Medium format versus digital sharpness

Q.G. de Bakker wrote:

Pitty that people rather believe their convenient consumeristic instant
gratification bull ****, rather than open their eyes and see for themselves.

What makes you believe that this isn't exactly what they do? Just because they
don't agree with you they are fooled, right?

Matthew Winn wrote:

That's 102 line pairs per millimetre, which is at least double what
I'd expect for average consumer film because the ability to retain
contrast down to the level of individual grains is poor. [...]

Shows how much nonsense is passed around in these digital vs film
discussions.
I'm sorry i got involved in one - i know i shouldn't and normally i don't.
Serves me right...

Quite normal, 'average' films, like Kodak's T-Max or Portra, the staple of
many photographers, resolve well beyond 102 lp/mm.
And laying that much down on film is not a tall order for even quite
'average' lenses either.

But that doesn't suit the how-good-digital-capture-is camp at all.
So they forget about the soft focus filters that are in front of their
sensors. they forget about the small size of their sensors, and still think
that because the file that comes out at the end contains so many pixels, it
must be very good.

And converseley, a 'sensor' that has a much higher resolution, is larger and
thus has more area to capture even more detail to boot, can not be as good,
because it isn't 'digital' to begin with.
Or rather would make these expensive dslrs look bad, and we can't have that,
can we? So we won't.

Dear, oh dear, oh dear...

On Sun, 7 Oct 2007 20:49:32 +0200, "Q.G. de Bakker"
wrote:

Quite normal, 'average' films, like Kodak's T-Max or Portra, the staple of
many photographers, resolve well beyond 102 lp/mm.
And laying that much down on film is not a tall order for even quite
'average' lenses either.

In which case the resolution of the system would be around 74 lp/mm.
Moreover, only the slowest traditional black and white films approach
that 102 lp/mm number. Portra 160NC is generally rated around 75 lp/mm.
Source:

http://aic.stanford.edu/sg/emg/libra...resolution.pdf

Finally, if you are trying to determine megapixels, with film that
implies scanning, with the inherent sampling losses of that technology.
If you don't scan and try to compare print quality, you have to try to
figure out how to equalize the analog vs. digital print processes.

Don't get me wrong. I'm not claiming that even the best 24x36mm dSLR
can match a 6x7 film camera. But trying to reduce the difference down
to a number of megapixels is not a simple exercise, nor to my mind is it
a useful exercise.

--
Michael Benveniste --
Spam and UCE professionally evaluated for $419. Use this email
address only to submit mail for evaluation.

On Sun, 7 Oct 2007 20:49:32 +0200, "Q.G. de Bakker"
wrote:

Matthew Winn wrote:

That's 102 line pairs per millimetre, which is at least double what
I'd expect for average consumer film because the ability to retain
contrast down to the level of individual grains is poor. [...]

Shows how much nonsense is passed around in these digital vs film
discussions.
I'm sorry i got involved in one - i know i shouldn't and normally i don't.
Serves me right...

Quite normal, 'average' films, like Kodak's T-Max or Portra, the staple of
many photographers, resolve well beyond 102 lp/mm.

I don't consider Portra an "average consumer film". Not many average
consumers keep their film refrigerated.

It depends what you mean by "resolve". Kodak's own MTF graph for
Portra 160NC-2 cuts off at about 80 lp/mm, at which point it's at 50%
for the blue-sensitive layer and 25% for red. Extrapolating, at 100
lp/mm the best the film can manage is about 35% in the blue-sensitive
layer. 160VC-2 is a little better, managing around 40% at 100 lp/mm
for blue although the red is down at 15%. This is what I meant by
the gradual roll-off of resolution meaning that an exact figure is
meaningless. If you're photographing a subject with low contrast,
then once you've taken the lens into account the detail that might
be retained by the film at 100 lp/mm is getting lost in the noise.

I've standardised on 160VC-2 -- I think it's the best colour negative
film Kodak have produced since Vericolor III S -- but it's clear that
the image is getting mushy at 80 lp/mm even with the best prime lenses
I have.

