Dynamic Range and DPI the BIG LIE!
Phil lippincott
I'm troubled by my perception that no one cares or holds scanner
manufacturers accountable for their false advertising.
I've read on this other discussion groups a number of people
declaring how
wonderful the dynamic range of their scanners were because the
manufacture specification says it's 3.6 density or 3.4 etc. I would like
to publicly announce these specification for dynamic range are totally
false! I wonder if anyone cares.
I am doing a scanner dynamic range density test of several different
manufacturer scanners. Polaroid Sprint Scan, Umax PowerLook 3000, Howtek
Drum scanners (three models), Imacon Flextight II, and Nikon Super Cool
scan.
My testing began with the Kodak standard scanner density step wedge
chart (positive transparency T-126). It is labeled for density responses
for each wedge. I validated the target densities by checking the values
on my X-Rite Status A densitometer. The Kodak values were all within +
or - .02 density of the printed values on the Kodak Target. I then
scanned on each scanner the target based on the histogram Dmin value and
Dmax values set to the step wedge back and white points.
I then took the US air force 1954 optical quality target and scanned
it on each scanner with the stated maximum resolution in DPI. This
target shows with numbered lines down to one part per millimeter optical
resolution.
WOW, I'm shocked the various scanner specifications have little to
no relevance to real optical resolution or color fidelity. I expected
them to be stretching the truth, but I had no idea. How about a reported
3.6 D dynamic range scanner who only can see a density of 2.2 status A
density. No CCD scanner can see a density greater than 2.7 D. Or a 3400
dpi CCD scanner that can only resolve optical quality of about 1300 DPI
for $7,200.
Does anyone care or notice that this is why professional quality
highlight and shadows cannot be seen on most CCD scanners ???
There was one notable stand out CCD scanner and that is the Imacon
Flextight II (made in Denmark). It has a real optical resolution of 4000
dpi (advertised for 5600 dpi) and a Dmax of 3.7 (advertised for 4.1).
Its price is $16,950. In scanners there is no free lunch. With the drum
scanner prices coming down, a HOWTEK 4500 is now $18.900 at AZTEK
Imaging. There just isn't any comparable price performance to these drum
scanners. The scanner saw all of the dynamic range of the test target a
true 3.6D the same as Ektachrome, FujiChrome and Agfa Rs-200 and the
optical quality is above 4000 dpi. The Howtek scanners were the only
ones tested that meet their specifications.
Another interesting note was that use of the Binuscan software
(Polaroid, Agfa, Umax) stretched the color response but did nothing to
improve the dynamic density range from the film.
I still wonder does anyone really care, because the pretenders like
HP, Microtek, Nikon, Agfa, el al keep saying their products do one thing
and they really do another. The public is buying them up with these wild
claims and these firms flourish. While professionals struggle to
reproduce film quality digitally with inferior tools, and firms like
Optonics, Scanview and Howtek struggle to have any understand the BIG
LIE! None of the pretenders hold a candle to a real Photo multiplier
(PMT) drum scanner..
Sincerely
Phil Lippincott
AZTEK IMAGING
aztek.net
phil.li...@worldnet.att.net
Fdecarlo
already knew that. Do scanner manufacturers lie about
their product's specs in order to sell more scanners?
Maybe, but they're the ones who can answer your
claims, so ask them about it. Maybe they can give you
some information about their methodology and how they
arrive at their specs.
I also take issue with your comment that "HP, Microtek,
Nikon, Agfa, el al keep saying their products do one thing
and they really do another. The public is buying them up
with these wild claims and these firms flourish".
What public are you referring to? How many posts have
you read in this newsgroup related to PMT drum scanners?
The vasy majority of posts (and therefore I think, the vast
majority of readers in this newsgroup) are average SOHO
users. They don't have several thousand dollars to spend
on a scanner, and they also don't need the dynamic ranges
these scanners provide. So why complain here about the
shortcomings of $7,000 scanners?
