FF sensors: is 80MP needed?
Ilya Zakharevich
what is the MP count starting from which the higher resolution of the
sensor does not help to *significantly* improve the total lens+sensor
resolution. Of course, this depends on the lens, so it makes sense to
discuss the diffraction-bounded region of f-stops.
Short answer: with a f/16 diffraction-bound lens, the total resolution
should start to level out with 30MP full frame sensor
(one similar in MTF-relative-to-pixel-count to current
generation of sensors; see P.S. for other possibilities).
One way to do this in experimental fashion, is to compare shots made
with an "ideal lens" and a real-life sensor, to shots made with an
"ideal sensor" and a real-life lens. Since in practice "defects"
contributed by the lens and the sensor combine, this would compare the
relative magnitude of these defects.
One can use the result of this comparison like this: of course,
increasing the resolution of sensor would bring some improvement in
the total resolution in a very large region (keywords are `cut-off
spacial frequency', and `Nyquist limit'). However, at some point the
benefits will start to level out; our target is to find this point.
We assume that it is close to the point where "defects" contributed by
the lens are comparable to "defects" contributed by the sensor.
First experiments (lens which provides much better resolution than
real sensor): it is easy if you believe the "short answer" above: any
resolution charts shot with a decent lens would be in this ballpark.
To avoid circular reasoning, one should do more; e.g., compare
resolution charts shot with half-frame 3MP sensor to those shot with
16 MP FF sensor; one can easily see that at 100% magnification, there
is little difference, thus the contribution of lens is negligible indeed.
Judging by these shots, one can give a simple model [**] of a digital
sensor of today; to provide a different resolution, just rescale.
The second scenario (ideal sensor, real lens) is also easy to obtain:
good film is an ideal sensor media when used in 4x5in (or larger)
format (at least with strongly stepped down lens). I use (wonderful)
data in
http://www.clarkvision.com/photoinfo/large_mosaics/
as a source of "high-quality" image (digital mosaic), comparing it to
a "low-quality" (1/2 ;-) image (4x5in Velvia shot made with aperture
f/64).
The comparison shows:
A) Resolution of the film shot is very similar to the resolution of
high-quality shot scaled down to 16 MPixel JPEG [*].
B) Resolution of the film shot is very similar to a (simulated) image
taken with 30 MPixel digital camera [**]
[Of course, the possibility to redo the shot on a 30MP FF digital
camera assumes a presence of f=22mm tilt lens with good
performance at f/16. Performance at f/16 is not a big deal; but
I suspect tilt-f=22mm lens is not available; one would need a
tilt sensor to compensate).
[This assumes that the film shot Roger did has no significant out-of-focus
and blur areas. Since he says it took 6hours to set up this shot, I assume
this holds.]
===========================
Since 4x5in format is very easy to extrapolate (up to f/32 there is no
significant dependence on the quality of the lens design), this allows
to predict "digital equivalents" of other shots (of course, to do this
one needs to assume that the film shot is in ideal focus; I hope we
can trust Roger on this; he says he was waiting for 5 hours for ideal
conditions to make this shot ;-):
f/96: 15 MPixel sensor;
f/64: 30 Mpixel sensor;
f/48: 60 Mpixel sensor;
f/32: lower than 120 Mpixel sensor (abberrations start to feel);
f/22: about 120 Mpixel sensor (assuming lens resolution
saturates near f/22, and ).
===========================
One can also rescale the numbers above to 24x36mm formfactor with
typical DSLR-type fixed-focal-length lenses:
f/22: 15 MPixel sensor;
f/16: 30 Mpixel sensor;
f/11: less than 60 Mpixel sensor;
f/8: about 60 Mpixel sensor (assuming lens resolution
saturates near f/8).
With highest-quality rangefinder-type lenses (AFAIK, these are
available with f about 45mm and above):
f/22: 15 MPixel sensor;
f/16: 30 Mpixel sensor;
f/11: 60 Mpixel sensor;
f/8: less than 120 Mpixel sensor;
f/5.6: about 120 Mpixel sensor (assuming lens resolution
saturates near f/5.6).
With high-quality zoom lenses:
f/22: 15 MPixel sensor;
f/16: less than 30 Mpixel sensor;
f/11: about 30 Mpixel sensor (assuming lens resolution
saturates near f/11).
(All these number assume some "theoretically sound" behaviour of MTF
in the region N steps down the optimal resolution of the lens:
2 steps) Practically diffraction-bound resolution;
1 step) Slightly less than diffraction-bound resolution;
0 steps) About the same as diffraction-bound resolution when
closed 1 step more.
Of course, any particular lens with show some variations, but AFAIK,
the variations are negligible)
===========================
Here is some more info on how I interpret Roger's short. Velvia at
1/64 is more or less an ideal sensor; it has practically no noise, and
practically ideal MTF. So the film shot is equivalent to, e.g., f/16
24x36mm shot made with an ideal sensor.
The "high-resolution" shot is made at f/11; since we shrank it 2.5x, the
diffraction doesn't matter; likewise for focussing errors. So, essentially,
it is "an ideal image" used with the given sensor. So one can see that
an ideal (diffraction-, abberration-, and focus-error-less) shot made
with 30 MPixel sensor is very similar to f/16 shot made with an ideal sensor.
===========================
[*] I took the crop of high quality image, reduced it to 29% linear size,
then enlarged it to 345% (to get the initial size). Then I compared
it with the film quality image. [Replacing 29% with 25% or 33%
produces images which differ significantly from the film image:
worse looking and better looking correspondingly.]
[**] To simulate a shot made with an "ideal lens" and a digital sensor
similar in construction to today's sensors: take a "reasonable
quality" image; scale it down several times; you get a very high
quality image. Now blur it to reduce MTF at high spacial frequencies.
My experiments show that the following convolution matrix produces
results very similar to real shots (tested with resolution chart
shots from DPReview): [assumes fixed-width fonts; should be divided
by 13.5]:
1 1 1
1 5.5 1
1 1 1
(Convolution with this function reduces the MTF 3 times at 2/3 of
cut-off-frequency of the sensor.)
Now: this is what I did: I took the crop of high quality image,
reduced it to 40% linear size, then enlarged it to 250% (to get
the initial size). Then I compared it with the film quality image.
Enjoy,
Ilya
P.S. Of course, only the "sensors of today's construction" are
covered. And with high MP count, there is no need to
(significant!) decreases in resolution and "sensibility" implied
by anti-aliasing filters (AAF *are* needed with the minuscule MP
counts of todays's sensors). E.g., one can easily show that
about 40MP FF sensors, there would be no need in AAF (at least
when used with SLR-type lenses). This would improve the
resolution [#] similar (?) to about 1.6x increase of MP count.
So with such a sensor, one gets (for SLR fixed-focus lenses):
f/22: 10 MPixel sensor;
f/16: 20 Mpixel sensor;
f/11: less than 40 Mpixel sensor;
f/8: about 40 Mpixel sensor (assuming lens resolution
saturates near f/8).
Similarly, with rangefinder-type lenses, the beginning-of-leveling-out
point would be about 80MP (needed at f/8 and f/5.6 shots).
[#] (Actually, removing AAF would also *significantly* increase
sensitivity [measured via the exposition needed to achieve a
certain visible S/N level], but this deserves a separate message...)
Rich
Ilya Zakharevich wrote:
> With proliferation of half-frame 10MP cameras, a natural question is:
> what is the MP count starting from which the higher resolution of the
> sensor does not help to *significantly* improve the total lens+sensor
> resolution. Of course, this depends on the lens, so it makes sense to
> discuss the diffraction-bounded region of f-stops.
>
> resolution [#] similar (?) to about 1.6x increase of MP count.
>
> So with such a sensor, one gets (for SLR fixed-focus lenses):
>
> f/22: 10 MPixel sensor;
> f/16: 20 Mpixel sensor;
> f/11: less than 40 Mpixel sensor;
> f/8: about 40 Mpixel sensor (assuming lens resolution
> saturates near f/8).
Makes sense. Most camera lenses are not that great when it comes to
their optical quality. Otherwise, they would maximize their resolution
at their widest opening, there would be no overbearing aberrations to
require the lens be stopped down to f8.
>
> Similarly, with rangefinder-type lenses, the beginning-of-leveling-out
> point would be about 80MP (needed at f/8 and f/5.6 shots).
>
> [#] (Actually, removing AAF would also *significantly* increase
> sensitivity [measured via the exposition needed to achieve a
> certain visible S/N level], but this deserves a separate message...)
You'll find that professional optical sources are more "honest" about
what is needed to support high megapixel count cameras with small
pixels than consumer camera makers. You need a lens capable of at
least 100lp/mm to support a 5um pixel size. I doubt 1/4 of the lenses
offered by most mfgs meet this spec. So, what is someone going to do
with 80 megapixels?
frederick
<snip>
You can theorise as much as you like, but if you just observe results,
then you can see that 6-8mp aps-c dslrs meet or exceed 35mm colour
transparency, 16/17mp 36x24 sensor is near 645, 20+mp larger format 20+
mp(PhaseOne sensor) exceeds 645, 39mp PhaseOne comes very close to 4x5.
If you want to use monochrome film - a material that has incomparably
inconvenient performance compared with even the cheap digital cameras
that they now sell at the supermarket - and you want to take photos only
of test charts, then you can possibly prove that better resolution is
possible with film. For people who want to take photographs, that is
now seldom true and in any case irrelevant.
Ilya Zakharevich
Rich
<rande...@gmail.com>], who wrote in article <1160802577....@m7g2000cwm.googlegroups.com>:
> You'll find that professional optical sources are more "honest" about
> what is needed to support high megapixel count cameras with small
> pixels than consumer camera makers. You need a lens capable of at
> least 100lp/mm to support a 5um pixel size. I doubt 1/4 of the lenses
> offered by most mfgs meet this spec. So, what is someone going to do
> with 80 megapixels?
Note that I consider exactly the opposite problem: now which lens
would "support" a given sensor (BTW, I have no idea what "support"
will mean), but what sensor will get "almost all" out of the given
lens.
E.g., assuming 5um pixel size, this would be about 35MP FF sensor.
The charts show that it will get "practically all" the resolution out
of a decent zoom lens (maybe even with major postprocessing); one will
be able to get better results out of a "usual" fixed-focal SLR lens
with more pixel count, but not *much* better.
It would not be able to get enough data out of a "best resolution lens".
[Even assuming that "supporting a sensor" would make any sense (do
not see how, since MTF of a lens and of a sensor combine),
considering a lens which supports a sensor is, in most cases,
silly. Sensors are cheap (the semiconductor part of FF CMOS sensor
costs about $100 to produce; the total assembly may be more
expensive, but it would not be comparable to price of decent lenses...)]
Yours,
Ilya
Ilya Zakharevich
<lo...@sea.com>], who wrote in article <1160802542.137637@ftpsrv1>:
> You can theorise as much as you like, but if you just observe results,
> then you can see that 6-8mp aps-c dslrs meet or exceed 35mm colour
> transparency
Irrelevant. You are discussing lens+sensor vs lens+film. I'm
discussing lens+sensor vs lens+ideal_sensor.
> 39mp PhaseOne comes very close to 4x5.
IMO, this makes absolutely no sense without discussing the lens
settings used. Note that my experiment shows that 4x5in USED WITH
f/64 would not benefit MUCH from more than about 20MP non-AA sensor;
probably would not benefit AT ALL from more than 30MP non-AA sensor.
With different f-stops, the situation is (proportionally) different.
Hope this helps,
Ilya
Greg "_"
>39mp PhaseOne comes very close to 4x5.
& and your proof exists < ? >
> If you want to use monochrome film - a material that has incomparably
> inconvenient performance compared with even the cheap digital cameras
> that they now sell at the supermarket
Right <smirk :^>
--
Reality-Is finding that perfect picture
and never looking back.
Matt Clara
news:1160802542.137637@ftpsrv1...
> Ilya Zakharevich wrote:
> <snip>
>
> You can theorise as much as you like, but if you just observe results,
> then you can see that 6-8mp aps-c dslrs meet or exceed 35mm colour
> transparency,
Sorry, fella, but I shoot weddings with both film and digital, and what you
say here simply isn't true.
frederick
Please post some scans from your 35mm negs.
frederick
> In article <1160802542.137637@ftpsrv1>, frederick <lo...@sea.com> wrote:
>
>> 39mp PhaseOne comes very close to 4x5.
>
> & and your proof exists < ? >
>
>> If you want to use monochrome film - a material that has incomparably
>> inconvenient performance compared with even the cheap digital cameras
>> that they now sell at the supermarket
>
> Right <smirk :^>
25 iso monochrome....
zzzzzz
frederick
> In article <1160802542.137637@ftpsrv1>, frederick <lo...@sea.com> wrote:
>
>> 39mp PhaseOne comes very close to 4x5.
>
> & and your proof exists < ? >
oops - missed that.
http://www.luminous-landscape.com/essays/Cramer.shtml
http://www.outbackphoto.com/artofraw/raw_28/essay.html
Father Kodak
wrote:
Matt,
How would you modify the frederick's statement, based on your
experience? E.g. would a 10 mp APC-C D SLR meet/exceed 35 mm color
transparency.
Father Kodak
Matt Clara
news:fv75j2h3va90d3jtr...@4ax.com...
I haven't used 10 mp, so I don't know, but I suspect it would lack in one
way, and that's the same way the 6-8 mp cameras do: upon enlargement they
begin to break down rather quickly. To me this speaks to overall levels of
infomation present in the two mediums, but it could be any number of
reasons. Enlargement, of course, isn't the reason digital is eclipsing
35mm, and even there it's good enough for some purposes. I still shoot my
formals as well as some select other shots with medium-format film precisely
because it can be blown up poster-size with little detraction to the image,
provided it's a sharp, well exposed negative, of course.
Greg "_"
Reasonably few B&W shooting photographers shoot 25 asa BW film on a
consistent, none I know personally.
Greg "_"
Have I told you I am not impressed with the work of his I have seen. I
am also not impressed with the quality of the tests of his I have seen
over the years.
David J. Littleboy
"Greg "_"" <grey_egg@greg_photo.com> wrote:
>> >> If you want to use monochrome film - a material that has incomparably
>> >> inconvenient performance compared with even the cheap digital cameras
>> >> that they now sell at the supermarket
>> >
>> > Right <smirk :^>
>> yawn.
>> 25 iso monochrome....
>> zzzzzz
>
> Reasonably few B&W shooting photographers shoot 25 asa BW film on a
> consistent, none I know personally.
That's because most of them have been discontinued. Panatomic X and Tech Pan
are history. (I think there's still an Agfa film in production, though.)
Panatomic-X is the reason I hate 35mm. At 11x14, Plus-X in 645 looks way
better than Panatomic X in 35mm, and is two stops faster. ISO 25 gets real
painful when you want to use either a red filter or a polarizer.
Nowadays, TMX 100 is close to what the ISO 25 films used to be in terms of
grain and resolution, although some people complain they don't like its
tonal rendition. TMX 100 in 6x7 will edge out the 5D, though.
