645D vs 1DSmk3 vs D3X

[ndevlin]

So disregarding the mathematics and the formula, print by print does the Pentax better the 35mm DSLR's?

Kevin.

As I said in the article, at 13x19 the print from an M9 is very, very close to that from the P40+ and and 645D. Under ideal use conditions, the same would be true for the D3x, and, depending on the lens, the 5DII and 1DsIII.

At sizes where you have hit the limits of the printer's resolutions, whatever tonal transition benefits one gains from the increased  information in the MF image are probably lost. At sizes bigger than 13x19, I expect that the differences will become more readily apparent to discerning viewers. It's important to emphasize that these differences are of a subtle scale.  These are not 'screaming-in-your-face' differences.  Michael's experiment with the P65+ and the G12 a couple of years ago demonstrated this quite acutely.

I'm still trying to come to grips with he bit-depth issue.  I've always thought there was more robustness to well-exposed MF files. There is no difference that I can find, however, between a P40 and a 645D file.  Indeed, the losslessly compressed RAW files off the 645D are a lot larger than the IIQ files from the P40+ (which means principally that Phase has better compression algorithms)- but there is no secret vault of extra data hiding in there.

That said, the difference between 12 and 14 bit may be real. It would be nice if someone did some sort of real-world test of this.

- N.

[LKaven]

Once I made a comparative test between a Sony a900 with zeiss 135mm (as good as any D3x with equivalent lens).

I find this hard to believe.  The extra stop of dynamic range in the D3x amounts to nothing in your opinion?

[ErikKaffehr]

Hi,

It doesn't matter until you need it. If your subject has a limited tonal range it won't be helped by extensive DR. Also, DR is quite good on the Pentax 645D.

Another point, I have something like 50000 images on my computer, but it's not easy to find samples that really require a lot of DR.

Best regards
Erik

I find this hard to believe.  The extra stop of dynamic range in the D3x amounts to nothing in your opinion?

[ziocan]

I find this hard to believe.  The extra stop of dynamic range in the D3x amounts to nothing in your opinion?

In 99% of photos it wont matter a zic.
But if for extra stop of DR you mean a chart representation of it, ala DxO, then I think is not even worth talking about it.

[eronald]

In 99% of photos it wont matter a zic.
But if for extra stop of DR you mean a chart representation of it, ala DxO, then I think is not even worth talking about it.

On the backs and the D3x you lose one stop of DR per stop ISO. So push the back to ISO 800, expose for the highlights, and you will notice your shadow noise and shadow details getting marginal - hair shadows are a good place to look.

Edmund

[ErikKaffehr]

Hi,

I'd suggest that A3 is to small to show real difference. There are to many variables in the kind of test you have down:

- Processing to taste
- Possible focusing errors
- Capture sharpening
- Uprezzing and output sharpening

I'd made some comparisons between my Alpha 900 and my Alpha 700 doing A2 size prints and the results were quite close. I one case essentially no one could tell them apart. Two other comparisons I could easily tell which was which. Small differences in contrast may make a picture sharper, but contrast is pretty much a processing parameter. When I looked at the files there was a huge difference.

Michael Reichmann did a comparison between P45+ and Canon G10 where no difference could be seen on 13x19" prints.

http://www.luminous-landscape.com/reviews/kidding.shtml

I'm pretty sure that the Pentax 645D would have and advantage over Nikon D3X at larger size prints, if both cameras were correctly focused that is.


Best regards
Erik

You betcha!

Once I made a comparative test between a Sony a900 with zeiss 135mm (as good as any D3x with equivalent lens) and a Phase P30 with 150mm mamiya.
processed  and sharpened to taste and printed in A3.
Showed the prints to 2 photographers that shoot only dslr, 2 art directors who normally works with high end photographers and an very fine architect.
Asked which one they preferred and one of the two 2 photographers could not tell the difference, the other 4 persons picked the Phase one print.
It was not a night and day difference, but it was there and could be seen.

From the files I have seen, IMO the Pentax is tangibly better than the phase P30 and will print better than any current DSLR.

[ErikKaffehr]

Bill,

I'm much surprised when looking at your Nikon D3X plot. It has very odd behavior right of the Nyquist limit. The Pentax image is much nicer. I guess this is coming from in camera JPEG processing.

I tried to upload my plots from Imatest, but they would not show up on the forum. Still learning to use the new forum software ;.(

Best regards
Erik

Eric,

Interesting results. I don't know which images you downloaded, but for some testing on own, I downloaded D3XhRES6048F for the D3x and 645DhRES7264F.JPG for the 645. My Imatest results are shown below. The 645 resolution varies in the horizontal and vertical axis, so I chose the vertical for testing.

The problem is that the 645 image does not appear to be sharpened and the D3X image appears oversharpened to my eye, even though Imatest reports slight undersharpening. My results for MTF50 are shown:

D3x: 2750 lw/ph unsharpened, and 2818 lw/ph with Imatest standard sharpening.
645: 3021 lw/ph unsharpened, and 3773 lw/ph with Imatest standard sharpening.

Sharpening is said to be less critical with cameras lacking a low pass filter, but the 645 does appear to need sharpening. The D3x has a slightly lower MTF at Nyquist, consistent with the low pass filter, but there appear to be a lot of artifacts beyond Nyquist, which I would consider undesirable.

I think that the characteristics of the two sensors are so different that it would be difficult to equalize the variables with Imatest, where oversharpening can lead to MTF50 well beyond Nyquist consistent with artifacts. Ideally, one would use deconvolution sharpening for the D3x, but I don't see a raw file. Looking at the rendered images, the 645 appears smoother, but the magnification is greater as Bart pointed out.

Regards,

Bill

[BernardLanguillier]

In 99% of photos it wont matter a zic.
But if for extra stop of DR you mean a chart representation of it, ala DxO, then I think is not even worth talking about it.

There are plenty of cases where a one stop extra DR can make or break an image:
- morning overcast landscape scenes with a very bright area in the clouds surrounding the sun and much darker skies around,
- summer sunlit villages in Southern France with deep shadows where detail matters,
- Architecture scenes
- rock concerts
- ...

In all of these cases lifting shadows significantly is often required and the amount of noise you see there has a major impact on the perceived quality of the photograph. The recent addition of single shot HDR on DxO shows the way, information in deep shadows is becoming a key differentiator between cameras because it is the most obvious manifestation of DR.

Unless I am mistaken, you appreciate the qualities of your phaseone back in those conditions and the D3x would provide the same advantage. At least is it what I have experienced again and again shooting tens of thousands of images with one, just like the other owners on this very site.

Cheers,
Bernard

[yaya]

At sizes where you have hit the limits of the printer's resolutions
- N.

Can you explain the above sentence please?

[fredjeang]

Size bloody matters!
I don't understand why it is so difficult to accept for some.


If this guy http://www.youtube.com/watch?v=pdfUCQjYPCQ  would try to do this work at this enlargement's sizes with a D3 or 1D MK2 it just wouldn't be possible with the today's tech without loosing too much quality and enter into unacceptable compromises.

Maybe a little less of reviews beleifs, brands supporting and lab measurements...

[ErikKaffehr]

Hi,

If the printer cannot resolve the detail you won't see it. It's a bit more complex than that, because you really need to sharpen for output, so an additional halo is added to the image.

The sharpening workflow is essentially

- Capture sharpening, and this should be quite agressive on a OLP-filtered camera
- "Creative sharpening"
- Resizing for output
- Output sharpening

After which the image is will be sent to the printer that does error diffusion dithering before ink hits the paper.

Just to explain that many things happen until the image is transferred to print...
Best regards
Erik

Can you explain the above sentence please?

[bjanes]

Bill,

I'm much surprised when looking at your Nikon D3X plot. It has very odd behavior right of the Nyquist limit. The Pentax image is much nicer. I guess this is coming from in camera JPEG processing.

