Is the difference of DR on MFDB vs 35mm dslr discernible on print?

[EricWHiss]

....
I have many other examples where all you can tell is whether the photographer did a good job or not - and you absolutely can not tell from the final image what cameras were used - be it digital, film, etc.

Definitely some good points Lisa.  I think I remember some nice images you posted recently that were taken with a camera phone?  

My choice is to have several cameras to work with.  Recently I have been re-exploring film too.  Everything has its place.

Actually to the original poster - film might an easy way for you to explore MF cameras.

[ErikKaffehr]

Hi,

I agree with all Bernard says. The only issue I'd add that there are sample variations. I'd suggest that you check out these three articles before
spending 30K on MFDB.

http://www.josephholmes.com/news-sharpmediumformat.html
http://www.josephholmes.com/news-medformatprecision.html
http://www.josephholmes.com/news-medformatprecision.html

What Joseph Holmes has found that many MFDBs and lenses have alignement issues and rental stuff is far worse than new stuff.

This problem is in no way specific to MFDBs, Lloyd Chambers (http://www.diglloyd.com) who does a lot of testing has seen this on several of his cameras.

In a sense this seems to me like a factor in favor of the new Leica S2, having a DSLR-type body with fixed sensor should make tolerances smaller and probably more resistant to wear.

Best regards
Erik

You can look at this from both additive and substractive standpoints.

On the plus side, the P65+ certainly offers the best single frame image quality in the world right now.

On the negative side:
- Perfect focus in the field is not always easy (depending on what camera you will use the back on),
- The Rugdness of the system, battery life,... make it less usable in the wild (like the real wild at -10C without power plugs for a few days),
- MF has less DoF than 35 mm which is mostly a problem for landscape,
- You are much more limited in terms of focal lenghts,
- The gap of image quality is small compared to the best DSLRs (how small will not be settled unless somebody does a pixel peeping type of comparison),
- There are much cheaper ways to achieve much higher resolutions as soon as stitching is considered a valid technique,
- ...

Finally, and most importantly, what works for me and my applications will probably not work for you so trying out the systems yourself is key.

Cheers,
Bernard

[tho_mas]

That's not the point. The point is that MFDB files are measured without an equivalent NR so that the comparision on DXO is pointless.

Well fine, show me another comparison that prooves your point.

Which point? I am only referring to the DXO thing. But I absolutely don't care which camera has the highest DR.

[ErikKaffehr]

Jim,

It seems, after a long not always sane discussion, that answer to your question is a firm and unequivocal maybe.

My 2-cents:

- A view camera employing the "scheimpflug" effect would certainly give an advantage in depth of field in many situations.
- The same effect are possible using TS-lenses
- There are economical "tilt" or "shift" options available from Arax, see this article:
    http://www.photoclubalpha.com/2009/01/09/t...ll-frame-dslrs/
- Picture processing pipeline plays a large role on the outcome
- Renting equipment to test is a good idea, but be aware of possible alignment issues
- Read this: http://www.josephholmes.com/news-sharpmediumformat.html

An additional comment is that it would be a great service to photographers in your position if an experienced tester posted a couple of full size "raw" images from a test shot.

There was a shoot out between different formats on this site a couple of years ago, when the P45 arrived. LL offered those images for download on DVD, I still have it around. I don't think it's on sales any more.

That article is here: http://www.luminous-landscape.com/essays/back-testing.shtml

Update: I set up a link collection on the issue here: http://83.177.178.241/ekr/index.php/photoa...vs-mfdb-vs-film

Best regards
Erik

I'm currently shooting landscape as a hobby, hoping that one day I can sell the ones I consider great. I'm currently using 1ds3 and have been wondering about whether to get an MFDB + view camera. The MFDB would give me a higher res and the view camera would give me better Dof on grand scheme type of shots.

- I'm wondering whether I should wait for 1ds4, hoping it would have a higher resolution + better DR (if that's even possible?) and use either the Canon TS lens or Cambo X2

Assuming (and please confirm?) that MFDB would have a much better DR than 1ds3 or future 1ds4 (speculative?), would the edge in DR show up in prints with the current technology? I'm using a Canon ipf6100 printer at the moment but the future might offer us better printing technology too.

I guess this is all technical and at the end of the day a nicely captured shots with 1ds3 would be quite nice too and I know a few professional landscape photographers who sell their photos for a living use 5d2.

Thanks for your input / comments / advice / suggestions / thoughts.