David J. Littleboy
Tokyo, Japan
Scott W
> I haven't used 10 mp, so I don't know, but I suspect it would lack
in one
> way, and that's the same way the 6-8 mp cameras do: upon enlargement they
> begin to break down rather quickly. To me this speaks to overall levels of
> infomation present in the two mediums, but it could be any number of
> reasons. Enlargement, of course, isn't the reason digital is eclipsing
> 35mm, and even there it's good enough for some purposes. I still shoot my
> formals as well as some select other shots with medium-format film precisely
> because it can be blown up poster-size with little detraction to the image,
> provided it's a sharp, well exposed negative, of course.
This seems like a bit of bait and switch to me, you seem to claim that
35mm film is better then an 8 MP DSLR but then you say you are using MF
film. Clearly MF will easily beat an 8 MP DLSR but I have yet to see a
color image from 35mm film that beats a 8 MP DSLR and the vast majority
of 35mm film scans fall far short of a 8 MP DSLR.
Scott
Roger N. Clark (change username to rnclark)
> In article <1160943104.100931@ftpsrv1>, frederick <lo...@sea.com> wrote:
>
>
>>Greg "_" wrote:
>>
>>>In article <1160802542.137637@ftpsrv1>, frederick <lo...@sea.com> wrote:
>>>
>>>
>>>>39mp PhaseOne comes very close to 4x5.
>>>
>>>& and your proof exists < ? >
>>
>>oops - missed that.
>>http://www.luminous-landscape.com/essays/Cramer.shtml
>>http://www.outbackphoto.com/artofraw/raw_28/essay.html
>
>
> Have I told you I am not impressed with the work of his I have seen. I
> am also not impressed with the quality of the tests of his I have seen
> over the years.
Yeah, I agree. At luminous-landscape, we saw 3 megapixels
beating 35mm , then it was 6 megapixels, then....
Seems like the latest digital beats whatever was out before
regardless of facts.
Kinda like a marketing department ;-).
Roger
Roger N. Clark (change username to rnclark)
> [A complimentary Cc of this posting was sent to
> frederick
> <lo...@sea.com>], who wrote in article <1160802542.137637@ftpsrv1>:
>
>>You can theorise as much as you like, but if you just observe results,
>>then you can see that 6-8mp aps-c dslrs meet or exceed 35mm colour
>>transparency
>
>
> Irrelevant. You are discussing lens+sensor vs lens+film. I'm
> discussing lens+sensor vs lens+ideal_sensor.
Theory is fine, but you must consider all factors.
One factor you have not considered, and is a source of
confusion in many of the film versus digital religious wars is
not just resolution, but tonality and signal-to-noise ratio.
Digital has a much higher signal-to-noise ratio and that
greatly influences perceived image detail. Digital
also has much greater dynamic range.
Roger
frederick
> In article <1160943104.100931@ftpsrv1>, frederick <lo...@sea.com> wrote:
>
>> Greg "_" wrote:
>>> In article <1160802542.137637@ftpsrv1>, frederick <lo...@sea.com> wrote:
>>>
>>>> 39mp PhaseOne comes very close to 4x5.
>>> & and your proof exists < ? >
>> oops - missed that.
>> http://www.luminous-landscape.com/essays/Cramer.shtml
>> http://www.outbackphoto.com/artofraw/raw_28/essay.html
>
> Have I told you I am not impressed with the work of his I have seen. I
> am also not impressed with the quality of the tests of his I have seen
> over the years.
Would you like to qualify your "lack of being impressed" by some
objective measure of your own, or reference to other qualified
comparisons that have been done - where the results may better suit your
apparent agenda.
Or are we better just to dismiss your opinion as hot air?
frederick
> Digital
> also has much greater dynamic range.
How dare you keep saying that - when believing the reverse is the raison
d'être for the quaint obsession some have for 35mm.
Roger N. Clark (change username to rnclark)
;-)
Ilya Zakharevich
> > Irrelevant. You are discussing lens+sensor vs lens+film. I'm
> > discussing lens+sensor vs lens+ideal_sensor.
> Theory is fine, but you must consider all factors.
One can never consider all factors. It is important to consider *key*
factors - of course.
> One factor you have not considered, [...] is not just resolution,
> but [...]
Since all I was considering was resolution, it is not surprising I did
not consider other factors. ;-) (Especially since I already addressed
"other factors" in my other posts; here I concentrated on some "new"
experiments.)
> Digital has a much higher signal-to-noise ratio
Such a blank statement is definitely wrong; you mean the same
exposition, and it is not applicable to my comparison. E.g, with the
current technology digital will give much worse noise than film if the
exposition of digital is 20x smaller than of film ;-). And note that
this is quite probably holds when you compare LF film to FF digital
(but I do not remember the details of the exposition you used...) -
count number of photons per "pixel" (I mean square with size
determined by the MTF curve).
> ... that greatly influences perceived image detail.
Film has practically no noise when correctly exposed with aperture
f/32 - or much smaller f-number.
> Digital also has much greater dynamic range.
IMO, this remains to be proven yet (though I suspect that the proof
*will* support your claim - I just *have not seen* it).
Your investigation is very interesting, but, IMO, it completely missed
the point. Essentially, you measured the noise of 6um square of film
vs 8um square of semiconductor. So, film's noise *at these extremely
high spacial frequencies* turns out to be much more than digital
sensors'; fine. But, as you probably saw it in many posts about your
experiments, it is the noise at much lower spacial frequencies what
many people consider the measure of dynamic range.
So how one could measure "true dynamic range"? Essentially, one
cannot separate dynamic range from resolution; e.g., in each
exposition zone, one could measure the S/N ratio at different spacial
frequencies, and find at which frequency S/N ratio goes above 3 (or
some other reasonable number).
One gets a curve of (thus defined) spacial resolution per degree of
underexposition; *this curve*, IMO, is the measure of dynamic range.
Again: I agree with you that such an investigation will *most
probably* give film much lower mark than digital sensor - but 10x20in
film is practical, and digital is not. BTW, the question I addressed
in my initial posting is also related to this: for which f-numbers
this advantage of film will disappear, since FF digital sensor may
behave as well as HUGE area of film.
Thanks,
Ilya
Ilya Zakharevich
<lo...@sea.com>], who wrote in article <1160974662.417746@ftpsrv1>:
> Would you like to qualify your "lack of being impressed" by some
> objective measure of your own
Already the second paragraph contains complete BS:
And with that smaller focal length is around a 2.5 times increase
in available depth of field.
Anyone knows that switching formats does not bring any change in DoF -
if one WANTS to produce identical images, one can.
Different formats produce non-identical results only because of
different "effective MP count", and lens quality at the used f-stop
(well, QE may also enter the picture, e.g., when you compare film to
digital). You just use f-stop "proportional" to the formfactor, and
everything else comes out to be identical.
Let me repeat:
The physical laws of scaling are the following: to produce the same
image from N times smaller sensor (linearly) one needs to:
a) have the same count of pixels;
b) have the same QE;
c) have the same readout noise;
d) have the same full well;
e) have the same exposure time;
f) use N times higher aperture (measured as an F-number, e.g., 1/45);
g) have the same "quality" of the lens (e.g, measured as quotient
of actual MTF of the lens to MTF of diffraction-limited lens)
Hope this helps,
Ilya
Matt Clara
news:1160972901.6...@b28g2000cwb.googlegroups.com...
You're right Scott, I should have mentioned that I have shot 35mm right
along side digital, plus I used to shoot nothing but 35mm, and 35mm blows up
better. I also think I get more keepers with 35mm, and fewer exposure
problems, but the ability to see those problems with the digital is
priceless and the fact that I don't have to pay for 15 rolls of development
and prints isn't priceless, I can see a very real savings there, though it's
traded for lots of time in front of a computer. As for your inability to
see that 35mm beats digital in most forms today, I suggest you simply don't
have the experience there, or you would. And I'm not talking about scans of
35mm--you don't have to scan it to get the job done, see.
--
Regards,
Matt Clara
www.mattclara.com
Scott W
>And I'm not talking about scans of
> 35mm--you don't have to scan it to get the job done, see.
But scanning is about the only way we have to compare, unless you want
to come to Hawaii with your prints. If you are trying to say that
optical prints show more detail then it should be easy to scan the
optical print.
The best that I have seen by far is a scan Max Perl did and whereas the
35mm shot showed more detail then an 8 MP camera when printed out they
were about the same quality due to the grain in the film scan.
Since I would not do a workflow that does not have the image in a
digital form at some point if it can't be scanned it is not of much
value to me. And since I have yet to see anyone demonstrate that a
workflow that does not scan the film can produce a better image when
when you scan film I am not going to take that on faith.
Scott
Neil Ellwood
About 40 or so years ago over a period of a couple of years I used
micro-neg pan developed in Dilute FX1a developer (designed by Geoffrey
W. Crawley and published in BJP) rated at 8asa and although it was slow
the results were pleasing but not for action. The grain could hardly be
seen.
--
Delete l to reply.
frederick
negative and transparency - and have a big portfolio of larger
(>10x8)prints that I can easily compare. Most of the larger prints I
made from 35mm used Fuji iso 50 and Cibachrome. I always carried two
slr bodies, one loaded with iso50 transparency for enlargements, the
other loaded with iso 200 negative film because iso 50 is practically
useless for handheld shooting in less than ideal light, yet ideal light
is seldom what you want for taking interesting pictures.
Scanned 35mm negative film printed with light-jet leaves the optical
process cibachrome prints in the dust for reproducing detail. Inkjet
prints direct from digital look better again, and are almost infinitely
more convenient.
I really couldn't care less about technical arguments - how the picture
looks is the most important factor. A few years ago, I realised that
snapshots taken from an inexpensive pocket sized 4mp P&S digital were
practically as good as what I could get with iso200 35mm. 35mm film is
dead - there is no point at all now that dslr cameras are affordable.
frederick
> [A complimentary Cc of this posting was sent to
> frederick
> <lo...@sea.com>], who wrote in article <1160974662.417746@ftpsrv1>:
>> Would you like to qualify your "lack of being impressed" by some
>> objective measure of your own
>
> Already the second paragraph contains complete BS:
>
> And with that smaller focal length is around a 2.5 times increase
> in available depth of field.
>
> Anyone knows that switching formats does not bring any change in DoF -
> if one WANTS to produce identical images, one can.
>
>
> Different formats produce non-identical results only because of
> different "effective MP count", and lens quality at the used f-stop
> (well, QE may also enter the picture, e.g., when you compare film to
> digital). You just use f-stop "proportional" to the formfactor, and
> everything else comes out to be identical.
>
"The only downside to my new system is working with depth of field. On
my 4x5 Linhof, I have a focusing gauge that allows me to quickly
determine the optimum f stop for each situation. With the Mamiya and the
zoom lenses, this is much harder to determine. On the plus side, a
composition that required a 200mm on the 4x5 needs only an 80mm for the
P45 sensor. And with that smaller focal length is around a 2.5 times
That comment "available depth of field" doesn't have to mean as you
assume it could.
Ilya Zakharevich
frederick
> > Anyone knows that switching formats does not bring any change in DoF -
> > if one WANTS to produce identical images, one can.
> Only when maximum DOF is the objective. Not when shallow DOF is sought.
When true, this is a part of what I said: one needs the same "lens
quality at the used f-stop". When/If you can open 35mm-formfactor
lens to the same entry pupil as the LF lens (with the same "optical
build quality"), they create identical images (only scaled
differently) - same diffraction, "same" abberations, same DoF. Of
course, this "If/when" is rarely possible...
> > Different formats produce non-identical results only because of
> > different "effective MP count", and lens quality at the used f-stop
> > (well, QE may also enter the picture, e.g., when you compare film to
> > digital). You just use f-stop "proportional" to the formfactor, and
> > everything else comes out to be identical.
> You have quoted selectively:
> "The only downside to my new system is working with depth of field. On
> my 4x5 Linhof, I have a focusing gauge that allows me to quickly
> determine the optimum f stop for each situation. With the Mamiya and the
> zoom lenses, this is much harder to determine. On the plus side, a
> composition that required a 200mm on the 4x5 needs only an 80mm for the
> P45 sensor. And with that smaller focal length is around a 2.5 times
> increase in available depth of field."
>
> That comment "available depth of field" doesn't have to mean as you
> assume it could.
I do not agree with you. Read: "on the plus side"; obviously, he
thinks that with smaller formfactor he will have larger depth of
field. (This is only true if he would use the same f-stop; but when
you convert to a more open f-stop needed to get the same diffraction
[measured in the subject space, i.e., in angular units], one gets
exactly the same DoF...)
Yours,
Ilya
frederick
same shutter speed. For sure a big drawback for large format is slow
shutter speeds, as although you can use a tripod, you can't stop things
from moving in the frame.
Ultimately you are of course correct - but it may mean that you can't
take a photo with even a moving snail in the frame ;-)
David J. Littleboy
"Matt Clara" <hey.w...@buzz.off> wrote:
>
> I haven't used 10 mp, so I don't know, but I suspect it would lack in one
> way, and that's the same way the 6-8 mp cameras do: upon enlargement they
> begin to break down rather quickly.
Perhaps you are upsampling in a poor way?
This is exactly the opposite result of just about everyone else; lots of
people are very happy with their insane enlargements from digital (much more
so than from film). I suspect that your lab may be messing up your digital
enlargements.
Greg "_"
Then again; maybe Matt has a tad more critical eye than you common folk.
David J. Littleboy
"Greg "_"" <grey_egg@greg_photo.com> wrote:
> "David J. Littleboy" <dav...@gol.com> wrote:
>> "Matt Clara" <hey.w...@buzz.off> wrote:
>> >
>> > I haven't used 10 mp, so I don't know, but I suspect it would lack in
>> > one
>> > way, and that's the same way the 6-8 mp cameras do: upon enlargement
>> > they
>> > begin to break down rather quickly.
>>
>> Perhaps you are upsampling in a poor way?
>>
>> This is exactly the opposite result of just about everyone else; lots of
>> people are very happy with their insane enlargements from digital (much
>> more
>> so than from film). I suspect that your lab may be messing up your
>> digital
>> enlargements.
>
No: Matt claims Matt gets better enlargements from 35mm, whereas everyone
else finds the digital to hold up better. Same eyes doing the comparisons in
both cases. Besides, we're talking about really really ugly prints that
you'd have to be blind not to see the problems with if you walk up for a
closer look.
frederick
Do you have an Epson R1800 in Tokyo?
Or do I have you confused with someone else?
David J. Littleboy
I have the R800 and the 2400 (although here they are called the PX-G900 and
the PX-5000, respectively).
> Or do I have you confused with someone else?
Probably not. Not a lot of Davids in Tokyo that do MF and 5D.
Roger N. Clark (change username to rnclark)
> I should have mentioned that I have shot 35mm right
> along side digital, plus I used to shoot nothing but 35mm, and 35mm blows up
> better.
> As for your inability to
> see that 35mm beats digital in most forms today, I suggest you simply don't
> have the experience there, or you would. And I'm not talking about scans of
> 35mm--you don't have to scan it to get the job done, see.
