Focal plane vs. leaf shutters in MF SLRs

Bill Hilton wrote:
From: (Bob Monaghan)

Do the math on a 300 dpi vs. Leica standard print (which is 392 dpi), viz:

"300 dpi" is not the right number. 300 ppi (pixels per inch) is a common
digital input file resolution but that gets rasterized by the printer software


If the input is 300DPI then how could the output being anything else?
GIGO? No?

Nick

"Bob Monaghan" wrote in message
...
SNIP
basically, one interpretation is that you have to sample at
least twice the rate of the maximum frequency you want to
characterize. For us, it means that a fixed sensor with say
100 pixels per mm can only respond at a 50 lpmm or lower
rate.

Correct, it's basic sampling theory (and it's not just a theory ;-) it's
confirmed in practice by the observed behavior of all sampling systems).
However, it is possible to confuse aliasing (caused by sampling higher
spatial frequencies than the Nyquist limit) with detail. Sometimes it isn't
too damaging and may mimick (pseudo-) detail. But when film is involved it
will result in more prominent graininess that there acually is. And if the
noise floor is raised the, inherently low contrast, finest detail will drown
in the noise.

So when a sensor that has say 124 pixels per mm is in a
DSLR, it can respond to 124/2 or 62 lpmm.

Theoretical maximum on sensor, yes. Also don't overlook that larger sensors
(in terms of mega pixels AND physical size) need less magnification, so they
keep their sensor resolution better when output. Sampling pitch, and
magnification to output, determine the absolute visible limitations for
resolution if all other factors were optimal, which they aren't.

Claims that it is recording 75 lpmm or any higher value have
to be challenged on the basis of the Nyquist theorem - you
aren't sampling fast enough to "see" a 75 lpmm image.

Correct. Aliasing maybe, actual detail no.

Even the maximum 62 lpmm would be hard to support, because
the effects of other system elements like the anti-aliasing low-pass
filter and optics and so on would likely reduce the system response
below that of the sensor alone.

Yes, although since these low-pass filters are not perfect 'brick-wall'
filters they mostly attenuate the modulation near the Nyquist frequency. The
interaction between lens and sensor (film or CCD/CMOS device), and in some
cases also a scanner's lens and sensor, will reduce especially fine detail
modulation. But then there is sharpening to compensate for the reduced
contrast at higher spatial frequencies.

Bart

Recently, Bart van der Wolf posted:
(Excellent expansions on BobM's explanation)

"Bob Monaghan" wrote in message

Even the maximum 62 lpmm would be hard to support, because
the effects of other system elements like the anti-aliasing low-pass
filter and optics and so on would likely reduce the system response
below that of the sensor alone.

Yes, although since these low-pass filters are not perfect
'brick-wall' filters they mostly attenuate the modulation near the
Nyquist frequency. The interaction between lens and sensor (film or
CCD/CMOS device), and in some cases also a scanner's lens and sensor,
will reduce especially fine detail modulation. But then there is
sharpening to compensate for the reduced contrast at higher spatial
frequencies.

The difficulty, at least for me, is that sharpening is an artifact in
itself. If you look at it at high magnification, sharpening "flattens" the
the values of pixels near higher contrast edges. The more sharpening, the
wider the band of flattened pixels. So, sharpening along with the
generally lower range of color gradations has an impact on the textures in
an image. This works for some images, but damages others.

Neil

Recently, one_of_many posted:

In article ,
dy (Bill Hilton) wrote:

From: (Bob Monaghan)

Do the math on a 300 dpi vs. Leica standard print (which is 392
dpi), viz:

"300 dpi" is not the right number. 300 ppi (pixels per inch) is a
common digital input file resolution but that gets rasterized by the
printer software to a much finer pitch, typical numbers are 720x720
up to 2,880 x 1,440
dpi (dpi
is correct here, ppi is correct on the input side). While you don't
actually get 2,880 dots/inch of resolution due to ink spreading on
the paper you do get a lot more than 300 dpi.

I think the acronym you want is SPI - Stuff Per Inch. Splats Per Inch,
Spritzes Per Inch, (bull)**** Per Inch - none of which makes a whole
lot of difference in the end as it concerns _our_ printers. What
difference does a few steps of the step or ink engine make when the
ink can't be resolved on the paper. It's like talking about a
400lp/mm lens over 2475 recording film: downsampling occurs first
intentionally in the printer then by analog accident in _any_ event
in the paper.

Just for the hell of it, would a bunch of you all print a calibrated
high-rez military target? Simplify it and make it monochrome (B&W).
Let's see what printers can really do.

Then show me the gallery that will hang it.

ROTFL Excellent commentary! And, an excellent suggestion to those that
think their inkjets are high resolution printers.

Neil

Recently, one_of_many posted:

In article . net,
"Neil Gould" wrote:

Recently, Q.G. de Bakker posted:

[...]

I was *not* suggesting that a Leica/Fuji collaboration would be for
a film camera. I tried to express that Fuji's large presence in the
world of film photographers would make it easier to market *any*
photographic product that the partnership made. Certainly, it would
be an easier sell than products from the Leica/Panasonic
collaboration, because Panasonic does not have as much of a
reputation among film photographers.

Is this not a moot point because Leica is working with Kodak and
Imacon for their new R8,9 digital back?

No, it's not moot. I think that there will be a *range* of products
available, with the digital back for the R9 falling at the top of their
price range. What might kill it (and may be adding to the time to bring
this product to market) is that the advantage of "digital specific" lenses
may outperform the R9/digital combo. So, it could well be that a $1,500
Lumix dZLR (tm) with a Leica digital lens could produce better images
than, say an R9/digital back/100 mm APO that someone paid $10k to have.

