Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Yep, I did. Sorry if that wasn't clear.But maybe you intended this in a more lite hearted fashion than it reads
to me. If so I resolve to not be offended and forgive.
Ah, okay. No problems then. I find myself occasionally amazed by what people will pay for--a friend had wedding pictures taken and to my eyes they were not acceptable, but he seemed quite happy with them.When I said digital prints were indistinguishable from photographic
prints I hope I said, or at least hinted, to the casual observer. The
majority of the people reading this can see the difference, but the
people who you took you photo's for: your family, friends, most clients
cannot tell.
The big problem is of course the lite coloured areas, which can look awful.
Ah, but there is a difference. In colour film, there are multiple (three or four) layers, each of which is sensitive to a different frequency range. This would be like having three separate CCDs in a camera, one for each colour (which is done in broadcast-quality video cameras, and in one of the Minolta digicams).If people are rating film resolution based on crystal count then the
estimate is too high. If three crystal colour types are present then the
actual count must be divided by three at least.
Actually, the camera makers have been cheating. They specify the mega pixels by CCD cell count. But each CCD only records one color. So a single crystal from a single layer and a CCD cell count is the same. I don't know how the crystal count is derived. But I'm looking forward to the CCD count to be a few order of magnitudes (2's) about crystal count then regardless of distribution pattern, digital wins in terms of accuracy.My only point to make at this time is that I do not believe a crystal, no
matter what colour depth it carries is the equivalent of a digital pixel.
Firstly the crystal only carries one colour while a digital pixel carries
several - up to 4 in the case of CMYK or 3 for RGB.
entropy!?! now you are making my brain hurt!!!From the perspective of entropy: the higher the randomness, the lower the
level of order. Lower order means less infomation in the system. The
information in a digital image is easily understood but film contains
less information than the sum of it's crystals the laws of entropy
dictate that it must. Although film may have a higher 'real' resolution
now you cannot equate a single crystal to a single digital pixel. It
takes a group of crystals to equal the information in a pixel.
Good to hear Chris and thanks for the clarification.Yep, I did. Sorry if that wasn't clear.But maybe you intended this in a more lite hearted fashion than it reads
to me. If so I resolve to not be offended and forgive.
Unfortunately for us our standards are set so high we can see the flaws in anything. This is a blessing for others but a curse to us. Most would probably think we were crazy to have such a discussion in the first place.Ah, okay. No problems then. I find myself occasionally amazed by whatWhen I said digital prints were indistinguishable from photographic
prints I hope I said, or at least hinted, to the casual observer. The
majority of the people reading this can see the difference, but the
people who you took you photo's for: your family, friends, most clients
cannot tell.
The big problem is of course the lite coloured areas, which can look awful.
people will pay for--a friend had wedding pictures taken and to my eyes
they were not acceptable, but he seemed quite happy with them.
Chris
Good point. Multiple overlapping layers is better. This would also aid in the positional accuracy of recorded information. It depends on how the resolution of film is measured. It would have to be based on the crystal content of a single layer.Ah, but there is a difference. In colour film, there are multiple (threeIf people are rating film resolution based on crystal count then the
estimate is too high. If three crystal colour types are present then the
actual count must be divided by three at least.
or four) layers, each of which is sensitive to a different frequency
range. This would be like having three separate CCDs in a camera, one
for each colour (which is done in broadcast-quality video cameras, and in
one of the Minolta digicams).
Chris,As for the analysis of film resolution, while it is true that there is a
random distribution and a random crystal size, it is highly likely that
these are bounded by certain max and min values, and that they obey a
statistical distribution, which could be measured. Because of this, they
would be susceptable to statistical analysis methods if anyone were to
actually take the time to do it.
Chris
Actually, the camera makers have been cheating. They specify the megaMy only point to make at this time is that I do not believe a crystal, no
matter what colour depth it carries is the equivalent of a digital pixel.
Firstly the crystal only carries one colour while a digital pixel carries
several - up to 4 in the case of CMYK or 3 for RGB.
pixels by CCD cell count. But each CCD only records one color. So a
single crystal from a single layer and a CCD cell count is the same. I
don't know how the crystal count is derived. But I'm looking forward to
the CCD count to be a few order of magnitudes (2's) about crystal count
then regardless of distribution pattern, digital wins in terms of
accuracy.