Your post is similar to ones I read when low-cost color inkjets
started making headway in the SOHO market. A $250 Epson
Stylus 800 produces color output which puts some dye sub
printers to shame. But other dye sub printers produce much
better output than the Epson. But does it make any sense to
tell people not to buy a $250 Epson inkjet because some
$8,000 dye sub printers are capable of much better output?
Frank
David Boll
: Hi
: I've read on this other discussion groups a number of people
: declaring how
: wonderful the dynamic range of their scanners were because the
: manufacture specification says it's 3.6 density or 3.4 etc. I would like
: to publicly announce these specification for dynamic range are totally
: false! I wonder if anyone cares.
There is a similar problem with low-end scanners. I have seen some dynamic
range reports that depend completely on the bit depth of the A/D converter.
For example, a 30-bit scanner (with a 10-bit A/D converter) could claim
a dynamic range of 3.0, since:
log10(2^10) = 3.01
This, of course, ignores the fact that these 10 bits are probably
uncompensated (no PRNU adjustment), and the bottom 1 or 2 are probably
noise anyway, and gamma correction, autoexposure both consume some of the
bit depth as well.
The claim "dynamic range=3.0" claim is impossible to test with a reflective
scan. Reflective originals (photos, whatever) typically have a dynamic range
less than 2.0. In fact, it is difficult (not quite impossible) to create
a reflective source > 2.0.
So, since even a 24-bit scanner can claim log10(2^8)=2.4 as a dynamic
range (which is likely sufficient for any reflective source), a better
test for low-end scanners is the signal/noise ratio or quantization test.
Here's a fun test to try on your scanner:
1) Prop the lid open by a couple of inches (if the document cover is
black, affix a sheet of white paper to it)
2) Preview-scan the bed. You should see a greyscale, ranging from
close-to-white near the lid hinge to close-to-black on the other
end. You may need to place a cloth over your scanner to block out
ambient light, in order to get the dark area as dark as possible.
3) Scan a narrow rectangle down the center of the bed, ranging from
fairly light to as dark as possible. Set your scanner to 150 ppi, greyscale.
4) Examine the resulting bitmap in PhotoShop (or PhotoPaint, or something
that allows a histogram option). Look at the dark area of the bitmap
to see if there are visible steps. Histogram the bitmap to see if there
are any gaps/spikes. A good scan will show few gaps in the histogram.
5) Repeat the above with a full-color scan, and check the histogram on
each channel.
If you see lots of gaps in the histogram (probably toward the dark end),
your scanner is not capable of fully populating even an 8-bit range.
Does this mean you have a bad scanner? Probably not - your photos likely
look just fine. It does mean that your scanner is not capable of delivering
8 bits of true image data (much less 10, or 12, or whatever the bit-depth
spec on your scanner is).
--
Dave Boll
(working, but not speaking, for) Hewlett Packard
Valburg
> Hi
> I'm troubled by my perception that no one cares or holds scanner
> manufacturers accountable for their false advertising.
> I've read on this other discussion groups a number of people
> declaring how
> wonderful the dynamic range of their scanners were because the
> manufacture specification says it's 3.6 density or 3.4 etc. I would
> like
> to publicly announce these specification for dynamic range are totally
>
> false! I wonder if anyone cares.
Of course we do. We simply haven't much in the way of choices!
And BTW, thanks for an informative post. Please feel free to duplicate
your efforts on the affordable products from Microtek, Umax, etc. Or
tell us where we can get equally accurate, reliable test results of this
(constantly changing) range of scanners we can afford, and we'll use
them in making our buying choices. But if all you're saying is that the
choices we in the SOHO market have are inferior to those the pre-press
people can afford, you're preaching to the converted.
Sure, I'd love to be able to afford a drum scanner, but I'm just one guy
doing photo restoration in a small market that simply won't pay enough
for the services I offer to allow that. Or even to allow me to farm the
scanning out at $30 a pop. I suspect that most in this newsgroup are a
similar boat.
