AA filter question - Joe?

Started Feb 9, 2012 | Discussions thread
bobn2 Forum Pro • Posts: 56,750
Re: The resolution limit is way out there...

Ray Ritchie wrote:

Joseph S Wisniewski wrote:

And, as the pixel pitches get tighter, it gets more and more practical to imagine moving the sensor in a small, square pattern to do the AA, instead of using liNbO3 slabs, so I'm betting the day of the AA filter you can switch off is drawing near. Picture a 60mp camera knowing that f8 has enough diffraction for complete anti-aliasing, f5.6 needs an additional 0.4 micron movement, f4 needs 0.7, on up to 3.3 at f1.4.

Another approach (other than moving the sensor) occurs to me, though it might not be as effective: suppose the sensor locations were perturbed from their nominal regular spacings in tiny amounts (maybe your 0.4 micron value would be about right - I haven't done the analysis) according to a pseudo-random pattern. This would be analogous to dithering of the sampling clock in the time domain, which is a known problem in digital communications. I believe the impact on the image spectrum would be to turn fine spectral lines (such as moire) into lower-level white noise - imagine "squashing" them down, but maintaining the same power levels. This would be a compromise, of course, as it would also slightly impact resolution, but it might be worth doing if it could eliminate the need for an AA filter. I haven't thought about whether/ how it might be switched on and off, though.

How would this be effected. The 'sensor locations' (by which you mean pixels, I guess) are etched into a rigid piece of silicon, so moving them individually means stretching or compressing the silicon between pixels. I can see how one might get a tiny movement with standing waves in the silicon, but get enough to have an effect and the chip will shatter, I would think. In any case, such movement would be regular, not random.

It might be the case that the 'preferred embodiment' of the shifting sensor idea would be to build piezo actuators into the sensor package itself, so only the chip moves, it's much lighter than the whole package and carrier pcb, so much easier to move. Quite a few sensors (the Panasonic ones, in particular) are installed using 'chip on board' technology, where the silicon chip of the sensor is glued directly to its carrier PCB. It might be that something can be done around that concept.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow