Pentax 645D compared to Nikon 3DX based on Imaging Resource's samples

[ErikKaffehr]

Hi!

Let's put it this way, my article is about what we can expect from Pentax 645D, not about uprezzing for printing. I don't think my choice of uprezzing falsifies the results of the comparison.

Diglloyd has noted that off axis performance on the 55/2.8 is bad. That would be not seen in my comparison as it looks at central detail. I have actually noted lack of sharpness on the "brush" on the right edge but attributed it to being out of focus.

Best regards
Erik

I'd like to see this in real prints...
While this might be true for certain images (probably only for certain details of certain images actually) I suspect it very much depends on the amount of uprezzing. In every image there is a certain limit (threshold) of uprezzing. If you uprez too high you are just introducing artefacts such like doubling of edges, a kind of ghosting in straight vertical or horizontal lines etc. (at least as long as we talk about bicubic smoother). You certainly won't enhance those pseudo details...?
Within reasonable values I think a higher enlargement is leveled out by higher printer resolution. But if the enlarged image is actually destroyed by too heavy uprezzing I seriously doubt that the higher printer resolution will level out the artefacts (i.e. "new" details that were actually not captured).
I do prefer a higher amount of uprezzing and printing at 300ppi over less uprezzing and printing at 200ppi on a Chromira RGB printer. However printing on an Epson I would prefer 240ppi over 360ppi with some critical images. Though mostly I print at 360ppi. But I can't imagine that I would gain anything printing at 480ppi... at least not when heavy uprezzing is required to meet a printer resolution that high.

[ndevlin]

I really think that 160mp'ish is going to be the sweet spot.

If, and only  if, we are talking about a FF 6x7 sensor, or something like that. The lenses and the AF systems already don't support the resolution at 40MP, much less 60 or 80. Smaller pixels suck. Nothing has quite matched what the big 9 micron sensors in the P25 and D3 can do. Images off those cameras, while not as enlargable, always have a tangibly different feel' to them, imho.

- N.

[ErikKaffehr]

Hi,

In my view resolution does not hurt. DR in the DxO sense will go down, but in my opinion  practical photography is dominated by shot noise and that is pretty independent on the number of pixels. With improved resolution we won't get much better detail so long as lenses and AF systems don't keep up with the smaller pixels, but we will have less aliasing artifacts.

To put it shortly, I don't think that small pixels suck, but they will not result in a sharper image. The improvement is less artifacts.

Best regards
Erik

If, and only  if, we are talking about a FF 6x7 sensor, or something like that. The lenses and the AF systems already don't support the resolution at 40MP, much less 60 or 80. Smaller pixels suck. Nothing has quite matched what the big 9 micron sensors in the P25 and D3 can do. Images off those cameras, while not as enlargable, always have a tangibly different feel' to them, imho.

- N.

[Leping]

Hi Erik,

The problem of your comparison, as well as many similar ones has been that, you put "enlarge-ability" of D3x image into the formula while did not consider the same aspect for the higher resolution side.

To me 645D images provide about the best upsampling experience over all Bayer type sensor based examples.  Usually no matter how careful you are and what method and software you use, you only need factors like 135% to turn DSLR images into really ugly.

I have been really impressed with the D3x image quality, but when I upsampled 645D's images to 30x40 at 300dpi, I see very low levels of demosaicing artifacts and still nearly pixel level details.  The D3x file totally broke down at this level.

If you email me I will send you some 100% unsampling samples in private.  Thanks.

P.S.: I just added an example, 645D and D3x imaging-resource images, both carefully blow up to 24" high at 300dpi.  See difference in small letters ("AUTHENTIC BEERS", etc.)  Yes this is pixel peeping and as you put a ruler on screen and see how big an inch is there you realized on prints the difference tend to diminish.  However, the difference will be there is you were printing 48" high, I am sure.

[ErikKaffehr]

Hi,

I don't really understand your point on the aspect ratio. I guess that what my scanned print samples show is that the P645D is superior to D3X in the aspects I looked at. I'm fully aware that the two cameras have different aspect ratios and the numerical figures are for the short size on each.

I'm glad that you are happy with your Pentax P645D, seems to be a very good camera to me. Lloyd Chambers is testing it and is quite unhappy with the new 55/2.8 lens (although he has two samples).

Best regards
Erik

The problem of your comparison, as well as many similar ones has been that, you put "enlarge-ability" of D3x image into the formula while did not consider the same aspect for the higher resolution side.

[Leping]

Yes we looked over a dozen 55mm/f2.8 examples, Japanese, Russian, and even imaging-resource (the "house" shot), and they are all the same -- out of the center 2/3 (or 1/2 as Lloyd said today) it is not sharp at all.

However, he did find the 75mm OK to the sensor, and online we could see good examples of corner-to-corner sharpness off the 45-85mm zoom and the 120mm macro lens.  We were also quite amused that there were no 55mm examples in the Pentax's own official 645 sample images.  Just as he said, how could Pentax released such a lens, and termed it "outstanding", is well beyond me.

On the other hand, I am a practical photographer and I do need DOF with MF in the real world.  I modified a 55mm P67 lens with fixed ~1 degree down tilt to help me in near-far type wide angle compositions just to deal with the DOF issue, for example.  With film f/16 to f/22 was the norm, and I was aware loss of sharpness due to diffractions.  However the softness can be at least partially corrected in sharpening, while the out-of-focus objects are hopeless.  For the same reason Ansel started the f/64 school and virtually turned every of their lenses pinhole.  When you look up close his large prints are not really sharp at all, but sharpness is only part of artistic expression, to the most.  I believe, select samples carefully (even the MF lens variations is smaller than that of 35mm to my experience), I can comfortably work with the 645D with many old FA and A lenses at f/11 to f/16 and make great prints that is beyond my 6x7 chrome level, technically.  The Velvia color pallet, the long exposure ability (star trails, light painting, etc.) is likely a totally different discussion though.

[ErikKaffehr]

Hi,

Yes I just checked out Diglloyd's article for the second time. The 75 lens is much better. I would be interested in your experience of Pentax 67 lenses on the 645D, because I have a handful of those (45, 90, 165/2.8, 300 and the fisheye).

Are you drum scanning Velvia? The reason I'm asking was that I was doing some testing on Velvia 67 vs. full frame digital last year and my Velvia scans were not that impressive. I was scanning on a Minolta Dimage Scan Multi Pro at 3200 PPI. Not high end digital but not exactly a cheap flatbed either...

Best regards
Erik

Yes we looked over a dozen 55mm/f2.8 examples, Japanese, Russian, and even imaging-resource (the "house" shot), and they are all the same -- out of the center 2/3 (or 1/2 as Lloyd said today) it is not sharp at all.

However, he did find the 75mm OK to the sensor, and online we could see good examples of corner-to-corner sharpness off the 45-85mm zoom and the 120mm macro lens.  We were also quite amused that there were no 55mm examples in the Pentax's own official 645 sample images.  Just as he said, how could Pentax released such a lens, and termed it "outstanding", is well beyond me.

On the other hand, I am a practical photographer and I do need DOF with MF in the real world.  I modified a 55mm P67 lens with fixed ~1 degree down tilt to help me in near-far type wide angle compositions just to deal with the DOF issue, for example.  With film f/16 to f/22 was the norm, and I was aware loss of sharpness due to diffractions.  However the softness can be at least partially corrected in sharpening, while the out-of-focus objects are hopeless.  For the same reason Ansel started the f/64 school and virtually turned every of their lenses pinhole.  When you look up close his large prints are not really sharp at all, but sharpness is only part of artistic expression, to the most.  I believe, select samples carefully (even the MF lens variations is smaller than that of 35mm to my experience), I can comfortably work with the 645D with many old FA and A lenses at f/11 to f/16 and make great prints that is beyond my 6x7 chrome level, technically.  The Velvia color pallet, the long exposure ability (star trails, light painting, etc.) is likely a totally different discussion though.