I agree with you but with a condition: it depends on how one
enlarges the digital image. I shot 35mm, 4x5 and digital along
side each other for a while before concluding which route
I would go. My first decision was whether or not to go scanned
film or stick with traditional enlarging. I did multiple tests,
and this one shows my experience (a comparison of cibachrome
prints of the same image, printed traditionally versus
a drum scanned film + lightjet print) (be sure to step
back 10 to 15 feet from your monitor):
http://www.clarkvision.com/imagedetail/digital_advantage.html
Besides better control of color and contrast, I could
precisely dodge and burn, further making better prints
with digital, plus sharper prints that could be enlarged
more.
With 6-mpixel DSLR, my experience was that it was not up to
drum scanned Velvia (50). Neither was 8 mpixel DSLRs, until
I learned to improve resolution:
Image Restoration
Using Adaptive Richardson-Lucy Iteration
http://www.clarkvision.com/imagedetail/image-restoration1
I routinely upsample and run the Richardson-Lucy algorithm
and get close to double the pixels, turning an 8-megapixel
DSLR image into 32 megapixels. You can do this with
digital because of the higher signal-to-noise ratio of
digital. You can't do this with film very well because
of the noise from grain. So with 8-megapixel DSLR images,
I get much sharper 16x24 inch prints than I ever did from
35mm film printed traditionally or from drum scanned images.
Roger
frederick
> "frederick" <lo...@sea.com> wrote:
>> David,
>> Do you have an Epson R1800 in Tokyo?
>
> I have the R800 and the 2400 (although here they are called the PX-G900 and
> the PX-5000, respectively).
>
>> Or do I have you confused with someone else?
>
> Probably not. Not a lot of Davids in Tokyo that do MF and 5D.
>
For the PX-G cartridges and K3 cartridges, is the Epson Japan
recommended price of ~ 1102 yen a good price, or is it possible to get
discounted cartridges on line etc?
Epson prices in some countries are more than double Japanese prices for
the same cartridges. The cartridges are region zone chipped - but a
simple workaround may be possible.
David J. Littleboy
"frederick" <lo...@sea.com> wrote:
>>
> Thanks - Tokyo's a big place though - and you never know...
> For the PX-G cartridges and K3 cartridges, is the Epson Japan recommended
> price of ~ 1102 yen a good price, or is it possible to get discounted
> cartridges on line etc?
I've not gone to the effort to look for discounted cartriges online; 1102 is
what I'm paying. Note that the gloss optimizer is 556 Yen, but all the
others are 1102. (There are minor savings to be had by buying a whole set,
but there's no sticker on the box at hand.)
> Epson prices in some countries are more than double Japanese prices for
> the same cartridges.
Ouch! That's painful.
> The cartridges are region zone chipped - but a simple workaround may be
> possible.
Best of luck.
Raphael Bustin
to rnclark)" <user...@qwest.net> wrote:
>I routinely upsample and run the Richardson-Lucy algorithm
>and get close to double the pixels, turning an 8-megapixel
>DSLR image into 32 megapixels. You can do this with
>digital because of the higher signal-to-noise ratio of
>digital. You can't do this with film very well because
>of the noise from grain. So with 8-megapixel DSLR images,
>I get much sharper 16x24 inch prints than I ever did from
>35mm film printed traditionally or from drum scanned images.
I'll agree with your conclusion if not the logic.
Not sold yet on the alleged merits/benefits of R-L
over any other sharpening method, or magically
"doubling" pixel counts by any means. I'm
really surprised that you and Bart keep trying
to sell us that Hocus-Pocus.
Not sold on the (major) adavantages of drum
scanning over an LS-8000, having compared
both very carefully.
OTOH, I'm constantly amazed at how captures
from a 10D can look really good at 16x24".
Really kind of uncanny.
I'm just back from Taos and Santa Fe. At Taos
I saw an exhibition of prints... 20x30", printed on
Somerset Velvet... and taken with a Nikon D70.
A bit oversharpened for my tastes, but they were
quite impressive nevertheless.
Oh, and I saw a nice Paul Strand collection at
the Georgia O'Keefe museum in Santa Fe.
rafe b
www.terrapinphoto.com
Roger N. Clark (change username to rnclark)
> On Mon, 16 Oct 2006 20:02:23 -0700, "Roger N. Clark (change username
> to rnclark)" <user...@qwest.net> wrote:
>
>>I routinely upsample and run the Richardson-Lucy algorithm
>>and get close to double the pixels, turning an 8-megapixel
>>DSLR image into 32 megapixels. You can do this with
>>digital because of the higher signal-to-noise ratio of
>>digital. You can't do this with film very well because
>>of the noise from grain. So with 8-megapixel DSLR images,
>>I get much sharper 16x24 inch prints than I ever did from
>>35mm film printed traditionally or from drum scanned images.
>
> I'll agree with your conclusion if not the logic.
>
> Not sold yet on the alleged merits/benefits of R-L
> over any other sharpening method, or magically
> "doubling" pixel counts by any means. I'm
> really surprised that you and Bart keep trying
> to sell us that Hocus-Pocus.
Its not hocus pocus. It is grounded on solid
math.
> Not sold on the (major) adavantages of drum
> scanning over an LS-8000, having compared
> both very carefully.
I agree. I didn't mean to imply drum scans are
a major advantage. While I believe a drum scan
is the top, but the difference is not that great from
new consumer scanners, and at this point I'll only
have a drum scan of a super 4x5. But then I
haven't shot a 4x5 this year, because, as you know,
I've been doing digital mosaics.
> OTOH, I'm constantly amazed at how captures
> from a 10D can look really good at 16x24".
> Really kind of uncanny.
>
> I'm just back from Taos and Santa Fe. At Taos
> I saw an exhibition of prints... 20x30", printed on
> Somerset Velvet... and taken with a Nikon D70.
>
> A bit oversharpened for my tastes, but they were
> quite impressive nevertheless.
>
> Oh, and I saw a nice Paul Strand collection at
> the Georgia O'Keefe museum in Santa Fe.
Cool. Wish I was there. I'm stuck in meetings.
Roger
frederick
>
>> Epson prices in some countries are more than double Japanese prices for
>> the same cartridges.
>
> Ouch! That's painful.
>
Japanese price:
1102 yen
1775 yen ~ MSRP in USA
2646 yen ~ MSRP in UK (includes tax - but would still be double
Japanese price with no VAT tax)
MSRP's are from Epson UK/USA sites. The cartridges are discounted by
various online resellers - but 10% or so is usually about the best you
can hope for. Some offer better "headline" discounts, but gouge on freight.
Thanks.
David J. Littleboy
>>
>> Not sold yet on the alleged merits/benefits of R-L over any other
>> sharpening method, or magically "doubling" pixel counts by any means.
>> I'm really surprised that you and Bart keep trying to sell us that
>> Hocus-Pocus.
>
> Its not hocus pocus. It is grounded on solid
> math.
Maybe. But the example Bart put up the other day definately showed artifacts
that were unjustified by the data. The tip of a leaf, where it came to a
sharp point, was made far brighter than any pixel values in the region would
justify it being. Like regular sharpening, this is something that simply
wasn't in the original data. It was calculating what _might have been
there_, not what actually was.
When viewed at an appropriate distance, sharpening halos actually do give
the impression of more detail, and I'm sure your more sophisticated
manipulations are even better. Sharpening halos have the problem that
there's an optimal viewing distance; they don't allow the viewer to walk up
to the print and see more detail. Maybe the R-L stuff doesn't have that
problem.
Anyway, if you want to talk about optimal rendition of information actually
captured, we can have a discussion. But "more resolution"? We really are
getting painfully close to "Ignobel prize" territory.
Roger N. Clark (change username to rnclark)
>>Digital has a much higher signal-to-noise ratio
>
> Such a blank statement is definitely wrong; you mean the same
> exposition, and it is not applicable to my comparison. E.g, with the
> current technology digital will give much worse noise than film if the
> exposition of digital is 20x smaller than of film ;-). And note that
> this is quite probably holds when you compare LF film to FF digital
> (but I do not remember the details of the exposition you used...) -
> count number of photons per "pixel" (I mean square with size
> determined by the MTF curve).
By ANY measure that is even close to similar between
digital and film, digital wins. If you want to define an
area, then fine, but digital gets the same area, which
means you average pixels drop the noise, and increase the
dynamic range too.
> Film has practically no noise when correctly exposed with aperture
> f/32 - or much smaller f-number.
Well, that is simply not true. Film has grain and grain
clumps. That translates to irregularities people see in
images, of variations measured by a densitometer.
And it has nothing to do with f/stop.
>> Digital also has much greater dynamic range.
>
> IMO, this remains to be proven yet (though I suspect that the proof
> *will* support your claim - I just *have not seen* it).
e.g.:
Dynamic Range and Transfer Functions of Digital Images
and Comparison to Film
http://www.clarkvision.com/imagedetail/dynamicrange2
See Figures 5 and 8. I n figure 8, not how noisy the film
characteristic curve is compared to the digital (blue dots).
> Your investigation is very interesting, but, IMO, it completely missed
> the point. Essentially, you measured the noise of 6um square of film
> vs 8um square of semiconductor. So, film's noise *at these extremely
> high spacial frequencies* turns out to be much more than digital
> sensors'; fine. But, as you probably saw it in many posts about your
> experiments, it is the noise at much lower spacial frequencies what
> many people consider the measure of dynamic range.
1) other digital cameras with 6-micron pixels have pretty
similar S/N (e.g. Canon 20D curves fall almost on top of the
1D Mark II curves, and all digital cameras with similar
sized pixels are well above film at similar spatial resolutions.
2) If you want to average larger areas of film to improve
the S/N and dynamic range, it is only fair to do the
same with digital. Digital will always win.
> So how one could measure "true dynamic range"? Essentially, one
> cannot separate dynamic range from resolution; e.g., in each
> exposition zone, one could measure the S/N ratio at different spacial
> frequencies, and find at which frequency S/N ratio goes above 3 (or
> some other reasonable number).
It doesn't matter how you define it, as long as you define
it equally for all media. Digital wins in all cases,
because if it wins on a pixel basis, it wins when you
average larger areas.
> One gets a curve of (thus defined) spacial resolution per degree of
> underexposition; *this curve*, IMO, is the measure of dynamic range.
>
> Again: I agree with you that such an investigation will *most
> probably* give film much lower mark than digital sensor - but 10x20in
> film is practical, and digital is not.
If you have to go to 10x20 inch film to beat a small digital
sensor in dynamic range, it is not worth much. In the time
one took to slide the 10x20 inch plate holder into place,
one could take 2 different exposures providing 15 stops of
dynamic range. Gee you could probably take 4 or 5 exposures,
getting greater than 25 stops. Not that one needs that
most of the time.
> BTW, the question I addressed
> in my initial posting is also related to this: for which f-numbers
> this advantage of film will disappear, since FF digital sensor may
> behave as well as HUGE area of film.
While your original post referenced my digital mosaics
web article, I feel you were quite off in the f/ratio
tables you posted. I routinely make 30x40 inch prints
from f/32 to f/64 4x5 Velvia transparencies. I've even
done 40 x 60 inch prints from f/45 4x5s. The detail is quite
impressive. My own testing, with is documented on my
web pages at http://www.clarkvision.com/imagedetail
puts 4x5 Velvia at 200 megapixel digital equivalent.
However, digital, with its high signal to noise ratios,
can be spatially improved with image deconvolution
thus 50 megapixel digital could be close to 4x5 in
total image quality.
Roger
Roger N. Clark (change username to rnclark)
> "Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
>
>>Raphael Bustin wrote:
>>
>>>Not sold yet on the alleged merits/benefits of R-L over any other
>>>sharpening method, or magically "doubling" pixel counts by any means.
>>>I'm really surprised that you and Bart keep trying to sell us that
>>>Hocus-Pocus.
>>
>>Its not hocus pocus. It is grounded on solid
>>math.
>
>
> Maybe. But the example Bart put up the other day definately showed artifacts
> that were unjustified by the data. The tip of a leaf, where it came to a
> sharp point, was made far brighter than any pixel values in the region would
> justify it being. Like regular sharpening, this is something that simply
> wasn't in the original data. It was calculating what _might have been
> there_, not what actually was.
Like any reconstruction method, knowing the blur function
accurately directly impacts the solution. In my experience,
it is better to be conservative to reduce artifacts.
The article:
Large Digital Mosaics as a Substitute for Large Format Film
http://www.clarkvision.com/photoinfo/large_mosaics
included a Richardson-Lucy deconvolution step, and I tried
to minimize the artifacts.
> When viewed at an appropriate distance, sharpening halos actually do give
> the impression of more detail, and I'm sure your more sophisticated
> manipulations are even better. Sharpening halos have the problem that
> there's an optimal viewing distance; they don't allow the viewer to walk up
> to the print and see more detail. Maybe the R-L stuff doesn't have that
> problem.
>
> Anyway, if you want to talk about optimal rendition of information actually
> captured, we can have a discussion. But "more resolution"? We really are
> getting painfully close to "Ignobel prize" territory.
Yes, and not prize territory. R-L and similar methods have
been used in the sciences for decades. R-L was a method of choice
in reconstructing the Hubble space telescope images before
the optics fix, for example.
No imaging system is perfect. They all have artifacts.
Flair in the lens system, focus and depth of field limitations.
Film often has >100% MTF at some frequencies. Digital
has optical blur filters, Bayer interpolation algorithms
and accutance enhancement algorithms (also called unsharp
mask), all producing artifacts.
Nothing is perfect; everything is a compromise in photography
(as in most things in life).
Roger
Raphael Bustin
<dav...@gol.com> wrote:
>
>"Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
>> Raphael Bustin wrote:
>>>
>>> Not sold yet on the alleged merits/benefits of R-L over any other
>>> sharpening method, or magically "doubling" pixel counts by any means.
>>> I'm really surprised that you and Bart keep trying to sell us that
>>> Hocus-Pocus.
>>
>> Its not hocus pocus. It is grounded on solid
>> math.
>
>Maybe. But the example Bart put up the other day definately showed artifacts
>that were unjustified by the data. The tip of a leaf, where it came to a
>sharp point, was made far brighter than any pixel values in the region would
>justify it being. Like regular sharpening, this is something that simply
>wasn't in the original data. It was calculating what _might have been
>there_, not what actually was.
Unless I'm wrong, R-L requires characterizing lens distortion.
But really now... are we going to do that over a continuum of
focal lengths, focus distances, and apertures? Will the
characterization be constant throughout the image circle?
Will we repeat this process for each lens in the kit, and
record all taking parameters, etc.?
I mean, seriously... how well can a lens (not to mention
a zoom lens) be characterized?
Roger's fox-eye comparisons haven't sold me. In particular,
Roger's choice of USM parameters is strange. For that
particular image sample -- after careful upsampling -- I'd
have used a much smaller radius and much higher
amount. (Eg. radius 1.0 and amount = 400%).
rafe b
www.terrapinphoto.com
Roger N. Clark (change username to rnclark)
> Unless I'm wrong, R-L requires characterizing lens distortion.
No, it is any distortion. You can even do motion blur.
