Do you think there will be lower diffraction limit with the Canon 5D2?

[NigelC]

Not sure I'm ready to exchange my 5D for the new model, even though there are some useful benefits, leaving aside the increase in resolution. Based on what I've seen of the 40D to 50D progression, I would expect it to be at least no worse on the high ISO noise front. But could there be a downside from the extra resolution, even assuming I can make use of it sufficiently often? Is it possible that the useable F-stop range will be lower than the original 5D, which is particularly good in this respect. Don't see any specific mention of this in most DSLR reviews. Interesting to look at the DPreview of the Sony A900, which felt the IQ of the 5D still stands up well, even some aspects of operation have fallen behind, especially, for me the LCD.

[Boris_Epix]

Mate: I really hated the 1Ds MK2. I got a 5D as a backup and liked it more. Resolution seemed to be pretty much comparable. What really annoyed me was that the 1Ds MK2 was making dull pictures. Eyes had no good reflections and looked mushy. 5D was not much different in this aspect but it seemed to have more pop and I could work the file more like I wanted it too. But still I didn't like the colors much. It often gave that artificial look.

Now with the 1Ds MK3 I really feel like the 14 bits compared to 12 bits makes a huge difference when retouching and making bigger adjustments. And suddenly the eyes also have better reflections (not on Nikon level but nice). For closeups and portraits I feel the eyes are most important. Resolutionwise the 1Ds MK3 is really a giant leap from the 5D. You can count nose-hairs and pores.

If the 5D Mark2 is on a similar level I think you can't make much wrong with it. But this really is about the magic bullet. Everyone focusses on a different aspect of a camera. That's the hunting for the magic bullet thing.

If high ISO is really what floats your boat and you don't sound like you will be needing the resolution often then possibly the Nikon D700 is a better choice.

But I've heard that the 5D Mark2 is lightyears above what the 1Ds MK3 delivers which is lightyears above what the original 1Ds (11 Megapixels) could deliver.

Sorry if I didn't make much sense... I wasn't trying to. Just random thoughts

[ejmartin]

Is it possible that the useable F-stop range will be lower than the original 5D, which is particularly good in this respect.

Why, do you think that a newer camera body will decrease the resolution of your lenses?

Diffraction is an optical property of lenses, independent of what sensor you put behind them.  Using higher resolution sensors provides, um, higher resolution.  That means in particular that it will resolve diffraction effects of the lens better too.

It's better to think of things in terms of the reverse of your question -- will there be a smaller range of f-stops over which the sensor is the limiting factor in resolution with the 5D2?  The answer to that question is yes.

[Panopeeper]

I would expect it to be at least no worse on the high ISO noise front

The 5DMkIII is better than the 5D by 0.3-0.5 EV @ ISO 1600 on a per pixel basis.

But could there be a downside from the extra resolution, even assuming I can make use of it sufficiently often? Is it possible that the useable F-stop range will be lower than the original 5D, which is particularly good in this respect

"Using the extra resolution" means for me presenting the result with the same resolution (ppi) as with the 5D, i.e. making larger prints, or cropping, etc. In that case the diffraction effect will be visible sooner. As Emil wrote: the higher resolution will resolve the diffraction effect too "better", i.e. it becomes more evident.

[Ray]

The 5DMkIII is better than the 5D by 0.3-0.5 EV @ ISO 1600 on a per pixel basis.

Gabor,
That's encouraging. My own tests comparing the 5D with the Nikon D3 implied a similar degree of difference in favour of the D3, both cameras having a similar pixel count.

Now, I know you don't like comparing equal size images, insisting that the reason you buy a camera with more pixels is to be able to produce larger prints. However, if for some strange reason one were to make prints as large as one's printer were capable of, irrespective of the pixel count of one's camera, would it be reasonable to suppose that the 5D MkII would then produce marginally less noise than the D3, at high ISO?

[Panopeeper]

if for some strange reason one were to make prints as large as one's printer were capable of, irrespective of the pixel count of one's camera, would it be reasonable to suppose that the 5D MkII would then produce marginally less noise than the D3, at high ISO?

Ray,

this is pure speculation. I can imagine that printing with an inkjet printer of for example 1440dpi at different pixel densities, for example 360dpi and 600dpi, the reduction can be ideal, much better than reducing the pixel count by any program (this is pure speculation on my part, but if I were to write the printer driver, this would be so).

Anyway, there is no such thing as reducing the noise by downresing an image; that is plain BS. The noise becomes less visible by eliminating details, which exhibit the noise. However, think of an important aspect of noise: dynamic range. Do you really believe, that the dynamic range can be increased by reducing the image size?

Anyway, here are two crops from a very noisy 50D image; NR is turned off (as much as it is possible in ACR), no sharpening, no nothing. One is 100%, the other is reduced to 75% linearly. What do you think of the noise reduction between them?

