Apparently size really does matter or does it . . .

[lars_bergquist]

Advertisement (gone after registration)

Dynamic range and tonal resolution are very, very important. I hope Leica can find more powerful processors for handling true 16 bit files, or for better tonal compression. The idea of no-linear tonal compression is good, but the 8-bit compression of the M8 seems to me excessive. The M8 and DRM are very good cameras, so they don't need to invent the wheel but only improving. Newer technologies will help here.

 

There has been a very extensive discussion of this. It seems to me that the demand for 16 bits is more due to quasi-religious numerology than real need. I think it has been pretty clearly demonstrated that with a linear 16 bit conversion you would gain nothing in the highlights and lose much in the shadows. Photography is one of the few fields where the odious old adage that "what you can't see won't hurt you" is actually true.

 

The old man from the Age of Tri-X

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[rosuna]

There has been a very extensive discussion of this. It seems to me that the demand for 16 bits is more due to quasi-religious numerology than real need. I think it has been pretty clearly demonstrated that with a linear 16 bit conversion you would gain nothing in the highlights and lose much in the shadows. Photography is one of the few fields where the odious old adage that "what you can't see won't hurt you" is actually true.

 

The old man from the Age of Tri-X

 

Due to this I think some kind of non-linear tonal compression is a good idea.

[sandymc]

I think it has been pretty clearly demonstrated that with a linear 16 bit conversion you would gain nothing in the highlights and lose much in the shadows.

 

Ummm, Lars, I may be misunderstanding you here - are you suggesting that a linear 16-bit image would be be "worse" in some way than the level compressed version of the same image? If so, you're going to have to explain......

 

Sandy

[carstenw]

That isn't right. The 8-bit mapping of the M8 favours the shadows, unlike normally encoded files, but the 16-bit linear files still have more detail in the shadows. They just have a lot more in the highlights.

 

I would also like to see a linear or encoded 16-bit file. I sometime imagine that I can see the difference, but it is hard to know when you have nothing to compare to. I don't see any reason for Leica not to add this possibility. I know that it would cut battery life and storage space, but I prefer to make the choice myself.

[Guest guy_mancuso]

I honestly don't care about battery life for a second nor storage space. There both cheap enough to replace more often. rather have my 16 bit files and have my camera act more like a MF back camera with all the image benefits i can squeeze out of it. I don't want a 35mm cmaera , i want one that acts like a big brother. If you understand my meaning

[mjh]

the only 6.8 micron sensor that I would trust for skin rendition would be the Phase One P30.

But the P 30’s sensor is the same Kodak KAF-31600 that’s in the Hasselblad H3D-31, so as far as the sensor is concerned, there shouldn’t be any difference. And of course, the 39 MP KAF-39000 is quite similar to the KAF-31600, except that it has no microlenses.

[mwilliamsphotography]

Advertisement (gone after registration)

Ed, I'm using almost exactly the on the job gear spread you outlined. H3D/31, H3D/39, CFV on a 230FE & 503CW, M8, DMR, and Canon 1DsMKII, 5D ... both recently sold and replaced with a 1DMKIII for now.

 

I cannot delve deeply into all the tech aspects as well as those that have already posted. All that I can relate are the actual results for commercial work and wedding photography.

 

Those experiences exactly mirror that of the Pro friend you quoted ...

 

" ... but the increased resolution afforded by the 39 MP back is so extraordinary and he would never part with it. Yet still loves his 16MP CWD ..."

 

Images from the two H3Ds look remarkably similar. The 31 camera is faster than the 39 and features a higher maximum ISO ... the 39 meg camera allows bigger enlargements or more severe crops, but not by a lot ... but it offers ISO 50 where the 31 does not.

 

While I love the CFV, I don't use it for any critical commercial work ... I could, but prefer the huge resolution advantage of the H3D ... on which I sometimes use the Zeiss V lenses via the CF adapter.

 

I personally prefer images from the two Leicas, over all the Canons, even when I use R glass on Canons. However, I need AF and excellent flash interface for a lot of event work.

 

As of a few days ago both H cameras will be going in for an upgrade to the H3D-II which further expands their performance on a number of fronts.

 

Sorry, no tech explainations, just practical experience having heavily shot all these cameras on a broad variety of works.