[Dick Roadnight]

An additional comment is that it would be a great service to photographers in your position if an experienced tester posted a couple of full size "raw" images from a test shot.

There was a shoot out between different formats on this site a couple of years ago, when the P45 arrived. LL offered those images for download on DVD, I still have it around. I don't think it's on sales any more.

Erik

Sometime near Christmas, when I have an H3D11-60, perhaps we can set up a comparison test between the P65+, the H3D11-60, 5*4 and DSLR pano - preferably on the right side of the pond.

Volunteers please.

[JeffKohn]

If you search the forums for "DXO" +  "medium format"  you'll find many sage posts explaining why DXO's measurements don't make a valid comparison between DSLR files and MFDB.  The crux is that most MFDB back makers handle noise processing outside the camera in the RAW software while most DSLR makers do noise reduction on the chip - before you even get the RAW file.  This results in an apparently cleaner DSLR file and better DXO numbers than MFDB.  But once the MFDB file is run through the software the same work is done to it.  Hence, without using the recommend software,  its easy for the wrong conclusion to be made.  If one goes to the trouble to develop the RAW files in the proprietary MFDB software then a much better result for MFDB files will be seen and the comparison to DSLR is much more fair (since they already had their noise reduction and file handing done).    Since DXO does not make RAW handling software for MFDB cameras, I would guess they don't care to correct their mistake since this is not important to them and in fact even helps them by falsely showing some DSLR's to be better cameras than they are.  Another way that DXO numbers are not useful is in the DR measurement - they use a standard that's technically correct for signal processing, but not very useful to photographers.  Where they to use a different and perhaps more appropriate threshold the results would be much different.    But again I'm not posting to get into a discussion or prove anything.  If you are curious about this, do your own research.

This is not accurate. The CMOS DSLR's are not performing on-chip noise reduction in the sense you imply (except maybe some of the Sony's, although I think they fixed that in firmware after enough people complained). The on-chip "noise reduction" on CMOS is at the photosite level to remove systemic noise such as patten noise. It is not the type of noise reduction that modifies ranges of pixels to reduce random noise, and it doesn't affect detail or resolution the way noise removal of the type you're referring to does, such as what raw converters  or post-processing plugins perform (which most definitely _does_ affect image detail). So you can't use this excuse to explain away the DxO results, sorry.

[ErikKaffehr]

Hi,

To my best knowledge CMOS-sesors use a technique called "correlated double sampling" which essentially measure the cell voltages after reset (that is prior exposure) and after exposure. This eliminates the noise from the reset circuitry. Sony does employ some spatial noise reduction on chip but it can be disabled. There is little evidence of spatial noise reduction with NR off at low ISO AFAIK. This cannot be done CCD because they work differently.

Our friend Panopeeper has pretty good evidence that there is something odd with Sony ARWs, but it's not noise reduction but rather DR-extension in his view. He has written a long article on it. Canon EOS 5DII does seem to have similar phenomena. Panopeeper did not test the Nikon 3DX because no one has been kind enough to send him test raws. (Actually I owe Gábor some test shoots, too, I never seem to get around to it. Sorry Gábor). I can feel that we have a little bit to much talk and to little action on this forum.

I would guess that the advantage of MFDBs over DSLRs is simple related to larger sensor which collects more photons and also some factors related to MTF which I don't want to discuss right now.

Best regards
Erik Kaffehr

This is not accurate. The CMOS DSLR's are not performing on-chip noise reduction in the sense you imply (except maybe some of the Sony's, although I think they fixed that in firmware after enough people complained). The on-chip "noise reduction" on CMOS is at the photosite level to remove systemic noise such as patten noise. It is not the type of noise reduction that modifies ranges of pixels to reduce random noise, and it doesn't affect detail or resolution the way noise removal of the type you're referring to does, such as what raw converters  or post-processing plugins perform (which most definitely _does_ affect image detail). So you can't use this excuse to explain away the DxO results, sorry.

[EricWHiss]

This is not accurate. The CMOS DSLR's are not performing on-chip noise reduction in the sense you imply (except maybe some of the Sony's, although I think they fixed that in firmware after enough people complained). The on-chip "noise reduction" on CMOS is at the photosite level to remove systemic noise such as patten noise. It is not the type of noise reduction that modifies ranges of pixels to reduce random noise, and it doesn't affect detail or resolution the way noise removal of the type you're referring to does, such as what raw converters  or post-processing plugins perform (which most definitely _does_ affect image detail). So you can't use this excuse to explain away the DxO results, sorry.