>
> But really now... are we going to do that over a continuum of
> focal lengths, focus distances, and apertures? Will the
> characterization be constant throughout the image circle?
>
> Will we repeat this process for each lens in the kit, and
> record all taking parameters, etc.?
>
> I mean, seriously... how well can a lens (not to mention
> a zoom lens) be characterized?
>
> Roger's fox-eye comparisons haven't sold me. In particular,
> Roger's choice of USM parameters is strange. For that
> particular image sample -- after careful upsampling -- I'd
> have used a much smaller radius and much higher
> amount. (Eg. radius 1.0 and amount = 400%).
You are welcome to download the image and try. Many have.
Only Bill Hilton has equaled the R-L results using
edge detection masks plus unsharp mask (I gotta add that
result to the web page sometime).
Roger
David J. Littleboy
"Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
> Ilya Zakharevich wrote:
>
>> Film has practically no noise when correctly exposed with aperture
>> f/32 - or much smaller f-number.
>
> Well, that is simply not true. Film has grain and grain
> clumps. That translates to irregularities people see in
> images, of variations measured by a densitometer.
> And it has nothing to do with f/stop.
It does, indirectly. The f stop tells you what the zero MTF point is
(1600/(f number)) which tells you the aperture to use with the densitometer
(or how strong noise reduction one can apply without losing detail, or the
maximum sensible ppi with which to scan).
When you handle the film appropriately for the (lack of) detail caused by
diffraction, the noise due to grain goes down the larger the f number.
Of course "practically no noise" doesn't mean zero, and digital will still
do a lot better, even at f/32.
> If you have to go to 10x20 inch film to beat a small digital
> sensor in dynamic range, it is not worth much.
Really.
> While your original post referenced my digital mosaics
> web article, I feel you were quite off in the f/ratio
> tables you posted. I routinely make 30x40 inch prints
> from f/32 to f/64 4x5 Velvia transparencies. I've even
> done 40 x 60 inch prints from f/45 4x5s. The detail is quite
> impressive.
It's pretty easy to calculate. f/45 projects an image with MTF 50 at 18
lp/mm onto the film. Since you should probably scan that at 3 pixels per
line pair, that's an (18 x 3 x 25.4) = 1371 ppi scan, which will make a nice
4x enlargement = 16x20. That's a 34MP scan. I doubt you can get much more
than that, since the MTF will be falling from 50% to 0% between 18 and 38
lp/mm with 25% at 27 lp/mm. I suppose you could calculate two pixels per
cycle at extinction and get (36 x 2 x 25.4) = 1829, which is still only 67
MP.
> My own testing, with is documented on my
> web pages at http://www.clarkvision.com/imagedetail
> puts 4x5 Velvia at 200 megapixel digital equivalent.
Not at f/45 it isn't.
> However, digital, with its high signal to noise ratios,
> can be spatially improved with image deconvolution
> thus 50 megapixel digital could be close to 4x5 in
> total image quality.
Especially if you are shooting the 4x5 at f/45!
Raphael Bustin
to rnclark)" <user...@qwest.net> wrote:
>You are welcome to download the image and try. Many have.
>Only Bill Hilton has equaled the R-L results using
>edge detection masks plus unsharp mask (I gotta add that
>result to the web page sometime).
I did exactly that, Roger. Upon downloading the
image, I upsampled in four or five steps, and then
sharpened with Amt=400-500, Radius=0.8-1.2.
I find my parameters much less garish, and much
more subtle than your USM sample. Yours does
show more "apparent" sharpness, but also more
artifacts. Bottom line: if faced with the same
upsampling reqiurement, I wouldn't dare sharpen
to the extent that you did with USM.
In both your USM sample and in your RL sample,
there is visible staircasing on some of the fox's
whiskers and facial hairs.
You must be aware that these are the two most
contentious and subjective issues in digital image
processing. RL gives a sharp fox, but it's still
not a real fox.
rafe b
www.terrapinphoto.com
Roger N. Clark (change username to rnclark)
> "Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
>
>>Ilya Zakharevich wrote:
>>
>>
>>>Film has practically no noise when correctly exposed with aperture
>>>f/32 - or much smaller f-number.
>>
>>Well, that is simply not true. Film has grain and grain
>>clumps. That translates to irregularities people see in
>>images, of variations measured by a densitometer.
>>And it has nothing to do with f/stop.
>
> It does, indirectly. The f stop tells you what the zero MTF point is
> (1600/(f number)) which tells you the aperture to use with the densitometer
> (or how strong noise reduction one can apply without losing detail, or the
> maximum sensible ppi with which to scan).
Yes, I see that point, but to maintain resolution, you
need much smaller aperture than you seem to think, as I'll shown below.
>
> When you handle the film appropriately for the (lack of) detail caused by
> diffraction, the noise due to grain goes down the larger the f number.
Only after the image detail is lost. There is image detail
right down to the Dawes limit, again see below. For Velvia,
that detail reaches down to the film grain clumps, which
cause noise.
> Of course "practically no noise" doesn't mean zero, and digital will still
> do a lot better, even at f/32.
Yes
>>While your original post referenced my digital mosaics
>>web article, I feel you were quite off in the f/ratio
>>tables you posted. I routinely make 30x40 inch prints
>>from f/32 to f/64 4x5 Velvia transparencies. I've even
>>done 40 x 60 inch prints from f/45 4x5s. The detail is quite
>>impressive.
>
> It's pretty easy to calculate. f/45 projects an image with MTF 50 at 18
> lp/mm onto the film. Since you should probably scan that at 3 pixels per
> line pair, that's an (18 x 3 x 25.4) = 1371 ppi scan, which will make a nice
> 4x enlargement = 16x20. That's a 34MP scan. I doubt you can get much more
> than that, since the MTF will be falling from 50% to 0% between 18 and 38
> lp/mm with 25% at 27 lp/mm. I suppose you could calculate two pixels per
> cycle at extinction and get (36 x 2 x 25.4) = 1829, which is still only 67
> MP.
This MTF50 would be losing a LOT of detail, detail important in
showing texture in a big print. See the demonstration
of sampling of the grass field, bottom image set at:
http://www.clarkvision.com/imagedetail/sampling1.html
The bottom frame is sampling 2 pixels/cycle at 50% MTF
Comparing the bottom frame (50% MTF) with sampling at the
top frame (3 pixels per cycle at the Dawes limit) is
a huger difference. There is even a difference between
3 pixels/cycle at the Dawes limit and 2 pixels/cycle at the
Rayleigh limit. And that detail is noticeable in 30x40 inch
prints. I wouldn't scan 4x5 f/45 Velvia at less than
3200 ppi, or image detail is lost; detail important in
showing texture in large prints.
>> My own testing, with is documented on my
>>web pages at http://www.clarkvision.com/imagedetail
>>puts 4x5 Velvia at 200 megapixel digital equivalent.
>
> Not at f/45 it isn't.
See above. You can't simply calculate 50% MTF and say
you've got it all. It's the MTF curve shape differences
that fed the film versus digital wars for years.
>>However, digital, with its high signal to noise ratios,
>>can be spatially improved with image deconvolution
>>thus 50 megapixel digital could be close to 4x5 in
>>total image quality.
>
> Especially if you are shooting the 4x5 at f/45!
Velvia film grain is still a contributor to limiting
image detail at f/45.
Roger
David J. Littleboy
> David J. Littleboy wrote:
>> "Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
>>
>>>Ilya Zakharevich wrote:
>>>
>>>
>>>>Film has practically no noise when correctly exposed with aperture
>>>>f/32 - or much smaller f-number.
>>>
>>>Well, that is simply not true. Film has grain and grain
>>>clumps. That translates to irregularities people see in
>>>images, of variations measured by a densitometer.
>>>And it has nothing to do with f/stop.
>>
>> It does, indirectly. The f stop tells you what the zero MTF point is
>> (1600/(f number)) which tells you the aperture to use with the
>> densitometer (or how strong noise reduction one can apply without losing
>> detail, or the maximum sensible ppi with which to scan).
>
> Yes, I see that point, but to maintain resolution, you
> need much smaller aperture than you seem to think, as I'll shown below.
What I think is going on here is that you don't like my pet and favorite
resolution estimate (800/(f number) = MT50 and 1600/(f number) =
extinction).
I went looking for a good reference for it, and didn't find one, so it may
be wrong. But assuming that it's a good estimate for the nonce...
>> When you handle the film appropriately for the (lack of) detail caused by
>> diffraction, the noise due to grain goes down the larger the f number.
>
> Only after the image detail is lost. There is image detail
> right down to the Dawes limit, again see below.
Well, yes. But the Ilya was talking about f/32, and you were talking about
f/45.
Either my estimate of resolution is _way_ off, or yours is.
> For Velvia,
> that detail reaches down to the film grain clumps, which
> cause noise.
Not at f/32 and f/45.
>>>While your original post referenced my digital mosaics
>>>web article, I feel you were quite off in the f/ratio
>>>tables you posted. I routinely make 30x40 inch prints
>>>from f/32 to f/64 4x5 Velvia transparencies. I've even
>>>done 40 x 60 inch prints from f/45 4x5s. The detail is quite
>>>impressive.
Note that you (Roger) are talking about f/45. I suspect you are (a) way
underestimating diffraction, and (b) way _overestimating_ what is required
to make an attractive looking print.
>> It's pretty easy to calculate. f/45 projects an image with MTF 50 at 18
>> lp/mm onto the film. Since you should probably scan that at 3 pixels per
>> line pair, that's an (18 x 3 x 25.4) = 1371 ppi scan, which will make a
>> nice 4x enlargement = 16x20. That's a 34MP scan. I doubt you can get much
>> more than that, since the MTF will be falling from 50% to 0% between 18
>> and 38 lp/mm with 25% at 27 lp/mm. I suppose you could calculate two
>> pixels per cycle at extinction and get (36 x 2 x 25.4) = 1829, which is
>> still only 67 MP.
>
> This MTF50 would be losing a LOT of detail, detail important in
> showing texture in a big print.
Probably. For smaller f numbers and scanning, 3 pixels at MTF50 is a good
estimate, since the film is going to be imposing a painful MTF term as well.
But at 18 lp/mm, there's still quite a bit beyond that.
Which is why I calculated for the extinction point. Scanning at 2 pixels per
cycle at the point of zero MTF should be fine, no???
> See the demonstration
> of sampling of the grass field, bottom image set at:
> http://www.clarkvision.com/imagedetail/sampling1.html
> The bottom frame is sampling 2 pixels/cycle at 50% MTF
> Comparing the bottom frame (50% MTF) with sampling at the
> top frame (3 pixels per cycle at the Dawes limit) is
> a huger difference. There is even a difference between
> 3 pixels/cycle at the Dawes limit and 2 pixels/cycle at the
> Rayleigh limit. And that detail is noticeable in 30x40 inch
> prints. I wouldn't scan 4x5 f/45 Velvia at less than
> 3200 ppi, or image detail is lost; detail important in
> showing texture in large prints.
I'm not sure what you are talking about here. At f/45, there's no
information in the image that hits the film (let alone the recorded image)
at over 1850 ppi (36 lp/mm).
I really think you are _way_ overestimating 4x5 at f/45.
>>> My own testing, with is documented on my
>>>web pages at http://www.clarkvision.com/imagedetail
>>>puts 4x5 Velvia at 200 megapixel digital equivalent.
>>
>> Not at f/45 it isn't.
>
> See above. You can't simply calculate 50% MTF and say
> you've got it all. It's the MTF curve shape differences
> that fed the film versus digital wars for years.
Scanning at 3 pixels per cycle at 50% MTF (for only one component of the
system!) is, in my experience a sensible thing to do, but I also calculated
scanning at 2 pixels per cycle at 0% MTF (for only one component of the
system!).
>>>However, digital, with its high signal to noise ratios,
>>>can be spatially improved with image deconvolution
>>>thus 50 megapixel digital could be close to 4x5 in
>>>total image quality.
>>
>> Especially if you are shooting the 4x5 at f/45!
>
> Velvia film grain is still a contributor to limiting
> image detail at f/45.
I guess I don't get it. You've got zero, zilch, nothing, nada in the image
on the film at 36 lp/mm. I thought Velvia was way better than that.
Roger N. Clark (change username to rnclark)
David,
I think the problem here regarding MTF and thinking
resolution is that there is more to images than simply
bar charts. For example, stars are smaller than the diffraction
limit but are still detected and recorded on film.
Sampling film by scanning is more than resolving
lines in bar charts. One needs to accurately portray
edges and small things like blades of grass, veins
in a leaf, hair on an animal, stars etc. One needs high
sampling to capture that detail. That has to do
with both phasing and sample (pixel) size to capture that
detail. If you scanned an f/45 4x5 of a star field at
1372 ppi the result would be pretty poor, whereas
3000+ ppi scans would appear pretty spectacular.
Roger
David J. Littleboy
"Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
> David J. Littleboy wrote:
>>
>> What I think is going on here is that you don't like my pet and favorite
>> resolution estimate (800/(f number) = MT50 and 1600/(f number) =
>> extinction).
>>
>> I went looking for a good reference for it, and didn't find one, so it
>> may be wrong. But assuming that it's a good estimate for the nonce...
>
> I think the problem here regarding MTF and thinking
> resolution is that there is more to images than simply
> bar charts. For example, stars are smaller than the diffraction
> limit but are still detected and recorded on film.
By the time a star's image gets to the image plane, it's not a point, it's
an image that looks exactly like the diffraction disk.
> Sampling film by scanning is more than resolving
> lines in bar charts. One needs to accurately portray
> edges and small things like blades of grass, veins
> in a leaf, hair on an animal, stars etc.
That's fine. But if those edges have all been smeared by diffraction,
they're not there in the image recorded on the film since they were not in
the image that hit the film. And (to the best of my understanding) talking
about diffraction as an MTF term in the system transfer function is quite
sensible and reasonable.
This is very odd. It sounds as though you don't, at the gut level, accept
that diffraction actually destroys detail.
> One needs high sampling to capture that detail.
Only if there is detail there to be captured.
> That has to do
> with both phasing and sample (pixel) size to capture that
> detail. If you scanned an f/45 4x5 of a star field at
> 1372 ppi the result would be pretty poor, whereas
> 3000+ ppi scans would appear pretty spectacular.
(Note that my claim is that 1372 is for practical purposes, and that 1830 is
the maximum ppi that makes scientific sense.)
To the best of my understanding, you couldn't tell the difference between an
1830 ppi scan and a 3000 ppi scan; all information above 36 lp/mm has been
lost due to diffraction.
Scott W
yours.
http://www.pbase.com/konascott/image/54355011/original
But what I do is to convert the image into a linear space before
applying the USM
and then back to sRGB after. For some odd reason PS is assuming that
the pixels are linear when doing USM, which is why you get many of the
odd artifacts, light really bright pixels.
Scott
Scott W
> This is very odd. It sounds as though you don't, at the gut level, accept
> that diffraction actually destroys detail.
I think Roger is believing he is seeing detail in the grass part of his
photos that is in
fact grain, this is pretty clear since the diffraction spot is lager
then some of the detail in the grass.
