New t3i or new lens

[Colin Rowe]

Originally Posted by Steve Bleasdale

Cheers guys, thanks for your input, there are a lot of people now saying that very same thing... A reduction in a little quality would mean a reduction in a liittle buisness? Got to be safe, thanks guys

Probably not Steve. if you are delivering HD, possibly, but not enough to worry about. If you are delivering SD DVD then nobody is going to know the difference. And the 3x is a very handy tool

[James Donnelly]

Originally Posted by Daniel Browning

Most photography and video with any camera made in the last 8 years or so is photon-shot-noise limited. So the amount of noise depends solely and entirely on the total number of photons detected by the image sensor. Modern sensors show the same quantum efficiency (QE, or percentage of light detected) between a huge variety of pixels sizes: from 80 square microns all the way to 3 sq. um. Many would have thought this was not possible because a relatively larger amount of area is lost to the metal between the photodiodes, but it turns out that gapless microlenses and other innovations have overcome that entirely. So whether you collect 50000 photons in 2 million large pixels or 5000 photons in 20 million small pixels, the total amount of light detected (and therefore the noise power from photon shot noise at any given spatial frequency) remains the same.
.

That is very insightful, so thanks for sharing your knowledge on this. I appreciate the time, and I know your main interest is in getting the truth out there.

I should have made it clear that my issue is not really the initial discussion about whether to buy a 70-200mm L lens in preference to a t3i, but what specific fidelity losses occur when shooting at 3x crop vs non cropped shooting only on the t3i itself. I am considering buying this camera, and find it strange that this argument is not settled by a proper full test, rather than postulations and hypotheses.

Since it's clear you know a lot about this subject, once again, I would love to get your thoughtful response on the main issue I, and many others I'm sure, are interested in here. I don't see you refuting a specific assertion made by myself and others (and in other threads) as follows.

As far as I know, no specification has been released by Canon about how the 1080p image is sampled from the full sensor. Studies have been made of the aliasing present which suggest vertical line skipping and a certain amount of binning/averaging in the horizontal, but not enough to eliminate horizontal aliasing. If proper binning was being used over the whole sensor, I think we could all subscribe to your statements about noise.

However, most real life accounts (some by real professionals), and certainly what I have pixel peeped from Vimeo, suggest that while there is no doubt a fidelity fall off in 3x crop, it is not as substantial as your claims would indicate.

Something else not mentioned before is that the 17-55, and other shorter lenses, will typically give you a better minimum focus distance than longer ones, making the crop mode attractive for macro videography.

[Daniel Browning]

Originally Posted by James Donnelly

I am considering buying this camera, and find it strange that this argument is not settled by a proper full test, rather than postulations and hypotheses.

I agree on the value of proper full tests.

Originally Posted by James Donnelly

Something else not mentioned before is that the 17-55, and other shorter lenses, will typically give you a better minimum focus distance than longer ones, making the crop mode attractive for macro videography.

That may be.

[Kin Lau]

Originally Posted by Daniel Browning

I disagree. The primary difference is that video does more of the processing in-camera, but other than that the comparison is quite similar.

I don't it can really be summarized as just "about 1920x1080 pixels worth of info" -- it's too complicated for that. In any case, but as I mentioned in the last post, the losses from poor sampling and resampling are separate from and in addition to the effects from cropping.

You're resampling in software, from a 21mp image. We _know_ that's not what happens, if that did happen, then we would not see line-skipping.

You're basically pixel-binning when you resample from 21mp - which also reduces noise, making this test invalid for video.

I have never seen any tests that show that Canon resamples at all in video mode, all the artifacts are consistent with line-skipping.