1.6 versus full frame for birds

[John Chardine]

There is a lot of discussion out there on this topic with the consensus being that a 1.6 or 1.5 crop factor camera is better for bird photography because of the "extra reach", "narrower angle of view", etc. You may disagree but I think this fairly reflects opinion out there.

I have just added a Canon 5D to my 50D and 40D (will sell) and so thought I would run a simple test on a stormy New Years Day (felicitations everyone!). The purpose of the test is to ask: how does final quality of an image prepared for BPN differ between the FF 5D and the 1.6 factor 50D, all other things being equal. Subject was a bird tree ornament about 3" high with a good amount of detail, distance to subject was about 20', AF point was centre of breast close to the two white patches. 500/4 was set up on tripod, mirror lockup, 2s self timer, cable release, manual metering, 1/500 @f5.6, flash with beamer ETTL no compensation, ISO 400. RAW images processed in ACR 4.6 (to eliminate effect of all in-camera settings).

Processing in Ps involved cropping each image similarly, resampling each to 800 high with Bicubic (not Bicubic sharper which might have introduced different sharpening amounts for the two images because they started out at very different sizes before resampling), bringing them together in a single image and sharpening the same amount- USM 50%, 0.3, 0. Save for Web and Devices at 200k.

Camera details are as follows:
5D: FF- 36 x 24 mm, 4368 x 2912
50D: 1.6-22.3 x 14.9 mm , 4752 x 3168

You can draw your own conclusions but here's what I take away from this:

1. Very little difference in IQ between the two, 5D marginally softer, which means that at least for web posting there is little advantage in the "extra reach" of a 1.6x body. A different test would have to be done for printing and the outcome may well be different.

2. More noise in the 50D (expected)

Comments?

[Doug Brown]

I think that putting more pixels on the bird is important when cropping and subsequently printing larger sized images. The frame rate on the 5D is too slow for flight, and so is the AF. Still, it's interesting to see how similar the two images look.

[John Chardine]

Forgot to mention how heavy the crop was Doug. Here is the original from the 5D.

I agree that the rubber likely meets the road when you do a heavy crop and print big- then the 50D may shine. However, really need to test to confirm.

[Ed Cordes]

Very interesting! Maybe it's my lap top monitor, but to my eye the 5D image(despite the large crop) looks sharper with more contrast. This is hard to believe as the 50D certainly puts more pixels on the subject - 15 mp none to little crop vs 12 mp with large crop! I'll bet that not downsizing for the web, but looking at print size detail would show more difference.

[Roger Clark]

John,
I agree with Ed, on my laptop screen, the 5D image looks sharper with more pixel to pixel contrast.

Regarding which is better, cropped versus uncropped sensor is not the correct question (there is a lot of confusion out there). The question of which gives more pixels on the subject for a given lens is dictated by pixel pitch (the spacing between pixel centers). The 50D has a pixel pitch of 4.7 microns, and the 5D has 8.2 microns. So one would expect 8.2/4.7 = 1.7 times more pixels on the subject with the 50D. With the pixel pitch of 6.4 microns on the 5D Mark II, the ratio would be only 6.4/4.7 = 1.36 times more pixels on the 50D versus 5DII. Then compare a FF 5DII to a 1.6x crop 30D and the pixel pitch is the same, so both cameras give the same pixels on the subject with the same focal length lens. Crop factor has nothing to do with pixels on subject.

Next is quality of pixels. On the 50D with 4.7 micron pixels, the pixel to pixel contrast can never be more than 40% at f/4 due to diffraction, whereas on the 5D it can be as high as 60%. You can see this effect (called the Modulation Transfer Function, or MTF) in Figure 8 at:
http://www.clarkvision.com/imagedeta...el.size.matter

Pixel pitch is given for many cameras in Table 2 at:
http://www.clarkvision.com/imagedeta...rmance.summary

Roger

[Ed Cordes]

Roger, thanks for all the technical detail. I have read your articles and still am not sure I understand the science. Am I correct in thinking that when I replace my 1D Mark 2 (hopefully soon) the replacement should have larger pixel pitch? I have been thinking of switching to FF, so in FF the diffraction effects are less than a smaller sensor? Maybe I have it all wrong? Of course any new camera will be 14 bit, but I am not sure if this is the great advantage the marketing guys say.

Please correct me if I am wrong.

Thanks
Ed

[Roger Clark]

Roger, thanks for all the technical detail. I have read your articles and still am not sure I understand the science. Am I correct in thinking that when I replace my 1D Mark 2 (hopefully soon) the replacement should have larger pixel pitch? I have been thinking of switching to FF, so in FF the diffraction effects are less than a smaller sensor? Maybe I have it all wrong? Of course any new camera will be 14 bit, but I am not sure if this is the great advantage the marketing guys say.

Please correct me if I am wrong.

Thanks
Ed

Hello Ed,
If you want to reduce diffraction effects, you need to increase the pixel pitch. Again, full-frame or crop is irrelevant. However, if you want more pixels and larger pixel pitch, then you need a larger sensor. If you get a full frame sensor and it's pixel pitch is small, the diffraction effects will be larger on a per-pixel basis.

