Sharpness: 5Ds vs. 5Ds R -- with amazing glass?

Started 2 months ago | Questions thread
MitchAlsup Veteran Member • Posts: 5,048
Re: 5Dsr for the win... provided have a plan for any moire (also available on 5Ds)

Dr_Jon wrote:

MitchAlsup wrote:

icor1031 wrote:

MitchAlsup wrote:

icor1031 wrote:

MitchAlsup wrote:

icor1031 wrote:

What's going on with this?

The sensor can ONLY capture the image presented by the lens. If the lens fails to produce dots as small as the pixels on the sensor, the the sensor simply captures the slight blur produced by the lens. That is, this is not a limitation of the sensor (especially when the lens is wide open) but a limitation of the optics in the lens and their ability to put up an image on the sensor with detail at the limit of the pixels on the sensor.

... I know that. But this is an 85mm Otus, and DXOMARK lists it as capturing almost twice as much data.

Are you suggesting that lens rental shot the image very stopped, perhaps at f/11 or such?

https://www.dxomark.com/Lenses/Compare/Side-by-side/Zeiss-Carl-Zeiss-Apo-Planar-T-Star-Otus-85mm-F14-ZE-Canon-on-Canon-EOS-5DS-R-versus-Zeiss-Carl-Zeiss-Distagon-T-STAR-Otus-55mm-F14-ZE-Canon-on-Canon-EOS-5DS-R__1383_1009_1241_1009

No, not al all.

twice as much data as what?

If you could find an MTF graph of the Otus maybe I could point out where the discrepancy lies.

https://www.lensrentals.com/blog/2014/10/quick-zeiss-85mm-otus-mtf-charts/

The MTF graph has Tangential and Sagittal directions. The tangential direction is the direction around the circle at constant radius, while the sagittal direction is along the radial axis.

The anti-aliasing filter in "normal" dSLR cameras is there to take an image with the resolution of a single pixel, and blur it out to a 2×2 pixel blur. This limits the moiré by removing the high frequency resolution components to below the Nyquist limit for sampling.And lens that produces a larger blur spot will (essentially) have whatever size blur spot the lens produces have another pixel of width and a pixel of height added to the original blur spot.

Although the AA filters only add an amount of blur, if you blurred it enough to lose all the aliasing/moire you'd lose too much detail. So for example:
https://www.dpreview.com/reviews/image-comparison?attr18=daylight&attr13_0=canon_eos5ds&attr13_1=canon_eos5dsr&attr13_2=canon_eos5ds&attr13_3=canon_eos5dsr&attr15_0=raw&attr15_1=raw&attr15_2=raw&attr15_3=raw&attr16_0=100&attr16_1=100&attr16_2=800&attr16_3=800&normalization=full&widget=1&x=-0.5924104672306111&y=0.3590082644628099

As I mentioned above (err, or below) See Figure 7 here:
http://www.strollswithmydog.com/resolution-model-digital-cameras-aa/#more-1723
Is good at showing the difference in blurring between the s/sr.

I do not think that anything I wrote here disagrees with the information in the links you posted.

What we see in the MTF graph is that in the center of the image, the lens puts up a round "blur spot" that is about 2.5-3.0 pixels wide on a 5Ds[R]--this corresponds that the 70% contrast transfer at the 0mm off axis on the horizontal axis. This is close to the Nyquist limit--which enables moiré to slip in.

As we proceed towards the 1/2 way point to the edge, the blur spot slowly increases to 3.5-4.0 pixels in both directions--this corresponds to the 50% contrat transfer at the 12mm off axis position. Since both the tangential and sagittal lines remain close the blur spot remains essentially round.

As we proceed towards the edge from the 1/2 way point, the tangential blur spot increases rather rapidly (heading towards 20% contrast transfer at the edge); while the sagittal contrast transfer remains at about 3.5-4.0 pixels wide. This implies an elongated blur spot, longer in the tangential direction while remaining tight in the sagittal direction. I would estimate that the blur spot at the edge of the field is about {3.5-4.0 pixels sagittal × 8-9 pixels tangential}. This means one is still seeing high resolution for those things radially aligned with the center of the frame, while there becomes some significant blurring of those things aligned tangentially to the center of the frame.

In the center of the image, this lens has no peers (currently) from about 1/2 way out and towards the edge, the Sigma ART F/1.4 is competitive in the MTF sense. Having used neither, I can't say about the color.

Quite non-usefully I would like to say the Art 50/1.4 is very nice for the out-of-focus stuff...

The kind of increase in the size of the blur spot is small enough that simple post processing sharpening should produce fabulous images. Go softly with a deft hand as trying to sharpen the tangential direction too much will over sharpen the sagittal direction, leading to artifacts.

If this were my choice, and my use of this lens and the 5Ds[R] was for landscapes I would go 5DsR accepting a few shots will have moiré. Most landscapes have opportunities for moiré, but moiré in landscapes seldom ruins and otherwise excellent image.

If I were a fashion photographer, I would be inclined to go 5Ds leaving a trifling of resolution on the table while avoiding (that messy) moiré and dealing with it in post processing.

I think I'd agree with you, as although I see aliasing effects I don't like only very very rarely on an sr the s does have an advantage. Also what raw processing tools you use matter, as some are miles better at automatically removing moire than others. Quite old now, but I think makes the point:
http://www.dpreview.com/forums/post/58034634

If the lens in question had lower MTF contrast transfer I would go for a 5DsR and hope the lens blur spot was big enough to avoid moiré and learn to live with it.

While this might be considered unflatteringly sharp (err, sorry Georgia, but a little bit of frequency separation and it'd be great, alas I need sharp for the example) I'm struggling to see anything in the hair that would really upset me...

5Dsr, Pixel level if click "original size"

This is one of those "it depends" things. I could probably pixel peep the image and find something that annoyed me, but the overall image is good enough to ignore those trifles.

-- hide signature --

Mitch

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