And laying that much down on film is not a tall order for even quite
'average' lenses either.

But that doesn't suit the how-good-digital-capture-is camp at all.
So they forget about the soft focus filters that are in front of their
sensors. they forget about the small size of their sensors, and still think
that because the file that comes out at the end contains so many pixels, it
must be very good.

I am wondering how much the theoretical performance of the latest high
resolution sensors is compromised by the capabilities of the lens.

--
Matthew Winn
[If replying by mail remove the "r" from "urk"]

"Matthew Winn" wrote:

If you're photographing a subject with low contrast,
then once you've taken the lens into account the detail that might
be retained by the film at 100 lp/mm is getting lost in the noise.

If you are photographing a subject with low contrast (as most of us a
there are very few things in real life that have the 1000:1 contast of
resolution test charts) the MTF performance will be very very different from
the published graphs, which only apply to 1000:1 contrast patterns. TMX100
resolves 200 lp/mm at 1000:1 contrast and 63 lp/mm at 1:1.6 contrast.
(Classic Velvia claims 80 lp/mm at 1:1.6 and 160 lp/mm at 1000:1, but it's
large MTF gain at 10 lp/mm is more likely why people love it than the high
end of the performance. This is a cheap shot at Velvia fans, but I actually
think that it's right. And it's the reason you really want to use Velvia50
at f/45 and f/64.)

There are very few films that can resolve much over 60 lp/mm at low
contrast.

To the best I can tell, lp/mm at 1000:1 is a lousy measure of performance
_for photographic purposes_. Even the 1:1.6 figures overstate the
capabilities of film, since that's for a low response that's not
photographically useful.

And laying that much down on film is not a tall order for even quite
'average' lenses either.

But that doesn't suit the how-good-digital-capture-is camp at all.
So they forget about the soft focus filters that are in front of their
sensors. they forget about the small size of their sensors, and still
think
that because the file that comes out at the end contains so many pixels,
it
must be very good.

I am wondering how much the theoretical performance of the latest high
resolution sensors is compromised by the capabilities of the lens.

The 5D has a Nyquist frequency of 60 lp/mm, and it's not all that hard to
provide decent contrast at 45 to 50 lp/mm. Images look great at f/16, maybe
a tad softer at f/22.

That's plenty to create significantly better looking images than 35mm film
can even dream of. If you take your eye off the test charts under your
microscope and actually look at some prints. (Cheapshot a QG, not you.)

The D2x (Nyquist frequency more like 90 lp/mm) fans _claim_ that they get
sharp images near wide open with their fast lenses, but I'd like to see a
careful comparison at similar f stop. However, the D2x has an even weaker AA
filter than the 5D, and I've seen aliasing in real-world D2x shots, so they
are getting over 90 lp/mm images at the sensor. Some of the time, at least.

David J. Littleboy
Tokyo, Japan

Michael Benveniste wrote:

In which case the resolution of the system would be around 74 lp/mm.

No, no. That is the resolution as found on exposed film.
None of that "system resolution" malarky.


Moreover, only the slowest traditional black and white films approach
that 102 lp/mm number. Portra 160NC is generally rated around 75 lp/mm.
Source:

Portra is not 'rated', but seen recording 140 lp/mm.
http://www.zeiss.com/C12567A8003B58B...256CEF002B0246

Finally, if you are trying to determine megapixels, with film that
implies scanning, with the inherent sampling losses of that technology.

That depends on the scanner.
You can't just assume that the thing performs that bad that you lose
resolution.

If you don't scan and try to compare print quality, you have to try to
figure out how to equalize the analog vs. digital print processes.

And that "prints" red herring again...
The question was how many MP there are in film. Not how you many of them you
can save or can squander while laying them down on paper.

Don't get me wrong. I'm not claiming that even the best 24x36mm dSLR
can match a 6x7 film camera. But trying to reduce the difference down
to a number of megapixels is not a simple exercise, nor to my mind is it
a useful exercise.

I agree.
I shouldn't have responded at all.
It's just sad to see entire new generations being brought up to believe the
greatest nonsense, unchallenged...

