D800 AA filter

Started Jan 19, 2012 | Discussions thread
bobn2 Forum Pro • Posts: 59,092
Re: D800 AA filter

apaflo wrote:

bobn2 wrote:

apaflo wrote:

For this application, if all other characteristics were equal, the plastic filter would produce significantly better results in many, if not most, circumstances. Of course all else is not equal, but the red curve is better in this application.

For aliasing, yes, the extra love lobe (oops, typo) of the sinc function which is the MTF of the birefringent filter obviously lets through more energy above Nyquist, and so more aliasing. However, the extra droop below Nyquist is significant, about two grades of lens quality equivalent.

For aliasing, yes... and that is what the discussion is about. Your last sentence of course has no meaning.

I would claim it has meaning (otherwise I wouldn't have written it). The point I was trying to get across is that the relative droop in contrast of the plastic filter versus the quartz one is about the same droop as caused by going a couple of grades down in price for the lenses. I think the reason that camera companies go for the much more expensive birefringent filter, as opposed the the cheaper (and according to you better) plastic filter is that making your top grade pro lenses perform like your competitors kit lenses is a hard sell, even if the aliasing artifacts are a bit worse.

Whatever the curve is below the Nyquist Limit (we are presuming 80lp/mm) can be reshaped with a software filter applied to the digital data.

But not without consequence. The SNR has been reduced at those frequencies. The amplitude response can be restored (to a point) but the lost SNR cannot be recovered. This is exactly what we do, apply a bit of sharpening to correct for the AA filter, at the cost of some noise.

It is essentially "without consequence" simply because the potential loss of SNR due to use of a high pass filter is less than the potential loss of SNR due to the aliasing.

it is still with consequence. The right comparison is the consequence of a low sample rate, AA filter cut and sharpening versus a higher sampling rate, without less cut at the same frequencies, less sharpening and better SNR. The solution to aliasing is raise the sample rate, not cut the AA filter.

Clearly which is actually more significant for any given image depends on the specific image.

There will always be detail at higher frequencies. The level will depend on the scene of course, and on the lens, and finally on the AA filter and the resolution of the sensor.

There will be no detail above the diffraction cutoff. Sub-diffraction sampling is the aim for a properly sampled camera system.

That's redundant to what was stated. The level of detail depends on a series of characteristics, including of course the lens and it's aperture.

diffraction is an absolute and depends only on the aperture. The concept of sub-diffraction sampling entails sampling beyond the diffraction cutoff for any aperture available on the system.

Sampling at above the Diffraction Limited Aperture rate is clearly common and beneficial. Sampling at above the Diffracton Cuttoff Frequency has no benefit as such, but it also is not necessarily detrimental (particularly if the lens has a variable aperture).

'at' is good enough, 'above' a bit easier to achieve, since the target is larger.

I'm not sure what you meant by "sub-diffraction sampling".

sampling such that the Nyquist frequency is greater than the diffraction cut-off point. I suppose it should be 'super diffraction', but 'sub-diffraction' is the term that has become common in the literature (it just means looking at it in the spatial rather than spatial frequency domain)

If you meant at a rate higher than the Diffraction Cuttoff Frequency, I agree. But even at 36MP on a full frame sensor, a D800 is not going to approach that with common lenses.

36MP is peanuts.

Producing sharp crispy artifacts can be avoided with appropriate balance of characteristics, though it won't always be perfect for every scene.

Agreed, what we are doing nowadays is inherently undersampled, and will be until we get to sub-diffraction sampled systems. Let's hope Eric Fossum's jots research bears fruit.

Young people might live that long...

That's unduly pessimistic. Pixels get that small, and a lot of the rules change, which makes things easier in some ways (and harder in others). Very small pixels don't have to be very good pixels to produce good results. I doubt whether Eric would be pursuing it if he thought it a no-hoper.

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