I also have a 1D Mark II and hope for an upgrade. I mainly skipped the 1DIII because 2 more megapixels is not much, and the 14-bit converters in the current generation of cameras (from any manufacturer) are only slightly better than the 12-bit systems. The 1DII is still a great camera. With the 50D and the 5DII, plus competition from Nikon and other manufacturers, I predict (actually I'm hoping here) the 1D Mark IV will be in the 16+ megapixel range (20 would be better), at least 10 frames/second and probably full frame. Probably 16-bit too, with much better performance. But overall performance in terms of dynamic range and noise will still be controlled by pixel size (pixel pitch). Pixel size (pitch) is the main factor in performance and image quality (not counting AF and frames/second). Larger pixels collect more light giving higher signal (and signal-to-noise ratio) and higher dynamic range. That must be balanced by wanting many pixels. The optimum in my opinion is in the 6 to 8 micron pixel pitch range. Five microns is pushing the lower limit and I think we are seeing some of the detrimental effects on the 50D (4.7 micron pitch) with higher noise and lower dynamic range. While this decrease in image quality is slight, if the camera manufacturers push for more pixels and smaller pitch, I think we will see more issues and complaints will rise. So get the largest pixel size with the most pixels you can afford.

I hope this helps.

Roger

[John Chardine]

Yes thanks Roger. The tech stuff is slowly sinking in.

You make an important point that sensor size (1.6, 1.5, 1.3, FF etc) and pixel size should be considered independently. The original idea of the post was to see if you take a big hit with a FF sensor in terms of capturing an image of a small bird at a real-life distance with a commonly used bird lens. The answer seems to be no- for web posts at least- and the reason is the size and quality of the pixels on the FF camera I tested. I strongly suspect that printing would show the same results.

In keeping with your emphasis on not confusing technical detail, I want to bring up the issue of pixel size versus pitch. Pitch is the on-centre distance between adjacent pixels- 8.2 microns for the 5D. However, this does not necessarily translate to pixel size because coverage by the pixels themselves or the microlenses above them may not be complete. You could have a big pitch and small pixel with lots of dead space between them. I understand that the 50D does as well as it does noise-wise (considering it's small pitch) because it has "gapless microlenses". Not sure if this is manufacturer hype or not.

So if I'm right, you can't just divide sensor width by number of pixels to get pitch, and assume that this also translates to pixel size. Or can you with most/all modern sensors?

[Ed Cordes]

John, thanks for starting this thread. I am learning a lot. Roger, thanks for the reply. It is much clearer now. I too am looking forward to the release of the 1D Mark IV. Based on current releases my guess is it will be FF, with around 20 mp, and 5-8 fps. However, I think it will have an option of selecting a smaller capture size with faster frame rate.

I too think we have reached the limit of mega pixels. File sizes are getting huge and storage as well as computer processing power needed are now large issues . You clearly show that cramming more pixels on a sensor by reducing pixel pitch. So, the manufacturers should quit conducting a mega pixel race and give us better quality pixels with larger pitch and better processing.

The current mega pixel race has always reminded me of the muscle cars of the '60s - throwing more and more horse power into a car with a chassis that can't corner or break with agility to control the HP when the car is not going in a straight line.

OK, I'm done rambling. Great thread.

[Roger Clark]

In keeping with your emphasis on not confusing technical detail, I want to bring up the issue of pixel size versus pitch. Pitch is the on-centre distance between adjacent pixels- 8.2 microns for the 5D. However, this does not necessarily translate to pixel size because coverage by the pixels themselves or the microlenses above them may not be complete. You could have a big pitch and small pixel with lots of dead space between them. I understand that the 50D does as well as it does noise-wise (considering it's small pitch) because it has "gapless microlenses". Not sure if this is manufacturer hype or not.

Hi John,
Yes, that is correct. Pixels must have dead space in between them so electrons do not leak into adjacent pixels. The active area sensitive to light divided by the area given by the pixel pitch (the pixel pitch squared) is called the fill factor. With the micro lens designs these days, the effective fill factor is probably above 90% and may be as high as 95%. Back in the old Canon 10D days, the fill factor was about 60 to 70%. So with modern cameras, the pixel pitch gives a close value for the active area. The models on my sensor performance page uses a half micron dead space between pixels (even assuming micro lenses) and the models fit observed data pretty well. As pixel size and pitch decreases, the fill factor must also decrease due to the need for the dead space, another strike against small pixels.

So if I'm right, you can't just divide sensor width by number of pixels to get pitch, and assume that this also translates to pixel size. Or can you with most/all modern sensors?

While technically true, it is a very close estimate for modern cameras.

Roger

[Roger Clark]

The current mega pixel race has always reminded me of the .....

Yes, we shouldn't get too wrapped up in megapixels, nor pixel size or pitch. There are many factors in choosing a camera for a job, including AF performance, frames per second, write speed to the card during action, weather resistance, and overall feel of operation (and probably many more). While the sensor is important it is only one of several important factors.

For example, for an upcoming trip I wanted to have a 5D mark II for landscapes. I'll still take my 1D Mark II for action. I wanted something more than my 30D backup. But if a 5D2 didn't come in in time, I was concerned about which camera to substitute: a 50D or 40D, or perhaps a 5D, or another 1DII (they are pretty cheap used). In good light the 50D is great, but I do a lot of low light imaging, so I was still considering the 40D. With the 5DII out, it didn't seem like a 5D would be the right way. I never could make a decision and still do not know which I would have chosen but fortunately my 5D2 came in. ;-)

Now I want the 1D Mark IV. Hmmm. They've got me where they want me: constant upgrades.

Happy New Year.
Roger