1Driver for Trust Computer Products Framegrabber, version 0.6.1 2------ --- ----- -------- -------- ------------ ------- - - - 3 4- ZORAN ------------------------------------------------------ 5 Author: Pauline Middelink <middelin@polyware.nl> 6 Date: 18 September 1999 7Version: 0.6.1 8 9- Description ------------------------------------------------ 10 11Video4Linux compatible driver for an unknown brand framegrabber 12(Sold in the Netherlands by TRUST Computer Products) and various 13other zoran zr36120 based framegrabbers. 14 15The card contains a ZR36120 Multimedia PCI Interface and a Philips 16SAA7110 Onechip Frontend videodecoder. There is also an DSP of 17which I have forgotten the number, since i will never get that thing 18to work without specs from the vendor itself. 19 20The SAA711x are capable of processing 6 different video inputs, 21CVBS1..6 and Y1+C1, Y2+C2, Y3+C3. All in 50/60Hz, NTSC, PAL or 22SECAM and delivering a YUV datastream. On my card the input 23'CVBS-0' corresponds to channel CVBS2 and 'S-Video' to Y2+C2. 24 25I have some reports of other cards working with the mentioned 26chip sets. For a list of other working cards please have a look 27at the cards named in the tvcards struct in the beginning of 28zr36120.c 29 30After some testing, I discovered that the carddesigner messed up 31on the I2C interface. The Zoran chip includes 2 lines SDA and SCL 32which (s)he connected reversely. So we have to clock on the SDA 33and r/w data on the SCL pin. Life is fun... Each cardtype now has 34a bit which signifies if you have a card with the same deficiency. 35 36Oh, for the completeness of this story I must mention that my 37card delivers the VSYNC pulse of the SAA chip to GIRQ1, not 38GIRQ0 as some other cards have. This is also incorporated in 39the driver be clearing/setting the 'useirq1' bit in the tvcard 40description. 41 42Another problems of continuous capturing data with a Zoran chip 43is something nasty inside the chip. It effectively halves the 44fps we ought to get... Here is the scenario: capturing frames 45to memory is done in the so-called snapshot mode. In this mode 46the Zoran stops after capturing a frame worth of data and wait 47till the application set GRAB bit to indicate readiness for the 48next frame. After detecting a set bit, the chip neatly waits 49till the start of a frame, captures it and it goes back to off. 50Smart ppl will notice the problem here. Its the waiting on the 51_next_ frame each time we set the GRAB bit... Oh well, 12,5 fps 52is still plenty fast for me. 53-- update 28/7/1999 -- 54Don't believe a word I just said... Proof is the output 55of `streamer -t 300 -r 25 -f avi15 -o /dev/null` 56 ++--+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 57 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 58 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 59 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 60 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 61 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 62 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 63 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 64 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 65 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 66 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 25/25 67 +-s+-+-+-+-+-+-+-+-+-+-+-+-+-s+-+-+-+-+-+-+-+-+-+-+- 68 syncer: done 69 writer: done 70(note the /dev/null is prudent here, my system is not able to 71 grab /and/ write 25 fps to a file... gifts welcome :) ) 72The technical reasoning follows: The zoran completed the last 73frame, the VSYNC goes low, and GRAB is cleared. The interrupt 74routine starts to work since its VSYNC driven, and again 75activates the GRAB bit. A few ms later the VSYNC (re-)rises and 76the zoran starts to work on a new and freshly broadcasted frame.... 77 78For pointers I used the specs of both chips. Below are the URLs: 79 http://www.zoran.com/ftp/download/devices/pci/ZR36120/36120data.pdf 80 http://www-us.semiconductor.philips.com/acrobat/datasheets/SAA_7110_A_1.pdf 81 82The documentation has very little on absolute numbers or timings 83needed for the various modes/resolutions, but there are other 84programs you can borrow those from. 85 86------ Install -------------------------------------------- 87Read the file called TODO. Note its long list of limitations. 88 89Build a kernel with VIDEO4LINUX enabled. Activate the 90BT848 driver; we need this because we have need for the 91other modules (i2c and videodev) it enables. 92 93To install this software, extract it into a suitable directory. 94Examine the makefile and change anything you don't like. Type "make". 95 96After making the modules check if you have the much needed 97/dev/video devices. If not, execute the following 4 lines: 98 mknod /dev/video c 81 0 99 mknod /dev/video1 c 81 1 100 mknod /dev/video2 c 81 2 101 mknod /dev/video3 c 81 3 102 mknod /dev/video4 c 81 4 103 104After making/checking the devices do: 105 modprobe i2c 106 modprobe videodev 107 modprobe saa7110 (optional) 108 modprobe saa7111 (optional) 109 modprobe tuner (optional) 110 insmod zoran cardtype=<n> 111 112<n> is the cardtype of the card you have. The cardnumber can 113be found in the source of zr36120. Look for tvcards. If your 114card is not there, please try if any other card gives some 115response, and mail me if you got a working tvcard addition. 116 117PS. <TVCard editors behold!) 118 Dont forget to set video_input to the number of inputs 119 you defined in the video_mux part of the tvcard definition. 120 Its a common error to add a channel but not incrementing 121 video_input and getting angry with me/v4l/linux/linus :( 122 123You are now ready to test the framegrabber with your favorite 124video4linux compatible tool 125 126------ Application ---------------------------------------- 127 128This device works with all Video4Linux compatible applications, 129given the limitations in the TODO file. 130 131------ API ------------------------------------------------ 132 133This uses the V4L interface as of kernel release 2.1.116, and in 134fact has not been tested on any lower version. There are a couple 135of minor differences due to the fact that the amount of data returned 136with each frame varies, and no doubt there are discrepancies due to my 137misunderstanding of the API. I intend to convert this driver to the 138new V4L2 API when it has stabilized more. 139 140------ Current state -------------------------------------- 141 142The driver is capable of overlaying a video image in screen, and 143even capable of grabbing frames. It uses the BIGPHYSAREA patch 144to allocate lots of large memory blocks when tis patch is 145found in the kernel, but it doesn't need it. 146The consequence is that, when loading the driver as a module, 147the module may tell you it's out of memory, but 'free' says 148otherwise. The reason is simple; the modules wants its memory 149contiguous, not fragmented, and after a long uptime there 150probably isn't a fragment of memory large enough... 151 152The driver uses a double buffering scheme, which should really 153be an n-way buffer, depending on the size of allocated framebuffer 154and the requested grab-size/format. 155This current version also fixes a dead-lock situation during irq 156time, which really, really froze my system... :) 157 158Good luck. 159 Pauline 160