cpia2.rst 8.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193
  1. The cpia2 driver
  2. ================
  3. Authors: Peter Pregler <Peter_Pregler@email.com>,
  4. Scott J. Bertin <scottbertin@yahoo.com>, and
  5. Jarl Totland <Jarl.Totland@bdc.no> for the original cpia driver, which
  6. this one was modelled from.
  7. Introduction
  8. ------------
  9. This is a driver for STMicroelectronics's CPiA2 (second generation
  10. Colour Processor Interface ASIC) based cameras. This camera outputs an MJPEG
  11. stream at up to vga size. It implements the Video4Linux interface as much as
  12. possible. Since the V4L interface does not support compressed formats, only
  13. an mjpeg enabled application can be used with the camera. We have modified the
  14. gqcam application to view this stream.
  15. The driver is implemented as two kernel modules. The cpia2 module
  16. contains the camera functions and the V4L interface. The cpia2_usb module
  17. contains usb specific functions. The main reason for this was the size of the
  18. module was getting out of hand, so I separated them. It is not likely that
  19. there will be a parallel port version.
  20. Features
  21. --------
  22. - Supports cameras with the Vision stv6410 (CIF) and stv6500 (VGA) cmos
  23. sensors. I only have the vga sensor, so can't test the other.
  24. - Image formats: VGA, QVGA, CIF, QCIF, and a number of sizes in between.
  25. VGA and QVGA are the native image sizes for the VGA camera. CIF is done
  26. in the coprocessor by scaling QVGA. All other sizes are done by clipping.
  27. - Palette: YCrCb, compressed with MJPEG.
  28. - Some compression parameters are settable.
  29. - Sensor framerate is adjustable (up to 30 fps CIF, 15 fps VGA).
  30. - Adjust brightness, color, contrast while streaming.
  31. - Flicker control settable for 50 or 60 Hz mains frequency.
  32. Making and installing the stv672 driver modules
  33. -----------------------------------------------
  34. Requirements
  35. ~~~~~~~~~~~~
  36. Video4Linux must be either compiled into the kernel or
  37. available as a module. Video4Linux2 is automatically detected and made
  38. available at compile time.
  39. Setup
  40. ~~~~~
  41. Use 'modprobe cpia2' to load and 'modprobe -r cpia2' to unload. This
  42. may be done automatically by your distribution.
  43. Driver options
  44. ~~~~~~~~~~~~~~
  45. .. tabularcolumns:: |p{13ex}|L|
  46. ============== ========================================================
  47. Option Description
  48. ============== ========================================================
  49. video_nr video device to register (0=/dev/video0, etc)
  50. range -1 to 64. default is -1 (first available)
  51. If you have more than 1 camera, this MUST be -1.
  52. buffer_size Size for each frame buffer in bytes (default 68k)
  53. num_buffers Number of frame buffers (1-32, default 3)
  54. alternate USB Alternate (2-7, default 7)
  55. flicker_freq Frequency for flicker reduction(50 or 60, default 60)
  56. flicker_mode 0 to disable, or 1 to enable flicker reduction.
  57. (default 0). This is only effective if the camera
  58. uses a stv0672 coprocessor.
  59. ============== ========================================================
  60. Setting the options
  61. ~~~~~~~~~~~~~~~~~~~
  62. If you are using modules, edit /etc/modules.conf and add an options
  63. line like this:
  64. .. code-block:: none
  65. options cpia2 num_buffers=3 buffer_size=65535
  66. If the driver is compiled into the kernel, at boot time specify them
  67. like this:
  68. .. code-block:: none
  69. cpia2.num_buffers=3 cpia2.buffer_size=65535
  70. What buffer size should I use?
  71. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  72. The maximum image size depends on the alternate you choose, and the
  73. frame rate achieved by the camera. If the compression engine is able to
  74. keep up with the frame rate, the maximum image size is given by the table
  75. below.
  76. The compression engine starts out at maximum compression, and will
  77. increase image quality until it is close to the size in the table. As long
  78. as the compression engine can keep up with the frame rate, after a short time
  79. the images will all be about the size in the table, regardless of resolution.
