The program has been tested on a Debian Linux system with an GeForce2
MX card and Brooktree 869 chip. If you run into trouble with a
different system, first look at the 'Config' page. Compare the PCI
Card and TV chip values with those of your system. (Either look
directly onto the card and read the chip labels, or examine the XFree
logfile with the closed source NVidia driver activated). When
reporting a bug, mention these values and please state exactly what
does not work and what happens when you try it.
So far, only Brooktree and Chrontel chips are supported; Chipsets
other than Brooktree or Chrontel (Philips, for example) could in
principle also be supported; contributions are welcome.
(Note: 'Unknown chips' in the I2C range A0-AF are EDID monitor
addresses and not TV chips).
Probably many, including:
* If you start with the -N or --nvdev option, and there is no running
closed source 'nvidia' X driver present, the system will crash due
to a bug in the NVidia kernel driver.
* Switching to TV mode might need several attempts, especially
for the large and huge Brooktree NTSC modes, and if the open-source
'nv' X driver is active.
* The Chrontel PAL modes have a significant distortion on the right
edge. Sometimes the whole right edge seems to be missing. The
NTSC modes are a lot better (also with respect to overscan size).
I have no idea if this is a fault of my Chrontel chip, the
Macrovision distortion, or wrong register programming.
* Not all of the predefined Brooktree modes work well. It should be
possible for all modes, except PAL 640x480 Large & Huge, to be displayed
as dualview on the monitor.
* Especially in the Brooktree Huge modes, there are CRT values that allow
doubleview, but there is not enough time to draw the hardware cursor
in these cases. That may cause a system freeze. (Disabling the
hardware cursor in XF86Config helps in those cases). Even without
hardware cursor, sometimes the system freezes.
* There are similar problems with the Chrontel PAL 800x600 Large mode:
I have sometimes experienced a system freeze, and the TV color flickers
since it seems not to be able to produce the data in time, so the
color information is shifted for half a pixel. (Color data is latched
on two subsequent clocks).
* The calculated Brooktree overscan compensation on the Status page
sometimes differs slightly from the one used in the calculation. I
have no idea why.
* The Chrontel overscan values are completely wrong. The calculations
have to be done in a different way.
* If the server dies, the client segfaults (should check eof condition