valoos - 07 January 2010 10:29 AM
So, the interpolation process goes on the camera side? I was sure that fps limit of this stuff is only from USB bandwidth limit. (from this reason was born my idea of cuting data, for example drop the color information)
So, the only one way to push more fps from it is cool down the sensor ... :D
So as you said, since camera CMOS sensor pixels are already in the bayer configuration, there is no way to drop the colors (other than skipping the pixels, thus decimating the image).
The CMOS sensor, as any other semiconductor, has the max operating speed rating and I guess cooling it down will probably help with noisy image at higher framerates, but that still does not solve the max running clock issue. Basically we are talking here about overclocking the CMOS sensor chip. I have already done this with 75fps and as you can see the image is usable but there are some artifacts already visible in there. Pushing it higher than this is possible (I experimented with this in my early driver development), but it introduces even more artifacts and makes the image pretty much unusable. Also, as the exposure time gets shorter you will need more incoming light to the sensor to get the normal brightness image. If you try to compensate for this by adjusting the gain value, this will add even more noise to the resulting image. So it is a trade off.
I suggest you get our CL-Eye SDK and test this with the supplied API. I haven’t tried cooling the CMOS sensor and running it at 75fps, but I’m very interested to hear about the results.
If you are willing to try this out, please post your findings here. We might be able to push even higher framerates (>75) out of the sensor if this works.
AlexP