[Leping]

Yes but very very few.  I found the best way for Velvia is Imacon (now Hasselblad) -- the diffused light source hide Velvia "popcorn" (actually "pepper" -- see my correction below) grains and like the 645D images the results withhold a lot of abuse (sharpening) even they look soft to begin with.  Drum scans use very harsh point light source and actually emphasize film grain, making it hard to work with.

Of course I can't afford an Imacon but I found (at least in SF Bay area where I used to live) I can rent for hours or a day, and productivity was restricted by film mounting and cleaning not scanning.  Thus I bought additional 6x7 and 4x5 film holders from B&H and keep mounting while scanning.  This way my cost is typically below $2.50 per ~450MB 16-bit scan.  Don't attempt to do anything (color and contrast correct, cropping, etc.) and just straight 16-bit with the level fixed at 0 and 255.  With 16-bit you have a lot of room later.  Of course never spend time burning CDs or DVDs, bring firewire drives and write straight.

These two works were both off my P67 300mm ED-IF.  I tested with a Canon 5DII and found all my P67 lenses resolves to the 21MP sensor limit at f/8.  The myth that MF lenses are not as sharp as 35mm ones is totally untrue.

[eronald]

I think we can all agree that the 645 outresolves the current Nikon, because that was the design goal.

It is a newer camera, pitted against a 2 year older camera! This usually means a resolution disadvantage!

Things will get interesting when the comparison matches generations.

Edmund

[ErikKaffehr]

Hi,

Great images! But not a surprise as I already have checked out your homepage.

Best regards
Erik

Yes but very very few.  I found the best way for Velvia is Imacon (now Hasselblad) -- the diffused light source hide Velvia "popcorn" grains and like the 645D images the results withhold a lot of abuse (sharpening) even they look soft to begin with.  Drum scans use very harsh point light source and actually emphasize film grain, making it hard to work with.

Of course I can't afford an Imacon but I found (at least in SF Bay area where I used to live) I can rent for hours or a day, and productivity was restricted by film mounting and cleaning not scanning.  Thus I bought additional 6x7 and 4x5 film holders from B&H and keep mounting while scanning.  This way my cost is typically below $2.50 per ~450MB 16-bit scan.  Don't attempt to do anything (color and contrast correct, cropping, etc.) and just straight 16-bit with the level fixed at 0 and 255.  With 16-bit you have a lot of room later.  Of course never spend time burning CDs or DVDs, bring firewire drives and write straight.

These two works were both off my P67 300mm ED-IF.  I tested with a Canon 5DII and found all my P67 lenses resolves to the 21MP sensor limit at f/8.  The myth that MF lenses are not as sharp as 35mm ones is totally untrue.

[mtomalty]

Leaping

What is the 'popcorn' grain you refer to when speaking of Velvia?

Mark
www.marktomalty.com

[eronald]

I think Digilloyd has proved, yet again that the full-system behavior of the 645 cameras is not always as good as their sensor. And that they work well as landscape cameras.

Edmund

[Dave Gurtcheff]

Lepig: Your two images blew me away. I do seascapes, but those two landsapes are in a different league.
Dave

[Leping]

Hi Mark,

My bad!  It is "pepper grain" not "popcorn".  I was indeed hungry but we don't put peppr on popcorn normally.  

I could not find the orginal page of the term now but here is a discussion on the issue:

http://photo.net/digital-darkroom-forum/00Ok5u

Sorry again,

Leping

P.S.: Dave: thank you for your very kind words, but the only thing hard in making such images was luck and patience, not "level" of the maker.

[BJL]

I tested with a Canon 5DII and found all my P67 lenses resolves to the 21MP sensor limit at f/8.  The myth that MF lenses are not as sharp as 35mm ones is totally untrue.