While digital and chemical base flim are both improving, I do believe
digital flim is progressing at a much more rapid pace. There are still
people that are convinced that vinyl "sounds" better than CD, and I will
never argue that and I can respect their opinion. Then there are people
that are convinced that vinyl is more accurate, that I'll have to argue.
At this point, I do believe digital is a few order of magnitudes behind,
but I won't expect it'll be too long before digital catches up and
surpass chemical film in strict accuracy measurement. As in what pleases
the eyes more, it's up to anybody's taste.
entropy!?! now you are making my brain hurt!!!From the perspective of entropy: the higher the randomness, the lower the
level of order. Lower order means less infomation in the system. The
information in a digital image is easily understood but film contains
less information than the sum of it's crystals the laws of entropy
dictate that it must. Although film may have a higher 'real' resolution
now you cannot equate a single crystal to a single digital pixel. It
takes a group of crystals to equal the information in a pixel.
Thanks for a stimulating discussion.
gordon
I just LOVE this thread. Guess we're all geeks here. With an avid interest, and appreciation, of all types of photography.Good point. Multiple overlapping layers is better. This would also aidAh, but there is a difference. In colour film, there are multiple (threeIf people are rating film resolution based on crystal count then the
estimate is too high. If three crystal colour types are present then the
actual count must be divided by three at least.
or four) layers, each of which is sensitive to a different frequency
range. This would be like having three separate CCDs in a camera, one
for each colour (which is done in broadcast-quality video cameras, and in
one of the Minolta digicams).
in the positional accuracy of recorded information. It depends on how
the resolution of film is measured. It would have to be based on the
crystal content of a single layer.
Of course, NOW that you reminded of how the CCD works we have to rethink
digital camera resolutions since some interpolation has been performed in
camera. It would seem to follow that the information content of a
digital image is less than and not equal to the sum of it's pixels.
Darn it !!! I was beginning to feel like we were getting some where :-o
Chris,As for the analysis of film resolution, while it is true that there is a
random distribution and a random crystal size, it is highly likely that
these are bounded by certain max and min values, and that they obey a
statistical distribution, which could be measured. Because of this, they
would be susceptable to statistical analysis methods if anyone were to
actually take the time to do it.
Chris
I wonder if film uses the same trick as the CCD, having more green
sensors to carry the detail since the human eye is more sensitive to
green. In photographic prints the blotchiness of the pure blue sky can
be abvious while other objects in the same image are sharp.
The one problem with a random distribution is that there is no actual
maximum and minimum peak at any one point. The location could contain no
crystals or all the crystals. Where randomness works in our favour is
that these two extremes are near enfinately unlikely. The likely crystal
distributions are close to even but an absolutely even distribution is
just as unlikely as the other two extremes. The result is that the
resolution at any one point can not be determined until after processing.
With digital I know (or at least suspect) in advance - even if just to
know that it is not as good as film - for now.
Indeed they are, both film and digital...P.S. Did I forget to mention anywhere around here that I also think
photographs are beautiful.
Nyquest rate only requires the sample rate is be > 2 times max resolution for a perfect reconstruction. For a repeating pattern of line pairs, it only requires > 2 pixels per pair. 2.x pixels per pair is sufficient if a perfect filter is available. Four may be a good real world requirement, but not the minimum requirement.it takes four pixels to adequately detect a line
pair. Why four? If you try to use one pixel for each white line and one
pixel for each black line, you're fine as long as the lines fall square
on the pixels; however, shift the lines by half a pixel, and all you
detect is a nice uniform grey. Four pixels per line pair fixes that,
though were starting to go into the area were contrast measurement might
have to be called upon. (Yikes!, not with pixels...)