[mwilliamsphotography]

Here's an example of "Available Resolution". A quick demo of a H3D/31 cull file from a wedding. Don't know how this will come across using web compressions ... but maybe will give some idea of the concept.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[lars_bergquist]

Ummm, Lars, I may be misunderstanding you here - are you suggesting that a linear 16-bit image would be be "worse" in some way than the level compressed version of the same image? If so, you're going to have to explain......

 

Sandy

 

Very well. It would mean that half the available bit-space would be wasted on simply the last f-stop before highlight burnout. Very little would be left for differentiating the shadows. Leica's (and Adobe's) scheme means that the conversion is more 'linear' in a visual sense, which I think is the important sense: The bit-space is more evenly distributed over the aperture gamut.

 

The scheme is that there is a table of available gray scale values. The sensor signal is coded as positions in this table. And this takes only eight bits. As long as the number of gray scale steps is all that we can discriminate between, nothing is lost, and much is gained, in processor time, writing speed and storage space.

 

The old man from the Age of Tri-X

[sandymc]

As long as the number of gray scale steps is all that we can discriminate between, nothing is lost, and much is gained, in processor time, writing speed and storage space.

 

Lars, ok, I agree with that - but I think that the issue is in the "all that we can discriminate between" part. Firstly, some people would claim to be able to discriminate quite a lot, although personally I can't tell the difference between a level compressed file and a linear file if they are "normally" processed. And I've tried - see my other thread re emulating an uncompressed M8 file from a DMG file.

 

But more important, with pretty much any raw developer, you can post process heavily, effectively expanding one part of the tonal range of the original image, to the point that you would be able to discriminate. Which is why I think a lot of people would prefer at least the option of the uncompressed file. It may save your ass at some point(!)

 

Sandy

[luigi bertolotti]

Lars, ok, I agree with that - but I think that the issue is in the "all that we can discriminate between" part. Firstly, some people would claim to be able to discriminate quite a lot, although personally I can't tell the difference between a level compressed file and a linear file if they are "normally" processed. And I've tried - see my other thread re emulating an uncompressed M8 file from a DMG file.

 

But more important, with pretty much any raw developer, you can post process heavily, effectively expanding one part of the tonal range of the original image, to the point that you would be able to discriminate. Which is why I think a lot of people would prefer at least the option of the uncompressed file. It may save your ass at some point(!)

 

Sandy

 

Well, Sandy and Lars, this issue of Leica's choice for a 8 bit unlinear compression Vs. a 16 bit linear RAW I think was extensively discussed, and rather in depth... discussions about can never have an end for the obvious reason that we KNOW how the M8 performs as is, we DO NOT KNOW how would it perform with a 16 (14...?) bit (linear or NOT) coding : we can only make speculations about, based on tests on other systems. My opinion, for what is worth, is that a system which carries on MORE information CAN anyway give something "more" if not necessarly something "better" : "more" in the sense that you have more numbers to work on through software, but given that "all that we can discriminate about" is imho a real factor to consider, it has to be proved that the "more" numbers, when processed, can result REGULARLY in a "better" picture TO SEE : i say REGULARLY for I think that in some extreme cases of scenes, a full 14-16 bit conversion surely could be a better solution

[lars_bergquist]

------------But more important, with pretty much any raw developer, you can post process heavily, effectively expanding one part of the tonal range of the original image, to the point that you would be able to discriminate. Which is why I think a lot of people would prefer at least the option of the uncompressed file. It may save your ass at some point(!)

 

Sandy

 

Sandy, this is not correct. The reason why actuaL 8-bit files do not take kindly to post-processing is that this can empty some brightness or colour levels, leading to posterizing or brightness "stepping". But what you get when developing a DNG file is NOT an 8-bit file. It is an 14-bit file. I can do no better than referring you to the LFI article that discussed this matter in depth (LFI 2/2007 p. 54; the pagina of the English edition may be different). But frankly, this removes this objection too. And what is left, but the magical power of the sacred number 16?

 

The old man from the Age of Analog Technology

 

P.S. I think that answers Luigi's objection too.

[sandymc]

Sandy, this is not correct. The reason why actuaL 8-bit files do not take kindly to post-processing is that this can empty some brightness or colour levels, leading to posterizing or brightness "stepping". But what you get when developing a DNG file is NOT an 8-bit file. It is an 14-bit file. I can do no better than referring you to the LFI article that discussed this matter in depth (LFI 2/2007 p. 54; the pagina of the English edition may be different). But frankly, this removes this objection too. And what is left, but the magical power of the sacred number 16?

