Just 10Mp??? What was Leica thinking?

[jaapv]

Trust me, over the past 15 years I have made a very good living as a professional, manipulating more scanned film and digitally captured images than than I care to admit in polite company. 😉 We are not there yet, but it's just over the horizon.

Digital certainly has the edge in the noise area, no argument there, but it still has a way to go in terms of exposure latitude. Note how I say latitude, not over all exposure range.

Let's hope Leica IS the horizon 🙂 I bow to your expertise, but I think you'll agree that latitude of the end print is one of the bottlenecks.

 

[Pherdinand]

Jaap, the idea with a wide latitude on the film is, that you have it all there recorded, and you can print the range you want. At least that's what it means to me.

 

[jaapv]

Jaap, the idea with a wide latitude on the film is, that you have it all there recorded, and you can print the range you want. At least that's what it means to me.

You're right, and that is the way RAW works as well, with the added bonus of exposing different elements of the image differently.If the latitude of RAW would be larger, as it will be in the future, that would give more scope in that respect. But it will also lead to sloppy shooting, as so many "advantages"of digital do. It may be a corny and well-worn cliche, but I firmly believe, both in film and on digital, that, ideally, the final print should be visualised at the moment of pressing the shutter, also in repect to the exposure.

On the subject of sloppy, I hate the term "digital" for sensor photography. The only truly digital medium is film, as a Halide crystal is either activated or not and, after developing, present or not, whereas a sensor is a collection of analog devices. But I fear I am, as we say in Holland, rowing against the stream here.🙁

 

[varjag]

If you look deep enough, all the world is discrete, being made from quantum compounds 🙂 In film though, a single halide crystal is far less in size than minimal resolving unit: the grain that one can see is not single crystals but enormous clumps of them.

 

[jaapv]

That is more philosophy than science, Varjag, as those two blend when we descend into subatomics.....

 

[Socke]

When we've reached a point where we have to proof the superiority of film on a quantum level, we might as well take a shot from Schrödingers cat 🙂

 

[varjag]

Black Schrödinger's cat in black hole? 😛

 

[fgianni]

When we've reached a point where we have to proof the superiority of film on a quantum level, we might as well take a shot from Schrödingers cat 🙂

You mean from inside the box?

Why not: you have a cat in a box, and a camera with the shutter connected to a button near the cat's paw, the longer the cat stays in the box, the more likely the cat has tripped the shutter and taken a picture of himself.

So we need a wave function that describes both the presence and absence of the cat's photograph.
With the added bonus that no cat dies (except of starvation if you leave them in the box long enough).

 

[jaapv]

Not just Erwin Puts, Erwin Schrödinger as well..... I side with
Stephen Hawking

 

[ampguy]

isn't that...

isn't that...

the nyquist theorum?

Unfortunately it can't. Quantization/Shannon sampling theorem tells us that the sampling interval (in our case, pixel size) should be choosen to be less or equal to half of the smallest interesting detail in image. I.e. you need 2 pixels to represent a line, and 4 pixels for pair of adjacent lines of same width. Quantization theorem is a fundamental part of information theory and has major consequences for signal transmission, sound encoding (e.g. you need 22KHz to capture typical 11KHz spectrum of human voice) and other everyday practical applications.

 

[Bob Ross]

Unfortunately it can't. Quantization/Shannon sampling theorem tells us that the sampling interval (in our case, pixel size) should be choosen to be less or equal to half of the smallest interesting detail in image. I.e. you need 2 pixels to represent a line, and 4 pixels for pair of adjacent lines of same width. Quantization theorem is a fundamental part of information theory and has major consequences for signal transmission, sound encoding (e.g. you need 22KHz to capture typical 11KHz spectrum of human voice) and other everyday practical applications.

I will bow to your math here, because my math degree warrantee expired in 1965...😀 Please check my numbers, if you would. With the 6.8µm pixel we would get max 147 pixels per mm (1/6.8µm) and at 4 pixels per line pair, we would get 36.75 line pairs per mm or 933.45 lpi (again max). How does this reconcile with the resolution charts on the various reviews? My own experience with an Oly E-1 that has the same size pixel would favor the two pixel per line pair approach give an efficiency of 78%. (147/2=73.5; 73.5 * 25.4=1866.9; 1866.9*78%=1456.18 = approximately what shows on rez charts on review sites). I realize that this is sort of mechanical vs. quantum, but can we guess at what the photographer might see?
BTW, I notice that you have a Contax II. That was my first 35mm (in 1961) given to me together with a 50mm and a high speed 28mm f/8.0..🙂 by someone who bought an M3.

 

[Bob Ross]

Socke, I realized after writing my blurb on inkjet, that youknow way more than I ever will about prepress and high end printers. Going from the darkroom trays to the inkjet for hobby printing has been an interesting transition.

 

[Socke]

Socke, I realized after writing my blurb on inkjet, that youknow way more than I ever will about prepress and high end printers. Going from the darkroom trays to the inkjet for hobby printing has been an interesting transition.