I tried to upload my plots from Imatest, but they would not show up on the forum. Still learning to use the new forum software ;.(

Erik,
 
I agree that the Imatest results for the D3x are odd and likely related to processing of the JPEGs, probably sharpening. The Pentax resolution series had a raw file (DNG), but I was unable to locate a raw file for the Nikon D3x series so that I could do my own processing.

The new software has a number of frustrations. If you use the insert image icon on the toolbar, it merely inserts an image tag in the file for linking to an image on your own server. It is easier to use the attachment option at the bottom of the screen, as shown below. For anything including text or line graphics, it is best to use GIF or PNG. Unfortunately, the image does not appear in the preview and you have no control over the location of the attachment on the screen.  

Regards,

Bill

[yaya]

Hi,

If the printer cannot resolve the detail you won't see it. It's a bit more complex than that, because you really need to sharpen for output, so an additional halo is added to the image.

The sharpening workflow is essentially

- Capture sharpening, and this should be quite agressive on a OLP-filtered camera
- "Creative sharpening"
- Resizing for output
- Output sharpening

After which the image is will be sent to the printer that does error diffusion dithering before ink hits the paper.

Just to explain that many things happen until the image is transferred to print...
Best regards
Erik

Thanks Erik, I think I just misread Nick's post and did not realise he was talking about down-scaling

I know the ins and outs of printing, having worked in the imaging/ pre-press industry for a long time in both R&D and field positions...

BTW there is no such thing as "capture sharpening" in MFDB...there might be some basic parameters tagged to the raw file that the raw converter will use (or not) but there is no sharpening that is done in the capture stage

Creative sharpening, IMO means nothing if it does not relate to the final output size and medium

Yair
Leaf Imaging

[ErikKaffehr]

Hi,

The sharpening workflow I described has been "invented" by the late Bruce Faser and advocated by Jeff Schewe, among others. Capture sharpening is not done in the camera but on raw import. The idea with capture sharpening is to "restore" contrast that has been lost in the capture process.

Creative sharpening is local enhancement of sharpness.

Output sharpening is done after resizing for print. This is the way Lightroom works.

Best regards
Erik

Thanks Erik, I think I just misread Nick's post and did not realise he was talking about down-scaling

I know the ins and outs of printing, having worked in the imaging/ pre-press industry for a long time in both R&D and field positions...

BTW there is no such thing as "capture sharpening" in MFDB...there might be some basic parameters tagged to the raw file that the raw converter will use (or not) but there is no sharpening that is done in the capture stage

Creative sharpening, IMO means nothing if it does not relate to the final output size and medium

Yair
Leaf Imaging

[bjanes]

I know the ins and outs of printing, having worked in the imaging/ pre-press industry for a long time in both R&D and field positions...

BTW there is no such thing as "capture sharpening" in MFDB...there might be some basic parameters tagged to the raw file that the raw converter will use (or not) but there is no sharpening that is done in the capture stage
Yair
Leaf Imaging

Yair,

I see that you are with Leaf imaging and presumably have some expertise with digital backs. I understand that MFDBs need less capture sharpening than dSLRs, since the former usually have no low pass filter. However, my Imatest analysis of images from the Pentax 645D indicate that some capture sharpening is needed. Please document your assertion with data from your backs.

Regards,

Bill

[Dennis Carbo]

I think what YAYA is saying is that MFDB actual raw files are not sharpened during capture by the back  - ever,
not that they dont need it applied by the capture software - just that the RAW data is actual RAW un-molested data.

I believe nikon and canon dslr Raws do get some minor processing of some sort in camera - I may be mistaken however ? 

[ondebanks]

AFAIK the only thing that is ever done to RAW files is that sometimes the electronic offset (bias) level is subtracted; Nikon does this, Canon does not. Canon is correct it its approach - "un-molested" data, as Dennis nicely puts it.

How could one sharpen a RAW image, when adjacent pixels are sampling different, independent colours?