Yes the per line resolution is lower at 45 degrees, but the lines of detection is closer together. For example, if you draw parallel horizontal lines to connect 2 rows of pixels, the lines are 1 pixel unit apart. If you draw diagonal lines, the samples on each line is squareroot of 2 (1.414) apart, but the lines are (squareroot of 2 ) 2 (.707) apart. The two dimensional resolution is conserved for a groups of diagonal lines.With a cartesian sensor matrix, the best resolution is when the lines are
parallel to an edge of the sensor matrix. BUT, resolution should
decrease as you rotate the resolution target so the lines are no longer
parallel with an edge, and reach a minimum resolution when the target is
lined up along the matrix diagonal.
In the audio world, there is a movement to go to 96 KHz (and 24 bits for the dynamic range, ...and so such things as digital volume controls don't hurt things) as a way to accurately reproduce music. Trusted reviewers (I trust them, anyway) claim that for the first time, digital audio actually sounds like its analog counterpart. Hearing is supposed to only go to 20 KHz, but 44.8 KHz (Nyquest rate) CD sampling doesn't seem to cut it - at least for all people. (Side note: when I switched to CDs from LPs, I was disappointed by the sound - even though I got a good CD player and the rest of my audio chain was the same. Yet, I stopped playing LPs because of the CD convenience factor and CDs were just SO cool.) Nyquest rate may be okay as a theoretical construct, but does it really work? We're not theoretical. And if there are 2.1 pixels, say, per line pair, what does the image of the resolution chart look like? Can you, as an observer, easily tell what the frequency of the target lines is supposed to be?Nyquest rate only requires the sample rate is be > 2 times max resolutionit takes four pixels to adequately detect a line
pair. Why four? If you try to use one pixel for each white line and one
pixel for each black line, you're fine as long as the lines fall square
on the pixels; however, shift the lines by half a pixel, and all you
detect is a nice uniform grey. Four pixels per line pair fixes that,
though were starting to go into the area were contrast measurement might
have to be called upon. (Yikes!, not with pixels...)
for a perfect reconstruction. For a repeating pattern of line pairs, it
only requires > 2 pixels per pair. 2.x pixels per pair is sufficient if a
perfect filter is available. Four may be a good real world requirement,
but not the minimum requirement.
Yep, I thought about that some more after I posted it, and I think you are right. However, I also suspect that now we're REALLY starting to get into the area of using a contrast threshold as a way to determine the frequency of the target lines pairs. The pixel centers may lie upon a black line (or a white line) but the corners may not; such pixels would be less of a pure black (or white). A resolution target is likely to look much less crisp. As above, the underlying target resolution may be determinable through technical analysis, but what do our photographs look like?Yes the per line resolution is lower at 45 degrees, but the lines ofWith a cartesian sensor matrix, the best resolution is when the lines are
parallel to an edge of the sensor matrix. BUT, resolution should
decrease as you rotate the resolution target so the lines are no longer
parallel with an edge, and reach a minimum resolution when the target is
lined up along the matrix diagonal.
detection is closer together. For example, if you draw parallel
horizontal lines to connect 2 rows of pixels, the lines are 1 pixel unit
apart. If you draw diagonal lines, the samples on each line is squareroot
of 2 (1.414) apart, but the lines are (squareroot of 2 ) 2 (.707)
apart. The two dimensional resolution is conserved for a groups of
diagonal lines.
Minor problem with your statement. While it is true that Nyquist says you need only > 2 times max resolution, what you've neglected is that lines/mm is not sine waves, but square waves. This means that to faithfully represent the square wave you need much higher sampling than twice the actual lines/mm value. If you were sampling sine waves, then a bit more than twice the max resolution would be sufficient.Nyquest rate only requires the sample rate is be > 2 times max resolution
for a perfect reconstruction. For a repeating pattern of line pairs, it
only requires > 2 pixels per pair. 2.x pixels per pair is sufficient if a
perfect filter is available. Four may be a good real world requirement,
but not the minimum requirement.
