2 SN9C1xx PC Camera Controllers
4 =============================
14 4. Overview and features
15 5. Module dependencies
18 8. Optional device control through "sysfs"
20 10. Notes for V4L2 application developers
21 11. Video frame formats
22 12. Contact information
28 Copyright (C) 2004-2006 by Luca Risolia <luca.risolia@studio.unibo.it>
33 SONiX is a trademark of SONiX Technology Company Limited, inc.
34 This software is not sponsored or developed by SONiX.
39 This program is free software; you can redistribute it and/or modify
40 it under the terms of the GNU General Public License as published by
41 the Free Software Foundation; either version 2 of the License, or
42 (at your option) any later version.
44 This program is distributed in the hope that it will be useful,
45 but WITHOUT ANY WARRANTY; without even the implied warranty of
46 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
47 GNU General Public License for more details.
49 You should have received a copy of the GNU General Public License
50 along with this program; if not, write to the Free Software
51 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
54 4. Overview and features
55 ========================
56 This driver attempts to support the video interface of the devices assembling
57 the SONiX SN9C101, SN9C102, SN9C103, SN9C105 and SN9C120 PC Camera Controllers
58 ("SN9C1xx" from now on).
60 The driver relies on the Video4Linux2 and USB core modules. It has been
61 designed to run properly on SMP systems as well.
63 The latest version of the SN9C1xx driver can be found at the following URL:
64 http://www.linux-projects.org/
66 Some of the features of the driver are:
68 - full compliance with the Video4Linux2 API (see also "Notes for V4L2
69 application developers" paragraph);
70 - available mmap or read/poll methods for video streaming through isochronous
72 - automatic detection of image sensor;
73 - support for built-in microphone interface;
74 - support for any window resolutions and optional panning within the maximum
75 pixel area of image sensor;
76 - image downscaling with arbitrary scaling factors from 1, 2 and 4 in both
77 directions (see "Notes for V4L2 application developers" paragraph);
78 - two different video formats for uncompressed or compressed data in low or
79 high compression quality (see also "Notes for V4L2 application developers"
80 and "Video frame formats" paragraphs);
81 - full support for the capabilities of many of the possible image sensors that
82 can be connected to the SN9C1xx bridges, including, for instance, red, green,
83 blue and global gain adjustments and exposure (see "Supported devices"
84 paragraph for details);
85 - use of default color settings for sunlight conditions;
86 - dynamic I/O interface for both SN9C1xx and image sensor control and
87 monitoring (see "Optional device control through 'sysfs'" paragraph);
88 - dynamic driver control thanks to various module parameters (see "Module
89 parameters" paragraph);
90 - up to 64 cameras can be handled at the same time; they can be connected and
91 disconnected from the host many times without turning off the computer, if
92 the system supports hotplugging;
96 5. Module dependencies
97 ======================
98 For it to work properly, the driver needs kernel support for Video4Linux and
101 The following options of the kernel configuration file must be enabled and
102 corresponding modules must be compiled:
108 To enable advanced debugging functionality on the device through /sysfs:
112 CONFIG_VIDEO_ADV_DEBUG=y
118 In addition, depending on the hardware being used, the modules below are
121 # USB Host Controller Drivers
123 CONFIG_USB_EHCI_HCD=m
124 CONFIG_USB_UHCI_HCD=m
125 CONFIG_USB_OHCI_HCD=m
127 The SN9C103, SN9c105 and SN9C120 controllers also provide a built-in microphone
128 interface. It is supported by the USB Audio driver thanks to the ALSA API:
134 # Advanced Linux Sound Architecture
140 CONFIG_SND_USB_AUDIO=m
144 # USB Multimedia devices
151 To use the driver, it is necessary to load the "sn9c102" module into memory
152 after every other module required: "videodev", "v4l2_common", "compat_ioctl32",
153 "usbcore" and, depending on the USB host controller you have, "ehci-hcd",
154 "uhci-hcd" or "ohci-hcd".
156 Loading can be done as shown below:
158 [root@localhost home]# modprobe sn9c102
160 Note that the module is called "sn9c102" for historic reasons, althought it
161 does not just support the SN9C102.
163 At this point all the devices supported by the driver and connected to the USB
164 ports should be recognized. You can invoke "dmesg" to analyze kernel messages
165 and verify that the loading process has gone well:
167 [user@localhost home]$ dmesg
169 or, to isolate all the kernel messages generated by the driver:
171 [user@localhost home]$ dmesg | grep sn9c102
176 Module parameters are listed below:
177 -------------------------------------------------------------------------------
179 Type: short array (min = 0, max = 64)
181 Description: Specify V4L2 minor mode number:
182 -1 = use next available
183 n = use minor number n
184 You can specify up to 64 cameras this way.
186 video_nr=-1,2,-1 would assign minor number 2 to the second
187 recognized camera and use auto for the first one and for every
190 -------------------------------------------------------------------------------
192 Type: bool array (min = 0, max = 64)
194 Description: Force the application to unmap previously mapped buffer memory
195 before calling any VIDIOC_S_CROP or VIDIOC_S_FMT ioctl's. Not
196 all the applications support this feature. This parameter is
197 specific for each detected camera.
