This file contains some additional information for the Philips webcams. E-mail: webcam@smcc.demon.nl Last updated: 2001-04-25 The main webpage for the Philips driver is http://www.smcc.demon.nl/webcam/. It contains a lot of extra information, a FAQ, and the binary plugin 'PWCX'. This plugin contains decompression routines that allow you to use higher image sizes and framerates; in addition the webcam uses less bandwidth on the USB bus (handy if you want to run more than 1 camera simultaneously). These routines fall under an NDA, and may therefor not be distributed as source; however, its use is completely optional. You can build this code either into your kernel, or as a module. I recommend the latter, since it makes troubleshooting a lot easier. The built-in microphone is supported through the USB Audio class. (Taken from install.html) When you load the module you can set some default settings for the camera; some programs depend on a particular image-size or -format. The options are: size Can be one of 'sqcif', 'qsif', 'qcif', 'sif', 'cif' or 'vga', for an image size of resp. 128x96, 160x120, 176x144, 320x240, 352x288 and 640x480 (of course, only for those cameras that support these resolutions). fps Specifies the desired framerate. Is an integer in the range of 4-30. palette Specifies the desired colour order that should be delivered by read() and mmap(). The string can be one of bgr24, rgb24, rgb32, bgr32, yuyv, yuv420, yuv420p. If the tool you use produces odd colours (more specificly, red and blue are swapped), try palette=bgr24 or palette=rgb24. fbufs This paramter specifies the number of internal buffers to use for storing frames from the cam. This will help if the process that reads images from the cam is a bit slow or momentarely busy. However, on slow machines it only introduces lag, so choose carefully. The default is 3, which is reasonable. You can set it between 2 and 5. mbufs This is an integer between 1 and 4. It will tell the module the number of buffers to reserve for mmap(), VIDIOCCGMBUF, VIDIOCMCAPTURE and friends. The default is 2, which is adequate for most applications (double buffering). Should you experience a lot of 'Dumping frame...' messages during grabbing with a tool that uses mmap(), you might want to increase if. However, it doesn't really buffer images, it just gives you a bit more slack when your program is behind. But you need a multi-threaded or forked program to really take advantage of these buffers. The absolute maximum is 4, but don't set it too high! Every buffer takes up 1.22 MB of RAM, so unless you have a lot of memory setting this to something more than 2 is an absolute waste. This memory is only allocated during open(), so nothing is wasted when the camera is not in use. power_save When power_save is enabled (set to 1), the module will try to shut down the cam on close() and re-activate on open(). This will save power and turn off the LED. Not all cameras support this though (the 645 and 646 don't have power saving at all), and some models don't work either (they will shut down, but never wake up). Consider this experimental. By default this option is disabled. compression (only useful with the plugin) With this option you can control the compression factor that the camera use to squeeze the image through the USB bus. You can set the parameter between 0 and 3: 0 = prefer uncompressed images; if the requested mode is not available in an uncompressed format, the driver will silently switch to low compression. 1 = low compression. 2 = medium compression. 3 = high compression. High compression takes less bandwidth of course, but it could also introduce some unwanted artefacts. The default is 2, medium compression. See the FAQ on the website for an overview of which modes require compression. The compression parameter only applies to the Vesta & ToUCam cameras. The 645 and 646 have fixed compression parameters. trace In order to better detect problems, it is now possible to turn on a 'trace' of some of the calls the module makes; it logs all items in your kernel log at debug level. The trace variable is a bitmask; each bit represents a certain feature. If you want to trace something, look up the bit value(s) in the table below, add the values together and supply that to the trace variable. Value Value Description Default (dec) (hex) 1 0x1 Module initialization; this will log messages On while loading and unloading the module 2 0x2 probe() and disconnect() traces On 4 0x4 Trace open() and close() calls Off 8 0x8 read(), mmap() and associated ioctl() calls Off 16 0x10 Memory allocation of buffers, etc. Off 32 0x20 Showing underflow, overflow and Dumping frame On messages 64 0x40 Show viewport and image sizes Off For example, to trace the open() & read() fuctions, sum 8 + 4 = 12, so you would supply trace=12 during insmod or modprobe. If you want to turn the initialization and probing tracing off, set trace=0. The default value for trace is 35 (0x23). Example: # modprobe pwc size=cif fps=15 power_save=1 The fbufs, mbufs and trace parameters are global and apply to all connected cameras. Each camera has its own set of buffers. size, fps, palette only specify defaults when you open() the device; this is to accommodate some tools that don't set the size or colour palette. You can change these settings after open() with the Video4Linux ioctl() calls. The default of defaults is QCIF size at 10 fps, BGR order. The compression parameter is semiglobal; it sets the initial compression preference for all camera's, but this parameter can be set per camera with the VIDIOCPWCSCQUAL ioctl() call. All parameters are optional.