I wrote "on chip" didn't I? and its happening there in one form or another isn't it?  In any case were DXO to compare S/N of cameras using thresholds practical to photographers and use images converted in the matched software, I don't doubt that the digital backs would triumph handily. This matches my real world tests.   You have your own real world tests right?  D

[marcs]

I've got a P65, balcony, and I'm in NYC until Saturday am.  I'll only participate if the end print is 30x40 or larger.  I abandoned the D3X several months ago.  

Still prefer the look of 4x5 color neg (of course conditions must be controlled so as to minimize all the things that can go wrong...).  Film is better for certain subjects/styles, digital for others.  Let's not forget that.

 

Bernard,

For some reason that never seems to happen. We don't have any P65 owners with access to a D3X or Canon D1sIII or Sony ALpha 900 and a balcony with a view?

Best regards
Erik

[Guillermo Luijk]

I wrote "on chip" didn't I? and its happening there in one form or another isn't it?  In any case were DXO to compare S/N of cameras using thresholds practical to photographers and use images converted in the matched software, I don't doubt that the digital backs would triumph handily.

What Jeff tried to explain you is that the so called NR performed by the CMOS electronics is totally different from the one any software will achieve over the RAW data. The first doesn't mean detail loss at all, the second does. So it's irrelevant how well the MFDB software performs NR, because it will also damage the captured information.

That is why correct comparisions are to be made using the same converter. If after conversion one camera has less noise, it has more DR. If after conversion one camera keeps more detail, then it has more detail. And that's it.

What is crazy is to compare DR (i.e. shadows noise) between cameras, and defend the approach of using different software developers with implicit different NR each. It could happen the MFDB software reduces dramatically noise, but also texture. Would you conclude then that the MDFB has more DR but the same degree of detail as the rival DSLR? No way. The correct conclusion would probably be: the MFDB has more noise, i.e. less DR, while also keeping more detail in the well exposed areas. And this can be achieved by performing the same neutral conversion on the two RAW files, not using the best software on each.

DxO provides the complete SNR response of every sensor, so you can calculate DR according to your preferred criteria (typ. 12dB), and see D3X still wins.

Regards

[ErikKaffehr]

Marc!

Thank's for your support. Unfortunately I'm in Sweden and Bernard in Japan I think. My suggestion was to make two identical identical photos of the same subject under the same conditions and make available to whoever is interested to print or evaluate. Something like I have on my site here:

http://83.177.178.241/ekr/index.php/photoa...-sony-alpha-900

Probably with original files more prominently displayed, less analysis and raw files included.

Best regards
Erik

I've got a P65, balcony, and I'm in NYC until Saturday am.  I'll only participate if the end print is 30x40 or larger.  I abandoned the D3X several months ago.  

Still prefer the look of 4x5 color neg (of course conditions must be controlled so as to minimize all the things that can go wrong...).  Film is better for certain subjects/styles, digital for others.  Let's not forget that.

[BernardLanguillier]

Thank's for your support. Unfortunately I'm in Sweden and Bernard in Japan I think. My suggestion was to make two identical identical photos of the same subject under the same conditions and make available to whoever is interested to print or evaluate. Something like I have on my site here:

Yep... Sat AM in NY will be hard to manage for me.

Cheers,
Bernard

[BernardLanguillier]

I've got a P65, balcony, and I'm in NYC until Saturday am.  I'll only participate if the end print is 30x40 or larger.  I abandoned the D3X several months ago.

Where you trying to do 30x40 prints from a single D3x file?

Cheers,
Bernard

[douglasf13]

DxO Mark apparently measures from the sensor.  Let us not forget that, due to internal reflections/flare, even the best primes only allow 11 stops or so of DR, which becomes quite a neutralizer.

[EricWHiss]

What Jeff tried to explain you is that the so called NR performed by the CMOS electronics is totally different from the one any software will achieve over the RAW data. The first doesn't mean detail loss at all, the second does. So it's irrelevant how well the MFDB software performs NR, because it will also damage the captured information.

That is why correct comparisions are to be made using the same converter. If after conversion one camera has less noise, it has more DR. If after conversion one camera keeps more detail, then it has more detail. And that's it.

What is crazy is to compare DR (i.e. shadows noise) between cameras, and defend the approach of using different software developers with implicit different NR each. It could happen the MFDB software reduces dramatically noise, but also texture. Would you conclude then that the MDFB has more DR but the same degree of detail as the rival DSLR? No way. The correct conclusion would probably be: the MFDB has more noise, i.e. less DR, while also keeping more detail in the well exposed areas. And this can be achieved by performing the same neutral conversion on the two RAW files, not using the best software on each.