Minor problem with your statement. While it is true that Nyquist saysNyquest rate only requires the sample rate is be > 2 times max resolution
for a perfect reconstruction. For a repeating pattern of line pairs, it
only requires > 2 pixels per pair. 2.x pixels per pair is sufficient if a
perfect filter is available. Four may be a good real world requirement,
but not the minimum requirement.
you need only > 2 times max resolution, what you've neglected is that
lines/mm is not sine waves, but square waves. This means that to
faithfully represent the square wave you need much higher sampling than
twice the actual lines/mm value. If you were sampling sine waves, then a
bit more than twice the max resolution would be sufficient.
Chris
repeating pattern is the key. For a single line pair, the limit of 4 in the original and joe's post would help. However, a typical spatial frequency measurement is a frequency measurement, not a position (phase) measurement. A typical test pattern uses a group of lines, not a single pair of line. As long as the number of cycles is longer than the reconstruction filter, it is possible to reconstruct the original frequency with nyquest samples.For a repeating pattern of line pairs, it only requires > 2 pixels per pair.
[joe, good explaination of sampling theory limit , sorry I missed it the first time around]I believe there is some confusion regarding sampling theory and
reconstruction as in applies to the CCD in our digital cameras. Below is
a cut and paste of a posting I made about two months ago on this same
subject.
**********************
Sampling theory is correct (by definition). What happens when you
asynchronously sample a signal at a rate ever so slightly higher than
twice the signal frequency? If you apply an infinite number of samples,
and wait an infinite amount of time, it is possible to accurately
reconstruct the signal. Nobody should have a problem with this concept.
*************************
At any rate, above is my input regarding the sampling/reconstruction
subject as it applies to the CCD.
Minor problem with your statement. While it is true that Nyquist saysNyquest rate only requires the sample rate is be > 2 times max resolution
for a perfect reconstruction. For a repeating pattern of line pairs, it
only requires > 2 pixels per pair. 2.x pixels per pair is sufficient if a
perfect filter is available. Four may be a good real world requirement,
but not the minimum requirement.
you need only > 2 times max resolution, what you've neglected is that
lines/mm is not sine waves, but square waves. This means that to
faithfully represent the square wave you need much higher sampling than
twice the actual lines/mm value. If you were sampling sine waves, then a
bit more than twice the max resolution would be sufficient.
Chris
I had a humbling experience. I played a test CD that contains clips like wobble tones in step frequency and noise levels. I found my ears are only good for 16.xkHz and -70dB noise. I check with other folks in my group and a few of us are in the same ball park. I think the 20kHz is overrated, and the 100dB+ SNR type high end equiptment are nothing more than status symbols. Until an avarage joe can "hear" 18k and -80dB, save your money and stick with CD. I am sure there are a few people with golden ears out there, but chances are, it's not you (Ed: a general you, not a personal you and I do know people that claims to hear over 20kHz and the alias noise from a CD ). Heck, even MP3 is good enough when I'm not paying attention.Hearing is supposed to only
go to 20 KHz, but 44.8 KHz (Nyquest rate) CD sampling doesn't seem to cut
it - at least for all people. (Side note: when I switched to CDs from
LPs, I was disappointed by the sound - even though I got a good CD player
and the rest of my audio chain was the same. Yet, I stopped playing LPs
because of the CD convenience factor and CDs were just SO cool.)
I can tell you that sampling theory works a lot better in audio world than video/graphics. Sound is more cyclical and less phase dependent. It was one of the biggest adjustment I had to make when switching from audio to video.Nyquest rate may be okay as a theoretical construct, but does it really work?
We're not theoretical. And if there are 2.1 pixels, say, per line pair,
what does the image of the resolution chart look like? Can you, as an
observer, easily tell what the frequency of the target lines is supposed
to be?
Pixels are fat samples, and in signal processing, it's sample-and-hold. That in itself, is a low pass filter. Sinc filter to be exact.Does Nyquest rate require point samples? (I honestly don't know.)
Pixels are fat samples.
I am not too hung up on black and white lines. And I suspect most digital camera's max recordable black and white line pairs is a bit lower than pixel_width/2. So the 4 pixels per pair may be closer to the truth afterall. And with enough points, contrast threshold is not necessary.Yep, I thought about that some more after I posted it, and I think you
are right. However, I also suspect that now we're REALLY starting to get
into the area of using a contrast threshold as a way to determine the
frequency of the target lines pairs. The pixel centers may lie upon a
black line (or a white line) but the corners may not; such pixels would
be less of a pure black (or white). A resolution target is likely to
look much less crisp. As above, the underlying target resolution may be
determinable through technical analysis, but what do our photographs look
like?