Matthew Winn wrote:

Quite normal, 'average' films, like Kodak's T-Max or Portra, the staple
of
many photographers, resolve well beyond 102 lp/mm.

I don't consider Portra an "average consumer film". Not many average
consumers keep their film refrigerated.

The one has nothing to do with the other.
First, resolution does not depend on colour balance. You do not need to
refrigerate film to keep the three B&W layers capable of resolving what they
do.

Second: yes, Portra is a run-of-the-mill film. Like i said: teh staple of
many photographers.

And there's a "third" too:
Why are you looking for "consumer" grade film to compare digital to?

It depends what you mean by "resolve". Kodak's own MTF graph for
Portra 160NC-2 cuts off at about 80 lp/mm,

Yep. That's the graph.

at which point it's at 50%
for the blue-sensitive layer and 25% for red. Extrapolating, at 100
lp/mm the best the film can manage is about 35% in the blue-sensitive
layer.

"Extrapolating"... ;-)
Let's suppose 35%. That means the detail is there in a quite visible, and
retrievable way.
Not that it is not there.

160VC-2 is a little better, managing around 40% at 100 lp/mm
for blue although the red is down at 15%. [...]
-- but it's clear that
the image is getting mushy at 80 lp/mm even with the best prime lenses
I have.

And that's the the big difference between film and digital. Though the
detail starts to "get mushy", it is still recorded on film.
Digital sensors do not do that at all: it's either there, or not at all.

I am wondering how much the theoretical performance of the latest high
resolution sensors is compromised by the capabilities of the lens.

Not at all.
The soft focus filter typically cuts-off at 50 lp/mm.
Only very bad lenses do not go beyond that.

David J. Littleboy wrote:

That's plenty to create significantly better looking images than 35mm film
can even dream of. If you take your eye off the test charts under your
microscope and actually look at some prints. (Cheapshot a QG, not you.)

:-)

David, that microscope is a scanner.
The result is a file with very usefull information. Information your digital
sensor just does not capture.

By the way: you have noticed the title of this thread, have you?
Why is it that digi-guerrileros like you keep dumbing down the competition
in your efforts to proof that your favourite is at least as good...?
;-)

"Q.G. de Bakker" wrote:

Moreover, only the slowest traditional black and white films approach
that 102 lp/mm number. Portra 160NC is generally rated around 75 lp/mm.
Source:

Portra is not 'rated', but seen recording 140 lp/mm.
http://www.zeiss.com/C12567A8003B58B...256CEF002B0246

Zeiss typically assumes a contrast differential of 2-5% for its
measurements,
which is meaningless for photographic purposes. This is the main reason
why
Zeiss lp/mm numbers are so high for its lenses. The study I cite used a
more
reasonable 30% criterion.

That depends on the scanner.
You can't just assume that the thing performs that bad that you lose
resolution.

I can assume the Nyquist-Shannon sampling theorem applies. So in order
to not lose resolution, your scanner needs to sample at twice the
resolution of the original image. So to get 75 lp/mm you would need to
scan at over 7500 dpi. And that assume a perfect optical system in the
scanner and perfect film flatness. It also assumes that the data and
sampling are "in phase."

As a demonstration of scanning loss even with an 8000 dpi drum scanner,
I suggest you look at the following article:
http://www.normankoren.com/Tutorials/Scan8000.html

--
Michael Benveniste --
Spam and UCE professionally evaluated for $250. Use this email
address only to submit mail for evaluation.

Michael Benveniste wrote:

Zeiss typically assumes a contrast differential of 2-5% for its
measurements, [...]

Uhm... It's you who are assuming.
Zeiss did test the films not in a lab, but by taking pictures under quite
normal circumstances.
Then they did something very few of us do, and look closely at the film and
counted line pairs.
Nothing is assumed.

What indeed is meaningless are all those assumptions that are tossed about
in defence of what inherently still is a poor imaging system: direct digital
capture.
It's getting there, but still has some way to go.

That depends on the scanner.
You can't just assume that the thing performs that bad that you lose
resolution.

I can assume [...]

Indeed!
;-)