I dug out a book on digital colour printing just to find out that screening angles, dot forms (they vary too), clustered or stochastic raster and halftone frequencies and their influence on what we see in a print are not mine to translate into english 🙂

It all starts with the often confused pixel per inch in scanning and dot per inch in printing. Then add to that that we want every dot of the 300 per inch representing a halftone value from 0 to 256 in colour!

At a pixeldepth (my poor translation) of one bit (either full colour or nothing) you need a 16x16 matrix to create 256 shades of a colour. Starting at 300dpi we are down to 18.75lpi. And that's just enough for greyscale images since the human eye can discern between some 150 greys.

Lasermaster used drums from Toshiba intended for A3+ copyers and exposed them with a more acurate laser and used a very fine grained toner to get to 1200dpi, I never believed the 1800 dpi printer was better than the 1200 dpi one, under a loupe (we call them "Fadenzähler" a loupe made to evaluate screening etc.) it didn't show a higher resolution.

We had two customers who bought recycled toner cartridges from the company they bought their copyer toner at. The cartridges looked the same and worked in the printer, up to 600dpi. At 1200 they got a real mess.

The modern inkjets print wet in wet, so they can create shades of a colour in one dot, unluckily the accuracy is severly limited in the vertical direction, the printhead traveling from side to side can be controlled much more acurate than the paper traveling under the printhead. Thus there is a limit to the number of halftones created by wet in wet printing since we can't hit the same place several times.
As far as I know, the Indigo uses 8 2 picolitre drops to create one pixel which means every pixel has 64 shades of the colour used, mixing colours you get a slightly higher count depending on the purity of the colour, i.E. 64 shades of either cyan, magenta, yellow or black. That's why they use seven instead of four colours, adding a light cmy to the normal cmy they can create more than 64 shades of any colour, although I don't know how much.

 

[Harry Lime]

Jaap, the idea with a wide latitude on the film is, that you have it all there recorded, and you can print the range you want. At least that's what it means to me.

Exactly.

The more information that is captured at the moment that the image is taken, the better off you are later on.

 

[Bob Ross]

Thanks for the information Socke. There was an article on Luminous Landscape about commercial printing and prepress. At least it introduced me to the various terms and concepts that live in the land of CYMK..🙂

 

[JohanV]

...and my daily car has been pushed into second place by my TR4 these summer days, which, even if cars have evolved dramatically the last 40 years, drives as well, but with more character. Morgans are still being sold.

Totally OT, I know, but do the people who love Leica have this nostalgic treat that makes us also go for classic cars?
I sold my Jag E type and MG TC last year, to buy, wait for it...
a brand new Morgan Roadster.
If I find the time, I'll open a thread about this somewhere else...

 

[varjag]

I will bow to your math here, because my math degree warrantee expired in 1965...😀 Please check my numbers, if you would. With the 6.8µm pixel we would get max 147 pixels per mm (1/6.8µm) and at 4 pixels per line pair, we would get 36.75 line pairs per mm or 933.45 lpi (again max). How does this reconcile with the resolution charts on the various reviews? My own experience with an Oly E-1 that has the same size pixel would favor the two pixel per line pair approach give an efficiency of 78%. (147/2=73.5; 73.5 * 25.4=1866.9; 1866.9*78%=1456.18 = approximately what shows on rez charts on review sites). I realize that this is sort of mechanical vs. quantum, but can we guess at what the photographer might see?

That's actually interesting. A simple example illustrating Shannon theorem (also known as Nyquist as ampguy pointed out) would be to imagine what happens when we have a line 1 pixel wide, and the boundary for line pair falls in the middle of pixel: we lose it. Hence the detailwidth/2 sampling (or pixel size) criterion.

Now, if you count line boundary as a line pair, it could be represented by 2 pixels. I have certain doubts on how correct is that though, since for a typical chart it should result in moire pattern at this level of detail.

Also, there is certain subjectiveness in chart measurement. With some of lines you might get lucky and they'll fall on the pixel rows perfectly, making impression the sensor "nearly" resolves this level of detail, while other lines (esp. in different spatial directions) might be far less clear. The rest is issue of enthusiasm and magazine's affiliation with camera manufacturer 🙂

BTW, I notice that you have a Contax II. That was my first 35mm (in 1961) given to me together with a 50mm and a high speed 28mm f/8.0..🙂 by someone who bought an M3.

It's a nice camera, although it is older than my father by a decade 🙂 Apparently it was used by press or wedding photographer, as serrated limbs are nearly polished down with fingers and aftermarket flash synch socket also seen some use. Picked it up fairly cheap, to replace a Kiev body that I wore out, amazing that it still works.

 

[jaapv]

Totally OT, I know, but do the people who love Leica have this nostalgic treat that makes us also go for classic cars?
I sold my Jag E type and MG TC last year, to buy, wait for it...
a brand new Morgan Roadster.
If I find the time, I'll open a thread about this somewhere else...

I know about Morgans, I sold mine of 20 years some time ago, as the dog didn't fit in it....

 

[JohanV]

You must have a huge dog, I even manage to fit my mother-in-law in there...

 

[jaapv]

A Komondor, pretty large, sadly deceased now....