M9: Does Phase One Know Something?

[jaapv]

Advertisement (gone after registration)

Well, that was all I was saying. There were multiple reasons for the smaller format. Optical, processsing, economical, for all I know engineering. As nobody has access to the minutes of the meeting where the final design parameters were set, who is to know what the proverbial final straw was? I have a feeling, stemming from the briefing in Hessenpark, that the optical issues were relatively easily overcome, whereas the real headache was in the IR filtering.

[sdai]

As I've posted in a previous post, the M8 and M9 generic profiles found in C1 4.8.2 (WIN) are completely different profiles but both created on July 17, 2006 according to their header information ... I've no idea what's going on there but it certainly could be a mistake, a prank, or a leak ... just anything you could think of.

 

If the "M9" is indeed a 1.33x camera using the same 6 micron Kodak sensor architecture, you'll see 13.57 MP in its spec.

 

Such a camera deserves the name M9 instead of M8.3 if it has a proper IR solution, and the refreshed electronics/software shared with the S2.

 

I highly doubt that we'll see a FF M anytime soon, because the IQ difference between a FF 24MP M and the 37.5MP S2 would be way too small and totally negligible for all practical use and reasons, which is exactly why Leica dropped the R10 in favor of S2.

[Advertisement]

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

[scaryink]

Is there any net gain in IQ when converting to 16bits if there is no additional information available within the increase bit size?

 

As an analogy it seems that in the equation 8x1=8. What am I gaining when computing 8x1x1x1 etc. Its the same information computed in a redundant fashion.

[jaapv]

The gain is in postprocessing.I'll use an analogy: If you convert to 8-bits, you put different values in little buckets standing next to one another. As soon as you start postprocessing you will be moving the values from one bucket to another. At the next processing move there are buckets with mixed values moved on etc. The end result is that a number of nuances is lost through lumping the values together in this way.

If you convert to sixteen bit, you create a large number of empty buckets between the buckets filled with your values. Now if you post process you simply fill the next empty bucket and do not contaminate the filled bucket down the line. Thus you prevent loss of data.

In practice this shows up as posterizing when postprocessing 8 bits, which does not happen as easily in 16 bits. Another indication of data loss through insufficient bit width ( I call it that to avoid the expression of bit depth which should be reserved for the capture process) is the comb-like appearance of the histogram after levels and curves. (although that can happen in 16 bits too)

[devils-advocate]

Or maybe this is just a big joke. Since PO sees the Leica S2 as competition they figured they play a joke on all the M users and Leica."

 

The Danes are a sphincter-constrictingly law abiding lot. Given that this has not insignificant legal dimensions to it, there would be no joke of this nature, except by some disgruntled ex-employee.

 

This is either the photographic iteration of the really, really, well-hung dog in the IKEA catalogue two years ago, or a sign that PO knows there is an M9 coming and allowed this to slip into an accessible part of their software. My vote is for the latter scenario.

 

This is not diminished by the fact that there is no unique content in the profile yet. PO would not make that until the camera 'went-final' which would likely be after its announcement.

[t024484]

Is there any net gain in IQ when converting to 16bits if there is no additional information available within the increase bit size?

 

As an analogy it seems that in the equation 8x1=8. What am I gaining when computing 8x1x1x1 etc. Its the same information computed in a redundant fashion.

 

Compression to 8 bits, as MJH said, is done to save time in writing the file to the memory card.

It is not a truncation, but a compression to 8 bits.

Expanding truncated 8 bits data back to 16 bits, gives no benefit at all.

But expanding data previously compressed to 8 bits back to 16 bits does a fairly good job in case of the M8 in retrieving the original data without adding too much noise.

 

Hans

[StS]

Advertisement (gone after registration)

(...)

If you convert to sixteen bit, you create a large number of empty buckets between the buckets filled with your values. Now if you post process you simply fill the next empty bucket and do not contaminate the filled bucket down the line. Thus you prevent loss of data.

(...)

 

Very nice way of explaining it, thank you.

 

Stefan

[mjh]

There were multiple reasons for the smaller format. Optical, processsing, economical, for all I know engineering. As nobody has access to the minutes of the meeting where the final design parameters were set, who is to know what the proverbial final straw was? I have a feeling, stemming from the briefing in Hessenpark, that the optical issues were relatively easily overcome, whereas the real headache was in the IR filtering.

I’m not sure. Even the latest FF DSLRs such as the D3X suffer from microlens-induced vignetting at short focal lengths. This can be overcome by stopping down to f8 or f11, or by activating the digital vignetting correction, and it is really only an issue for wide-angle lenses, but then the flange distance is nearly 19 mm longer than that of an M.