  80. At low alternate settings, the compression engine may not be able to
  81. compress the image enough and will reduce the frame rate by producing larger
  82. images.
  83. The default of 68k should be good for most users. This will handle
  84. any alternate at frame rates down to 15fps. For lower frame rates, it may
  85. be necessary to increase the buffer size to avoid having frames dropped due
  86. to insufficient space.
  87. ========== ========== ======== =====
  88. Alternate bytes/ms 15fps 30fps
  89. ========== ========== ======== =====
  90. 2 128 8533 4267
  91. 3 384 25600 12800
  92. 4 640 42667 21333
  93. 5 768 51200 25600
  94. 6 896 59733 29867
  95. 7 1023 68200 34100
  96. ========== ========== ======== =====
  97. Table: Image size(bytes)
  98. How many buffers should I use?
  99. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  100. For normal streaming, 3 should give the best results. With only 2,
  101. it is possible for the camera to finish sending one image just after a
  102. program has started reading the other. If this happens, the driver must drop
  103. a frame. The exception to this is if you have a heavily loaded machine. In
  104. this case use 2 buffers. You are probably not reading at the full frame rate.
  105. If the camera can send multiple images before a read finishes, it could
  106. overwrite the third buffer before the read finishes, leading to a corrupt
  107. image. Single and double buffering have extra checks to avoid overwriting.
  108. Using the camera
  109. ~~~~~~~~~~~~~~~~
  110. We are providing a modified gqcam application to view the output. In
  111. order to avoid confusion, here it is called mview. There is also the qx5view
  112. program which can also control the lights on the qx5 microscope. MJPEG Tools
  113. (http://mjpeg.sourceforge.net) can also be used to record from the camera.
  114. Notes to developers
  115. ~~~~~~~~~~~~~~~~~~~
  116. - This is a driver version stripped of the 2.4 back compatibility
  117. and old MJPEG ioctl API. See cpia2.sf.net for 2.4 support.
  118. Programmer's overview of cpia2 driver
  119. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  120. Cpia2 is the second generation video coprocessor from VLSI Vision Ltd (now a
  121. division of ST Microelectronics). There are two versions. The first is the
  122. STV0672, which is capable of up to 30 frames per second (fps) in frame sizes
  123. up to CIF, and 15 fps for VGA frames. The STV0676 is an improved version,
  124. which can handle up to 30 fps VGA. Both coprocessors can be attached to two
  125. CMOS sensors - the vvl6410 CIF sensor and the vvl6500 VGA sensor. These will
  126. be referred to as the 410 and the 500 sensors, or the CIF and VGA sensors.
  127. The two chipsets operate almost identically. The core is an 8051 processor,
  128. running two different versions of firmware. The 672 runs the VP4 video
  129. processor code, the 676 runs VP5. There are a few differences in register
  130. mappings for the two chips. In these cases, the symbols defined in the
  131. header files are marked with VP4 or VP5 as part of the symbol name.
  132. The cameras appear externally as three sets of registers. Setting register
  133. values is the only way to control the camera. Some settings are
  134. interdependant, such as the sequence required to power up the camera. I will
  135. try to make note of all of these cases.
  136. The register sets are called blocks. Block 0 is the system block. This
  137. section is always powered on when the camera is plugged in. It contains
  138. registers that control housekeeping functions such as powering up the video
  139. processor. The video processor is the VP block. These registers control
  140. how the video from the sensor is processed. Examples are timing registers,
  141. user mode (vga, qvga), scaling, cropping, framerates, and so on. The last
  142. block is the video compressor (VC). The video stream sent from the camera is
  143. compressed as Motion JPEG (JPEGA). The VC controls all of the compression
  144. parameters. Looking at the file cpia2_registers.h, you can get a full view
  145. of these registers and the possible values for most of them.
  146. One or more registers can be set or read by sending a usb control message to
  147. the camera. There are three modes for this. Block mode requests a number
  148. of contiguous registers. Random mode reads or writes random registers with
  149. a tuple structure containing address/value pairs. The repeat mode is only
  150. used by VP4 to load a firmware patch. It contains a starting address and
  151. a sequence of bytes to be written into a gpio port.