Well
while you guys were having a pleasant chat last night I was working on a
film vs digital simulation using Excel spreadsheet. Hope
you have access to Excel. I thought it would be easier just to E-mail it
to you to look out. I will also post this text part of the
message in the forum.
I will describe what the spreadsheet does.
The sheet contains two 20X20 and a 10X10 matrices:
On the upper right we have an image which is to be photographed. In the
example I am sending it is a sloppy 6. The image can
be changed to your liking to test various results.
The upper left is a simulation of randomly placed crystals. Each cell
has a 50/50 chance of containing a crystal and if it contains a
crystal it will be turned on (exposed) if it also lines up with a part of
the image on the upper right. To try a different crystal pattern
enter a space or backspace in an unocupied cell (no formula either). The
spreadsheet will recalculate the random placement of
crystals and display the new results.
The lower left represents a CCD type of sensor array. It looks like a
20X20 matrix but the formulas are arranged so as to behave
as a 10X10 matrix with each 'pixel' being a 2X2 square.
The 'image' in the upper right is the one I drew before entering the
formulas so it was not fixed to make the 'pixel' representation
look more accurate. There are images you can input that look worse than
this one but as long as the image is not too detailed the
pixel representation is quite consistent.
The 'crystal' representation is hit and miss. Some formula refreshes
show an excelent rendition of the image. In many the image
is unrecognisable.
I believe this illustrates the limitations of using randomly placed
sensors to store information. On the small scale you do not know
what you will get. It is interesting to note in this simulation that on
average there will be 20X20/2 = 200 crystals. The pixel matrix
always contains 10X10=100 pixels. I believe the pixel image is generally
more coherent while only having half the number of
sensors. Ratio of sensors to coherent information may be as high as 4 to
1 in favour of pixels. Maybe even higher.
Although illustrative there are some flaws with this analasis:
1 - It is a binary representation. Real images have colour depth.
2 - Real crystals are not restricted to a square matrix and can overlap
one another.
3 - Under present technology the relative comparison of pixel and crystal
sizes is much greater than in the simulation.
4 - The probability of the presence of a crystal is assumed to be 50/50
in this simulation which may not jive with reality.
I am sure there are other flaws here but it is an interesting experiment.
(the world is very warped with glasses and -6 diopter corrections)
With a slightly larger wire, or a tree limb, the 35mm camera would show
the actual color of the wire or limb.
With a digital camera, it'll probaby be purple.
That's my biggest complaint. I've now tried 3 digital cameras (Epson
3000z, Olympus 2500L, Nikon 990), and I can't seem to take outdoor photos
that include treetops against a bright sky, without seeing blue/purple
leaves and limbs. Same thing for powerlines against a bright sky -- lots
of purple.
Although I don't plan on taking photos of thin wires or powerlines, so I
could care less about them. However, I do care about taking photos of
landscapes, and the quality of the existing digicams is not up to my
expecations.
When digital cameras get around this blue/purple problem, give you the
same exposure/focus capabilities of 35mm cameras, at a reasonable price,
letting me print an 11x14 or larger print, then I'll be happy....
The 3.3MP images are detailed enough for me, with the exception of the
blue/purple fringing... That's my biggest complaint, and they all seem to
have this problem now.
If you have any doubts about 3 MP camers being equal to 35MM film, read
Discovery mag, August 2000, 'The chemistry of Photography' page 24 -27.
Film is not analog, its binary digital!!. It takes at least a 3x3 matrix
of crystals to give a 'gray scale'. That's a 9 division on the number of
crystals to give a 512 level gray scale. Cells (pixels) in CCDs and
CMOS sensors are analog with more possible bits per cell (the a-d
converter resolution).
The artical is wrong about the 'best' digital cameras, but we can forgive
them this time.
