Globo 4 - Digital camera

Principle

I'm working on this!

The camara electronics consist of three blocks:

The optical sensor
RAM storage
Alaris ASIC

The Optical sensor is, I suspect, an Omnivision OV6620 or a very similar device (Some of the arguments are: it has 48 connections, the resolution almost corresponds (356x292 instead of 320x240) and, most importantly, most of the commands I could detect, correspond to this sensor).

RAM: This is - once again - mostly guesswork. Y (mostly green) color information is stored into one of the RAMs, UV (red and blue difference signals) are stored in the other RAM.

The QV108-02 ASIC is a custom chip, developed by (or for) Alaris, which glues things together. It has the logic inside to attend to the parallel port in Compatibility Mode or in ECP mode, including the IEEE1284 identification and transaction negotiation. It also has the capability of generating the I2C control signals for the optical sensor chip, to control its intensity, brightnes and many other parameters. And lastly, it generates the complex control signals for the RAM to generate (at least) three compression modes: YUV422, YVU91 and VGPixel.

Here is an argument against the sensor being an OV6620: The Omnivision chip can handle 8 bits transfers too, needing only one RAM. There must be a reason why the designers decided to include a second memory chip, with the extra cost, and power consumption.

Optical sensor

Have a good look at how the pixels are obtained from the sensor chip. The spec says it has a resolution of 356 pixels horizontally, and 292 vertically. This is not entirely true - particularly in color (and b/w mode isn't accesible it seems). Of each square of 4 pixels, two are designated to capture green light, one red, and one blue.

This does compensate for the eye's sensitivity: it's about twice as sensitive for green as for blue/red. The output of the sensor chip can be either directly these R, G, and B values, or the chip itself can combine them into Y, U and V values, even with different coefficients to cover several standards.

So, the actual resolution is less than the 356 specified, but we cannot actually half it (to 178), as there is actually different intensity information available for each pixel. Later on, particularly after sending it using the (best) of the compressed norms, we will actually take the mean of the color for the four pixels. (One of the other modes, actually reduces the color information 4 more times!)

20744

mail