Regards,
Mitch Valburg
hlee
i was very impressed some time ago with the hp 4c scanner .. you could actually see
detail right down to the least significant bits ... now if you could just fix the
software you would have a great package .. at that time the 4c was rated by a
magazine below several far inferior scanners (my opinion) .. you could not resolve
anywhere near even several least significant bits in these "winning" scanners ....
the magazine actually published the scanned photos from each unit .. i called to
find out what in the world they could possible find better about these units ... it
was color !!! ... not quality, not resolution, not lack of noise but color ... my
ex wife buys new cars based on color ... i explained that color saturation (what hp
4c was down rated for) is a simple adjustment but that there was nothing anyone
can do about noise ... they said that most of their readers would have trouble even
correcting saturation .. i said "good grief" ... and phil i do care ... harry
David Boll wrote:
> Phil lippincott (phil.li...@worldnet.att.net) wrote:
> : Hi
> : I've read on this other discussion groups a number of people
> : declaring how
> : wonderful the dynamic range of their scanners were because the
> : manufacture specification says it's 3.6 density or 3.4 etc. I would like
> : to publicly announce these specification for dynamic range are totally
> : false! I wonder if anyone cares.
>
Tony Sleep
fdec...@aol.com (Fdecarlo) wrote:
> PMT drum scanners are better than CCD scanners? We
> already knew that. Do scanner manufacturers lie about
> their product's specs in order to sell more scanners?
> Maybe, but they're the ones who can answer your
> claims, so ask them about it. Maybe they can give you
> some information about their methodology and how they
> arrive at their specs.
Phil is correct, at least as far as density range is concerned.
Manufacturers quoted D is invariably theoretical, not measured. From my
own tests, a D>2.1 is good going, and that from scanners whose
manufacturers claim 3.0-3.6, based purely on what might be possible
within the bit depth, if everything else was perfect. CCD noise soon
annihilates the ideal, in real life.
> I also take issue with your comment that "HP, Microtek,
> Nikon, Agfa, el al keep saying their products do one thing
> and they really do another. The public is buying them up
> with these wild claims and these firms flourish".
Well, Phil is trying to promote his own solutions, obviously:-) But there
probably is a gulf between what people think these devices can achieve and
what they really do, in terms of accurate transposition from film-digital.
At least there was a gulf between what I needed and hoped for, and what I
found I got for my 1,400GBP.
Yes, it's true that expectations are *the* major variable, and that the
majority of users will be non-critical and seek to do no more than inkjet
snaps, and they will be happy as Larry. But there is an overlap with
professional-standard users who are in need of a bridge between film and
digital, and they (we!) are not well served by the overselling of immature
and transitional technologies. The PC press approaches this stuff as SoHo
kit, the Photo press doesn't seem to have much idea: it's difficult to
judge what is good enough, and for what, especially when the learning
curve is inventing itself!
Digital still has a long way to go, but IMO is closer as a prospect for
everyman than it is for the 'serious' pro-am user who knows and uses what
film can achieve and thought they might find it for less than the cost of
a new car.
Tony Sleep
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
http://www.cix.co.uk/~tsphoto/
Tony Sleep - online portfolio & exhibit + film scanner info & comparisons
Fdecarlo
>From: half...@cix.compulink.co.uk ("Tony Sleep")
>Date: 8/14/98 5:29 PM Pacific Daylight Time
>Message-id: <ExpG0...@cix.compulink.co.uk>
>
>In article <199808120710...@ladder03.news.aol.com>,
>fdec...@aol.com (Fdecarlo) wrote:
>
>> PMT drum scanners are better than CCD scanners? We
>> already knew that. Do scanner manufacturers lie about
>> their product's specs in order to sell more scanners?
>> Maybe, but they're the ones who can answer your
>> claims, so ask them about it. Maybe they can give you
>> some information about their methodology and how they
>> arrive at their specs.
>
>Phil is correct, at least as far as density range is concerned.
>Manufacturers quoted D is invariably theoretical, not measured. From my
>own tests, a D>2.1 is good going, and that from scanners whose
>manufacturers claim 3.0-3.6, based purely on what might be possible
>within the bit depth, if everything else was perfect. CCD noise soon
>annihilates the ideal, in real life.