Yes it is possable to capture a bright reflection off a thin wire that
would show up in a 35mm film, but you could not tell much from it because
you would not have enough information about the wire, only that it was
there. In the digital world, you wouldn't see the wire until the number
of pixels went up 2 or 3 fold, but when you did, you would know more
than that it was just there.
Yes it will be another few years before digital surpass film completely
and by a wide enough margin that film is left to history buffs, but will
come and sooner, not later. We are standing on the edge now!!!
I just noticed your post here from 3 days ago...
Apparently, some of the purple fringing problem has to due with a
"microlens" over each pixel within the CCD. Phil Askey explained it a
while back here:
http://www.dpreview.com/forums/read.asp?forum=1001&message=278022&query=microlens+purple
By the way, I did review photos from my 35mm camera (a Nikon N4004S,
using an inexpensive Sigma 28-80mm Zoom Lens), and was unable to find any
that exhibited the purple fringing problem that I see in photos from the
Digital Cameras that I've tried so far (in the last couple of months,
I've tried an Epson 3000z, an Olympus 2500L, and a Nikon 990).
However, in fairness to the Digital Cameras, most photos come out looking
great. Only the outdoor photos that I take, including treelines against
a bright sky, seem to be unacceptable in quality to me. Others don't
seem to notice or mind it as much as I do.
(the world is very warped with glasses and -6 diopter corrections)
With a slightly larger wire, or a tree limb, the 35mm camera would show
the actual color of the wire or limb.
With a digital camera, it'll probaby be purple.
That's my biggest complaint. I've now tried 3 digital cameras (Epson
3000z, Olympus 2500L, Nikon 990), and I can't seem to take outdoor photos
that include treetops against a bright sky, without seeing blue/purple
leaves and limbs. Same thing for powerlines against a bright sky -- lots
of purple.
Although I don't plan on taking photos of thin wires or powerlines, so I
could care less about them. However, I do care about taking photos of
landscapes, and the quality of the existing digicams is not up to my
expecations.
When digital cameras get around this blue/purple problem, give you the
same exposure/focus capabilities of 35mm cameras, at a reasonable price,
letting me print an 11x14 or larger print, then I'll be happy....
The 3.3MP images are detailed enough for me, with the exception of the
blue/purple fringing... That's my biggest complaint, and they all seem to
have this problem now.
If you have any doubts about 3 MP camers being equal to 35MM film, read
Discovery mag, August 2000, 'The chemistry of Photography' page 24 -27.
Film is not analog, its binary digital!!. It takes at least a 3x3 matrix
of crystals to give a 'gray scale'. That's a 9 division on the number of
crystals to give a 512 level gray scale. Cells (pixels) in CCDs and
CMOS sensors are analog with more possible bits per cell (the a-d
converter resolution).
The artical is wrong about the 'best' digital cameras, but we can forgive
them this time.
Yes it is possable to capture a bright reflection off a thin wire that
would show up in a 35mm film, but you could not tell much from it because
you would not have enough information about the wire, only that it was
there. In the digital world, you wouldn't see the wire until the number
of pixels went up 2 or 3 fold, but when you did, you would know more
than that it was just there.
Yes it will be another few years before digital surpass film completely
and by a wide enough margin that film is left to history buffs, but will
come and sooner, not later. We are standing on the edge now!!!
That's why you don't hear complaints about the D1 with purple fringe,
larger sensor. Give manufactures a few more years and they will fix the
problem after people complain enough.
I just noticed your post here from 3 days ago...
Apparently, some of the purple fringing problem has to due with a
"microlens" over each pixel within the CCD. Phil Askey explained it a
while back here:
http://www.dpreview.com/forums/read.asp?forum=1001&message=278022&query=microlens+purple
By the way, I did review photos from my 35mm camera (a Nikon N4004S,
using an inexpensive Sigma 28-80mm Zoom Lens), and was unable to find any
that exhibited the purple fringing problem that I see in photos from the
Digital Cameras that I've tried so far (in the last couple of months,
I've tried an Epson 3000z, an Olympus 2500L, and a Nikon 990).
