking in the Hauppauge 1.2a CD download I eventually found it 
in Drivers/Driver73/Hcw73bda.sys. It was at a file offset of 471424 with 
a length of 25098. Since it has a different length to the normal 
dvb-fe-tda10048-1.0.fw, I had to adjust the tda10048 kernel module to 
use it. Unfortunately, I have not noticed any difference using the newer 
firmware.

Do you know of any good references to the architecture of the HVR-1900 
and the chips used in it? Maybe looking though the device data sheets 
would give me further clues.

How should I probe the devices with linux to discover the exact devices 
used?

Gary