 

Well, Lars, there I have to respectfully disagree with you. The sequence of events is:

 

1. The 8-bit compressed DNG, which has 256 levels

 

2. The decompressed 14-bit raw file, which has 14-bit data, but still only 256 levels

 

3. The demosaiced image, which can be in whatever format - this has many levels, but the ONLY reason why it has more than 256 levels is the pixel interpolation - once demosaiced, only 1/3 of the pixel data is "real", the rest is interpolated.....

 

So your image, although its got lots of levels, is still based only on the original 256 levels, 14-bit data or not. And I'd really prefer a lot more...

 

Sandy

[lars_bergquist]

Sandy, if you can see more than 256 levels of brightness, then report to the nearest competent perception research lab. You will be well received, to say the least. In any case, the problem was never with the 256 levels as such. It was with a purely numerical effect (akin to rounding off after a division) which could collapse one or several levels into another one, leading to LESS THAN 256 levels. A 14 bit representation of the 256 levels does preclude this just as well in practice as a theoretical 16 bit representation does.

—I would contend that data that you cannot use are, hm, well, useless. If you have a zillion levels you cannot see, and a trillion of details you cannot discern, does that make your image any better? Remains only a cozy gut feeling.

—Carl Sagan was once asked if the thought there was life elsewhere in the universe. He answered that he couldn't say, there was not enough evidence. —"But don't you have a gut feeling, prof. Sagan?" Sagan's reply is famous:

—"But I try to avoid thinking with my gut."

 

The old relic from the Age of Reason

[sandymc]

Sandy, if you can see more than 256 levels of brightness, then report to the nearest competent perception research lab.

 

No Lars - in fact, if you take a look a few posts back I specifically said that I couldn't see the difference, with "normal" processing. But I sure can post process sufficiently to be able to see a difference in the final result. Point is, you lose nothing, other perhaps a bit a speed and storage space with linear files. And under certain circumstances, can gain something. So I'd like the choice, preferable as a menu selection, so that I can chose speed and size versus the ability to heavily post process.

 

Sandy

[rosuna]

A moderate tonal compression can be a good idea because: 1) the human eye cannot perceive millions of color/luminosity alternatives; 2) the pseudo-logarithmic way of archiving tonal information is ineficient. The point is not to go too far in this compression.

 

This is that Nikon does:

 

Is the Nikon D70 NEF (RAW) format truly lossless?

 

It is a 12 to 9.4 bits compression scheme (683 values instead of the 256 from the 8-bits output of the M8).

 

These are the practical differences (from 12-bits uncompressed to 9.4 compressed):

 

Jeffrey Friedl’s Blog » A Qualitative Analysis of NEF Compression

 

I suppose the differences would be more evident from uncompressed 14-bits to compressed 8-bits. In any case, it depends on the particular picture that is compressed. A safe margin would be enough.

[luigi bertolotti]

Sandy, this is not correct. The reason why actuaL 8-bit files do not take kindly to post-processing is that this can empty some brightness or colour levels, leading to posterizing or brightness "stepping". But what you get when developing a DNG file is NOT an 8-bit file. It is an 14-bit file. I can do no better than referring you to the LFI article that discussed this matter in depth (LFI 2/2007 p. 54; the pagina of the English edition may be different). But frankly, this removes this objection too. And what is left, but the magical power of the sacred number 16?

 

The old man from the Age of Analog Technology

 

P.S. I think that answers Luigi's objection too.

 

Lars, I have understood the principles of processing into M8 simply reading the posts in this forum... and didn't read the often quoted LFI article... (I studied too much digital signal processing at the University...so now I prefer to learn in "sparse" way... ). Now I seem to understand that until now maybe I have missed a final point... you say: ".. what you get when developing a DNG file is NOT an 8-bit file. It is an 14-bit file..": so the processing is (tell me if I mistake):

 

- CCD signal=16 bit (14 effective, for basic processing needs that "eat" 2 bits)

- Coded in 256 levels = 8 bit

- Processed through the compression algorithm (the famous "curve", logarithmic if I remember well)

- REMAPPED into a 14 bit space (always with 256 levels)

- And THIS is the file (DNG) processed by C1 or LR . So, the "developer session" of, say, LR works onto this 14 bit pixels.

 

Sorry for the trivial question... is it so ? I wasn't sure about the last two points...

 

Thank you a lot if You are so kind to answer me...