However, in fairness to the Digital Cameras, most photos come out looking
great. Only the outdoor photos that I take, including treelines against
a bright sky, seem to be unacceptable in quality to me. Others don't
seem to notice or mind it as much as I do.
(the world is very warped with glasses and -6 diopter corrections)
With a slightly larger wire, or a tree limb, the 35mm camera would show
the actual color of the wire or limb.
With a digital camera, it'll probaby be purple.
That's my biggest complaint. I've now tried 3 digital cameras (Epson
3000z, Olympus 2500L, Nikon 990), and I can't seem to take outdoor photos
that include treetops against a bright sky, without seeing blue/purple
leaves and limbs. Same thing for powerlines against a bright sky -- lots
of purple.
Although I don't plan on taking photos of thin wires or powerlines, so I
could care less about them. However, I do care about taking photos of
landscapes, and the quality of the existing digicams is not up to my
expecations.
When digital cameras get around this blue/purple problem, give you the
same exposure/focus capabilities of 35mm cameras, at a reasonable price,
letting me print an 11x14 or larger print, then I'll be happy....
The 3.3MP images are detailed enough for me, with the exception of the
blue/purple fringing... That's my biggest complaint, and they all seem to
have this problem now.
If you have any doubts about 3 MP camers being equal to 35MM film, read
Discovery mag, August 2000, 'The chemistry of Photography' page 24 -27.
Film is not analog, its binary digital!!. It takes at least a 3x3 matrix
of crystals to give a 'gray scale'. That's a 9 division on the number of
crystals to give a 512 level gray scale. Cells (pixels) in CCDs and
CMOS sensors are analog with more possible bits per cell (the a-d
converter resolution).
The artical is wrong about the 'best' digital cameras, but we can forgive
them this time.
Yes it is possable to capture a bright reflection off a thin wire that
would show up in a 35mm film, but you could not tell much from it because
you would not have enough information about the wire, only that it was
there. In the digital world, you wouldn't see the wire until the number
of pixels went up 2 or 3 fold, but when you did, you would know more
than that it was just there.
Yes it will be another few years before digital surpass film completely
and by a wide enough margin that film is left to history buffs, but will
come and sooner, not later. We are standing on the edge now!!!
That's why you don't hear complaints about the D1 with purple fringe,
larger sensor. Give manufactures a few more years and they will fix the
problem after people complain enough.
I just noticed your post here from 3 days ago...
Apparently, some of the purple fringing problem has to due with a
"microlens" over each pixel within the CCD. Phil Askey explained it a
while back here:
http://www.dpreview.com/forums/read.asp?forum=1001&message=278022&query=microlens+purple
By the way, I did review photos from my 35mm camera (a Nikon N4004S,
using an inexpensive Sigma 28-80mm Zoom Lens), and was unable to find any
that exhibited the purple fringing problem that I see in photos from the
Digital Cameras that I've tried so far (in the last couple of months,
I've tried an Epson 3000z, an Olympus 2500L, and a Nikon 990).
However, in fairness to the Digital Cameras, most photos come out looking
great. Only the outdoor photos that I take, including treelines against
a bright sky, seem to be unacceptable in quality to me. Others don't
seem to notice or mind it as much as I do.
(the world is very warped with glasses and -6 diopter corrections)
With a slightly larger wire, or a tree limb, the 35mm camera would show
the actual color of the wire or limb.
With a digital camera, it'll probaby be purple.
That's my biggest complaint. I've now tried 3 digital cameras (Epson
3000z, Olympus 2500L, Nikon 990), and I can't seem to take outdoor photos
that include treetops against a bright sky, without seeing blue/purple
leaves and limbs. Same thing for powerlines against a bright sky -- lots
of purple.
Although I don't plan on taking photos of thin wires or powerlines, so I
could care less about them. However, I do care about taking photos of
landscapes, and the quality of the existing digicams is not up to my
expecations.
When digital cameras get around this blue/purple problem, give you the
same exposure/focus capabilities of 35mm cameras, at a reasonable price,
letting me print an 11x14 or larger print, then I'll be happy....
The 3.3MP images are detailed enough for me, with the exception of the
blue/purple fringing... That's my biggest complaint, and they all seem to
have this problem now.
