[interested_observer]

I was doing some reading, and had come across a couple of threads here that had linked a post from Ken Rockwell. Now, I admit that I usually spend very little time on his site, however this page did get me thinking -

• The Full-Frame Advantage

In a nut shell, he made several observations (and provided images), that images from full frame sensors using crummy lenses, are better than cropped sensors using the very best lenses. This is essentially due to full frame sensors having larger pixels that are better able to gather the light.

So, that got me thinking this week. Rockwell used 2 cameras - a Canon 5D (FF/CMOS) with 8.2 micron pixels, and a Nikon D200 (cropped/CCD) with 6.1 micron pixels. Bottom line the FF pixels were about 25% larger. You can go to the link above and see the images and the visual effect.

Pentax, does not have a full frame body. So, I took a look at the pixel sizes on the bodies that I had available - a K100D (CCD) at 7.87 micron pixels and a K5 (CMOS) at 4.75 micron pixels. The K100D has approximately the same pixel size as the Canon 5D. The K100D's pixels are about 40% larger than the K5's. Now there is a slight difference in resolution between the two bodies - K100D at 3000x2008 and the K5 at 4928x3264, but that can be taken in to account (to a degree).

What I did was, to use a single lens for both shots (as opposed to a crummy lens vs a great lens) - in this case a Contax Carl Zeiss 85mm f2.8 Sonnar on both bodies, setting on a tripod, each focusing on the same point 165 feet away (measured via Google Earth's tape measure). Set the ISO to 200, using f8 and focused to infinity (and did get focus confirmation on each body using manual focus). The K100D shutter speed was 4 seconds, while the K5 was shot at 2.8 seconds. I shot both in RAW, and took 100% crops of the same area, then trimmed to a size that would be accepted for a post here.

So how does fat pixels compare to tiny pixels actually fair? - I know, the light could have been better, but it was dark by the time I got home. Anyway, I wanted to see how the K5 did in the evening with the ambient low light that I usually have.

I adjusted for physical size (so I compressed one image), however I did upload the image before compression for a full comparison...

Attached Images

     

[spade111]

Nice one Interested Observer!

I will reserve the few thoughts I have regarding comparative methodology and simply applaud your effort here. Could you possibly confirm that the first photo is the K100d, the second the K5, and the third is a full uncompressed photo? It is difficult for me to make this out in your post - and I'm not seeing the labels.

What were your impressions from the experiment?

[demp10]

So you are comparing two different sensor technologies CCD and CMOS, different resolutions (pixel size), one at its native ISO (CCD @ 200) and the other pushed one stop (CMOS native @ 100), using natural variable light with different exposures, one at 4 sec and the other at 2.8.

With so many variables it is next to impossible to draw conclusions on the effect of pixel size on the sensor.

Generally speaking, a sensor with larger pixels will have less problems with diffraction, or to put it differently, diffraction limitations will start showing at a small f-stop (larger number). Typically for k-7 or K-5 the early signs of diffractions begin at around F9 and become serious at F16 or smaller.

A more reveling test will be under studio conditions with controlled light, using resolution targets, shooting at base ISO and over a range of F-stops to see when diffraction kicks in for each sensor.

[interested_observer]

Wow - that was a quick reply. Yes, the first is the K100, and the last 2 are from the K5. I just received the K5 a few days ago. But, due to work have not really had any time to try it out - getting home way too late, and just crashing.

On the full size images, the K5 is Very nice. I have them up side by side using Bibble Labs 5, and the K100 does give a good account of itself, however the additional resolution of the K5 does have more detail. That said, the lack of extremely fine detail on this one cactus does look better on the K100's image.

I actually thought that the approach I tried would have been a bit easier to execute. Trying to preserve the pixels in near their original state is more difficult that I anticipated. I anticipated that since the sensors were essentially the same physical size, that would help, especially in terms of the angle of view - but resolution difference overwhelms that aspect.

I will say, that in another part of the image, you can see the cactus needles on one cactus in particular - due to the lighting angle.

The K100 being a CCD does do a good job in terms of detail and color with reasonably low noise. The K5 from my limited experience matches the CCD, the noise level is excellent, as its performance in low light.
___________________

Demp10 - The linked site had similar differences and saw (at least to me) some striking differences. I would have though that in terms of physical pixel size that I would have possibly seen somewhat similar differences. Anyway, I wanted to try out the new body and this seemed a reasonable first try in an area that I had shot in before.

---

Last edited by interested_observer; 02-03-2012 at 10:29 PM.

[spade111]

I agree with your assessment - that the needles are a bit more - sort of more contrasty in the k100 photo - even compared to the blown up k5.

I like the way the k100d deals with the shadows - it looks more histogrammatically correct compared to the k5's obviously deeper dynamic range. I wonder if that subtle variation in then the apparent contrast (due to the shadow handling) explains more the difference in the two images?

[Anvh]

Demp10 - The linked site had similar differences and saw (at least to me) some striking differences

But the question is are those differences caused by the pixel size or by other factors which aren't the same?