So your sensor outresolves your lens

[Roger Calixto]

I've noticed my 7D outresolves my sigma 120-400. I got much "sharper" images on my 30D. That brings me to the question:


How has the increased pixel density changed your workflow or evaluation of images? Or has it not even mattered?



I used to evaluate my images at 100% and choose the sharpest shot of a group for starters. I remember reading about senors outresolving lenses from one of Micheal's essays where he talks about how big an image needs to be for different printing sizes but he never got into how this would affect our day to day. From what I've read via Google on this topic, "pixel peeping" just isn't the way to go. Any comments?

[Jeremy Payne]

I don't really understand ... can you post some examples showing this phenomenon?

Same lens + better sensor = worse image?  Is that what you are saying?

I can see how a step-up in sensor might enable you to utilize a 'finer' lens to greater advantage than with the older body, but I'm having a hard time understanding how the 7D would underperform the 30D with the same lens.

[Ray]

I think Roger is trying to say that his 30D shot with the Sigma lens at 100% on the monitor appears sharper than the 7D shot at 100%, which of course would be about 1.4x larger in each dimension.

If Roger were to downsize the 7D image, using bicubic sharper, to the same file size as the 30D shot, he would find it is at least as sharp and probably slightly sharper than the 30D shot.

[Roger Calixto]

Ray is right, that is what I meant. Reading up on this POTN topic got me thinking about workflow (post#38 especially). There is an extremely technical article posted here on LL about sensor resolution that i haven't read in full, but may be of interest to anyone who wants more info.

[Jeremy Payne]

I think Roger is trying to say that his 30D shot with the Sigma lens at 100% on the monitor appears sharper than the 7D shot at 100%, which of course would be about 1.4x larger in each dimension.

If Roger were to downsize the 7D image, using bicubic sharper, to the same file size as the 30D shot, he would find it is at least as sharp and probably slightly sharper than the 30D shot.

Are they so much less sharp at the pixel level that making selects using 1-1 renderings on the monitor is really that difficult?

I guess I'll go look at some test images ... got a free afternoon ...  

[Jeremy Payne]

Are they so much less sharp at the pixel level that making selects using 1-1 renderings on the monitor is really that difficult?

I guess I'll go look at some test images ... got a free afternoon ...  

Can you post examples that show what you are talking about as well?

[Roger Calixto]

So, now that I've spent some more time reading up on this, it seems I've stuck my foot in my mouth. When looking for reasons why my shots were sharper from my 30D with the same lens, I came across a post talking about how the 7D may outresolve certain lenses (here and here) and kind of ran with the idea.

It turns out, now that I've had some free time to read more, it's technique. The post I mentioned above from POTN actually begins with the same problem I have. Upon upgrading to the 7D a user found his shots to be softer at 100% than with his previous camera. Same as me. The explanation the whole topic gives (now that I've read it all) is that since the pixel density is higher and the pitch smaller, smaller movements are visible. So, given that I shot say at 1/500 on my 30D and had good results the same setting on my 7D may not provide such nice results (zoomed to 100%) since micro-movements are more visible.

Looks like I get the golden foot award for this one. Could a Mod change the topic title to:
7D shots softer than 30D with same lens:
technique problems

Sorry for the misunderstanding!

[bjanes]

It turns out, now that I've had some free time to read more, it's technique. The post I mentioned above from POTN actually begins with the same problem I have. Upon upgrading to the 7D a user found his shots to be softer at 100% than with his previous camera. Same as me. The explanation the whole topic gives (now that I've read it all) is that since the pixel density is higher and the pitch smaller, smaller movements are visible. So, given that I shot say at 1/500 on my 30D and had good results the same setting on my 7D may not provide such nice results (zoomed to 100%) since micro-movements are more visible.

If you want to judge sharpness, you should look at the image at the same magnification. The 100% view on the 7D is effectively a higher magnification. As the reply by RD Kirk on your linked post indicates, the size of the blur limitation with a given shutter speed is the same regardless of pixel density. However, if you want to make use of the increased resolution of the 7D, you have to decrease that blur circle by using a higher shutter speed. The same considerations apply to diffraction. A high pixel density sensor becomes diffraction limited at a larger aperture, so this limits stopping down if you want to achieve maximum resolution. However, for a given aperture the image from the higher density sensor will be no worse than what you would get with a lower pixel density.

The 7D has a pixel spacing of 4.3 microns and the spacing for the 30D is 6.4 microns. If you use Nathan Myhrvold's criterion that the camera will achieve diffraction limited resolution when the Airy disk diffraction formula 2.44 * N * Lambda is equal to the effective pixel size, where N is the f/number and lambda is the wave length of the light. Doing the math for green light (550 nm) gives N = 4.54 for the 7D and 6.70 for the 30D.


 

[fike]

I don't think the original question was wrong at all.  

When you move up to the higher resolution sensors, particularly when they have a more densely packed sensor ( I upgraded from a 30D to 50D), you need to scale back your expectations for 100% viewing quality because the pixel density reveals the limits of your lens resolution. In this case, the 7D has smaller pixel spacing (as BJanes mentions), so it will be more susceptible to diffraction and an apparent lack of focus.

There are three ways that I have changed my workflow:

1) I use a maximum magnification of 50% to judge sharpening and sharpness
2) I no longer print everything at 240 DPI.  Now, I generally print at 300 DPI
3) I don't crop as aggressively as one might think you could based on earlier camera generations

[ErikKaffehr]

Hi,

Technically the posting #38 is quite true. It seems that something like the sharpest lenses at relatively large aperture are needed to achieve maximum performance. It is also true both aliasing and AA-filtering issues are reduced. Per pixel image quality is reduced with small pixels but image quality in print may be similar, and larger prints are possible.