[yaya]

Hi,

The sharpening workflow I described has been "invented" by the late Bruce Faser and advocated by Jeff Schewe, among others. Capture sharpening is not done in the camera but on raw import. The idea with capture sharpening is to "restore" contrast that has been lost in the capture process.

Creative sharpening is local enhancement of sharpness.

Output sharpening is done after resizing for print. This is the way Lightroom works.

Best regards
Erik

Adobe's workflow is only one of many...what you call Capture sharpening is a set of parameters applied to the preview in the raw converter (e.g. Lightroom). These will be provided by the software provider, sometimes after working/ testing with some files and sometimes as a "reasonable compromise"

In some workflows (for example in Leaf Capture or in the old Creo scanning software, oXYgen) the sharpening parameters will adjust automatically according to the output size if you work with a preset or you can fine tune them manually. Every preset comprises of 5 different parameters in addition to 3 channels to choose from. This way your "Output sharpening" is done WHILE resizing and not AFTER...you get a 100% preview of the scaled AND sharpened image...it'll only be applied to the image once the raw file is exported.

Naturally, those presets are geared towards commercial CMYK print and are created during extensive testing and years of field experience. I assume you already know that in general, a printed image requires more sharpening than an image that is viewed on a screen etc.

You can download Leaf Capture and some raw files from our website and try it for yourself

Cheers, yair

[yaya]

Yair,

I see that you are with Leaf imaging and presumably have some expertise with digital backs. I understand that MFDBs need less capture sharpening than dSLRs, since the former usually have no low pass filter. However, my Imatest analysis of images from the Pentax 645D indicate that some capture sharpening is needed. Please document your assertion with data from your backs.

Regards,

Bill

Hi Bill,
I don't recall any such assertion that I've made here and surely you are not expecting me to answer for Pentax, right?

But yes I think that the general consensus is that MFDB images will typically be sharper than those taken with DSLRs due to lack of AA filtering, pixel/ sensor design and sensor-specific characteristics (Dalsa Vs Kodak etc.). This of course is subject to many variables and should not be taken as a 100% fact.
If you would like to see some Leaf raw files you can find a few on our website. If you are after a specific type or subject/ conditions/ camera used please contact me directly off-line and I'll help you

Cheers

Yair
ysh@leaf-photography.com +44 77 8992 8199

[eronald]

Yair -

 It's a theory -
 The idea is to deconvolve the input in a first pass (focus errors, lens aberrations, sensor cover glass, body flare), to use esthetic and physiological sharpening in a second pass for emphasis (eg clarity), and then before printing factor in a sharpening that will compensate for the printer's loss of sharpness at the printed scale.
 So, it's an approach, a flowchart. You can like it or you can dump it. However it has little to do with Adobe, or Leaf, or Pentax, or Nikon, aside from the fact that some manufacturers like Hasselblad and Canon are better than others at measuring their equipment in-house and providing deconvolution algorithms for their cameras and lenses and sensors. I still remember how I almost fell off my chair when using the Canon algorithms in  DPP the first time - it was as if someone had handed me a new camera.

Edmund

Adobe's workflow is only one of many...what you call Capture sharpening is a set of parameters applied to the preview in the raw converter (e.g. Lightroom). These will be provided by the software provider, sometimes after working/ testing with some files and sometimes as a "reasonable compromise"

In some workflows (for example in Leaf Capture or in the old Creo scanning software, oXYgen) the sharpening parameters will adjust automatically according to the output size if you work with a preset or you can fine tune them manually. Every preset comprises of 5 different parameters in addition to 3 channels to choose from. This way your "Output sharpening" is done WHILE resizing and not AFTER...you get a 100% preview of the scaled AND sharpened image...it'll only be applied to the image once the raw file is exported.

Naturally, those presets are geared towards commercial CMYK print and are created during extensive testing and years of field experience. I assume you already know that in general, a printed image requires more sharpening than an image that is viewed on a screen etc.

You can download Leaf Capture and some raw files from our website and try it for yourself

Cheers, yair