DxO provides the complete SNR response of every sensor, so you can calculate DR according to your preferred criteria (typ. 12dB), and see D3X still wins.

Regards

It's very frustrating discussing this topic over and over with people who seem intent to win an argument based only on numbers when the answers are easy to see with your own eyes.  If you only had both cameras at your disposal you would clearly see that MFDB have big advantages in "usable" DR at base, while DSLR have advantages at higher ISO settings.    Many are quite misinformed about DR and why many cameras come up with high DR but low usable DR.  
Please read here: http://www.imatest.com/docs/dynamic.html#dr

I used this software to compare DR of my Leica DMR against a canon 1D3.  The canon had 12.7 stops DR according to the standard definition, while the leica had 12.6 stops - so the canon won right?  But using the imatest guidelines for "usable"  DR - see the other values in the charts top right - the Leica won by almost two stops. While the canon 1d3 had 12.7 stops according to definition (and probably close to what DXO is measuring)  it only had a little more than 8.5 usable @ the .25 ratio, and actually in the most stringent measurement only 6.6 stops of DR.  The Leica had not dropped as much from the max and was still showing 10 stops of DR at the .25 ratio measurement.    While I haven't taken the time to test my new Canon 5DmkII or my P20, I can see side by side shots show the P20 has a fair bit more usable.  What I'm saying is that while DXO measurements may be accurate according to definition they are not usable for photographers an only present an idealized case which is not good for photographers.  Better measurements can be made and the cameras do not compare the same!  

Now when I said I did my own testing, I hope you believe I was really testing!  It's a lot of work to do this right and the results only confirm what I was seeing.  Now I just trust my eyes.

[attachment=16575:Stouffer...7_Step_2.jpg]

[BernardLanguillier]

I used this software to compare DR of my Leica DMR against my canon 1D3.  The canon had 12.7 stops DR according to the standard definition, while the leica had 12.6 stops. But using the imatest guidelines for "usable"  DR - see the high value in the charts - the Leica won by almost two stops.

How is "usable DR" defined?

Cheers,
Bernard

[JeffKohn]

It's very frustrating discussing this topic over and over with people who seem intent to win an argument based only on numbers when the answers are easy to see with your own eyes.

I never made any assertions about the "usable" DR of prints. I simply pointed out that your excuse for why the D3x comes out on top in the DxO test was inaccurate.

Setting aside for a moment the definition of usable DR, I can actually find it quite believable that the MFDB have some advantage in DR. However, I think the difference is much smaller than many MFDB owners claim. The conventional wisdom that MFDB's have a huge advantage in DR may have been true a few years ago; but the CMOS sensors in DSLR's have been improving at a faster pace than the MF CCD's.

[ErikKaffehr]

Yes, that's a good point.

There are many factors involved in perception of image quality.

Best regards
Erik

DxO Mark apparently measures from the sensor.  Let us not forget that, due to internal reflections/flare, even the best primes only allow 11 stops or so of DR, which becomes quite a neutralizer.

[Anders_HK]

How is "usable DR" defined?

Cheers,
Bernard

Simple; one make appointment with a Leaf, Phase-One or other MFDB dealer in ones area and run own tests and test in actual shooting. If tests are run on basis of testing/proving within capabilities of a DSLR, then one might go blind to the difference.    However, if running tests to see the actual capabilities of high image quality of the back in normal through extreme photographic conditions - at low iso, then... it is likely similar to driving little circles with a Ferrari around a Toyota. On other hand, for high ISO, obvious a DSLR prevails.

Anyone with 20+MP DSLR that will make such test?  

Rgds
A

[EricWHiss]

I never made any assertions about the "usable" DR of prints. I simply pointed out that your excuse for why the D3x comes out on top in the DxO test was inaccurate.

Setting aside for a moment the definition of usable DR, I can actually find it quite believable that the MFDB have some advantage in DR. However, I think the difference is much smaller than many MFDB owners claim. The conventional wisdom that MFDB's have a huge advantage in DR may have been true a few years ago; but the CMOS sensors in DSLR's have been improving at a faster pace than the MF CCD's.

DR of prints? That's a whole different topic!   No I never suggested the DX3 would come out on top, only that if it did on DXO or other site, but I do feel that the information would be not very useful information.