Patch: gspca-pac73xx-fixes-v2.patch

From: Hans de Goede <hidden>
Date: 2008-08-22 20:30:35

Hi All,

I made a little mistake with the previous patch (forgot to hg add the new
pac_common.h, so it didn't make it into the patch) please use this version and
don't forget to hg add pac_common.h :)

##

I'm proud to present the results of spending way too much hours (days)
working on pac73xx results:

-add documentation (based on trial and error) for some registers
-some preparations for adding autogain_n_exposure functionality:
  -remove some old (broken, disabled in current version) autogain cruft
  -add working average lumination reading
-various pac7311 fixes so that a pac7311 cam will actually work
  (tested with a trust wb 3400-T: 093a:260e)
-disable brightness and colors controls for 7311, brightness was actually
  changing a gain register (I'll do a seperate patch adding gain and exposure
  controls, I already know which registers to poke for this).
-fix contrast control for 7311
-add hflip and vflip controls for 7311
-only send a very minimal jpeg header, tinyjpeg in libv4l can already handle
  jpeg's without a huffman header, and I've added the special Pixart
  quantization table to tinyjpeg, this saves some decoding time and more
  important makes us more flexibles with regards to the quantization table.
  I have strong indications that the magic marker in the bitstream after
  each DCT block actually codes which quantization table to use for that
  block. I believe this as with my 7302 testcam, some blocks get decoded
  looking actually blocky, just like what happens with the 7311 when you
  use the wrong quantization table, and exactly these blocks have a marker
  of 0x48 instead of 0x44. Now if only I knew which quantization table to use
  for those blocks :(
-add proper SOF detection instead of relying on the JPEG EOF marker. This
  code was taken from the pac207 subdriver which uses the some SOF marker,
  and is moved to a header file shared by both drivers. Also check the JPEG
  EOF from the previous frame for additional robustness.

Signed-off-by: Hans de Goede <redacted>

Regards,

Hans

Attachments

Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help