Scott
David J. Littleboy
"Scott W" <bip...@hotmail.com> wrote in message
news:1161103539.4...@b28g2000cwb.googlegroups.com...
> David J. Littleboy wrote:
>> This is very odd. It sounds as though you don't, at the gut level, accept
>> that diffraction actually destroys detail.
> I think Roger is believing he is seeing detail in the grass part of his
> photos that is in
> fact grain, this is pretty clear since the diffraction spot is lager
> then some of the detail in the grass.
Yep. I did too; I couldn't make any sense out of his examples that he
thought were showing more detail; I'm just being slower in figuring that
out.
I suspect that the problem here (especially with the R-L stuff) is that he
thinks about images as though they were star pattern. But it's a lot easier
to "resolution enhance" star patterns since to a very large degree you know
what you are looking for. But since you don't have anywhere near that good
an idea what pictorial images are, the resolution enhancement game doesn't
apply. (This is naively expressed, and the math probably claims that each
point in a pictorial image acts like a low brightness star; but I suspect
that you have to handle a small number of (extremely intense) true points
very differently from smoothly changing gradients with an occasional
low-contrast edge.)
Greg "_"
"David J. Littleboy" <dav...@gol.com> wrote:
> "Greg "_"" <grey_egg@greg_photo.com> wrote:
> > Then again; maybe Matt has a tad more critical eye than you common folk.
>
> No: Matt claims Matt gets better enlargements from 35mm, whereas everyone
> else finds the digital to hold up better.
A -soo David san- then one must qualify the digital & as well as the
film, asa 50 film versus a 2mp sensor :)
>Same eyes doing the comparisons in
> both cases.
One blind don't count ?
>Besides, we're talking about really really ugly prints that
> you'd have to be blind not to see the problems with if you walk up for a
> closer look.
for giggles define ugly :)
> David J. Littleboy
> Tokyo, Japan
Greg "_"
Raphael Bustin <ra...@speakeasy.net> wrote:
> Oh, and I saw a nice Paul Strand collection at
> the Georgia O'Keefe museum in Santa Fe.
Paul Strand IMOP is the man, worked in LF Platinum
influenced-inspired me early on, and next to AA is seriously
underrated.
David J. Littleboy
"Greg "_"" <grey_egg@greg_photo.com> wrote:
>
>>Besides, we're talking about really really ugly prints that
>> you'd have to be blind not to see the problems with if you walk up for a
>> closer look.
>
> for giggles define ugly :)
That's easy: looks significantly worse in an A/B comparison with a print
from the next format up.
Scott W
My criteria for the technical quality of a good print is that if I were
to cut out a 4 x 6 piece it would look good as a 4 x 6 print. I have
seen a lot of enlargements that might look ok when viewed from a
distance but if you were to cut out a 4 x 6 piece anyone looking at it
would agree it was crap. Obviously this is not looking at the
composition of the photo but just the sharpness and level of visible
noise.
Scott
David J. Littleboy
I'm still trying to decide if I'm that fussy. On a gallery wall, people will
walk up to within 10 inches or so, so 220 ppi (5D at 13x19) is just on the
edge of problematic. (200 ppi from cameras with stronger AA filters can be
seen to be lacking in detail at 10".) But handed an 8x10 or A4, some people
will put their noses on it.
Roger N. Clark (change username to rnclark)
At first look, your image does appear pretty close.
I put your image into photoshop and diffed it with the R-L
image (Figure 6 from the web page) and found a strange
odd-even pattern. Then looking at the original two
images (yours and Figure 6), the up sampling artifacts
are quite apparent, and your image has squarish pixels
double the size of the Figure 6 pixels. Put the two
images in photoshop and look at the eye at 300%
What up-sampling algorithm did you use?
Roger
Roger N. Clark (change username to rnclark)
> On Mon, 16 Oct 2006 22:53:07 -0700, "Roger N. Clark (change username
> to rnclark)" <user...@qwest.net> wrote:
>
>
>
>>You are welcome to download the image and try. Many have.
>>Only Bill Hilton has equaled the R-L results using
>>edge detection masks plus unsharp mask (I gotta add that
>>result to the web page sometime).
>
>
>
> I did exactly that, Roger. Upon downloading the
> image, I upsampled in four or five steps, and then
> sharpened with Amt=400-500, Radius=0.8-1.2.
>
> I find my parameters much less garish, and much
> more subtle than your USM sample. Yours does
> show more "apparent" sharpness, but also more
> artifacts. Bottom line: if faced with the same
> upsampling reqiurement, I wouldn't dare sharpen
> to the extent that you did with USM.
>
> In both your USM sample and in your RL sample,
> there is visible staircasing on some of the fox's
> whiskers and facial hairs.
Rafe,
Yes, I agree with you on the staircase problems.
But those are not due to R-L, those are due
to the up-sampling algorithm. The up-sampling
was a simple cubic spline in Photoshop. I have
not been happy with cubic cpline but haven't seen
anything better. Have you?
> You must be aware that these are the two most
> contentious and subjective issues in digital image
> processing. RL gives a sharp fox, but it's still
> not a real fox.
Yes, I agree. But the fox wasn't "real" before any processing
either. The camera optics smeared the detail, and the
raw converter interpolated RGB pixels to make the
RGB tif, producing other artifacts along the way.
Roger
Scott W
to linear space used good old bicubic for up sampling then used USM and
then translated back to sRGB. USM works like crap when working on sRGB
but does ok when working in a linear space. A good test of this is
using USM on uniform noise, when doing it on an sRGB image the over all
brightness increases, clearly this should not be the case. If first
convert to linear space and then use USM and then convert back the over
all brightness does not change. This is a really silly bug in PS and
it would be easy for them to fix it. The halos that come from using
USM are also not nearly as bad when using it in linear space.
Scott
Roger N. Clark (change username to rnclark)
The diffraction spot in an image CAN appear smaller than
the full diffraction spot. A diffraction spot has
a narrow peak. With high contrast film, like Velvia,
you can lose the bottom of the diffraction pattern
leaving only the peak. You can try this yourself:
generate a diffraction pattern, then slice off the lower end.
The higher you slice, the smaller the remaining spot
appears.
While diffraction limits detail, the diffraction limit (Dawes,
or 0% MTF) does NOT mean no detail. Example: two point sources,
or two lines close together have a broadened combined diffraction
pattern. But the fact that the pattern IS broadened
is information from which you can recover information
about the sources. The separation of stars in
telescopes can be determined when they are closer than
the Dawes limit (the 0% MTF point).
Second in the equation of image detail is
sampling and the phase of the sampling
by the scanner pixels. If you choose pixels spaced at just above
the Dawes limit, in theory you get all the information,
but only if the image detail falls perfectly on the scanner pixels.
Statistically it does not. A good demonstration of that
is detection of an edge that is not aligned to the pixel
grid. The higher you sample that edge, the closer you
can come to recording the information recorded by the
optical system, including sampling above the 0% MTF.
0% MTF means you can't resolve a grid of lines
at that frequency; it does not mean you can't
infer/compute or even visually see
two closely spaced lines or points.
Phase effects are illustrated in the grid at
(first figure on the page):
http://www.clarkvision.com/imagedetail/sampling1.html
Below 3x Nyquist sampling phasing errors cause image
artifacts. The difference here and in
scanned film, or digital cameras, is the sampling is
finite, and that changes the problem from the Nyquist
theorem.
So, then in the grass field image on the above web page,
it you stand back 15 feet or more, and tell me which
image looks the sharpest. I know for a fact, from
being at the scene, that the grass field at the top of
the image has long blades of grass sticking up. The
curvy-linear strands of bright posts are those grass
blades below the diffraction limit. The image is a
combined effect of grain clumps limiting resolution, and
diffraction spots. So then the question is what
ppi is needed to record that detail. There are 4
levels shown on the figure. Which one do you think
is adequate?
Roger
Roger N. Clark (change username to rnclark)
That's what R-L does. It is an iterative process.
Perhaps try reading one of the technical papers:
http://www.stsci.edu/stsci/meetings/irw/proceedings/whiter_damped.dir/whiter_damped.html
This page smears images then reconstructs them:
http://www.adass.org/adass/proceedings/adass97/prukschm.html
Algorithms like R-L have been in development for decades
with many scientific papers written on them. The theory
is well developed and they work well in practice.
But like any algorithm, it needs good data and should
not be abused.
Roger
Ilya Zakharevich
Roger N. Clark (change username to rnclark)
> >>Digital has a much higher signal-to-noise ratio
> > Such a blank statement is definitely wrong; you mean the same
> > exposition, and it is not applicable to my comparison. E.g, with the
> > current technology digital will give much worse noise than film if the
> > exposition of digital is 20x smaller than of film ;-). And note that
> > this is quite probably holds when you compare LF film to FF digital
> > (but I do not remember the details of the exposition you used...) -
> > count number of photons per "pixel" (I mean square with size
> > determined by the MTF curve).
> By ANY measure that is even close to similar between
> digital and film, digital wins.
In what I investigated, I studied the case when digital will be able
to win EVEN IF is the situation NOT IN ANY WAY close to film. E.g.,
4x5in film with f/64 vs FF sensor with f/1.
> > Film has practically no noise when correctly exposed with aperture
> > f/32 - or much smaller f-number.
> Well, that is simply not true. Film has grain and grain
> clumps. That translates to irregularities people see in
> images, of variations measured by a densitometer.
> And it has nothing to do with f/stop.
Just the opposite. E.g., the variations measured by a densitometer
will be very different depending on the area over which it averages
the density. And with f/64 the area of e.g., diffraction circle is so
large that the variations will be negligible.
> > IMO, this remains to be proven yet (though I suspect that the proof
> > *will* support your claim - I just *have not seen* it).
>
> e.g.:
> Dynamic Range and Transfer Functions of Digital Images
> and Comparison to Film
> http://www.clarkvision.com/imagedetail/dynamicrange2
> See Figures 5 and 8. I n figure 8, not how noisy the film
> characteristic curve is compared to the digital (blue dots).
As I discussed below, this investigation has very little to do with
what most people will call dynamic range. A measurement of
high-spacial-frequency-Noise is not a measurement of dynamic range.
> It doesn't matter how you define it, as long as you define
> it equally for all media. Digital wins in all cases,
> because if it wins on a pixel basis, it wins when you
> average larger areas.
Remember claims "one needs 16MP sensor to get similar performance to
the film"? It was very fine - just wrong. Your argument is very
convincing, but still "just an argument"; there is A LOT of
assumptions needed to true for your argument to work (e.g. the same
spectrum of noise).
As I said, until it is ACTUALLY measured, there is very little hope to
establish it "just by theoretizing".
> > One gets a curve of (thus defined) spacial resolution per degree of
> > underexposition; *this curve*, IMO, is the measure of dynamic range.
> >
> > Again: I agree with you that such an investigation will *most
> > probably* give film much lower mark than digital sensor - but 10x20in
> > film is practical, and digital is not.
>
> If you have to go to 10x20 inch film to beat a small digital
> sensor in dynamic range, it is not worth much. In the time
> one took to slide the 10x20 inch plate holder into place,
> one could take 2 different exposures providing 15 stops of
> dynamic range. Gee you could probably take 4 or 5 exposures,
> getting greater than 25 stops. Not that one needs that
> most of the time.
> While your original post referenced my digital mosaics
> web article, I feel you were quite off in the f/ratio
> tables you posted. I routinely make 30x40 inch prints
> from f/32 to f/64 4x5 Velvia transparencies. I've even
> done 40 x 60 inch prints from f/45 4x5s. The detail is quite
> impressive.
The details at the particular f/64 shot I investigated correspond to
14MP JPEG. Frankly speaking, I investigated only the crops you
provided on this web page; are they in slightly-out-of-focus area
then, or what?
> My own testing, with is documented on my
> web pages at http://www.clarkvision.com/imagedetail
> puts 4x5 Velvia at 200 megapixel digital equivalent.
I did not investigate 4x5 Velvia. I investigated 4x5 Velvia with f/64
lens. It corresponds to about 15x lower MP count.
Hope this helps,
Ilya
Ilya Zakharevich
Ilya Zakharevich
<nospam...@ilyaz.org>], who wrote in article <eh4iun$ci9$1...@agate.berkeley.edu>:
> [A complimentary Cc of this posting was NOT [per weedlist] sent to
> > By ANY measure that is even close to similar between
> > digital and film, digital wins.
>
> In what I investigated, I studied the case when digital will be able
> to win EVEN IF is the situation NOT IN ANY WAY close to film. E.g.,
> 4x5in film with f/64 vs FF sensor with f/1.
Of course, I meant f/16. (Do I have dust under the "6" key? ;-)
Sorry,
Ilya
Ilya Zakharevich
frederick
<lo...@sea.com>], who wrote in article <1161034115.40328@ftpsrv1>:
> > I do not agree with you. Read: "on the plus side"; obviously, he
> > thinks that with smaller formfactor he will have larger depth of
> > field. (This is only true if he would use the same f-stop; but when
> > you convert to a more open f-stop needed to get the same diffraction
> > [measured in the subject space, i.e., in angular units], one gets
> > exactly the same DoF...)
> Yes - but he also gets increased DOF at the same f-stop , same FOV, and
> same shutter speed.
Still, I can't agree with you. The only reason why DOF increases is
that in-focus areas contain about 2.5x less linear detail, thus you
may go further from the focus plane until the defocus circle of
confusion reaches this size.
I can hardly call it "same something" - although, of course, on paper
you are 100% correct - you just did not mention how much blurrier the
image is. ;-)
> Ultimately you are of course correct - but it may mean that you can't
> take a photo with even a moving snail in the frame ;-)
Now will start to violently agree on what the other guy saiz... ;-)
Yours,
Ilya
Paul Rubin
> Yes, I agree with you on the staircase problems. But those are not
> due to R-L, those are due to the up-sampling algorithm. The
> up-sampling was a simple cubic spline in Photoshop. I have not been
> happy with cubic cpline but haven't seen anything better. Have you?
I thought that fancier upconverters used some interpolation involving
the sinc function.
frederick
rafe b
"Greg "_"" <grey_egg@greg_photo.com> wrote in message
news:grey_egg-7F14E6...@news.isp.giganews.com...
> In article <emj8j2tmkttbs6t54...@4ax.com>,
> Raphael Bustin <ra...@speakeasy.net> wrote:
>
>> Oh, and I saw a nice Paul Strand collection at
>> the Georgia O'Keefe museum in Santa Fe.
>
> Paul Strand IMOP is the man, worked in LF Platinum
> influenced-inspired me early on, and next to AA is seriously
> underrated.
Odd collection of prints. Most of them were small.
There was nothing larger than 11x14, as I recall.
The collection was arranged/sponsored by Aperture.
Some seemed a bit too large to be contact prints, but
too small to be enlargements. Many (most) were of
subjects in New Mexico. Many were dark and of
very poor (limited) tonality. Some showed faint
chemical stains.
Composition and sharpness were uniformly excellent.
I share your respect for Strand's work. Always a
pleasure to look at.
rafe
Paul Rubin
> Some seemed a bit too large to be contact prints, but
> too small to be enlargements.