 

Anyway, if a FF sensor had been an option back then, would the bandwidth requirements have deterred Leica from using it? True, 18 megapixels rather than just 10 would have increased writing times by 80 percent, but wouldn’t that have been accepted gladly as the price for a (higher resolution) FF sensor? And there would have been options: Leica could have offered a virtual 1.33x crop option much like Nikon does now, even with a CCD (which has to be read-out in its entirety, but after that, the cropped data could have been thrown away). And they could have offered lower resolution raw formats like Canon does.

[wlaidlaw]

Compression to 8 bits, as MJH said, is done to save time in writing the file to the memory card.

It is not a truncation, but a compression to 8 bits.

Expanding truncated 8 bits data back to 16 bits, gives no benefit at all.

But expanding data previously compressed to 8 bits back to 16 bits does a fairly good job in case of the M8 in retrieving the original data without adding too much noise.

 

Hans

 

Hans if you were talking about 8 bit linear to 16 bit linear this would be true. But in Leica's case the 8 bits are logarithmically compressed and I believe that when you expand to 16 bits linear there is a gain, especially in shadow detail.

 

Wilson

 

PS I will have to go back and re-read the LFI article to make sure I have got this correct. Perhaps Michael would like to comment.

[jaapv]

to be fair, Michael , I dont know,but I am convinced a holistic explanation must be closer to the truth. occams razor.

[Printmaker]

It is possible that Leica was testing a camera's sensor response and sent a copy to PhaseOne for profiling. The fact that it was mistakenly included in a software package does not mean that the camera is ready for prime time. All this proves is the M9 is in development and that Leica is still working with PhaseOne for RAW processing.

 

Looking at all the excitement, Leica just got a great marketing report and valuable sales projection. They can't be too upset.

[jaapv]

I would like to add that I really love this forum. There are not many internet forums where a frivolous thread like this would morph into a mature discussion.

[t024484]

Hans if you were talking about 8 bit linear to 16 bit linear this would be true. But in Leica's case the 8 bits are logarithmically compressed and I believe that when you expand to 16 bits linear there is a gain, especially in shadow detail.

 

Wilson

 

PS I will have to go back and re-read the LFI article to make sure I have got this correct. Perhaps Michael would like to comment.

 

Lots have been written on this forum about the Square Root Compression that Leica Is using.

Hans

[farnz]

... True, 18 megapixels rather than just 10 would have increased writing times by 80 percent, but wouldn’t that have been accepted gladly as the price for a (higher resolution) FF sensor? ...

Michael,

 

I would have thought the same but in reality I don't think that this necessarily holds true; there's a close parallel with the Digilux 2. When shooting RAW files the D2 takes a long time to write the files to disk and opportunities can be missed, so now many D2 users exclusively shoot jpegs to avoid the write delay at the price of higher quality (raw) files.

 

Pete.

[scaryink]

So there was no way to increase the buffer size and keep the 16 bit depth? I have also thought it would be helpful to provide some sort of pixel binning to allow much higher iso at lower rez if need be. The files really do hold up well when enlarging.

 

I cant see how a slight increase in resolution will be that significant in practical application. The image size seems to be quite a sweet spot for most applications - especially seeing as more product is delivered exclusively for the web.

 

Again dynamic range, high bit depth and broad iso range are the really important points to me. Sure its not as sexy as ff 25mp chips, but it should would take the kit to the next level of all around usability.

 

I am constantly surprised at the range that negative film provides and would be thrilled with 10mp and the confidence of shooting with a similar dr.

[nagui]

Phase one is, may be, not happy about the new relationship with leica. So someone decide to give us a clue by capture one

 

Nag

Edited August 3, 2009 by nagui

[jonoslack]

All this proves is the M9 is in development and that Leica is still working with PhaseOne for RAW processing.

 

That's what I'd have said, and after the rumours of bad blood that has to be good news for the future.

[Jamie Roberts]

That's what I'd have said, and after the rumours of bad blood that has to be good news for the future.

 

I'd have said that too, Jono, except Phase isn't exactly known for doing things like profiles for in-development cameras ahead of time Rather the opposite: they only release a profile once the camera is actually released (and many times over the years they haven't been ready for the release!).

[jonoslack]

I'd have said that too, Jono, except Phase isn't exactly known for doing things like profiles for in-development cameras ahead of time Rather the opposite: they only release a profile once the camera is actually released (and many times over the years they haven't been ready for the release!).

 

Perhaps things are changing . . . they've beaten both Adobe and Apple to the punch with the Olympus Pen.

[Advertisement]

Hallo,

Look here: Leica M

[charlesphoto99]

Obviously a mistake. Somebody released the wrong build. Developers do not do "pranks."

 

Releasing a body dubbed M8.3 would be a nail in the coffin at least marketing wise. The next iteration will be 9 no matter the feature set.