The illuminator uses an 830nm laser diode. There is no modulation - output level is constant. Output power measured at the illuminator output is around 60mW (Using Coherent Lasercheck). The laser is temperature stabilised with a small peltier element mounted between the illuminator and the aluminium mounting plate. This element can both heat and cool the laser to maintain a constant temperature, presumably in order to stabilise the laser's output wavelength. on the sample tested, at room temperature, the laser is being slightly heated.
The depth sensing appears somewhat sensitive to the relative position of the illuminator and sensor. Even a gentle bend of the aluminium plate causes significant disturbance of the depth image.
Loosening the illuminator screws and moving the illuminator slightly has the following effects : rotation gradually narrows the visible width of the frame, down to about 10% - the image appearance is the same, the left & right edges just get progressively croppped to black. Horizontal panning alters just the depth values without changing the overall scene appearance. Vertical panning shifts the visible part (as limited by rotation) of the frame horizontally, again there is no change in appearance of the visible part of the scene.
There is a non-resettable 102°C thermal cutout attatched to the outside of the illuminator housing - this is connected in series with the main 12V supply to the kinect, so clearly a product-safety feature. This positioning suggests that it may be to detect fault conditions causing the laser to heat its enclosure, e.g. if the front window is damaged, or possibly a peltier-induced meltdown distorting the housing and causing laser light to escape other than through the pattern optics. The raw laser power is definitely eye-hazardous if not spread out by the pattern-generating optics. Apart from the laser diode, the illuminator is likely to contain a temperature sensor and a photodiode for laser output power feedback - the latter may be integrated in the laser diode can.
Depth Image sensor
The depth sensor uses a monochrome image sensor. Looking at the signals from the sensor, resolution appears to be 1200x960 pixels at a framerate of 30Hz. The camera's I2C control interface perfoms a one byte transaction every frame - this could be something like gain setting or average level sensing. No change in this data was seen during random hand-waving in front of the sensor. The camera has an IR-pass filter at the laser wavelength - tests with various light sources show minimal sensitivity to visible and 950nm sources.
This chip is for audio processing, and has nothing to do with the depth system. Markings on chip are 88AP1-BJD2 P2G2750A/2 1024 AOP CX08 88AP102. Possibly one of Marvell's ARMADA series ARM chips, which has members with clock rates from 400MHz to 1GHz. External clock is 26MHz. Connected to the chip are a Winbond 25Q16B 16mbit, quad SPI flash, a custom-marked 8 pin device marked H102338 XBOX1001 X851716-006 GEPP , and a 512mbit DDR2 SDRAM. The SPI flash contains the firmware, which apperas to resemble 32 bit ARM code. The Marvell is also responsible for cryptographic authentication of the Kinect to the Xbox (to prevent clones). It would make sense for the custom 8-pin device to be the authentication chip containing the RSA key and certificate, similar to what Apple does with "Made for iPod" devices with a custom auth chip.
The fan only comes on when the internal temperature exceeds about 70°C, measured by thermistor "RT3" near the fan connector on the camera board. Once on it stays on for at least a few tens of minutes (maybe until powerdown). The fan is powered by the 5V supply from the USB connector.
- RGB Camera - MT9v112 Datasheet, Flyer
- Depth Camera - MT9M001 Datasheet, Flyer
- USB HUB uPD720114 Product page
- USB Audio interface TAS1020B 
- Accelerometer KXSD9 
Standalone operation of image board
All the image and depth sensor functionality is on the front PCB, and it should be possible to operate this without the rear board, using an external power supply to provide the +5V, 3.3V and +1.8V rails.
It has been confirmed that connecting the USB pins directly to a PC USB port works, with only the power connections from the rear PCB remaining connected.
Unfortunately the peak current draw puts it slightly above what can be 'legally' drawn from a single USB connector, due to the current draw of the Peltier element. Total power = 10mA @5V, 200mA @1.8v, 900mA @3.3V, = 3.4W, which would pull about 750mA at 5V assuming 90% converter efficiency and a 5V supply from the USB port. In practice cable voltage drops on the USB supply would mean near 800-850mA.
It may be feasible to limit this, e.g. by adding a series resistor to the peltier to limit its maximum power draw. It may be possible to operate without the peltier - this is something still to be investigated.
Image board connector signals
Connector on PCB appears to be Tyco FP series Mating connector Tyco 4-174639-4 It appears that part 4-175638-4 is the same connector supplied on tape/reel, with a lower min order qty of 1500 vs. 10K for the 4-174639-4 part
Pins 1,3,5 : +3.3V 500-900mA depending on Peltier temperature
Pins 2,4,6,9,12 : Ground
Pin 7 : +1.8v, 200mA
Pin 8 : USB DP
Pin 10 : USB DM
Pin 11 : +5V 10mA directly from USB connector. The fan, when on, is also powered by this supply.
Pin 13 : Input to sensor board, pulled weakly to +3.3v. Rear board pulls low until USB connection established. Can be left unconnected.
Pin 14 : Output from sensor board - pulses high for about 70ms after pin 13 rising edge. Can be left unconnected.