645D vs 1DSmk3 vs D3X

[mhecker*]

Hi All,

See http://www.imaging-resource.com/PRODS/645D/645DA.HTM

The 645D betters the 35mm champs.   
Look, download and see for yourself.

Enjoy,
Miles

[Bart_van_der_Wolf]

The 645D betters the 35mm champs.   
Look, download and see for yourself.

Have you spotted the on sensor magnification differences? Makes everything else a moot point ...

Cheers,
Bart

[ErikKaffehr]

Hi Bart,

What do you mean? Can you explain a little?

Best regards
Erik

Have you spotted the on sensor magnification differences? Makes everything else a moot point ...

Cheers,
Bart

[nazdravanul]

Hi Bart,

What do you mean? Can you explain a little?

Best regards
Erik

Download the full files for yourself (you see only selected crops in the review) and check out the obvious framing/magnification differences. That is one ridiculous comparison.

[Bart_van_der_Wolf]

Hi Bart,

What do you mean? Can you explain a little?

Hi Erik,

The 645D image magnification is much higher, hence for that reason alone it will allow higher detail resolution and run a lower risk of aliasing. The 645D image magnification is in the order of 50% larger than the 1Ds3, and the D3x magnification is some 8% larger than the 1Ds3. The absence of an AA-filter may have created some pseudo detail, we don't know for sure. Finally, the 1Ds3 and D3x image crops do not seem to be optimally sharpened, but because of the earlier issues it doesn't matter that much, a resolution comparison is already flawed.

If the purpose is to demonstrate that higher on-sensor magnification results in higher image quality, then the obvious has been demonstrated.

Cheers,
Bart  

[ErikKaffehr]

Hi,

I guess I see your point.


I downloaded two images and upsized to 50x70 cm at 200 PPI using Bicubic Smoother. After that image-adjust-Match colors was applied. I addition Smart Sharpen 0.5, 150 was applied on the Nikon image.

The Pentax image is on the left and the Nikon image is on the right.

I also downloaded a 200 ISO image for the Pentax 645D and a 100 ISI image from the D3X and developed in CS5 resp Lightroom with default settings and calculated LW/PH for 50% MTF using Imatest:

Pentax 645D: 4605 LW/PH
Nikon D3X: 3057 LW/PH

Would I scale up the Nikon to the same number of pixels as the Pentax it would have around 4200 LW/PH. I don't know which lens was used but assume that it was the 24-70/2.8

Best regards
Erik

Hi Erik,

The 645D image magnification is much higher, hence for that reason alone it will allow higher detail resolution and run a lower risk of aliasing. The 645D imagnification is in the order of 50% larger than the 1Ds3, and the D3x magnification is some 8% larger than the 1Ds3. The absence of an AA-filter may have created some pseudo detail, we don't know for sure. finally, the 1Ds3 and D3x image crops do not seem to be optimally sharpened, but because of the earlier issues it doesn't matter that much, a resolution comparison is already flawed.

If the purpose is to demonstrate that higher on-sensor magnification results in higher image quality, then the obvious has been demonstrated.

Cheers,
Bart  

[mhecker*]

I just downloaded and measured the 2 full images, 645D vs D3x. I optimumally sharpened them myself.

The difference in vertical framing measured to the same points on the still life is slightly less than 2% in favor of the 645D.
If you want to believe that accounts for the big difference in micro detail we see, I think you're kidding yourself.

IMO the 645D under the conditions of the test is producing a file with significantly more micro detail and better tonal transitions than either of the 35mm cameras.  YMMV.

I will make 24x30 prints and see how visible the difference is in that venue.
I invite others to do the same and make their own evaluation.

[ErikKaffehr]

Hi,

If you do match color the images will get much closer in tonality...

Best regards
Erik

I just downloaded and measured the 2 full images, 645D vs D3x. I optimumally sharpened them myself.

The difference in vertical framing measured to the same points on the still life is slightly less than 2% in favor of the 645D.
If you want to believe that accounts for the big difference in micro detail we see, I think you're kidding yourself.

IMO the 645D under the conditions of the test is producing a file with significantly more micro detail and better tonal transitions than either of the 35mm cameras.  YMMV.

I will make 24x30 prints and see how visible the difference is in that venue.
I invite others to do the same and make their own evaluation.

[nazdravanul]

I just downloaded and measured the 2 full images, 645D vs D3x. I optimumally sharpened them myself.