Amazing. Just like when I pulled my new 17" NEC monitor out of
the box and out came a 15" screen. Another example of supply side
economics at their finest.
So if these dynamic range specs are theoretical (and therefore
meaningless), are you aware of any flatbed scanner testing that
has been done to come up with real numbers? I know Phil
tested some high-end scanners but it would be great to see the
same results for soho models. I bet it'd be a real eye-opener.
Thanks very much for the post, Tony. I was really hung up
with quoting manufacturer's specs to people. Not any more.
Frank
Nathan Wong
> I'm troubled by my perception that no one cares or holds scanner
> manufacturers accountable for their false advertising.
<lot's of info snipped>
> There was one notable stand out CCD scanner and that is the Imacon
> Flextight II (made in Denmark). It has a real optical resolution of 4000
> dpi (advertised for 5600 dpi) and a Dmax of 3.7 (advertised for 4.1).
> Its price is $16,950. In scanners there is no free lunch. With the drum
> scanner prices coming down, a HOWTEK 4500 is now $18.900 at AZTEK
> Imaging. There just isn't any comparable price performance to these drum
<snip>
I knew this started to sound like an advertisment. Sure enough, IT WAS!
Dude, for what it's worth we don't want to see or hear your ads in a
newsgroup. Go buy yourself some ad space in a magazine or something.
David Boll
: i was very impressed some time ago with the hp 4c
: to the least significant bits ... now if you could
: at that time the 4c was rated by a magazine below several
: anywhere near even several least significant bits in these "winning" scanners
I remember that review quite well myself. I was surprised that they were
judging image quality with a qualitative estimate of color accuracy.
Color accuracy is an important componant of image quality, but I would
add sharpness, noise, quantization, neutral-axis preservation, and probably
a couple others to that list.
Another important point is that it is very possible for knowledgeable
observers to disagree on a comparitive image quality rating, based on
unequal weighting of the componants of image quality. If I value sharpness
above all else, and you value color accuracy, we would certainly differ
in our opinions regarding an image that was tack-sharp but slightly off
in color.
: i explained that color saturation (what hp
: 4c was down rated for) is a simple adjustment
: but that there was nothing anyone can do about noise
The 4c (and its successors, the 6xxx series) were/are excellent in their
ability to deliver low-noise images. They produce somewhere between 10 and
11 bits per pixel of true image data (the variance is mostly due to the
lamp). I have often thought that we (HP) should turn this into an
advertising spec. The problem (according to Marketeers) is that the spec
is too similar to the bit-depth spec used today, so that a spec of
30 bpp true image data
appears inferior to a spec of
36 bit color
It is hard to know what to do about this!
The engineer side of me would like to spec scanners on (say) MTF, S/N,
and bit depth of true image data. However, specs like these are very
difficult to sell.
--
Dave Boll
(working, but not speaking, for) Hewlett-Packard
Fdecarlo
If HP scanners can deliver better SNR, they should do some comparitive
benchmarks and publish the results. As it is, the only references people
have are meaningless specs and other marketing hype.
Frank
>Subject: Re: Dynamic Range and DPI the BIG LIE!
>From: dwb...@gr.hp.com (David Boll)
<snip>
michel_...@hp.com
On 12 Aug 1998 05:05:21 GMT, Phil lippincott <phil.li...@worldnet.att.net> wrote:
> Hi
> manufacturers accountable for their false advertising.
> declaring how
> wonderful the dynamic range of their scanners were because the
> manufacture specification says it's 3.6 density or 3.4 etc. I would like
> to publicly announce these specification for dynamic range are totally
> false! I wonder if anyone cares.
> manufacturer scanners. Polaroid Sprint Scan, Umax PowerLook 3000, Howtek
> Drum scanners (three models), Imacon Flextight II, and Nikon Super Cool
> scan.
> My testing began with the Kodak standard scanner density step wedge
> chart (positive transparency T-126). It is labeled for density responses
> for each wedge. I validated the target densities by checking the values
> on my X-Rite Status A densitometer. The Kodak values were all within +
> or - .02 density of the printed values on the Kodak Target. I then
> scanned on each scanner the target based on the histogram Dmin value and
> Dmax values set to the step wedge back and white points.