If they were platinum prints they were all contact prints by
definition.
Roger N. Clark (change username to rnclark)
While I agree that using a large area to measure
dynamic range will reduce noise, but
it does not reduce noise seen in a print.
Kodak started this trick of averaging large areas
to get their data on film, including the characteristic
curve, and resolution, because film is pretty poor
at high spatial frequencies due to noise from film grain.
So let's say you average all pixels in a 10mm diameter
area, well do the same with digital pixels.
While film hits a floor beyond which there is no
information, digital keeps on digging signal
out of the noise, even below 1 photon per pixel per
exposure. (e.g. see:
http://www.clarkvision.com/photoinfo/night.and.low.light.photography ).
You can see dynamic range limitations in Figure 5 at:
http://www.clarkvision.com/imagedetail/dynamicrange2/index.html
Note the black round circle on the left in each image.
Your eye is effectively averaging pixels to see the subject.
It is easily seen in the digital, faintly seen in the
print film, and no hint of it in the slide film.
That will not change regardless of averaging some area.
Then look at Figure 8. Average blocks of pixels together.
It should be obvious to you that no matter how you average,
if you average the same block of pixels, each signal
will reduce in noise proportional (square root) to the number
of pixels. The least noisy sensor will remain the least
noisy sensor (digital) and show the greatest dynamic range.
Roger
Ilya Zakharevich
Roger N. Clark (change username to rnclark)
> While I agree that using a large area to measure
> dynamic range will reduce noise, but
> it does not reduce noise seen in a print.
References? It is hard to believe this claim... And it contradicts
your summary of night.and.low.light.photography.
> Kodak started this trick of averaging large areas
> to get their data on film, including the characteristic
> curve, and resolution, because film is pretty poor
> at high spatial frequencies due to noise from film grain.
Velvia 50 claims S/N ratio of 110 when averaged over a circle of
diameter 48um. (Do not remember which zone, but probably one
corresponding to 18%gray with normal exposure?) This beats or is on
par with almost any digital. I think it is related to f/32 or more
closed apertures.
> So let's say you average all pixels in a 10mm diameter
> area, well do the same with digital pixels.
I see that you are still under impression that averaging noise of
different physical origin will produce the same improvement. No, it
depends on spectrum of the noise. And until someone measures it,
theoretizing is futile.
> While film hits a floor beyond which there is no
> information, digital keeps on digging signal
> out of the noise, even below 1 photon per pixel per
> exposure. (e.g. see:
> http://www.clarkvision.com/photoinfo/night.and.low.light.photography ).
Sorry, I see no reference on film there.
> You can see dynamic range limitations in Figure 5 at:
> http://www.clarkvision.com/imagedetail/dynamicrange2/index.html
> Note the black round circle on the left in each image.
> Your eye is effectively averaging pixels to see the subject.
No, it is your scaling-down software which averaged i, not eye;
without data which way it was scaled down, it is still not "clean
enough" data... But if we knew the scaling algorithm, one could
indeed deduce something from these data.
> Then look at Figure 8. Average blocks of pixels together.
> It should be obvious to you that no matter how you average,
> if you average the same block of pixels, each signal
> will reduce in noise proportional (square root) to the number
> of pixels.
Right; obviously, your knowledge of DSP is still kinda rusty. Your
description is very plausible with digital (when the noise is [assumed
to be] white); we KNOW that the noise of film is NOT white, so it WILL
reduce differently than sqrt(scale).
> The least noisy sensor will remain the least noisy sensor (digital)
This is extremely PROBABLE; but the key question (when comparing to
larger format film) is HOW MUCH better. And: you will STILL be
operating with "1-pixel"-wide strip of film; not relevant to dynamic
range. One MUST start with 2-dim data to have a clean investigation.
> and show the greatest dynamic range.
Such a claim looks very plausible *theoretically*, but I do not see
any "experimental" data to be ABSOLUTELY SURE. IMO, one MUST do the
calculations before doing a claim like this.
Thanks,
Ilya
Greg "_"
"rafe b" <ra...@foobar.com> wrote:
I saw his large exhibit at the National Gallery I think after he
passed away. In any event I have the Gallery book. At first he used LF
cameras of various sizes, he was younger than Ansel Adams and E Weston,
he was well thought of by Alfred Steiglitz. The early prints he made
were either handcoated platinum or a variety of the then available
platinum papers that were commonly available to photographers. Making
ones own photo paper is not an easy task I know from limited personal
experience.
My goal is to get back to attempting it though. In latter life he shot
35mm and had silver prints done for him he did not print much of his own
work in latter life. He did will his imagery to Aperture (2 recall)
,.....so they periodically offer his reprinted imagery for sale.
My favorite images are the industrial images and of Rancho De Taos
Mission, maybe you did not see these, incredible close ups of metallic
machines, gears and such with great luminous scale- very sharp and
cleanly printed. Also I like the smaller images of Rancho De Taos
Mission before it became a photo tourist spot. When in Taos 1998 I
stopped there determined to make and image from an angle I had not seen.
The Vantage spot I took had me recognized by the state art council as
Maryland Contemporary artist in 2000, the image was displayed at the
Governors Residence in an exhibit for 3 months. So maybe something rubs
off when one is "Exposed" to master photographers work.
Greg "_"
Valid point. Most likely he did not make enlarged negatives.....
but you never know.
Scott W
> Velvia 50 claims S/N ratio of 110 when averaged over a circle of
> diameter 48um. (Do not remember which zone, but probably one
> corresponding to 18%gray with normal exposure?) This beats or is on
> par with almost any digital. I think it is related to f/32 or more
> closed apertures.
I must be missing something here, a S/N ratio of 110 is not even 7
stops, well below the range of a digital camera.
But looking at the curves for Velvia 100 (they don't make 50 amy more)
I can't believe that
you would get a range of 110, at the low exposure end it gets really
ugly with the colors not
tracking. I think you are going to get closer to 30 to 1 give or take
a bit.
Scott
Ilya Zakharevich
<bip...@hotmail.com>], who wrote in article <1161222934.3...@b28g2000cwb.googlegroups.com>:
> > Velvia 50 claims S/N ratio of 110 when averaged over a circle of
> > diameter 48um. (Do not remember which zone, but probably one
> > corresponding to 18%gray with normal exposure?) This beats or is on
> > par with almost any digital. I think it is related to f/32 or more
> > closed apertures.
> I must be missing something here, a S/N ratio of 110 is not even 7
> stops, well below the range of a digital camera.
It has nothing to do to "stops". It is just the noise in the image of
gray card. E.g., a digital camera with full well of 55K will have S/N
ratio for pixel's luminance about 1/100 at 18% of the full well.
BTW, due to enormous contrast of Velvia, density-noise of 1/110
correspond to luminance noise of something in the range of 1/200
.. 1/300 well above what ANY digital camera can produce. [Of course,
this number 110 comes from the manufacturer, and I see no reason to
trust this number. On the other hand, I saw no alternative data.]
Hope this helps,
Ilya
Roger N. Clark (change username to rnclark)
Scott,
S/N ratio is not related to dynamic range.
Take for instance the 1D Mark II with a full well
of 80,000 electrons and read noise of less than 4 electrons.
The dynamic range is 80000/4 = 20,000 (limited by the
A/D). But the maximum signal-to-noise ratio one can
get is the square root of the number of electrons,
so sqrt(80000) = 243 per pixel.
Roger
David J. Littleboy
"Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
> Scott W wrote:
>>
>> I must be missing something here, a S/N ratio of 110 is not even 7
>> stops, well below the range of a digital camera.
>>
>> But looking at the curves for Velvia 100 (they don't make 50 amy more)
>> I can't believe that
>> you would get a range of 110, at the low exposure end it gets really
>> ugly with the colors not
>> tracking. I think you are going to get closer to 30 to 1 give or take
>> a bit.
> Scott,
> S/N ratio is not related to dynamic range.
Huh?
> Take for instance the 1D Mark II with a full well
> of 80,000 electrons and read noise of less than 4 electrons.
> The dynamic range is 80000/4 = 20,000 (limited by the
> A/D). But the maximum signal-to-noise ratio one can
> get is the square root of the number of electrons,
> so sqrt(80000) = 243 per pixel.
Huh??? I thought that dynamic range was (max signal)/(noise floor)???
David J. Littleboy
dav...@gol.com
Tokyo, Japan
Roger N. Clark (change username to rnclark)
> [A complimentary Cc of this posting was sent to
> Roger N. Clark (change username to rnclark)
> <user...@qwest.net>], who wrote in article <4536611F...@qwest.net>:
>
>>While I agree that using a large area to measure
>>dynamic range will reduce noise, but
>>it does not reduce noise seen in a print.
>
> References? It is hard to believe this claim... And it contradicts
> your summary of night.and.low.light.photography.
This should be intuitively obvious. When you make
a print from film, the f/ratio of the lens
has nothing to do with the film grain that gets
enlarged by the enlarger lens. The page you reference
is for digital, not film. Even with digital,
noise in pixels does not change with the f/ratio
of the lens used to make the image, unless you
post process average.
>>Kodak started this trick of averaging large areas
>>to get their data on film, including the characteristic
>>curve, and resolution, because film is pretty poor
>>at high spatial frequencies due to noise from film grain.
>
> Velvia 50 claims S/N ratio of 110 when averaged over a circle of
> diameter 48um. (Do not remember which zone, but probably one
> corresponding to 18%gray with normal exposure?) This beats or is on
> par with almost any digital. I think it is related to f/32 or more
> closed apertures.
Using the Canon 1D Mark II example of 80,000 electrons
full well, ISO 50, and 8.2 micron pixels, one would get
about 0.18 * 80,000 electrons on an 18% gray card,
or 14,400 electrons, with a S/N per pixel of 120.
For an area 48 microns in diameter, 1810 square microns,
would be equivalent to 34.3 pixels or 5.2x better S/N.
One pixel gets sqrt(1440) = S/N = 120, so 5.8x better would
be ~700, which handily beats your Velvia S/N 110. What is your
reference for such a high claim for film? Astronomers who
did photometry of stars years ago generally never claimed
better than about 2% (S/N ~50). Film irregularities
was an additional source of noise.
>>So let's say you average all pixels in a 10mm diameter
>>area, well do the same with digital pixels.
>
> I see that you are still under impression that averaging noise of
> different physical origin will produce the same improvement. No, it
> depends on spectrum of the noise. And until someone measures it,
> theoretizing is futile.
Try reading. It is well demonstrated that digital sensors
are well described by Poisson counting statistics
from photo-electrons. That gives a square root dependence.
Add the read noise of a few electrons (which would be
negligible in your 18% gray card example). See equation
below.
>>While film hits a floor beyond which there is no
>>information, digital keeps on digging signal
>>out of the noise, even below 1 photon per pixel per
>>exposure. (e.g. see:
>>http://www.clarkvision.com/photoinfo/night.and.low.light.photography ).
>
> Sorry, I see no reference on film there.
Right. The page demonstrates digital can record
image information at less than 1 photon per pixel
per frame.
>>You can see dynamic range limitations in Figure 5 at:
>>http://www.clarkvision.com/imagedetail/dynamicrange2/index.html
>>Note the black round circle on the left in each image.
>>Your eye is effectively averaging pixels to see the subject.
>
> No, it is your scaling-down software which averaged i, not eye;
> without data which way it was scaled down, it is still not "clean
> enough" data... But if we knew the scaling algorithm, one could
> indeed deduce something from these data.
No it is not.
>>Then look at Figure 8. Average blocks of pixels together.
>>It should be obvious to you that no matter how you average,
>>if you average the same block of pixels, each signal
>>will reduce in noise proportional (square root) to the number
>>of pixels.
>
> Right; obviously, your knowledge of DSP is still kinda rusty. Your
> description is very plausible with digital (when the noise is [assumed
> to be] white); we KNOW that the noise of film is NOT white, so it WILL
> reduce differently than sqrt(scale).
Here you go again, you are starting your personal attacks.
My digital signal processing is not rusty; I use it
professionally every day analyzing imaging spectroscopy
data from spacecraft. If you can find a specific
problem with an analysis, please let me know.
Accusations do not do any good.
We do know that digital camera sensors have noise that
is well modeled with the equation:
N = (P + r^2 + t^2)^(1/2),
where N = total noise in electrons,
P = number of photons (electrons converted from photons),
r = read noise in electrons, and
t = thermal noise in electrons.
Noise from a stream of photons, the light we all see and
image with our cameras, is the square root of the number
of photons, so that is why the P in equation 2 is not
squared (sqrt(P)2 = P). The level of r is typically
below 4 up to about 15. Thermal noise is typically
a small fraction of an electron per second at room temperature,
decreasing with lower temperatures.
>>The least noisy sensor will remain the least noisy sensor (digital)
>
> This is extremely PROBABLE; but the key question (when comparing to
> larger format film) is HOW MUCH better. And: you will STILL be
> operating with "1-pixel"-wide strip of film; not relevant to dynamic
> range. One MUST start with 2-dim data to have a clean investigation.
See above.
>>and show the greatest dynamic range.
>
> Such a claim looks very plausible *theoretically*, but I do not see
> any "experimental" data to be ABSOLUTELY SURE. IMO, one MUST do the
> calculations before doing a claim like this.
It is not theory, it is measured. Multiple cameras are shown
on these web sites:
http://www.clarkvision.com/imagedetail/index.html#sensor_analysis
http://www.astrosurf.org/buil/20d/20dvs10d.htm
and related information:
http://spiff.rit.edu/classes/phys559/lectures/gain/gain.html
http://www.photomet.com/library_enc_fwcapacity.shtml
and a google search will turn up other information, including
sensor manufacturer's data.
The bottom line is, with noise dominated by photon statistics,
it is the best one can do. You can't get a system better
than that, and digital cameras have that.
Digital has higher QE than film, and film
is not photon noise limited, suffering from other noise
sources, including fogging, grain distribution irregularities,
and emulsion irregularities. It is no wonder people in this
newsgroup who have experience with both film and digital
see the much higher S/N in digital images.
Roger
Roger N. Clark (change username to rnclark)
Not only doesn't it help, but is totally absurd.
To get S/N 300 one needs 90,000 converted photons,
and with film's ~1% QE, that would equate to the
need of ~9 million photons per grain clump,
and that is well above what one would receive
in full sunlight.
Roger
Scott W
THO wrote:
> In article <1161222934.3...@b28g2000cwb.googlegroups.com>,
> "Scott W" <bip...@hotmail.com> wrote:
>
> > But looking at the curves for Velvia 100 (they don't make 50 amy more)
>
This may be but I could not find the data sheets for it. I don't
believe they are producing it any more but there may still be some on
shelves.
Scott
Roger N. Clark (change username to rnclark)
> "Roger N. Clark (change username to rnclark)" <user...@qwest.net> wrote:
>
>>Scott W wrote:
>>
>>>I must be missing something here, a S/N ratio of 110 is not even 7
>>>stops, well below the range of a digital camera.
