Home

Quickcam Web driver for Linux

This page exists to summarize the results of my reverse engineering of the Logitech Quickcam Web. I have my own variant quickcam-web.tar.gz of Georg's quickcam driver which is working with my cam in both compressed and uncompressed modes. The decompression code is included. It takes some MMX optimized routines from libmpeg2 which are also included. Note: You may need to choose between MMXEXT or MMX by changing the value of vo_mm_accel at the end of decode-lvc.c. Set it to 0 for no MMX.

Cristiano de Michele has added the necessary patches to the qce-ga driver, to get full video4linux support. His quickcam web driver based on qce-ga works with minimal changes to the official qce-ga driver and will hopefully be included soon. It doesn't support compressed mode, though. This is one reason why it is so much slower.
Quickcam Web (as seen by its twin brother)

Links to other sites

Most of the remaining text is probably black magic to you. But if not, read on :)

Components

The Logitech Quickcam Web consists of a STV0610 for USB communication and the imaging chip VV6410. The VV6410 has good public documentation, but the STV0610 seems to be kept secret.

STV0610 ASIC

The STV0610 works similar to the STV0600, but has several more registers. Here is a complete register dump taken just after plugging in the cam. Spaces mark that the registers are not giving out values. The dump was produced by qcdump which is included in the tar.gz package I mentioned above.

Range 0x400 - 0x600

 400:   34 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00
 410:
 420:   a2 07 06 00 00                                  
 540:   00 04 00 00 00 00 00 00  00 04 00 00 00 00 00 00
 550:   00 04                                           
 560:   40 00 80 80                                     

This seems to be some registers shadowed from 0x1xxx. They always contain the same value; changing one changes the other, too. The Logitech driver writes to these register set and writes a "1" to register 0x1704 afterwards. For explanation of the contents see the 0x1xxx area.

Range 0x1400-0x1424

1400:   34 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00
1410:
1420:   a2 07 06 00 00    ; I2C_ADDR, I2C_LENGTH I2C_RW, REG23, I2C_TXCOUNT

This is the I2C area. The registers 0x1400-0x140f contain the i2c register numbers, the register 0x1410-0x141f their values.
Register 0x1420 contains the address of the I2C partner, 0x1421 the number of valid register/value pairs minus 1. Register 0x1422 is 1 for write, 3 for read commands. The registers 0x1400-0x1422 are normally written in one big bunch for an I2C command.
The purpose of 0x1423 is still unknown: sometimes it is set to 4 sometimes to 5, I don't know.
The register 0x1424 gives the number of succesfully transmitted i2c register values.
When reading two I2C register the results are written to 0x1410 and 0x1412, instead of 0x1410-0x1411. Don't know why.

Range 0x1440-0x1424

1440:   00 00 12 00 00 00 ; ISO_ENABLE, ?,?, SCAN_RATE, ?, LED

0x1440 turn ISO stream (image data stream) on(1) or off(0)
0x1443 SCAN_RATE (?)
0x1445 turn LED on(1) or off(0)

Range 0x1500-0x15ff

1500:   00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00

1520:   00 00 00 00 00 00 00 00  00 00 00               
1530:   00 00 00 00 00 00 00 00  00 00 00 00            
1540:   00 04 00 00 00 00 00 00  00 04 00 00 00 00 00 00
1550:   00 04 00 00 00 00 00 00  00                     
1560:   40 00 80 80 00 ff 0a 01  00 00 00 00            

1580:   00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00
1590:   00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00
15a0:   00 00 00 00 00 00 00                            

15c0:   00 ff 03 00 84 00 00 00  0f 00                  

0x1580 - 0x15a6 Various compression quality parameters?
According to Georg Acher:
0x15c1/0x15c2 Max. ISO packet size L/H
0x15c3 Y-Control, 1: 288 lines, 2: 144 lines
0x1680?? X-Control, 0xa: 352 columns, 6: 176 columns
I don't know anything about the remaining registers. I may fill in more later.

Range 0x1600-0x16ff

1600:   41 3f 00 00 00 00 00 00  00 3f 2a 7f 0a 00 01 02
1610:   00 00 00 c9    00 00 40  1f 00 40 1f 00 80 bb 00
1620:   00 00 00                                        
1630:   00 01                                           

I don't know anything about these registers.

Registers 1701, 1704, b000

            1        4
1700:      00       00                                  

0x1704 is written when a 0x400-0x5ff register is accessed.
b000:   00                                              

????

Registers 0xe000-0xefff

e000:   12 01 00 01 ff ff ff 08  6d 04 50 08 00 01 00 01
e010:   00 01 09 02 b6 00 03 01  00 80 32 09 04 00 00 02
e020:   ff ff ff 00 07 05 81 01  00 00 01 07 05 82 03 01
e030:   00 10 09 04 00 01 02 ff  ff ff 00 07 05 81 01 ff
e040:   03 01 07 05 82 03 01 00  10 09 04 01 00 00 01 01
e050:   00 00 09 24 01 00 01 27  00 01 02 0c 24 02 01 01
e060:   02 00 01 00 00 00 00 09  24 06 02 01 02 43 00 00
e070:   09 24 03 03 01 01 00 02  00 09 04 02 00 01 01 02
e080:   00 00 09 05 83 01 00 00  01 00 00 09 04 02 01 01
e090:   01 02 00 00 07 24 01 03  01 01 00 1d 24 02 01 01
e0a0:   02 10 07 40 1f 00 11 28  00 80 3e 00 22 56 00 00
e0b0:   7d 00 44 ac 00 80 bb 00  09 05 83 01 64 00 01 00
e0c0:   00 07 25 01 01 00 00 00  04 03 09 04 0e 03 43 00           ......C.
e0d0:   61 00 6d 00 65 00 72 00  61 00 00 00 00 00 00 00  a.m.e.r. a.
e0e0:   00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00
....
e1f0:   00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00

Registers 0xe000-0xe0c8 is the USB info block. This contains the same as /proc/bus/usb/001/*. This 512 bytes repeats over and over up to 0xefff. I haven't tested whether you can write to this space.

Registers 0xf000-0xffff

f000:   00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00
....
fff0:   00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00

As you can see they are completely zero. They aren't accessed by the Logitech driver.

Compression

The Logitech Quickcam Web can deliver their images in compressed format. The image format is a kind of JPEG, but without headers describing the encoding. Instead the encoding is fixed. Every block starts with a bit selecting the quantizer scale and 10 bits representing the zeroth element of the DCT matrix. After this the huffman codes follow. Here is the huffman table. The last element is marked by the huffman code 0110. Before every fourth block (i.e. also before the first) there is a four bit number that has to do with the quantizer scale, but it is always 7 in my experiments. The file decode-lvc.c contains the code interpreting the huffman blocks and sending the result through an Inverse DCT, do a YUV transformation and return the RGB pixels.
Send mail to hoenicke at informatik dot uni-oldenburg dot de. This page is reachable via http://www.hoenicke.de.vu/quickcam

Last updated on 32-Jan-2002