------------------------------------------------------------------------------- Readme for Linux device driver for the OmniVision OV511 USB to camera bridge IC ------------------------------------------------------------------------------- Author: Mark McClelland Homepage: http://alpha.dyndns.org/ov511 NEW IN THIS VERSION: o Improvements to sensor detection code o Added "i2c_detect_tries" and "aperture" parameters o proc filesystem status support o read() fixed partially o code cleanups and minor fixes INTRODUCTION: This is a preliminary version of my OV511 Linux device driver. Currently, it can grab a frame in color (YUV420) at 640x480 or 320x240 using either vidcat or xawtv. Other utilities may work but have not yet been tested. Any camera using the OV511/OV511+ and the OV7610/20/20AE CCD should work. The driver only detects known cameras though, based on their custom id number. If you have a currently unsupported camera, the ID number should be reported to you in the kernel logs. Please send me the model, manufacturer and ID number and I will add it to the detection code. In the meantime, you can add to the code yourself in the function ov511_probe(). WHAT YOU NEED: - If you want to help with the development, get the chip's specification docs at http://www.ovt.com/omniusbp.html - A Video4Linux compatible frame grabber program (I recommend vidcat and xawtv) vidcat is part of the w3cam package: http://www.hdk-berlin.de/~rasca/w3cam/ xawtv is available at: http://www.in-berlin.de/User/kraxel/xawtv.html HOW TO USE IT: You must have first compiled USB support, support for your specific USB host controller (UHCI or OHCI), and Video4Linux support for your kernel (I recommend making them modules.) Next, (as root) from your appropriate modules directory (lib/modules/2.3.XX): insmod usb/usbcore.o insmod usb/usb-uhci.o insmod usb/ohci-hcd.o insmod misc/videodev.o insmod usb/ov511.o If it is not already there (it usually is), create the video device: mknod /dev/video c 81 0 Sometimes /dev/video is a symlink to /dev/video0 You will have to set permissions on this device to allow you to read/write from it: chmod 666 /dev/video chmod 666 /dev/video0 (if necessary) Now you are ready to run a video app! Both vidcat and xawtv work well for me at 640x480. [Using vidcat:] vidcat -s 640x480 > test.jpg xview test.jpg [Using xawtv:] You must make some modifications to the source and compile it before you use it. (Note: this may not be applicable to versions other than 3.06) In src/Xawtv.ad, change xawtv.tv.width to 640 and xawtv.tv.height to 480. Next, in src/grab-v4l.c, change SYNC_TIMEOUT from 1 to 2. Then, from the main xawtv directory: make clean ./configure make make install Now you should be able to run xawtv. Right click for the options dialog. If you get a scrambled image it is likely that you made a mistake in Xawtv.ad. Try setting the size to 320x240 if all else fails. FAQ: Q: "Why does the picture have noise and look grainy" A: This is a problem at low light levels, and may be also due to subtle bugs in the code. The cause is most likely the OV7610 settings we are currently using. I am looking into this problem. Q: "The driver sometimes says `Failed to read OV7610 ID.' What is the deal?" A: The I2C code that allows the OV511 to communicate with the camera chip is a bit flaky right now. This message means that the I2C bus never got initialized properly, and the camera will most likely not work even if you disable this warning. Try unloading/reloading the driver or unplugging/re- plugging the camera if this happens. Q: "Why do you bother with this phony camera detection crap? It doesn't do anything useful!" A: The main purpose of only supporting known camera models is to force people with new camera models to tell me about them, so I can assemble the list above, and so the code can know what CCD chip you have. Right now, nearly all of the cameras use the OV7610 and consequently I have not put support for other ones in, so the value of the detection code is questionable. Eventually though, new CCDs might appear and we will be fortunate to have the detection. MODULE PARAMETERS: You can set these with: insmod ov511 NAME=VALUE There is currently no way to set these on a per-camera basis. NAME: autoadjust TYPE: integer (boolean) DEFAULT: 1 DESC: The camera normally adjusts exposure, gain, and hue automatically. This can be set to 0 to disable this automatic adjustment. Note that there is currently no way to set these parameters manually once autoadjust is disabled. NAME: debug TYPE: integer (0-6) DEFAULT: 3 DESC: Sets the threshold for printing debug messages. The higher the value, the more is printed. The levels are cumulative, and are as follows: 0=no debug messages 1=init/detection/unload and other significant messages 2=some warning messages 3=config/control function calls 4=most function calls and data parsing messages 5=highly repetitive mesgs NAME: fix_rgb_offset TYPE: integer (boolean) DEFAULT: 0 DESC: Some people have reported that the blue component of the image is one or so lines higher than the red component. This is only apparent in images with white objects on black backgrounds at 640x480. Setting this to 1 will realign the color planes correctly. NOTE: This is still experimental and very buggy. NAME: snapshot TYPE: integer (boolean) DEFAULT: 0 DESC: Set to 1 to enable snapshot mode. read() will block until the snapshot button is pressed. Note that this does not yet work with most apps, including xawtv and vidcat. NOTE: See the section "TODO" for more info. WORKING FEATURES: o Color streaming/capture at 640x480 and 320x240 o YUV420 color o Monochrome o Setting/getting of saturation, contrast and brightness (no hue yet; only works with OV7610, not the OV7620 or OV7620AE) o proc status reporting EXPERIMENTAL FEATURES: o fix_rgb_offset: Sometimes works, but other times causes errors with xawtv and corrupted frames. o Snapshot mode (only works with some read() based apps; see below for more) o read() support TODO: o Fix the noise / grainy image problem. o Get compression working. It would be a nice addition as it improves frame rate quite a bit. OmniVision wouldn't tell me how the algorithm works, so we can't really work on that yet. Please kindly inform OmniVision that you would like them to release their specifications to the Linux community. o Get 160x120 working o YUV422 (and other color modes) o Fix read(). It only works right now if you run an mmap() based app like xawtv or vidcat after loading the module and before using read(). Apparently there are some initialization issues. o Get snapshot mode working with mmap(). o Fix fixFrameRGBoffset(). It is not stable yet with streaming video. o Get hue (red/blue channel balance) adjustment working (in ov511_get_picture() and ov511_set_picture()) o Get autoadjust disable working o V4L2 support (Probably not until it goes into the kernel) o Fix I2C initialization. Some people are reporting problems with reading the 7610 registers. This could be due to timing differences, an excessive I2C clock rate, or a problem with ov511_i2c_read(). o Get rid of the memory management functions (put them in videodev.c??) o Setting of contrast and brightness not working with 7620 o Driver/camera state save/restore for when USB supports suspend/resume o Multiple cameras reportedly do not work simultaneously o Problems with OHCI HOW TO CONTACT ME: You can email me at mmcclelland@delphi.com . Please prefix the subject line with "OV511: " so that I am certain to notice your message. CREDITS: The code is based in no small part on the CPiA driver by Johannes Erdfelt, Randy Dunlap, and others. Big thanks to them for their pioneering work on that and the USB stack. Thanks to Bret Wallach for getting camera reg IO , ISOC, and image capture working. Thanks to Orion Sky Lawlor and Kevin Moore for their work as well.