[Panopeeper]

Ray,

btw, if I could justify it, I would not hesitate buying a 5DMkII (AND the very best lenses to fully utilize the resolution; for example not that crappy 16-35mm MkII).

ADDED

I put the read noise values of some cameras in a chart, among others the 5D2. Unfortunately, the D3 does not write the masked pixels in the raw file and I don't have dark frames.

Read Noise Chart (The server is goofy)

[Daniel Browning]

Is it possible that the useable F-stop range will be lower than the original 5D?

No. The original 5D has more F-stops that are pixel-blur-limited. The 5D2 reduces that blur. The net result is that for any f/stop, the 5D2 will have more resolution, though diffraction will cause the the law of diminishing returns to take effect. There is very slight, but visible diffraction at f/5.6 in the 1Ds Mark III (and 5D Mark II, by extension), but even at f/22 it still has more detail than the 5D. It would never have less detail, of course.

[marcmccalmont]

[quote name='Panopeeper' date='Nov 11 2008, 04:29 PM' post='236281']
The 5DMkIII is better than the 5D by 0.3-0.5 EV @ ISO 1600 on a per pixel basis.

Could you explain the difference between per pixel noise/DR and overall sensor noise/DR?
In other words if pixels performance was the same would the higher resolution MKII have better overall noise/DR?
Marc

[Panopeeper]

Could you explain the difference between per pixel noise/DR and overall sensor noise/DR?
In other words if pixels performance was the same would the higher resolution MKII have better overall noise/DR?

First of all, caution with the term "per pixel". A single pixel does not have any noise. One can talk about noise only relating to a group of pixels (some area of the image).

Example in the attachment: a smooth (even though it does not look like smooth), evenly lit, unicolored spot is selected from a raw image and the noise is measured as the standard deviation.

a. In case of the 40D, the red channel's average values are at the 7.52nd EV of the dynamic range (from saturation), and the noise is 12% (compared to the normalized output value).

b. With the 50D (a completely different image, but that has nothing to do with it), the selection average is at the DR 7.63 EV, only slightly more in the shadows than the above example, but the noise is 15%.

If these samples are accepted, then they show that the 50D's noise level @ ISO 200 is somewhat higher than that of the 40D, around this depth of shadows.

This is the pixel level noise, here measured on the non-demosaiced raw data.

The sensor noise is IMO a muddy term. Who on earth cares for the "total sensor noise" or for any average of that?

As Ray and some others are understanding it (please correct me if I am wrong), they are trying to convert images of different sizes (i.e. of different pixel counts) into some common denominator and quantify the noise on the result. This approach ignores not only the problem of image deterrioration by resizing, but the fact, that reducing the image size simply hides the noise, while hiding the details as well.

So, if you want to use your 5DMkII as a 10 Mpix camera, then yes, you can evaluate the result by resizing the image and comparing it to a Nikon D3.

My point was, that reducing the image size and sweeping the noise under the carpet does not increase the dynamic range, as truely lower noise would do.

[ErikKaffehr]

Hi,

Resolution will be better on 5DII than on 5D. How much better depends on the lens. You will not find a single case whatever aperture or lens you use that will be less sharp from 5DII than from 5D.

On the other hand, to extract the inherent sharpness and image sharpness from the 5DII will take excellent lenses and technique, essentially the same as with the 1DsIII. Regarding noise I would guess that 5DII would be a little better than the 1DsIII, because there has probably been some development in keeping noise down since 1DsIII. My guess is that any ISO:s above 1600 will be fake.

There is no "diffraction limit",  IMHO, diffraction creeps in at small apertures. My guess (based on cameras with similar pixel pitch) is that loss of resolution will be measurable at f/11 and be significant beyond f/16, that is if you are shooting low ISO, use a decent tripod over solid ground, use mirror lockup and cable release and self timer.

Erik

Not sure I'm ready to exchange my 5D for the new model, even though there are some useful benefits, leaving aside the increase in resolution. Based on what I've seen of the 40D to 50D progression, I would expect it to be at least no worse on the high ISO noise front. But could there be a downside from the extra resolution, even assuming I can make use of it sufficiently often? Is it possible that the useable F-stop range will be lower than the original 5D, which is particularly good in this respect. Don't see any specific mention of this in most DSLR reviews. Interesting to look at the DPreview of the Sony A900, which felt the IQ of the 5D still stands up well, even some aspects of operation have fallen behind, especially, for me the LCD.

[ErikKaffehr]

Gabor,

Is the 16-35 really that bad? I understand that it is pretty ugly in the corners at 16mm, but what alternatives are around? I guess that most people still want automatic aperture and possibly even auto focus.
I'm sort of considering to buy an Sony Alpha 900, BTW, my choise for that camera would probably be a Sigma 12-24, trying to find a decent sample, but I'm not that much in for ultra wides.

