[Leica] Re: leica digital back-initial test results

[jcb at visualimpressions.com (JCB)]

At 11:50 AM 4/10/2005, Christopher Williams wrote:

>I've read a few reviews from professionals who are now saying that
>a FF 16mp chip now surpasses lens quality. Meaning they are now seeing the
>limits in 35mm designed lenses.


It is impossible to make pixels small enough to surpass the MTF of good 
35mm film lenses. Low pass filters, built into the sensor intended for 
"film lens" use, reduces the MTF of the lens to match the MTF requirement 
of five o nine micron pixels. Good 35mm film lenses can resolve far greater 
than the five-nine micron spacing therefore Mr. Nyquist rears his ugly 
head. Digital lenses are designed to meet the reduced MTF requirement of 
digital sensors. Film lenses are designed to get all they can out of a lens 
since film, being a homogenous dispersion, has no MTF requirements.

Superb digital images are "created" by the camera firmware and subsequent 
host software, not the sensor itself. This is why certain software packages 
produce better pictures than others.

A sensor is a grid. Film lenses are not designed to lay down an image with 
resolution matching the grid spacing. Actually, a lens has to be designed 
to produce an image roughly four times less than the grid MTF in order to 
not produce ugly artifacts. Digital lenses are designed this way. Film 
lenses require low pass filters at the sensor. Camera firmware and host 
software can resolve any left over artifacts.

There's a lot of shuckin' and jivin' (interpolation) going on behind the 
scenes that digital camera makers don't tell. There's many decades of 
software expertise in adjusting digital images in order to produce 
outstanding images. Nasa, JPL, and other users have been dealing with 
digital sensors in space, medicine, etc, for a very long while. The magic 
you see in digital imaging is from the firmware/software. Yes, good lenses 
and sensors produce better pictures. But it's still the firmware/software 
that help it to actually be better. A true case of making a silk purse out 
of a sow's ear. Truly raw pixel by pixel images are ugly! When raw is 
selected on a camera, it actually means raw after the first level of 
firmware has corrected each pixel for level, interpolated data into dead 
pixels, and put the whole image into the correct color space. Then you get 
out the so called raw image for PS or whatever to massage. Host software 
knows neither the idiosyncrasies of each pixel nor the overall color 
balance of the array. Internal camera firmware does. If a camera simply 
produced a pixel data stream directly off of the sensor, you would spend a 
lifetime attempting to correct it just so that what you see looks somewhat 
like what you photographed. The image processors in cameras are VERY 
powerful. RISC processors running at 200-300 MHz.

As I said, there's a heap of shuckin' and jivin' goin' on before any pixel 
sees the outside world.

JB