If you have any doubts about 3 MP camers being equal to 35MM film, read
Discovery mag, August 2000, 'The chemistry of Photography' page 24 -27.
Film is not analog, its binary digital!!. It takes at least a 3x3 matrix
of crystals to give a 'gray scale'. That's a 9 division on the number of
crystals to give a 512 level gray scale. Cells (pixels) in CCDs and
CMOS sensors are analog with more possible bits per cell (the a-d
converter resolution).
The artical is wrong about the 'best' digital cameras, but we can forgive
them this time.
Yes it is possable to capture a bright reflection off a thin wire that
would show up in a 35mm film, but you could not tell much from it because
you would not have enough information about the wire, only that it was
there. In the digital world, you wouldn't see the wire until the number
of pixels went up 2 or 3 fold, but when you did, you would know more
than that it was just there.
Yes it will be another few years before digital surpass film completely
and by a wide enough margin that film is left to history buffs, but will
come and sooner, not later. We are standing on the edge now!!!
Those glasses I got really have changed a lot about what I thought about
camera lenses.
That's why you don't hear complaints about the D1 with purple fringe,
larger sensor. Give manufactures a few more years and they will fix the
problem after people complain enough.
I just noticed your post here from 3 days ago...
Apparently, some of the purple fringing problem has to due with a
"microlens" over each pixel within the CCD. Phil Askey explained it a
while back here:
http://www.dpreview.com/forums/read.asp?forum=1001&message=278022&query=microlens+purple
By the way, I did review photos from my 35mm camera (a Nikon N4004S,
using an inexpensive Sigma 28-80mm Zoom Lens), and was unable to find any
that exhibited the purple fringing problem that I see in photos from the
Digital Cameras that I've tried so far (in the last couple of months,
I've tried an Epson 3000z, an Olympus 2500L, and a Nikon 990).
However, in fairness to the Digital Cameras, most photos come out looking
great. Only the outdoor photos that I take, including treelines against
a bright sky, seem to be unacceptable in quality to me. Others don't
seem to notice or mind it as much as I do.
(the world is very warped with glasses and -6 diopter corrections)
With a slightly larger wire, or a tree limb, the 35mm camera would show
the actual color of the wire or limb.
With a digital camera, it'll probaby be purple.
That's my biggest complaint. I've now tried 3 digital cameras (Epson
3000z, Olympus 2500L, Nikon 990), and I can't seem to take outdoor photos
that include treetops against a bright sky, without seeing blue/purple
leaves and limbs. Same thing for powerlines against a bright sky -- lots
of purple.
Although I don't plan on taking photos of thin wires or powerlines, so I
could care less about them. However, I do care about taking photos of
landscapes, and the quality of the existing digicams is not up to my
expecations.
When digital cameras get around this blue/purple problem, give you the
same exposure/focus capabilities of 35mm cameras, at a reasonable price,
letting me print an 11x14 or larger print, then I'll be happy....
The 3.3MP images are detailed enough for me, with the exception of the
blue/purple fringing... That's my biggest complaint, and they all seem to
have this problem now.
If you have any doubts about 3 MP camers being equal to 35MM film, read
Discovery mag, August 2000, 'The chemistry of Photography' page 24 -27.
Film is not analog, its binary digital!!. It takes at least a 3x3 matrix
of crystals to give a 'gray scale'. That's a 9 division on the number of
crystals to give a 512 level gray scale. Cells (pixels) in CCDs and
CMOS sensors are analog with more possible bits per cell (the a-d
converter resolution).
The artical is wrong about the 'best' digital cameras, but we can forgive
them this time.
Yes it is possable to capture a bright reflection off a thin wire that
would show up in a 35mm film, but you could not tell much from it because
you would not have enough information about the wire, only that it was
there. In the digital world, you wouldn't see the wire until the number
of pixels went up 2 or 3 fold, but when you did, you would know more
than that it was just there.
Yes it will be another few years before digital surpass film completely
and by a wide enough margin that film is left to history buffs, but will
come and sooner, not later. We are standing on the edge now!!!