198 0 = do not force memory unmapping
199 1 = force memory unmapping (save memory)
201 -------------------------------------------------------------------------------
203 Type: uint array (min = 0, max = 64)
205 Description: Timeout for a video frame in seconds before returning an I/O
206 error; 0 for infinity. This parameter is specific for each
207 detected camera and can be changed at runtime thanks to the
208 /sys filesystem interface.
210 -------------------------------------------------------------------------------
214 Description: Debugging information level, from 0 to 3:
215 0 = none (use carefully)
217 2 = significant informations
218 3 = more verbose messages
219 Level 3 is useful for testing only, when only one device
220 is used. It also shows some more informations about the
221 hardware being detected. This parameter can be changed at
222 runtime thanks to the /sys filesystem interface.
224 -------------------------------------------------------------------------------
227 8. Optional device control through "sysfs" [1]
228 ==========================================
229 If the kernel has been compiled with the CONFIG_VIDEO_ADV_DEBUG option enabled,
230 it is possible to read and write both the SN9C1xx and the image sensor
231 registers by using the "sysfs" filesystem interface.
233 Every time a supported device is recognized, a write-only file named "green" is
234 created in the /sys/class/video4linux/videoX directory. You can set the green
235 channel's gain by writing the desired value to it. The value may range from 0
236 to 15 for the SN9C101 or SN9C102 bridges, from 0 to 127 for the SN9C103,
237 SN9C105 and SN9C120 bridges.
238 Similarly, only for the SN9C103, SN9C105 and SN9120 controllers, blue and red
239 gain control files are available in the same directory, for which accepted
240 values may range from 0 to 127.
242 There are other four entries in the directory above for each registered camera:
243 "reg", "val", "i2c_reg" and "i2c_val". The first two files control the
244 SN9C1xx bridge, while the other two control the sensor chip. "reg" and
245 "i2c_reg" hold the values of the current register index where the following
246 reading/writing operations are addressed at through "val" and "i2c_val". Their
247 use is not intended for end-users. Note that "i2c_reg" and "i2c_val" will not
248 be created if the sensor does not actually support the standard I2C protocol or
249 its registers are not 8-bit long. Also, remember that you must be logged in as
250 root before writing to them.
252 As an example, suppose we were to want to read the value contained in the
253 register number 1 of the sensor register table - which is usually the product
254 identifier - of the camera registered as "/dev/video0":
256 [root@localhost #] cd /sys/class/video4linux/video0
257 [root@localhost #] echo 1 > i2c_reg
258 [root@localhost #] cat i2c_val
260 Note that "cat" will fail if sensor registers cannot be read.
262 Now let's set the green gain's register of the SN9C101 or SN9C102 chips to 2:
264 [root@localhost #] echo 0x11 > reg
265 [root@localhost #] echo 2 > val
267 Note that the SN9C1xx always returns 0 when some of its registers are read.
268 To avoid race conditions, all the I/O accesses to the above files are
270 The sysfs interface also provides the "frame_header" entry, which exports the
271 frame header of the most recent requested and captured video frame. The header
272 is always 18-bytes long and is appended to every video frame by the SN9C1xx
273 controllers. As an example, this additional information can be used by the user
274 application for implementing auto-exposure features via software.
