Exposure strategy with M8 non-linear encoding

[dhoelscher]

Advertisement (gone after registration)

The Feb. 2007 issue of LFI details in fair completeness the Leica strategy of taking sensor data, doing the A/D conversion, and multiplying by 4 and taking the square root, transforming 14 bit data into 8 bit non-linear space.

 

 

Of course, we all recognize forum members (was it Scott Kirkpatrick first?) who figured this out quite a long time ago and published this on the forum.

 

 

My question is this: how does knowledge of this mechanism influence our choice of correct exposure for M8 images?

 

Michael Reichmann has famously stated the "expose to the right" rule for "conventional" cameras that take 12 bit data and don't non-linearly encode - like Canons and Nikons - because most of the "bits" are in the upper tones (I'm poorly summarizing this).

 

Knowing where the "juicy" richness of tonal separations are in this different strategy (like in the shadows now according to the LFI article), do we still "expose to the right" or something else ????? I'm just a little confused with how to get the "best" data exposure-strategy wise with the M8 - which seems to be a bit of a different beast than the Canons/Nikons of yore (and of great recent interest!) ...

 

 

 

Yes, I understand we still have to consider highlight clipping - because once sensor wells (the "buckets") are full - there is no more opportunity to capture light level differences - so we don't want to clip highlights of course --- but would the M8 be "happier" with a more "correct" exposure, or underexposure (realizing about confusing data with the noise floor)?

 

I await wisdom from those a lot more experienced and smarter than poor I ....

[deltoid1]

The Feb. 2007 issue of LFI details in fair completeness the Leica strategy of taking sensor data, doing the A/D conversion, and multiplying by 4 and taking the square root, transforming 14 bit data into 8 bit non-linear space.

 

 

Of course, we all recognize forum members (was it Scott Kirkpatrick first?) who figured this out quite a long time ago and published this on the forum.

 

 

My question is this: how does knowledge of this mechanism influence our choice of correct exposure for M8 images?

 

Michael Reichmann has famously stated the "expose to the right" rule for "conventional" cameras that take 12 bit data and don't non-linearly encode - like Canons and Nikons - because most of the "bits" are in the upper tones (I'm poorly summarizing this).

 

Knowing where the "juicy" richness of tonal separations are in this different strategy (like in the shadows now according to the LFI article), do we still "expose to the right" or something else ????? I'm just a little confused with how to get the "best" data exposure-strategy wise with the M8 - which seems to be a bit of a different beast than the Canons/Nikons of yore (and of great recent interest!) ...

 

 

 

Yes, I understand we still have to consider highlight clipping - because once sensor wells (the "buckets") are full - there is no more opportunity to capture light level differences - so we don't want to clip highlights of course --- but would the M8 be "happier" with a more "correct" exposure, or underexposure (realizing about confusing data with the noise floor)?

 

I await wisdom from those a lot more experienced and smarter than poor I ....[/quote

 

Easy enough to test right? Take a pic of a subject, like of a bright window inside a room, Then get heavy-handed with the file, trying to raise the shadows, See if they band.

[barjohn]

I appologize in advance if this is a stupid question but what does LFI stand for (I know it is a Leica publication but I haven't been able to find it and where can one find this artical? I have tried to find it on the Leica web site and it is not published anywhere I could find. Any help please? I would like to try understand their algorithm.

[35mmSummicron]

LFI = Leica Fotographie Internationale

and from my understanding, Leica does not publish the magazine.

 

as for the original question, i recently shot some work with 2 stops overblown highlights and brought them back 100% in C1 Pro. granted i processed them in black and white so any kind of solarizing is not noticable, but my point is i think M8 DNG files are INCREDIBLY robust and VERY flexible.

 

so in terms of shooting i would say shoot DNG and try not to blow out high lights as with other digital cameras, but i think you will find that generally, M8 files will give you headroom to recover.

 

/a

 

I appologize in advance if this is a stupid question but what does LFI stand for (I know it is a Leica publication but I haven't been able to find it and where can one find this artical? I have tried to find it on the Leica web site and it is not published anywhere I could find. Any help please? I would like to try understand their algorithm.

[chetccox]

John -

 

The above post tells you what it stands for, but in regard to the articles it is a paid subscription magazine and is not available digitally, at least to my knowledge. In this day and age you would think they would publish both a digital and a paper edition. I would subscribe in an instant if I didn't have to worry about where in the world the paper edition would find me - and when!

[LichMD]

LFI International

 

Leica Fotografie International

 

Need to order the article, it doesn't appear to be downloadable.

[tashley]

Advertisement (gone after registration)

Thanks for that - I just ordered the last three issues, looks very interesting!

 

Tim

[scott kirkpatrick]

The Feb. 2007 issue of LFI details in fair completeness the Leica strategy of taking sensor data, doing the A/D conversion, and multiplying by 4 and taking the square root, transforming 14 bit data into 8 bit non-linear space.

 

Of course, we all recognize forum members (was it Scott Kirkpatrick first?) who figured this out quite a long time ago and published this on the forum.

 

My question is this: how does knowledge of this mechanism influence our choice of correct exposure for M8 images?

 

Yes, I understand we still have to consider highlight clipping - because once sensor wells (the "buckets") are full - there is no more opportunity to capture light level differences - so we don't want to clip highlights of course --- but would the M8 be "happier" with a more "correct" exposure, or underexposure (realizing about confusing data with the noise floor)?

 

the history -- Back in October, I stuck to my guns in saying that a fixed 10MP raw file had to mean that there was a simple transformation going on taking the 14bit data into 8 when an annoyingly long list of true believers insisted that Leica Would Never Do That to us. It was a Russian programmer, Andrej Kolev, working on his own raw file development package, who simply decoded the DNG file of the HP lady at Photokina, plotted up its lookup table, and observed that the linear pixel intensities were the square of the values in the DNG file, divided by four. I guess he has moved on to solve other problems now; haven't heard from him lately.

 

Jamie Roberts likes to point out that the best thing about the Kodak chip in the M8 is that it has significanly lowered the noise floor, so there is more room to play with. Each pixel can hold 60,000 electrons before they spill over, rather than the 30-40,000 electrons that is more typical. Its noise characteristics are like those of other chips, but there is about a stop more headroom. You can use that extra dynamic range in lots of ways, so I doubt there is one rule (like "expose to the right") that will suit everybody. Using the standard aperture priority exposures from the M8, I find I am often adding one or two thirds of a stop of exposure in C1, so I think the camera is set up to protect the highlights from overexposure. If you want to end up with black and white exposures with strong blacks, that's fine. And brilliant landscapes and artificially lit scenes work well with the default exposure. I'm trying to get brilliance in the midtones, and that might require pushing exposures up a bit...

 

 

 

scott