On the other hand, oversampling with a factor two would need four times the pixels. So if 10 MP is a good fit for the sensor you would need 40 MPixels for achieving twice the frequency so you could get rid of the AA-filter.

Speaking of outresolving the sensor or lens is not really correct. THe question is simply how much contrast the lens has at the pixel pitch of the sensor. If contrast at pixel level is high you are going to have aliasing. Some folks like it and some don't. Me standing in between, I hate it when I see it, I don't see it often. AA-filtering reduces aliasing and also reduces contrast at the pixel level. Technically there seems to be a consensus it is needed but it seems that few photographers see real issues with AA-filter less cameras.

It takes excellent technique to take any decent camera to it's limits. So without steady tripod, MLU and very good focusing I wouldn't be that concerned about aliasing and AA-filters.

Best regards
Erik

Ray is right, that is what I meant. Reading up on this POTN topic got me thinking about workflow (post#38 especially). There is an extremely technical article posted here on LL about sensor resolution that i haven't read in full, but may be of interest to anyone who wants more info.

[ErikKaffehr]

Hi,

The Swedish monthly Foto makes some decent tests. Normally they used Sigma 1-50/2.8 as a reference lens for APS-C size sensors, to achieve maximum performance on both the 5D and 7D they needed something sharper, like the EOS 50/2.8 Macro and they needed to stop down to below f/8 (somewhere between f/5.6 and f/8). They found about the same for 4/3. They found that one of the lenses having an aperture of f/4.5 was diffraction limited at f/5.6.

And yes, I'd say that "Bjanes" is absolutely right, even if I'd suggest that Nathan Myhrvolds criterium is a bit tight.

Image below illustrates loss of resolution when stopping down beyond optimum. It was measured on a Konica Minolta 7D and an Sony Alpha 100 a couple years ago.

http://83.177.178.7/ekr/images/stories/difractionlimit.gif

Best regards
Erik

If you want to judge sharpness, you should look at the image at the same magnification. The 100% view on the 7D is effectively a higher magnification. As the reply by RD Kirk on your linked post indicates, the size of the blur limitation with a given shutter speed is the same regardless of pixel density. However, if you want to make use of the increased resolution of the 7D, you have to decrease that blur circle by using a higher shutter speed. The same considerations apply to diffraction. A high pixel density sensor becomes diffraction limited at a larger aperture, so this limits stopping down if you want to achieve maximum resolution. However, for a given aperture the image from the higher density sensor will be no worse than what you would get with a lower pixel density.

The 7D has a pixel spacing of 4.3 microns and the spacing for the 30D is 6.4 microns. If you use Nathan Myhrvold's criterion that the camera will achieve diffraction limited resolution when the Airy disk diffraction formula 2.44 * N * Lambda is equal to the effective pixel size, where N is the f/number and lambda is the wave length of the light. Doing the math for green light (550 nm) gives N = 4.54 for the 7D and 6.70 for the 30D.

[Roger Calixto]

I don't think the original question was wrong at all.  

When you move up to the higher resolution sensors, particularly when they have a more densely packed sensor ( I upgraded from a 30D to 50D), you need to scale back your expectations for 100% viewing quality because the pixel density reveals the limits of your lens resolution. In this case, the 7D has smaller pixel spacing (as BJanes mentions), so it will be more susceptible to diffraction and an apparent lack of focus.

There are three ways that I have changed my workflow:

1) I use a maximum magnification of 50% to judge sharpening and sharpness
2) I no longer print everything at 240 DPI.  Now, I generally print at 300 DPI
3) I don't crop as aggressively as one might think you could based on earlier camera generations

Fike, thanks! That is actually exactly the information I was looking for. Nice to know that a change in workflow IS merited. I also noticed from the POTN thread plus some other reading that the performance of the 7D at higher ISO is been improved noisewise. So can I interpret this in a practical sense as the possibility to increase shutter speed to minimize micromovement? My 30D after 800 wasn't so nice to work with. With my 7D I should be able to go higher to compensate (per se) for the smaller pixel pitch (given the same glass in the same light), correct?

[bjanes]

And yes, I'd say that "Bjanes" is absolutely right, even if I'd suggest that Nathan Myhrvolds criterium is a bit tight.

Image below illustrates loss of resolution when stopping down beyond optimum. It was measured on a Konica Minolta 7D and an Sony Alpha 100 a couple years ago.

http://83.177.178.7/ekr/images/stories/difractionlimit.gif

Best regards
Erik

Eric,

You are correct that Myhrvold's criterion that the Airy disc should be no larger than the diagonal of the pixel spacing is conservative. In his excellent review of diffraction on Cambridge in Color, Sean McHugh uses twice the pixel spacing for a Bayer array. since demosaicing is done for a 2 x 2 pixel area. Photozone.de has some excellent lens reviews and it is interesting to compare Klaus's results with theory. For example, he reviews the Zeiss Makro Planar 50 mm f/2.0 on the Nikon D200, which has a pixel spacing of 6.1 microns. Optimal MTF 50 in the center is obtained at f/4 where the Airy disc is 5.4 microns for green light. Myhrvold's formula gives a diffraction limit of f/6.4 where the Airy disc diameter is 8.5 microns.

Regards,

Bill