Even using special lenses on the sensors to better capture the angular
rays at the edges of the frame is a compromise, because film lenses vary
with regard to those angles. Film is not as sensitive to those angular
rays in part due to its 3D nature compared to the "cell wells" of digital
sensors. So, while I'm curious about the upcoming R9 back, I'm equally
skeptical about its performance.

Neil

"one_of_many" wrote in message
news
In article ,
dy (Bill Hilton) wrote:
SNIP
"300 dpi" is not the right number. 300 ppi (pixels per inch) is
a common digital input file resolution but that gets rasterized by
the printer software to a much finer pitch, typical numbers are
720x720 up to 2,880 x 1,440 dpi (dpi is correct here, ppi is
correct on the input side). While you don't actually get 2,880
dots/inch of resolution due to ink spreading on the paper you
do get a lot more than 300 dpi.

Correct, the ppi is the number of smallest picture elements that translate
to spatial frequency information in the output, the printer driver *dithers*
that in order to approximate continuous tone color resolution.

SNIP
What difference does a few steps of the step or ink engine make
when the ink can't be resolved on the paper.

Most of the ink shouldn't be resolved on paper, because we try and
mix/blend/dither into 'continuous' tones not directly offered by the inkset.
We DO want that dithering to be finer than the visual resolution at the
given viewing distance.

Just for the hell of it, would a bunch of you all print a calibrated
high-rez military target? Simplify it and make it monochrome
(B&W). Let's see what printers can really do.

My HP printer can resolve(!) detail between 9.5 and 12.7 cycles/mm on glossy
paper. Cycles are more relevant than B&W lines, unless you are into line
drawings.
It can be rather easily tested with a target like:
http://www.xs4all.nl/~bvdwolf/main/d...0mm_600ppi.png
(2.1MB)
http://www.xs4all.nl/~bvdwolf/main/d...0mm_720ppi.png
(2.8MB)
Depending on the software used, you may need to convert to RGB before
printing in color.

I also have targets available for camera lens+film (+scanner) testing, they
are a bit less demanding on the printer. I've posted their links before, but
if requested I'll post them again with instructions.

Just print at the indicated ppi(!), measure the central blur diameter and
calculate: 120/pi/mm=cycles/mm, where pi is the number 3.14159265..., and mm
is the blur diameter in millimetres. If the diameter is more of an ellipse
than a circle, then the horizontal/vertical resolution is not the same in
both directions (e.g. paper transportation inaccuracy).

If you are into line drawings / text, a target like described he
http://www.ddisoftware.com/qimage/quality/ will also help to determine the
limits, but B&W lines use a lot more ink, and thus bleed easier.

Bart

"Bart van der Wolf" wrote in message
...
SNIP

My HP printer can resolve(!) detail between 9.5 and 12.7 cycles/mm on
glossy
paper.

P.S. for the mathematically challenged ;-): that is between 483 and 645 ppi,
so printing at 600/720ppi on an inkjet printer does provide more detail than
many people think, despite the dithering at 2400/4880 dpi(!).

Bart

"Neil Gould" wrote in message
ink.net...
SNIP
ROTFL Excellent commentary! And, an excellent suggestion to those that
think their inkjets are high resolution printers.

Sorry to disappoint you, they are (see my other post in this thread), and
you can test it yourself...

Now, if you were to address dithered highlight uniformity instead of
resolution, than we have a different subject, but on resolution you should
really try it.

Bart

From: (one_of_many)

I think the acronym you want is SPI - Stuff Per Inch. Splats Per Inch,
Spritzes Per Inch, (bull)**** Per Inch

Splats or Spritzes would work on the output end ... LOL.

none of which makes a whole lot
of difference in the end as it concerns _our_ printers.

Really? You can't see the difference on a print of a good file printed at 720
dpi vs 1,440 dpi on glossy paper? Your eyesight must not be very good because
everyone else can.

Then show me the gallery that will hang it.

You want to see galleries that hang LightJet prints? Not hard to find, most of
the best galleries in the west do this already. Start with the Ansel Adams
Galley in Yosemite and go from there ...

Bill

"Neil Gould" wrote in message
ink.net...
SNIP
The difficulty, at least for me, is that sharpening is an
artifact in itself.

It can be, but it should only attempt to restore contrast losses in the
imaging chain (most prominent in fine detail). The imaging chain also
includes losses in the output medium.

If you look at it at high magnification, sharpening "flattens"
the the values of pixels near higher contrast edges. The more
sharpening, the wider the band of flattened pixels. So,
sharpening along with the generally lower range of color
gradations has an impact on the textures in an image. This
works for some images, but damages others.

It may, but sharpening is also a bit of an art, and its success does depend
on the image data you start with. I see a lot of poor sharpening examples on
the web, I just hope most professionals do better than that.

For those using Photoshop, there is an easy method to reduce visual haloing
and totally avoid clipping due to sharpening. It requires USM sharpening of
a duplicate layer (preferably with an edge mask and prior noise reduction)
and blending adjustments such as in this example:
http://www.xs4all.nl/~bvdwolf/main/downloads/Non-clipped-sharpening.png.
Those can be used as a starting point (for digicam files use a very small
radius, large amount, for scans the radius depends on the resolution) and
the effect can be tuned for various output media by varying the layer's
opacity (exaggerate the sharpening amount and reduce opacity to your
liking).

Bart