>>>
>>>But looking at the curves for Velvia 100 (they don't make 50 amy more)
>>>I can't believe that
>>>you would get a range of 110, at the low exposure end it gets really
>>>ugly with the colors not
>>>tracking. I think you are going to get closer to 30 to 1 give or take
>>>a bit.
>
>>Scott,
>>S/N ratio is not related to dynamic range.
>
> Huh?
That is correct, see below.
>>Take for instance the 1D Mark II with a full well
>>of 80,000 electrons and read noise of less than 4 electrons.
>>The dynamic range is 80000/4 = 20,000 (limited by the
>>A/D). But the maximum signal-to-noise ratio one can
>>get is the square root of the number of electrons,
>>so sqrt(80000) = 243 per pixel.
>
>
> Huh??? I thought that dynamic range was (max signal)/(noise floor)???
Yes, that is correct. But noise floor is not noise
in the image. The maximum signal-to-noise ratio
is the square root of the maximum signal in electrons.
Example: in the canon 1D Mark II (similar pixels to the 5D),
the max signal is about 80,000 electrons, and the noise floor
is about 4 electrons. The dynamic range is 80,000/4
=20,000 (limited by the 12-bit A/D to about 11.6 bits).
But the maximum signal to noise ratio in any one
pixel can be no more than 80000/sqrt (80,000+4^2) = 293.5.
At 1/10 full well, the S/N drops to 8000/sqrt(8000+4^2)
= 89.3, and so on down to the read noise. At ten
converted photons, the S/N = 10 /sqrt(10+4^2) = 1.96.
Roger
Roger N. Clark (change username to rnclark)
> Have you (or Roger) looked at this:
> http://www.general-cathexis.com/
I have never seen this before. Do you know of any
independent reviews? Are you using it?
Roger
frederick
instruction manual is going to take more than few minutes for people
like me to achieve results.
I'm as sure as I can be that it's "safe" software.
I posted about this in rec.photo.digital.slr-systems, hoping for some
feedback. The author of the program signs his name in forums finishing
with "Author of SAR Image Processor and anomic sociopath".
He has posted at times some remarkable real examples of upsampling.
Ilya Zakharevich
Roger N. Clark (change username to rnclark)
> >>While I agree that using a large area to measure
> >>dynamic range will reduce noise, but
> >>it does not reduce noise seen in a print.
> > References? It is hard to believe this claim... And it contradicts
> > your summary of night.and.low.light.photography.
> This should be intuitively obvious.
Might be. However, since it is wrong, this does not matter... ;-)
> When you make a print from film, the f/ratio of the lens has
> nothing to do with the film grain that gets enlarged by the enlarger
> lens.
You assume one must "focus" the enlargement lens so that the grain is
seen. Matching f-stop of the enlargement lens to one of the shot will
average the grain without jeopardizing the details.
> > Velvia 50 claims S/N ratio of 110 when averaged over a circle of
> > diameter 48um. (Do not remember which zone, but probably one
> > corresponding to 18%gray with normal exposure?) This beats or is on
> > par with almost any digital. I think it is related to f/32 or more
> > closed apertures.
>
> Using the Canon 1D Mark II example of 80,000 electrons
> full well, ISO 50, and 8.2 micron pixels, one would get
> about 0.18 * 80,000 electrons on an 18% gray card,
> or 14,400 electrons, with a S/N per pixel of 120.
This is luminance S/N. Fuji says about density S/N, so the advertized
"effective luminance S/N" of Velvia 50 is much higher than 120 - but I
covered this in other posts.
> For an area 48 microns in diameter, 1810 square microns,
> would be equivalent to 34.3 pixels or 5.2x better S/N.
I'm not speaking of film and digital used at the SAME formfactor. On
the opposite, I assume FF of film to be 4x or 8x larger than of digital.
> One pixel gets sqrt(1440) = S/N = 120, so 5.8x better would
> be ~700, which handily beats your Velvia S/N 110. What is your
> reference for such a high claim for film? Astronomers who
> did photometry of stars years ago generally never claimed
> better than about 2% (S/N ~50). Film irregularities
> was an additional source of noise.
We discussed this already a couple of years ago, did not we? E.g.,
the film technology changed A LOT from '40s - '50s.
Anyway, I use the table and definitions in
http://creekin.net/films.htm
http://www.kodak.com/US/en/motion/students/handbook/sensitometric6.jhtml
(One of them is dead now, but might be available on some archives.) I
did not find any independent measurements, so one is forced to believe
Fuji's numbers...
> >>So let's say you average all pixels in a 10mm diameter
> >>area, well do the same with digital pixels.
> > I see that you are still under impression that averaging noise of
> > different physical origin will produce the same improvement. No, it
> > depends on spectrum of the noise. And until someone measures it,
> > theoretizing is futile.
> Try reading.
LOL! Roger, your condescending tone - when you do not understand a
bit in some topic - is really hilarious!
> It is well demonstrated that digital sensors are well described by
> Poisson counting statistics from photo-electrons.
In some ranges - yes. In dark areas - not. Just look in *your*
example night shots.
And FILM is NOT white at all.
> >>You can see dynamic range limitations in Figure 5 at:
> >>http://www.clarkvision.com/imagedetail/dynamicrange2/index.html
> >>Note the black round circle on the left in each image.
> >>Your eye is effectively averaging pixels to see the subject.
> >
> > No, it is your scaling-down software which averaged i, not eye;
> > without data which way it was scaled down, it is still not "clean
> > enough" data... But if we knew the scaling algorithm, one could
> > indeed deduce something from these data.
> No it is not.
Sorry, can't parse this.
> >>Then look at Figure 8. Average blocks of pixels together.
> >>It should be obvious to you that no matter how you average,
> >>if you average the same block of pixels, each signal
> >>will reduce in noise proportional (square root) to the number
> >>of pixels.
> > Right; obviously, your knowledge of DSP is still kinda rusty. Your
> > description is very plausible with digital (when the noise is [assumed
> > to be] white); we KNOW that the noise of film is NOT white, so it WILL
> > reduce differently than sqrt(scale).
>
> Here you go again, you are starting your personal attacks.
> My digital signal processing is not rusty
Obviously, you think so. But your postings speak otherwise.
> I use it professionally every day analyzing imaging spectroscopy
> data from spacecraft. If you can find a specific problem with an
> analysis, please let me know. Accusations do not do any good.
Obviously, there is some impendance mismatch in our communications. I
did not intend any *accusation*. [I will not answer immediately.]
> >>and show the greatest dynamic range.
> >
> > Such a claim looks very plausible *theoretically*, but I do not see
> > any "experimental" data to be ABSOLUTELY SURE. IMO, one MUST do the
> > calculations before doing a claim like this.
>
> It is not theory, it is measured. Multiple cameras are shown
> on these web sites:
> http://www.clarkvision.com/imagedetail/index.html#sensor_analysis
>
> http://www.astrosurf.org/buil/20d/20dvs10d.htm
>
> and related information:
> http://spiff.rit.edu/classes/phys559/lectures/gain/gain.html
>
> http://www.photomet.com/library_enc_fwcapacity.shtml
>
> and a google search will turn up other information, including
> sensor manufacturer's data.
... As I said, this is pixel noise. It has little relationship to
dynamic range for film.
> The bottom line is, with noise dominated by photon statistics,
> it is the best one can do. You can't get a system better
> than that, and digital cameras have that.
This is again said wrong. One cannot get a system better that that
ASSUMING THE SAME QE. E.g., my CURRENT estimates of "equivalent ISO"
of "an ideal sensor" is about 30x higher than those of current cameras
(my old estimates were "about 7x higher"; but it turned out that
taking into account AA filter makes an enormous effective decrease of
S/N ratio).
> Digital has higher QE than film, and film
> is not photon noise limited, suffering from other noise
> sources, including fogging, grain distribution irregularities,
> and emulsion irregularities. It is no wonder people in this
> newsgroup who have experience with both film and digital
> see the much higher S/N in digital images.
Why do you repeat again and again things not related to the topic at
hand? What you say holds ONLY if film vs digital is compared in the
same or close formfactor. I do not.
Hope this helps,
Ilya
Ilya Zakharevich
> > It has nothing to do to "stops". It is just the noise in the image of
> > gray card. E.g., a digital camera with full well of 55K will have S/N
> > ratio for pixel's luminance about 1/100 at 18% of the full well.
> >
> > BTW, due to enormous contrast of Velvia, density-noise of 1/110
> > correspond to luminance noise of something in the range of 1/200
> > .. 1/300 well above what ANY digital camera can produce. [Of course,
> > this number 110 comes from the manufacturer, and I see no reason to
> > trust this number. On the other hand, I saw no alternative data.]
> Not only doesn't it help, but is totally absurd.
> To get S/N 300 one needs 90,000 converted photons,
> and with film's ~1% QE, that would equate to the
> need of ~9 million photons per grain clump,
> and that is well above what one would receive
> in full sunlight.
Roger, get some clue about what we are talking about. What do you
think is the normal exposure of 18% gray in 50ISO? Now multiply by
the area of the circle 48um in diameter...
Hope this helps,
Ilya
Roger N. Clark (change username to rnclark)
As usual it doesn't. Your claim 1/300 noise/signal
for film is with zero, nada, zilch data and then claim
higher than any digital. Yet I showed real data for
a digital camera that gives S/N on an 18% gray card
of 120 per 8.2 micron pixel, so in a 48 micron diameter
pixel, the S/N would be over 600, double your film claim.
Anyone who has shot with film and digital knows the
digital is much higher signal to noise ratio.
Your claim is unsubstantiated, without fact, and
demonstrably false in multiple ways.
Roger
Roger N. Clark (change username to rnclark)
> [A complimentary Cc of this posting was sent to
> Roger N. Clark (change username to rnclark)
> <user...@qwest.net>], who wrote in article <453702F2...@qwest.net>:
>
>>>>While I agree that using a large area to measure
>>>>dynamic range will reduce noise, but
>>>>it does not reduce noise seen in a print.
>
>>>References? It is hard to believe this claim... And it contradicts
>>>your summary of night.and.low.light.photography.
>
>>This should be intuitively obvious.
>
> Might be. However, since it is wrong, this does not matter... ;-)
No it is not.
>> When you make a print from film, the f/ratio of the lens has
>>nothing to do with the film grain that gets enlarged by the enlarger
>>lens.
>
> You assume one must "focus" the enlargement lens so that the grain is
> seen. Matching f-stop of the enlargement lens to one of the shot will
> average the grain without jeopardizing the details.
Well, you will also smear detail. I have done a lot
of darkroom work in my time. You are grasping at straws.
>>>Velvia 50 claims S/N ratio of 110 when averaged over a circle of
>>>diameter 48um. (Do not remember which zone, but probably one
>>>corresponding to 18%gray with normal exposure?) This beats or is on
>>>par with almost any digital. I think it is related to f/32 or more
>>>closed apertures.
>>
>>Using the Canon 1D Mark II example of 80,000 electrons
>>full well, ISO 50, and 8.2 micron pixels, one would get
>>about 0.18 * 80,000 electrons on an 18% gray card,
>>or 14,400 electrons, with a S/N per pixel of 120.
>
> This is luminance S/N. Fuji says about density S/N, so the advertized
> "effective luminance S/N" of Velvia 50 is much higher than 120 - but I
> covered this in other posts.
Incorrect.
>>For an area 48 microns in diameter, 1810 square microns,
>>would be equivalent to 34.3 pixels or 5.2x better S/N.
>
> I'm not speaking of film and digital used at the SAME formfactor. On
> the opposite, I assume FF of film to be 4x or 8x larger than of digital.
So you want to compare 35 mm to an 8x smaller sensor
which would equal a 3 x 4.5 mm sensor? Pretty funny
that you have to stoop to such extremes to show film's
superiority!
>>One pixel gets sqrt(1440) = S/N = 120, so 5.8x better would
>>be ~700, which handily beats your Velvia S/N 110. What is your
>>reference for such a high claim for film? Astronomers who
>>did photometry of stars years ago generally never claimed
>>better than about 2% (S/N ~50). Film irregularities
>>was an additional source of noise.
>
> We discussed this already a couple of years ago, did not we? E.g.,
> the film technology changed A LOT from '40s - '50s.
So has digital, and even in the last couple of years
since you've been gone.
> Anyway, I use the table and definitions in
>
> http://creekin.net/films.htm
Link does not work.
> http://www.kodak.com/US/en/motion/students/handbook/sensitometric6.jhtml
>
> (One of them is dead now, but might be available on some archives.) I
> did not find any independent measurements, so one is forced to believe
> Fuji's numbers...
Granularity is not S/N ratio. Hint: it is a log scale.
>>>>So let's say you average all pixels in a 10mm diameter
>>>>area, well do the same with digital pixels.
>
>>>I see that you are still under impression that averaging noise of
>>>different physical origin will produce the same improvement. No, it
>>>depends on spectrum of the noise. And until someone measures it,
>>>theoretizing is futile.
>
>>Try reading.
>
> LOL! Roger, your condescending tone - when you do not understand a
> bit in some topic - is really hilarious!
You didn't read.
>>It is well demonstrated that digital sensors are well described by
>>Poisson counting statistics from photo-electrons.
>
> In some ranges - yes. In dark areas - not. Just look in *your*
> example night shots.
1) You are not talking about dark areas.
2) The level at which digital is dominated by read noise
and thermal noise is far below what film can record.
You consistently try and pull a fast one trying to prove
some point by citing some irrelevant detail. Try
keeping the problem between the two media the same,
or your comparison is simply not valid.
> And FILM is NOT white at all.
To quote the Kodak article you cited above, regarding film grain:
"Randomness is a necessary condition for the phenomenon."
> Obviously, there is some impendance mismatch in our communications. I
> did not intend any *accusation*. [I will not answer immediately.]
Obviously!
>>>>and show the greatest dynamic range.
>>>
>>>Such a claim looks very plausible *theoretically*, but I do not see
>>>any "experimental" data to be ABSOLUTELY SURE. IMO, one MUST do the
>>>calculations before doing a claim like this.
>>
>>It is not theory, it is measured. Multiple cameras are shown
>>on these web sites:
>>http://www.clarkvision.com/imagedetail/index.html#sensor_analysis
>>
>>http://www.astrosurf.org/buil/20d/20dvs10d.htm
>>
>>and related information:
>>http://spiff.rit.edu/classes/phys559/lectures/gain/gain.html
>>
>>http://www.photomet.com/library_enc_fwcapacity.shtml
>>
>>and a google search will turn up other information, including
>>sensor manufacturer's data.
>
> ... As I said, this is pixel noise. It has little relationship to
> dynamic range for film.
You are the one who asked for data to be absolutely sure.
Don't dismiss it when it is given.
>>The bottom line is, with noise dominated by photon statistics,
>>it is the best one can do. You can't get a system better
>>than that, and digital cameras have that.
>
> This is again said wrong. One cannot get a system better that that
> ASSUMING THE SAME QE. E.g., my CURRENT estimates of "equivalent ISO"
> of "an ideal sensor" is about 30x higher than those of current cameras
> (my old estimates were "about 7x higher"; but it turned out that
> taking into account AA filter makes an enormous effective decrease of
> S/N ratio).