Best regards
Erik

Ray,

btw, if I could justify it, I would not hesitate buying a 5DMkII (AND the very best lenses to fully utilize the resolution; for example not that crappy 16-35mm MkII).

ADDED

I put the read noise values of some cameras in a chart, among others the 5D2. Unfortunately, the D3 does not write the masked pixels in the raw file and I don't have dark frames.

Read Noise Chart (The server is goofy)

[marcmccalmont]

First of all, caution with the term "per pixel". A single pixel does not have any noise. One can talk about noise only relating to a group of pixels (some area of the image).

Example in the attachment: a smooth (even though it does not look like smooth), evenly lit, unicolored spot is selected from a raw image and the noise is measured as the standard deviation.

a. In case of the 40D, the red channel's average values are at the 7.52nd EV of the dynamic range (from saturation), and the noise is 12% (compared to the normalized output value).

b. With the 50D (a completely different image, but that has nothing to do with it), the selection average is at the DR 7.63 EV, only slightly more in the shadows than the above example, but the noise is 15%.

If these samples are accepted, then they show that the 50D's noise level @ ISO 200 is somewhat higher than that of the 40D, around this depth of shadows.

This is the pixel level noise, here measured on the non-demosaiced raw data.

The sensor noise is IMO a muddy term. Who on earth cares for the "total sensor noise" or for any average of that?

As Ray and some others are understanding it (please correct me if I am wrong), they are trying to convert images of different sizes (i.e. of different pixel counts) into some common denominator and quantify the noise on the result. This approach ignores not only the problem of image deterrioration by resizing, but the fact, that reducing the image size simply hides the noise, while hiding the details as well.

So, if you want to use your 5DMkII as a 10 Mpix camera, then yes, you can evaluate the result by resizing the image and comparing it to a Nikon D3.

My point was, that reducing the image size and sweeping the noise under the carpet does not increase the dynamic range, as truely lower noise would do.

So one should not hold out hope that a MKII would capture a wider Dynamic Range scene than the original 5D even though there are more/higher usable ISO settings? Just thinking out loud.
Marc

[Panopeeper]

Is the 16-35 really that bad? I understand that it is pretty ugly in the corners at 16mm

A disclaimer is required here: I tested ONE copy of it.

I ordered a 16-35mm f/2.8L MkII and and EF-S 17-55mm f/2.8 IS at the same time (with the agreement, that I will send back at least one of them). I tested them on a 40D, i.e. cropping; thus I can not say anything to the corners.

Flaring was bad, CA was bad, sharpness was not bad but not better than the 17-55mm or the Tamron 28-75mm (I still have that excellent copy, though the 15-55mm is already welded to the 40D). All that for a horrendeous price.

If I bought the 5DMkII (I don't plan to, so this is really hot air), I would try for example the new Zeiss 21mm. I mean, if I get one of the best cameras around, then I want to have one ofthe best lenses as well. The Nikon 17-35mm has an excellent reputation, with an adapter.

[Panopeeper]

So one should not hold out hope that a MKII would capture a wider Dynamic Range scene than the original 5D even though there are more/higher usable ISO settings? Just thinking out loud.

Hold on; I have not posted anything like that. I only disputed the comparion based on "image noise". I posted above:

The 5DMkIII is better than the 5D by 0.3-0.5 EV @ ISO 1600 on a per pixel basis

The per pixel basis is the "real thing", that translates to captured highlights and "pullable" shadows, i.e. the DR of the 5DMkII is greater that that of the 5D, at least @ ISO 1600.

I don't have raw images by the 5DMkIIunder ISO 1600 suitable for measurement (the requirement of containing smooth, unicolored spots is not so onerous, but "evenly lit" is a real problem). Thus I can niot say anything re lower ISOs, except that the read noise of the MkII - measured on the masked pixels - is much lower than that of the 5D at lower ISOs as well; that is an indication, that the noise will be lower. My estimation is, that the 5DMkII will have a DR about 1/3 EV lower than the D3 - and that would be GREAT!

Note, that at least 1 EV of the DR remain unused in many cases due to underexposure because of misinterpreting the in-camera indicators (in cleartext: due to less than ideal exposure). Thus 1/3 EV is nothing. If one can do perfect ETTR, that means much more than 1/3 EV difference in the DR. Though this does not work for example in sports shooting, the 5D/5DMkII are not for that anyway.

[PhotoTurbo]

Absolutely.  A sensor with higher pixel desity doesn't increase the diffraction of any given lens, it just does a better job of showing it when you're pixel-peeping at 100%.

100% on an image from a 21MP sensor is much more magnified than 100% on an image from from a 13MP sensor.  If you "zoom out" on the 21MP image to the same magnification as the 13MP image, the images will basically look the same.