> I then took the US air force 1954 optical quality target and scanned
> it on each scanner with the stated maximum resolution in DPI. This
> target shows with numbered lines down to one part per millimeter optical
> resolution.
> WOW, I'm shocked the various scanner specifications have little to
> no relevance to real optical resolution or color fidelity. I expected
> them to be stretching the truth, but I had no idea. How about a reported
> 3.6 D dynamic range scanner who only can see a density of 2.2 status A
> density. No CCD scanner can see a density greater than 2.7 D. Or a 3400
> dpi CCD scanner that can only resolve optical quality of about 1300 DPI
> for $7,200.
> Does anyone care or notice that this is why professional quality
> highlight and shadows cannot be seen on most CCD scanners ???
> Flextight II (made in Denmark). It has a real optical resolution of 4000
> dpi (advertised for 5600 dpi) and a Dmax of 3.7 (advertised for 4.1).
> Its price is $16,950. In scanners there is no free lunch. With the drum
> scanner prices coming down, a HOWTEK 4500 is now $18.900 at AZTEK
> Imaging. There just isn't any comparable price performance to these drum
> true 3.6D the same as Ektachrome, FujiChrome and Agfa Rs-200 and the
> optical quality is above 4000 dpi. The Howtek scanners were the only
> ones tested that meet their specifications.
> Another interesting note was that use of the Binuscan software
> (Polaroid, Agfa, Umax) stretched the color response but did nothing to
> improve the dynamic density range from the film.
> I still wonder does anyone really care, because the pretenders like
> and they really do another. The public is buying them up with these wild
> reproduce film quality digitally with inferior tools, and firms like
> Optonics, Scanview and Howtek struggle to have any understand the BIG
> LIE! None of the pretenders hold a candle to a real Photo multiplier
> (PMT) drum scanner..
>
> Sincerely
> Phil Lippincott
> AZTEK IMAGING
> aztek.net
> phil.li...@worldnet.att.net
Cool Scan II (LS20). Could you please let me know which Nikon you tested and how it came out in your test?
Best regards,
Michel Billard
--------------------------------------------------------------------
Posted using Reference.COM http://WWW.Reference.COM
FREE Usenet and Mailing list archive, directory and clipping service
--------------------------------------------------------------------
J.A. McCulloch
> I then took the US air force 1954 optical quality target and scanned
> it on each scanner with the stated maximum resolution in DPI. This
> target shows with numbered lines down to one part per millimeter optical
> resolution.
Perhaps I'm wrong, but ...
Doesn't the UASF 1954 show resolution in LPI (Lines Per Inch)?
If so ... it takes two (2) dots to reproduce one (1) line. One dot for
the line itself, and one dot for the background. OTW there is just
greyish tones = not resolved.
Therefore, if the target LPI is measured (resolved) at (say) 1500 LPI,
then the scanner is 3000 DPI.
jame...@my-dejanews.com
> I still wonder does anyone really care, because the pretenders like
> HP, Microtek, Nikon, Agfa, el al keep saying their products do one thing
> and they really do another. The public is buying them up with these wild
> claims and these firms flourish. While professionals struggle to
> reproduce film quality digitally with inferior tools, and firms like
> Optonics, Scanview and Howtek struggle to have any understand the BIG
> LIE! None of the pretenders hold a candle to a real Photo multiplier
> (PMT) drum scanner..
Despite the interminable perpetuation of the PMT v. CCD scanner debate (which
I thought was resolved years ago - Scanview, in particular, cited above, seem
to think so) I am amazed by the suggestion that the world's scanner
manufacturers are conspiring to mislead customers about the specification of
their products. Although I am not qualified to enter in to technical debate
over the methodology of your testing, I have no doubt that most or all of the
manufacturers would provide a full validation for their specifications - in
our own product development cycle design and spec validation are required at
the prototype stage, AND must be repeated on full production units.