This does not make sense. A blur filter simply redistributes
the photons, it doesn't reduce QE 30x. Again a claim
with no data, nada, zilch. Since you were gone for
a while, a lot of progress has been made.
Try reading:
Digital Cameras: Counting Photons, Photometry,
and Quantum Efficiency
http://www.clarkvision.com/imagedetail/digital.photons.and.qe
including the references at the end which include other
published references to QE. Reference 3 has graphs of
QE for many sensors. None are as low as you claim.
>>Digital has higher QE than film, and film
>>is not photon noise limited, suffering from other noise
>>sources, including fogging, grain distribution irregularities,
>>and emulsion irregularities. It is no wonder people in this
>>newsgroup who have experience with both film and digital
>>see the much higher S/N in digital images.
>
> Why do you repeat again and again things not related to the topic at
> hand? What you say holds ONLY if film vs digital is compared in the
> same or close formfactor. I do not.
Yes, we know that now. You want to make the film camera
8x larger than the digital camera to give film an
advantage. Case closed, we can all go home now.
Pretty funny.
> Hope this helps,
As usual, not.
Roger
Alan LeHun
> Yet I showed real data for
> a digital camera that gives S/N on an 18% gray card
> of 120 per 8.2 micron pixel, so in a 48 micron diameter
> pixel, the S/N would be over 600, double your film claim.
>
Shouldn't it be roughly the same? Or are we talking about some sort of
"exotic" ratio?
I don't claim to understand /all/[1] that you're talking about, btw.
[1] FVVO "all" ranging from greater than 0 to 100.
--
Alan LeHun
Ilya Zakharevich
Roger N. Clark (change username to rnclark)
> > You assume one must "focus" the enlargement lens so that the grain is
> > seen. Matching f-stop of the enlargement lens to one of the shot will
> > average the grain without jeopardizing the details.
>
> Well, you will also smear detail. I have done a lot
> of darkroom work in my time.
??? If the circle of confusion of enlargement lens is much more than
the cloud size, and much less that the circle of confusion of the
shot, the details won't be smeared, AND the grain will be averaged.
> You are grasping at straws.
Sorry, but I must reverse the charges...
> > This is luminance S/N. Fuji says about density S/N, so the advertized
> > "effective luminance S/N" of Velvia 50 is much higher than 120 - but I
> > covered this in other posts.
> Incorrect.
Can't parse this.
> > I'm not speaking of film and digital used at the SAME formfactor. On
> > the opposite, I assume FF of film to be 4x or 8x larger than of digital.
>
> So you want to compare 35 mm to an 8x smaller sensor
> which would equal a 3 x 4.5 mm sensor?
I have no idea why you put these words in my mouth... Why one would
want to do this? If you would remember what this thread is about, I'm
discussing possible differences between LF film and a (hypothetical)
"good" digital sensor in, e.g., 35mm formfactor.
> Pretty funny that you have to stoop to such extremes to show film's
> superiority!
Again, it looks like you behave agressively when you feel yourselves
cornered. Roger, there is absolutely no shame in being wrong. Just
accept it. ;-)
[Just think: did I ever said anything about film superiority?]
> > We discussed this already a couple of years ago, did not we? E.g.,
> > the film technology changed A LOT from '40s - '50s.
> So has digital, and even in the last couple of years
> since you've been gone.
I have no idea what you are talking about. Here we discuss manufacturer's
claims about noise of current film. What has "progress of digital" do
with it?! And I do not go too far, again have no idea what you are
talking about.
> > http://www.kodak.com/US/en/motion/students/handbook/sensitometric6.jhtml
> >
> > (One of them is dead now, but might be available on some archives.) I
> > did not find any independent measurements, so one is forced to believe
> > Fuji's numbers...
>
> Granularity is not S/N ratio. Hint: it is a log scale.
Again, this does not parse.
> > In some ranges - yes. In dark areas - not. Just look in *your*
> > example night shots.
> 1) You are not talking about dark areas.
Yes, I WAS talking about ALL THE ZONES, and that one needs measure the
spectrum of noise in all of them.
> 2) The level at which digital is dominated by read noise
> and thermal noise is far below what film can record.
You claim so. One can BELIEVE it - or one can DOUBT it. Given that
there is no hard data on the spectrum of noise of film (when converted
to linear scale) in dark areas, one can't be SURE.
Having your Figure 3 of
http://www.clarkvision.com/imagedetail/dynamicrange2/ produced with a
downsampler-by-averaging (and having the [linear] brightness increased
about 50x so that one can easily work with 8-bit software) would be
close to such a proof.
Well, in fact maybe not. The contrast is too low to be absolutely
sure (due to reflections off the patterns in the dark part). If you
would have put some objects inside the tube, this could help a lot...
The tube diameter is about 3mm on film, so one could investigate the
image even after A LOT of averaging...
From the images as given on the webpage, one can tell that the zone I
(1/16 sec) is easily distingushed from zone 0 *on large scale*; but
one cannot go below this. E.g., Assuming, e.g., that zone -2 is
"distingushable" from zone -1 on large scale (i.e., S/N ration is
better than 3/1 on very low frequencies), this would give about
dynamic range of about 2000/1. With 80K sensor, we are talking about
digital noise of 6e; this much less than the read noise of current
cameras on the setting with such a huge full well.
> You consistently try and pull a fast one trying to prove
> some point by citing some irrelevant detail.
Again, I reverse the charges.
> > And FILM is NOT white at all.
> To quote the Kodak article you cited above, regarding film grain:
> "Randomness is a necessary condition for the phenomenon."
Obviously, you have very little experience with probability theory...
There is a long way to go from noise being random to noise being white.
> > ... As I said, this is pixel noise. It has little relationship to
> > dynamic range for film.
> You are the one who asked for data to be absolutely sure.
> Don't dismiss it when it is given.
Nevertheless, I feel free to dismiss data which is known to be IRRELEVANT.
> > This is again said wrong. One cannot get a system better that that
> > ASSUMING THE SAME QE. E.g., my CURRENT estimates of "equivalent ISO"
> > of "an ideal sensor" is about 30x higher than those of current cameras
> > (my old estimates were "about 7x higher"; but it turned out that
> > taking into account AA filter makes an enormous effective decrease of
> > S/N ratio).
> This does not make sense. A blur filter simply redistributes
> the photons, it doesn't reduce QE 30x.
It does not decrease the "honest to goodness" QE indeed.
But I'm discussing "effective QE": one which determines visible noise
(i.e., noise filtered through the MTF of the eye/brain; of course,
this measure depends on the viewing angle of a pixel). By
redistributing photons, signal level is decreased; this affects S/N
ratio, thus the "effective QE".
> Yes, we know that now. You want to make the film camera
> 8x larger than the digital camera to give film an
> advantage.
Nope; apparently, you reading ability suffers a lot at the end of the
day... I discuss layout of a 35mm sensor which would provides the
same details as a LF shot (at least LF shots made at some particular
f-stops).
Hope this helps,
Ilya
Roger N. Clark (change username to rnclark)
Hi Alan,
The area of the 8.2 (assume square) micron pixel
= 8.2*8.2 = 67.24 sq. microns.
The area of a 48 micron diameter spot =
24*24*3.14 = 1809 sq. microns.
Ratio of the areas: 1809/67.2 = 26.9, and this equals
the increase in number of photons averaged in the digital
sensor. The signal-to-noise increase is square root 26.9
= 5.2. 5.2 * 120 = 624.
The claim for film was S/N =300, so the digital gives double
the claim. However, if one looks at the definition of
granularity by Kodak, one finds it is for density =1.
RMS granularity = RMS variations of the density
of repeated readings in a 48 micron diameter spot times 1000.
An RMS granularity number for fine grained film (e.g. Velvia)
is around 8, which when converted to density is 0.008. So we
have density variations of 0.008 out of 1. Converting
to linear space: S/N = 10^1 / (10^1.008 - 10^1)
= 10 /(10.186-10) = 54. An S/N ~ 54 is consistent to
astronomical photometry with film not getting better than
about 1%.
So we have digital over a 48 micron spot with a S/N of
>600, about 12 times that of film.
Roger
Roger N. Clark (change username to rnclark)
> [A complimentary Cc of this posting was sent to
> Roger N. Clark (change username to rnclark)
> <user...@qwest.net>], who wrote in article <45386F4B...@qwest.net>:
>>8x larger than the digital camera to give film an
>>advantage.
>
> Nope; apparently, you reading ability suffers a lot at the end of the
> day... I discuss layout of a 35mm sensor which would provides the
> same details as a LF shot (at least LF shots made at some particular
> f-stops).
Well, I guess I read what you wrote, not what you thought you wrote:
Ilya Zakharevich wrote (October 19,2006):
> I'm not speaking of film and digital used at the SAME formfactor. On
> the opposite, I assume FF of film to be 4x or 8x larger than of digital.
Maybe you should define what you mean more precisely.
Let me infer your position from what you have written for others
who may still be reading:
Your comparison is like saying a 35mm film camera
can take a better picture than a cell phone camera,
but if we jam a whole bunch of pixels in the cell phone
camera, it can equal the 35mm film camera. Well, in theory
if you only count numbers of pixels, yes, but then reality
sinks in.
Your wasting our time. Like I've said before from your
posts. If you are correct go into business and make the
cameras. You'll make millions and win a Nobel prize
along the way for the new physics you invent.
Roger
Ilya Zakharevich
Roger N. Clark (change username to rnclark)
> >>Yes, we know that now. You want to make the film camera
> >>8x larger than the digital camera to give film an
> >>advantage.
> > Nope; apparently, you reading ability suffers a lot at the end of the
> > day... I discuss layout of a 35mm sensor which would provides the
> > same details as a LF shot (at least LF shots made at some particular
> > f-stops).
> Well, I guess I read what you wrote, not what you thought you wrote:
Well, I guess you read what I wrote, but did not grasp it... ;-)
> > I'm not speaking of film and digital used at the SAME formfactor. On
> > the opposite, I assume FF of film to be 4x or 8x larger than of digital.
What do you think is relationship of sizes of LF and FF sensors?
> Your comparison is like saying a 35mm film camera
> can take a better picture than a cell phone camera,
A year ago, I was sure it is true. However, now I'm not. The
theoretical bounds on performance of digital sensors depends VERY
STRONGLY on the spectrum of photon noise (which I now address in a
different thread).
> but if we jam a whole bunch of pixels in the cell phone
> camera, it can equal the 35mm film camera.
If the photon noise is white, then this is possible (at least
theoretically). Then a sensor with QE close to 1 would be equivalent
to about 60x^2 larger area of Velvia 50 as far as visible noise is
concerned.
But anyway, I'm not discussing FF film vs 2mm digital; I'm discussing
LF film vs FF digital.
> Your wasting our time.
Do not think so. At least YOU are learning things from these
discussions; other people may learn something too.
Ilya
Ilya Zakharevich
Ilya Zakharevich
<nospam...@ilyaz.org>], who wrote in article <ehe829$1554$1...@agate.berkeley.edu>:
> > but if we jam a whole bunch of pixels in the cell phone
> > camera, it can equal the 35mm film camera.
>
> If the photon noise is white, then this is possible (at least
> theoretically). Then a sensor with QE close to 1 would be equivalent
> to about 60x^2 larger area of Velvia 50 as far as visible noise is
> concerned.
Oups, this 60x is a way off; I already forgot what I was thinking
about when I wrote this ;-). Even if Velvia 50ISO has noise similar
to 40000ISO ideal sensor (and I'm still not sure in this), this would
mean 800x gain in noise-restricted-sensitivity; thus one needs 800
times smaller area to capture the equivalent amount of photons, so a
28x smaller sensor.
[Aha, 60 is sqrt(4000), so I initialy divided by 10ISO instead of 50ISO?!]
Sorry,
Ilya
Ilya Zakharevich
Roger N. Clark (change username to rnclark)
> Link does not work.
Well, I SAID it does not work NOW. That's how internet works; don't
you yet know what to do in such a case?
OK, if you do not, here is the algo:
a) I went to
http://web.archive.org/web/*/http://creekin.net/films.htm
and had choosen the latest preserved copy of this page.
b) The I took a long enough phrase on this preserved copy ("RB-3 and
later renamed HD 200, Royal Gold and Supra 400"), and looked for
it on Google. It gives
http://www.cacreeks.com/films.htm
Could you preserve this recipe for future reference? BTW, I use
Firefox with extensions "Gcache", "WayBack", "Context Search" and "Web
Search Plus" (among 60 others :-( ), which make it a breeze...
Hope this helps,
Ilya
Ilya Zakharevich
Roger N. Clark (change username to rnclark)
> the claim. However, if one looks at the definition of
> granularity by Kodak, one finds it is for density =1.
> RMS granularity = RMS variations of the density
> of repeated readings in a 48 micron diameter spot times 1000.
> An RMS granularity number for fine grained film (e.g. Velvia)
> is around 8, which when converted to density is 0.008. So we
> have density variations of 0.008 out of 1. Converting
> to linear space: S/N = 10^1 / (10^1.008 - 10^1)
> = 10 /(10.186-10) = 54. An S/N ~ 54 is consistent to
> astronomical photometry with film not getting better than
> about 1%.
You might be right that one must read the Kodak definition as
"absolute variation of (log_base10-)density", while I was reading it
as "relative variation of linear density". This would decrease the
numbers log(10) times w.r.t. what I said.
E.g., the Velvia number would go about 145 (assuming contrast 3).
But anyway, your calculations are still wrong - your numbers are still
in non-linear space: when you divide by 2.3, this is the variation in
"linear density" space; which is non-linear w.r.t. linear luminance.
So you need to multiply by contrast, which IIRC is about 3 for Velvia.
For negatives, like Fuji Pro 160S/C: the RMS number is claimed to be
3, and contrast about 1.6; this gives the claimed linear S/N ratio of
1000/3/log(10)*1.6, about 230:1. Again, I have no idea whether one
can believe this...
Hope this helps,
Ilya
David J. Littleboy
"Ilya Zakharevich" <nospam...@ilyaz.org> wrote:
>
> For negatives, like Fuji Pro 160S/C: the RMS number is claimed to be
> 3, and contrast about 1.6; this gives the claimed linear S/N ratio of
> 1000/3/log(10)*1.6, about 230:1. Again, I have no idea whether one
> can believe this...
FWIW, these films scan _very_ nicely at 4000 ppi in a Nikon 8000. IMHO,
Provia 100F grain is less objectionable, but Fuji Pro 160S is very nice
stuff if you like Velvia's electric greens.
Also FWIW, I find that for pictorial photography, it's the 1:1.6 contrast
that best predicts what the usable photographic detail is going to be, so
there's not a lot of practical difference between the better films. (My
opinion is that 8 or 9x is the limit for enlargement from film; if I need
much more than 12 x 18, I don't shoot 645. (I noise reduce at 4000 ppi and
downsample to something in the 2200 to 2700 ppi range to produce images that
print nicely at 300 ppi.)) Thanks for reminding me of this link; it's nice
to see it's been updated.
http://www.cacreeks.com/films.htm
David J. Littleboy
Tokyo, Japan