Comparing at f/8 might not tell the whole story: maybe it is another myth, but it is often said that "all lenses are equally sharp at f/8", perhaps because for most good primes (aberration effects are far less than diffraction at f/8 and above. (Maybe less true for LF lenses.) So I think that the claims of higher resolution (l/mm) for good 35mm format lenses refers to their optimum resolution, occurring below f/8.

But on the other hand, I suppose that the great majority of high resolution photography done in formats like 6x7 and larger is done at aperture ratios of f/8 and up, where the lenses are close to "diffraction limited" (which is a good thing, despite many opinions to the contrary!) The ability to more often use lenses at aperture ratios high enough to be more or less diffraction limited is one optical advantage of a larger format.

And if you are seeing "sensor limited" resolution at f/8 and 21MP, there is clearly room for increasing overall resolution in 35mm format by moving to more, smaller photosites, like those of the recent 16-18MP sensors of some recent DSLRs.

[ErikKaffehr]

Hi,

From what I have seen from my Pentax 67 lenses I'm not really of the impression that they have really good MTF at 80 lp/mm (corresponding to 6 micron pitch). Using Imatest I estimated 2271 LW/PH for my Pentax 67 with 90/2.8 lens using Velvia and 2890 LW/PH for the Sony Alpha. This pretty much corresponds to my visual impression of scanned Velvia from the P67. I also tried microphotography using an enlarger lens and got results very close to the scanned slides.

I general I see a lot of color fringing on scanned Velvia with all lenses I have (45/4.5, 90/2., 165/2.8 and 300/4). I did not test the P67 lenses on a Sony 900 or Canon 5DII body. But from what I have seen from slides I would not expect that good resolution.

The article here contains a lot of samples: http://echophoto.dnsalias.net/ekr/index.php/photoarticles/16-pentax67velvia-vs-sony-alpha-900

Best regards
Erik

Comparing at f/8 might not tell the whole story: maybe it is another myth, but it is often said that "all lenses are equally sharp at f/8", perhaps because for most good primes (aberration effects are far less than diffraction at f/8 and above. (Maybe less true for LF lenses.) So I think that the claims of higher resolution (l/mm) for good 35mm format lenses refers to their optimum resolution, occurring below f/8.

But on the other hand, I suppose that the great majority of high resolution photography done in formats like 6x7 and larger is done at aperture ratios of f/8 and up, where the lenses are close to "diffraction limited" (which is a good thing, despite many opinions to the contrary!) The ability to more often use lenses at aperture ratios high enough to be more or less diffraction limited is one optical advantage of a larger format.

And if you are seeing "sensor limited" resolution at f/8 and 21MP, there is clearly room for increasing overall resolution in 35mm format by moving to more, smaller photosites, like those of the recent 16-18MP sensors of some recent DSLRs.

[Leping]

Hi Erik,

Very interesting post, but I think you did not pick the sharpest P67 lenses: from my tests the 45mm is not as sharp as the 55mm/f4 (new style), the 90mm  and the 300mm/f4 non ED IF are not the sharpest as well (the 300mm ED IF is extremely sharp and the 55-100mm zoom easily beats the 90mm at f/8, and sorry I did not own the 165/2.8 )  

A while ago I posted a test chart shot showing my copy of the P67 90-180mm zoom resolved to approximately the 5DMKII's sensor limit at 135mm f/8:

http://forums.dpreview.com/forums/readflat.asp?forum=1021&message=31210966

and after tests (on film) I put away and sold most of my P67 prime lenses since the two zooms are indeed that good (except for visual focusing uncertainties with the 90-180mm since the f/5.6 maximam aperfure).

Thanks,

Leping

[happyman]

....you only need factors like 135% to turn DSLR images into really ugly...

Cough.

[Leping]

Cough.

I should have said Bayer sensor based DSLRs with AA filter, ugly looking on screen at 100%.  One other common problem is thickening of single pixel details into a couple of pixels -- you usually don't realize until checking at the same shot with MF gear or scanned film.

Thanks.