On the subject of optical resolution, a scanner with 3048 CCD sensors per
objective inch of an original, being stepped across an image at 1/3048 inch
per line sample is correctly described as a 3048x3048dpi scanner (this spec
comes from the UMAX PowerLook 3000, which you mention). These "maximum
optical resolution" scans do not contain either identical adjacent pixels or
captured lines of pixels.
If there is any argument over resolution and the "marketeering" of data, then
the main issue is the description of a "600x1200dpi" scanner (thats optical x
mechanical), as a "1200dpi scanner". Even in this case, 1/1200th of an inch
can be resolved in one axis of the scanner.
On the subject of Dmax, there are plenty of grey issues: very few
manufacturers truly represent Dmax and Drange in their marketing materials,
more commonly referring to just "density". Yes, we could just quote the
maximum theoretical density determined by the bit depth as the Dmax value -
BUT many of these scanner models operate internally in 8-bit/ch mode when
scanning 24-bit images, only the more professional models use high-bit (10,
12, 14, 16/ch) internal processing, gamma correction etc. Yes, true Dmax is
reliant on lamp intensity - but a Dmax of 4.1 is no use if you can't read
0.1-1.2! In our own case, there are plenty of our scanner models with quoted
Dmax of *below* the theoretical maximum for the internal bit-depth, and we do
in fact have very precise Dmax-achieved figures, on a batch-by-batch basis,
which also take in to account the SNR and other image noise. As also
mentioned in other posts, some bits are lost to extraneous data and noise, in
conversion etc.
One note, though: there is one (low-end, mainly PC orientated) scanner
manufacturer currently marketing a 30-bit, "3.3 density" scanner. Go figure -
they're the exception, not the norm.
No-one doubts that PMT-based scanners can still achieve higher Dmax's than a
CCD scanner (>3.7ish, at the moment). Still, professional reprographic
companies throughout the world (not scientists) can now use CCD-based
scanners for 90%+ of their scanning, with far higher productivity and greatly
reduced TCO. The optical design of a PMT device inevitably gives a smaller
objective sample area, hence a "sharper" image - but the same design also
gives very poor responses in some colour ranges, between sharply contrasting
colour values etc.
> Another interesting note was that use of the Binuscan software
> (Polaroid, Agfa, Umax) stretched the color response but did nothing to
> improve the dynamic density range from the film.
The Binuscan software - more specifically the RecoTech portion of the image
processing engine - is designed to protect the absoloute Dmax and Dmin data
captured from a "raw" scan, before these sensitive colour ranges are destroyed
by bit-rate conversions, curve and level tool applications, gamma corrections
etc. The same engine also regenerates missing data in areas of subtle colour
change affected by too low a bit rate or the poor responses in some colour
ranges of even high-quality CCDs.
There are various places you can read about RecoTech, including
http://www.umax.co.uk/binuscan/ and http://www.binuscan.com/ - its
significance is that an entire image can be radically corrected without loss
of any detail. Scanner driver software inevitably post-processes image data,
resulting in loss of some of the sensitive info; but if you've ever seen a
true "raw scan" and spent literally hours trying to correct it, you'll know
just how valuable Binuscan's technologies are.
While we're listing the many, many CCD scanner manufacturers who "must"
allegedly) have been misleading us for so many years, we should recognise a
few less "easy" targets as well: Leaf, Kodak, Scitex, Linotype-Hell,
Scanview...
For UMAX customers that may suddenly be feeling they've been "lied" to, I will
try and publish a detailed explanation of technical issues behind defining a
scanner's resolution, Dmax, colour fidelity and gamut etc as soon as possible
in http://www.umax.co.uk/support/. I hope other manufacturers will also
contribute, via this group or their own documentation.
James Deveson
UMAX UK.
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member Forum
sho...@ibm.net
>I will try and publish a detailed explanation of technical issues
>behind defining a scanner's resolution, Dmax, colour fidelity and
>gamut etc as soon as possible in http://www.umax.co.uk/support/. I
>hope other manufacturers will also contribute, via this group or their
>own documentation.