276 The following table describes the frame header exported by the SN9C101 and
279 Byte # Value or bits Description
280 ------ ------------- -----------
281 0x00 0xFF Frame synchronisation pattern
282 0x01 0xFF Frame synchronisation pattern
283 0x02 0x00 Frame synchronisation pattern
284 0x03 0xC4 Frame synchronisation pattern
285 0x04 0xC4 Frame synchronisation pattern
286 0x05 0x96 Frame synchronisation pattern
287 0x06 [3:0] Read channel gain control = (1+R_GAIN/8)
288 [7:4] Blue channel gain control = (1+B_GAIN/8)
289 0x07 [ 0 ] Compression mode. 0=No compression, 1=Compression enabled
290 [2:1] Maximum scale factor for compression
291 [ 3 ] 1 = USB fifo(2K bytes) is full
292 [ 4 ] 1 = Digital gain is finish
293 [ 5 ] 1 = Exposure is finish
295 0x08 [7:0] Y sum inside Auto-Exposure area (low-byte)
296 0x09 [7:0] Y sum inside Auto-Exposure area (high-byte)
297 where Y sum = (R/4 + 5G/16 + B/8) / 32
298 0x0A [7:0] Y sum outside Auto-Exposure area (low-byte)
299 0x0B [7:0] Y sum outside Auto-Exposure area (high-byte)
300 where Y sum = (R/4 + 5G/16 + B/8) / 128
308 The following table describes the frame header exported by the SN9C103:
310 Byte # Value or bits Description
311 ------ ------------- -----------
312 0x00 0xFF Frame synchronisation pattern
313 0x01 0xFF Frame synchronisation pattern
314 0x02 0x00 Frame synchronisation pattern
315 0x03 0xC4 Frame synchronisation pattern
316 0x04 0xC4 Frame synchronisation pattern
317 0x05 0x96 Frame synchronisation pattern
318 0x06 [6:0] Read channel gain control = (1/2+R_GAIN/64)
319 0x07 [6:0] Blue channel gain control = (1/2+B_GAIN/64)
321 0x08 [ 0 ] Compression mode. 0=No compression, 1=Compression enabled
322 [2:1] Maximum scale factor for compression
323 [ 3 ] 1 = USB fifo(2K bytes) is full
324 [ 4 ] 1 = Digital gain is finish
325 [ 5 ] 1 = Exposure is finish
327 0x09 [7:0] Y sum inside Auto-Exposure area (low-byte)
328 0x0A [7:0] Y sum inside Auto-Exposure area (high-byte)
329 where Y sum = (R/4 + 5G/16 + B/8) / 32
330 0x0B [7:0] Y sum outside Auto-Exposure area (low-byte)
331 0x0C [7:0] Y sum outside Auto-Exposure area (high-byte)
332 where Y sum = (R/4 + 5G/16 + B/8) / 128
333 0x0D [1:0] Audio frame number
334 [ 2 ] 1 = Audio is recording
335 0x0E [7:0] Audio summation (low-byte)
336 0x0F [7:0] Audio summation (high-byte)
337 0x10 [7:0] Audio sample count
338 0x11 [7:0] Audio peak data in audio frame
340 The AE area (sx, sy, ex, ey) in the active window can be set by programming the
341 registers 0x1c, 0x1d, 0x1e and 0x1f of the SN9C1xx controllers, where one unit
342 corresponds to 32 pixels.
344 [1] The frame headers exported by the SN9C105 and SN9C120 are not described.
349 None of the names of the companies as well as their products will be mentioned
350 here. They have never collaborated with the author, so no advertising.
352 From the point of view of a driver, what unambiguously identify a device are
353 its vendor and product USB identifiers. Below is a list of known identifiers of
354 devices assembling the SN9C1xx PC camera controllers:
419 The list above does not imply that all those devices work with this driver: up
420 until now only the ones that assemble the following image sensors are
421 supported; kernel messages will always tell you whether this is the case (see
422 "Module loading" paragraph):
426 HV7131D Hynix Semiconductor, Inc.
427 MI-0343 Micron Technology, Inc.
428 OV7630 OmniVision Technologies, Inc.
429 OV7660 OmniVision Technologies, Inc.
430 PAS106B PixArt Imaging, Inc.
431 PAS202BCA PixArt Imaging, Inc.
432 PAS202BCB PixArt Imaging, Inc.
433 TAS5110C1B Taiwan Advanced Sensor Corporation
434 TAS5130D1B Taiwan Advanced Sensor Corporation
436 Some of the available control settings of each image sensor are supported
437 through the V4L2 interface.
439 Donations of new models for further testing and support would be much
440 appreciated. Non-available hardware will not be supported by the author of this
444 10. Notes for V4L2 application developers
445 =========================================
446 This driver follows the V4L2 API specifications. In particular, it enforces two
449 - exactly one I/O method, either "mmap" or "read", is associated with each
450 file descriptor. Once it is selected, the application must close and reopen the
451 device to switch to the other I/O method;
453 - although it is not mandatory, previously mapped buffer memory should always
454 be unmapped before calling any "VIDIOC_S_CROP" or "VIDIOC_S_FMT" ioctl's.
455 The same number of buffers as before will be allocated again to match the size
456 of the new video frames, so you have to map the buffers again before any I/O
459 Consistently with the hardware limits, this driver also supports image
460 downscaling with arbitrary scaling factors from 1, 2 and 4 in both directions.
461 However, the V4L2 API specifications don't correctly define how the scaling
462 factor can be chosen arbitrarily by the "negotiation" of the "source" and
463 "target" rectangles. To work around this flaw, we have added the convention
464 that, during the negotiation, whenever the "VIDIOC_S_CROP" ioctl is issued, the
465 scaling factor is restored to 1.