The difference in vertical framing measured to the same points on the still life is slightly less than 2% in favor of the 645D.
If you want to believe that accounts for the big difference in micro detail we see, I think you're kidding yourself.

IMO the 645D under the conditions of the test is producing a file with significantly more micro detail and better tonal transitions than either of the 35mm cameras.  YMMV.

I will make 24x30 prints and see how visible the difference is in that venue.
I invite others to do the same and make their own evaluation.

That is very strange. According to my own measurements of the same shots (the still life), the bottles in the 645D shot are at least 15-17% bigger than in the 35mm DSLR shots. I double checked my measurements  ? Can you please double check yours ?

Also, as far as lenses are concerned (check exif for the shots) , they used the 75mm for the Pentax, which I understand is a superb lens (no direct experience with that) , and the Sigma 70 Macro for DSLRs - apparently only a very good lens (no direct experience either) so I wish they had used a Zeiss MP, to give a true account of sensor potential, not lenses  (I bet even the 50 Makro-Planar  Zeiss -not to mention the 100 -  will kick the living hell out of the Sigma in terms of both micro-contrast and color gradation, for example )


PS I do have extensive direct personal experience with Zeiss versus cream of the crop Canikon lenses.

[Bart_van_der_Wolf]

I just downloaded and measured the 2 full images, 645D vs D3x. I optimumally sharpened them myself.

The difference in vertical framing measured to the same points on the still life is slightly less than 2% in favor of the 645D.
If you want to believe that accounts for the big difference in micro detail we see, I think you're kidding yourself.

I don't see how you could arrive at such conclusions. When I measure the length of the upper wooden bar (with the 19 strands of yarn on nails) at the top right, I measure 3094 pixels for the 645D and 2295 pixels for the D3x, a difference of +34.8% for the 645D. When subject matter is sampled by 34.8% more pixels, it obviously displays more detail. The  length on the 1Ds3 image is 2100 pixels, so the difference is +47.3% for the 645D. It doesn't require rocket science, denser sampling or larger magnification helps ...

Cheers,
Bart

[nazdravanul]

I don't see how you could arrive at such conclusions. When I measure the length of the upper wooden bar (with the 19 strands of yarn on nails) at the top right, I measure 3094 pixels for the 645D and 2295 pixels for the D3x, a difference of +34.8% for the 645D. When subject matter is sampled by 34.8% more pixels, it obviously displays more detail. The  length on the 1Ds3 image is 2100 pixels, so the difference is +47.3% for the 645D. It doesn't require rocket science, denser sampling helps ...

Cheers,
Bart

Thank you, for an even more precise measurement .

[elliot_n]

Download the full files for yourself (you see only selected crops in the review) and check out the obvious framing/magnification differences. That is one ridiculous comparison.

Why ridiculous?

In the test image I'm looking at ('still life' - beer bottles, fabric swatches etc.) the vertical framing is well matched. Obviously, objects in the Pentax image are bigger - it's a 40MP image vs. a 24MP image. You're free to upsize one, or downsize the other.

The Pentax image looks impressive.

[tsjanik]

..... . When subject matter is sampled by 34.8% more pixels, it obviously displays more detail. The  length on the 1Ds3 image is 2100 pixels, so the difference is +47.3% for the 645D. It doesn't require rocket science, denser sampling or larger magnification helps ...

Cheers,
Bart

Bart:

I'm not sure why you object to this, it is not a pixel-to-pixel comparison, but rather image-to-image.  If I were to conduct such a test, I would compare images with a similar angle of view, exactly what IR has done given the constraints of 4:3 vs 3:2 formats.  Of course the 645D has more pixels, isn’t that the point of 40MP?

[Bart_van_der_Wolf]

Bart:

I'm not sure why you object to this, it is not a pixel-to-pixel comparison, but rather image-to-image.  If I were to conduct such a test, I would compare images with a similar angle of view, exactly what IR has done given the constraits of 4:3 vs 3:2 formats.  Of course the 645D has more pixels, isn’t that the point of 40MP?

Sure, but why do people get worked up about the improved resolution when comparing smaller 21-24.5 MP sensors with a larger 40 MP one , and one of the cameras without AA-filter, and sub par sharpening of the smaller sensor results? What kind of comparison is that? I could have told what the outcome would be without even looking at such a comparison.

