Sunday, January 22, 2012

More On Meidum Format

OK, so perhaps this is a way to show the difference. Here are two huge enlargements of a tiny section of image. SO enlarged that virtually no detail is visible in either image, but what you are seeing is each square is a pixel, and look at the difference in tonal variation, even both images had similar overall contrast, highlights and shadows. This isn't noise you are looking at - it is genuine detail as confirmed by careful inspection of lower magnification views of the images show.

The first is the Canon 1Ds3

and the second is the Pentax 645D

This implies it isn't the number of pixels that counts, it is the quality of the pixels. Is it possible that removal of the fuzzy filter will so improve the quality of the dslr pixels as to make up the difference - I simply don't know. But what you are seeing here at 1500% magnfication is why you can see the difference in ordinary size prints.

This even has implications for stitching - I can easily equal the pixel count of the Pentax by stitching two or three overlapping Canon 5D2 images, but the inability to resolve low contrast areas will remain.

And I haven't even begun to look at shadow areas in which the differences between the two cameras is far greater. That Pentax, with all its limitations (no live view, shallow depth of field) may just be what I'm looking for.

Below are 100% crops of the same area, Canon first, Pentax second.


Anonymous said...

Interesting look at the subject, George. I don't know that I'll ever have a MF digital (do have film MF), but I think your research basically supporting Mark's thesis (tho not his entire train of thought) makes a good case for bigger is better. If I had the money, I'd seriously look at digital MF. Will await your decision!



Alexander S. Kunz said...

I'm curious about the nature of the comparison. Are both of the images converted from raw data? Was it the same raw converter? Was noise reduction and particularly COLOR noise reduction disabled? I found that the latter makes a BIG difference in Lightroom/ACR. Color noise reduction at the default setting ("25") really eats a lot of detail, and the second picture from the Canon suspiciously looks as if it received that unfortunate treatment.

George Barr said...

Alexander's questions about the source and processing of the images is entirely relevant and the short answer is I don't know. They were taken from the ephotozine review,

but it doesn't seem to say whether those images in particular were raw or jpeg, or in which raw processor they were run through.

The only way to really know the difference would be to do the test myself, and perhaps I will get a chance at some point. The comment about colour noise suppression is interesting though I presume that whatever processing the images had was the same.


Royce Howland said...

We certainly could do a side-by-side shoot that would answer questions like these. Could be fun. Though not every variable can be independently controlled, without upsetting some other variable. (Witness every comparative review posted on the web, which gets criticized by somebody for flawed methodology.)

I've been meaning to do some Canon full frame vs. 645D comparisons myself for some time. But whenever I go out to try it, I get preoccupied with shooting real scenes with whatever preferred system, and run out of time to shoot the tests... :)

Nagaraj said...

It is a well known fact that the Signal to Noise ratio for a pixel on the imaging chip reduces as the pixel size reduces. This is pure physics. The amount of noise captured by a pixel is independent of its size whereas the signal(in other words the detail) is proportional (more or less) to its size. So the signal to noise ratio is higher for a larger pixel than a smaller pixel. Everything else like dynamic range, detail in the shadow areas etc. is a consequence of this fact. For a given level of technological sophistication all other things being equal, the size of the pixel on the imaging chip determines how good the native(RAW) file would be rendered eventually by the camera. No technological (r)evolution can surpass this limit.

Alexander S. Kunz said...

Nagaraj - the pixel pitch of the 1DS3 (6.3µm)and the 645D (5.93µm) is pretty close, and the 1DS3 actually has slightly larger pixels. That can not be the explanation for George's observation.

Nagaraj said...

Alexander: Yep my bad. I presumed that 645D has a larger pixel pitch. Either the technology/electronics in 645D to render the RAW file is superior to 1DS3 or the post processing of the RAW files in both cases was not the same.
Either way very interesting to see what it will eventually turns out to be. I will eagerly await the conclusions.

Unknown said...

The 645D out resolved the best I could do with MF 6x7 film: Rollie ATP 1.1 with high resolution developer (negs scanned on a Nikon 9000ED) with the same P67 lens.