467 This driver supports two different video formats: the first one is the "8-bit
468 Sequential Bayer" format and can be used to obtain uncompressed video data
469 from the device through the current I/O method, while the second one provides
470 "raw" compressed video data (without frame headers not related to the
471 compressed data). The compression quality may vary from 0 to 1 and can be
472 selected or queried thanks to the VIDIOC_S_JPEGCOMP and VIDIOC_G_JPEGCOMP V4L2
473 ioctl's. For maximum flexibility, both the default active video format and the
474 default compression quality depend on how the image sensor being used is
475 initialized (as described in the documentation of the API for the image sensors
476 supplied by this driver).
479 11. Video frame formats [1]
480 =======================
481 The SN9C1xx PC Camera Controllers can send images in two possible video
482 formats over the USB: either native "Sequential RGB Bayer" or compressed.
483 The compression is used to achieve high frame rates. With regard to the
484 SN9C101, SN9C102 and SN9C103, the compression is based on the Huffman encoding
485 algorithm described below, while the SN9C105 and SN9C120 the compression is
486 based on the JPEG standard.
487 The current video format may be selected or queried from the user application
488 by calling the VIDIOC_S_FMT or VIDIOC_G_FMT ioctl's, as described in the V4L2
491 The name "Sequential Bayer" indicates the organization of the red, green and
492 blue pixels in one video frame. Each pixel is associated with a 8-bit long
493 value and is disposed in memory according to the pattern shown below:
495 B[0] G[1] B[2] G[3] ... B[m-2] G[m-1]
496 G[m] R[m+1] G[m+2] R[m+2] ... G[2m-2] R[2m-1]
498 ... B[(n-1)(m-2)] G[(n-1)(m-1)]
499 ... G[n(m-2)] R[n(m-1)]
501 The above matrix also represents the sequential or progressive read-out mode of
502 the (n, m) Bayer color filter array used in many CCD or CMOS image sensors.
504 The Huffman compressed video frame consists of a bitstream that encodes for
505 every R, G, or B pixel the difference between the value of the pixel itself and
506 some reference pixel value. Pixels are organised in the Bayer pattern and the
507 Bayer sub-pixels are tracked individually and alternatingly. For example, in
508 the first line values for the B and G1 pixels are alternatingly encoded, while
509 in the second line values for the G2 and R pixels are alternatingly encoded.
511 The pixel reference value is calculated as follows:
512 - the 4 top left pixels are encoded in raw uncompressed 8-bit format;
513 - the value in the top two rows is the value of the pixel left of the current
515 - the value in the left column is the value of the pixel above the current
517 - for all other pixels, the reference value is the average of the value of the
518 pixel on the left and the value of the pixel above the current pixel;
519 - there is one code in the bitstream that specifies the value of a pixel
520 directly (in 4-bit resolution);
521 - pixel values need to be clamped inside the range [0..255] for proper
524 The algorithm purely describes the conversion from compressed Bayer code used
525 in the SN9C101, SN9C102 and SN9C103 chips to uncompressed Bayer. Additional
526 steps are required to convert this to a color image (i.e. a color interpolation
529 The following Huffman codes have been found:
530 0: +0 (relative to reference pixel value)
533 1110xxxx: set absolute value to xxxx.0000
538 110001: ??? - these codes are apparently not used
540 [1] The Huffman compression algorithm has been reverse-engineered and
541 documented by Bertrik Sikken.
544 12. Contact information
545 =======================
546 The author may be contacted by e-mail at <luca.risolia@studio.unibo.it>.
548 GPG/PGP encrypted e-mail's are accepted. The GPG key ID of the author is
549 'FCE635A4'; the public 1024-bit key should be available at any keyserver;
550 the fingerprint is: '88E8 F32F 7244 68BA 3958 5D40 99DA 5D2A FCE6 35A4'.
555 Many thanks to following persons for their contribute (listed in alphabetical
558 - Luca Capello for the donation of a webcam;
559 - Philippe Coval for having helped testing the PAS202BCA image sensor;
560 - Joao Rodrigo Fuzaro, Joao Limirio, Claudio Filho and Caio Begotti for the
561 donation of a webcam;
562 - Dennis Heitmann for the donation of a webcam;
563 - Jon Hollstrom for the donation of a webcam;
564 - Nick McGill for the donation of a webcam;
565 - Carlos Eduardo Medaglia Dyonisio, who added the support for the PAS202BCB
567 - Stefano Mozzi, who donated 45 EU;
568 - Andrew Pearce for the donation of a webcam;
569 - John Pullan for the donation of a webcam;
570 - Bertrik Sikken, who reverse-engineered and documented the Huffman compression
571 algorithm used in the SN9C101, SN9C102 and SN9C103 controllers and
572 implemented the first decoder;
573 - Mizuno Takafumi for the donation of a webcam;
574 - an "anonymous" donator (who didn't want his name to be revealed) for the
575 donation of a webcam.
576 - an anonymous donator for the donation of four webcams.