A comparison with another say 45mp sensor would be much more interesting, especially given their price differences (although there are more differences than just that).

Cheers,
Bart

[mhecker*]

My measurements are based on % of image height, not pixel to pixel!

People shoot MF and LF to get more film or silicon on the image being shot.
Pixel for pixel, or millimeter for millimeter, 35mm may have an advantage.
That is in IMO a moot point. 

For small prints, the difference is probably minimal, it might be invisible.
For making large prints, 20"x25", 24"x30" or even 30"x40" the files I see being produced by the 645D
will yield a visibly better end product than the files produced by  the other two cameras.

That's all that matters to me.
I am not a pixel peeper.   YMMV.   

[bcooter]

I was in a large retailer yesterday in their pro department.  Which was empty of customers.  Asked one of the sales guys I know if they're selling any medium format and they said yea, mostly Hasselblad, nearly all to advanced amateurs.

So maybe Pentax doesn't have to worry about tethering, because their market doesn't have to tether.

The sales guy did say that he hopes the Pentax puts some pressure on Phase and Hasselblad to get their prices lower, though he's not sure if that will change anything.

On a side note, I went down one floor to the not pro department and the Leica area was empty, the Canon shelf had a few customers, Nikon a few more, the video area had a waiting line.

I don't know what this means, since it is the Christmas sales period, but  . . .

IMO

BC

[tsjanik]

Sure, but why do people get worked up about the improved resolution when comparing smaller 21-24.5 MP sensors with a larger 40 MP one , and one of the cameras without AA-filter, and sub par sharpening of the smaller sensor results? What kind of comparison is that? I could have told what the outcome would be without even looking at such a comparison.

A comparison with another say 45mp sensor would be much more interesting, especially given their price differences (although there are more differences than just that).

Cheers,
Bart

Sure a comparison of the 645D to a MFDB would be interesting.  Mark and Nick have already posted such a comparison here.  Although they were widely criticized for their technique, the report told me what I was interested in: that there was little to choose between the Pentax and Phase files.  As you point out, everyone expects the 645D to have more resolution than the D3x, what wasn’t known is how the 645D compares at higher ISO, or the general output from the sensor or how the old 645 lenses perform on a digital sensor.  If the 645D files were mediocre, whatever the cause, the D3x would be a better choice even for a landscape photographer.  I think IR rightly points out that the cost of the Pentax makes it a viable choice for anyone considering a D3x or 1Ds III, depending on the application.  For landscapes, the Pentax seems a clear winner, especially if you have a closet full of Pentax lenses. 

[ndevlin]

I don't see how you could arrive at such conclusions. When I measure the length of the upper wooden bar (with the 19 strands of yarn on nails) at the top right, I measure 3094 pixels for the 645D and 2295 pixels for the D3x, a difference of +34.8% for the 645D. When subject matter is sampled by 34.8% more pixels, it obviously displays more detail. The  length on the 1Ds3 image is 2100 pixels, so the difference is +47.3% for the 645D. It doesn't require rocket science, denser sampling or larger magnification helps ...

Maybe I am misreading you, but isn't the obvious reason that there are more pixels in same section the Pentax image because....there are more pixels in the Pentax sensor? 

The point is not to compare pixel-to-pixel, but rather similarly frame subject with similar perspective lens.

- N.

[Bart_van_der_Wolf]

Maybe I am misreading you, but isn't the obvious reason that there are more pixels in same section the Pentax image because....there are more pixels in the Pentax sensor? 

Hi Nick,

Bingo, exactly my point, but the resolution difference is not really surprising then, is it?

The point is not to compare pixel-to-pixel, but rather similarly frame subject with similar perspective lens.

Indeed, but what possible revelation could come from such a test? It's all rather predictable. The conclusion that a sensor array of that size with decent optics has its benefits is not exactly a surprise. Were the testers expecting something different?

Cheers,
Bart

[ErikKaffehr]

Hi,

The impression from Nick's and Mark's test I got was:

- The 39 MPixels are for real
- Old lenses work
- The Pentax 645D is competitive in performance with systems twice the price

Best regards
Erik

Hi Nick,

Bingo, exactly my point, but the resolution difference is not really surprising then, is it?

Indeed, but what possible revelation could come from such a test? It's all rather predictable. The conclusion that a sensor array of that size with decent optics has its benefits is not exactly a surprise. Were the testers expecting something different?

Cheers,
Bart