Thanks. I for one will appreciate the insights. I found what I think
is a good one-sentence qualitative definition for dynamic range in an
article in the June 1998 Desktop Publishers Journal: "Dynamic range
defines how dark the darkest part of a photograph can be, and how light
the brightest can be, while still showing detail in both." However
attempts to learn more about factors in its calculation and what a
specified value means have led to confusion, particularly where many try
to define it in terms of or relate its value it to the 0-4.0 photo
density standards. (It doesn't help that one term -- Dmax -- appears to
have different meanings in photo density vs scanner specs.)
--
Bob
http://www.jps.net/shomler/gallery.htm
jame...@my-dejanews.com
> Thanks. I for one will appreciate the insights. I found what I think
> is a good one-sentence qualitative definition for dynamic range in an
> article in the June 1998 Desktop Publishers Journal: "Dynamic range
> defines how dark the darkest part of a photograph can be, and how light
> the brightest can be, while still showing detail in both." However
> attempts to learn more about factors in its calculation and what a
> specified value means have led to confusion, particularly where many try
> to define it in terms of or relate its value it to the 0-4.0 photo
> density standards. (It doesn't help that one term -- Dmax -- appears to
> have different meanings in photo density vs scanner specs.)
Its a good one-line description; in reality the aim when designing the
scanners is to make sure the light source, CCD and optical design can reach a
high enough Dmax while maintaining a Dmin around 0.1-0.2; then electronically
to make sure the internal bit-depth is high enough to capture smooth
gradations (& minimum noise) in the very sensitive highlight and shadow
areas. Getting a higher Dmax is simple - the medical imaging industry has
been using laser light sources to scan Dmax 4.6+ for years, but the average
X-ray has a Dmin of 1.0+!
Optical density is a pretty well defined physical measure - its easier to
understand as the opacity of an area of a transmissive film than to think
about reflective originals. Again, density targets are pretty easy to
manufacture, with patches of incrementally more opaque film. The real area of
contention is that near the Dmax of a scanner the 3.3 and the 3.4 patch may
appear to be almost identical - there's no ready agreement as to when the
differences are genuinely due to the scanner's capabilities, and when the
background CCD noise level is so great as to make the differences
insignificant. There's another system of test targets which uses sheets of
polariser, offset at a variable angle, allowing you to minutely increase or
decrease the opacity of the combined sheets... I'll try and show some
examples of Dmax measurement in the tech note, when I get a chance to write
it!
Cheers
James D.
UMAX UK
hlee
really shocked at the poor quality of many cheepy scanners .. yesterday a friend got several low end machines
to evaluate .. most were soooo bad that i would feel sorry for anyone that bought one .. it would be nice to
nip this trend in the bud ... if we had a workable spec that would relate actual performance .. maybe we should
all get together and develop a "standard performance test" ... how can the honest scanner manufacturers survive
when stuff is being sold for $70 with a $30 rebate that claim 36 bit performance ... actually the $40 cost is
about right for these toys .. but of course the customer thinks they are getting a usable machine .. i just
tested the "worst scanner in the world" .. i sure hope its the worst .. chromatic aberration was equal to the
size of a typed character .. focus was awful ... yet there is probably someone that would be happy with it, not
knowing how much better it could be and having no honest comparison or spec ... the 30 and 36 bit stuff is
getting out of hand .. not many come anywhere near the spec in usable depth .. i would love to see a uniform
spec to which scanners could be compared .. i dont think its a matter of people not caring .. its not knowing
what we should expect .. no one knowledgeable looking over the shoulder of the manufacturers (except maybe this
forum) .. i dont think the manufacturers have a good guide line to abide by even if they dont intend deceit
.. i would like to see a copy of your work if its available .. harry
----------------------------------------------
michel_...@hp.com wrote:
> On 12 Aug 1998 05:05:21 GMT, Phil lippincott <phil.li...@worldnet.att.net> wrote:
>
> > Hi
>
> > I'm troubled by my perception that no one cares or holds scanner
>
> > manufacturers accountable for their false advertising.
>
> > I've read on this other discussion groups a number of people
>
> > declaring how
>
> > wonderful the dynamic range of their scanners were because the
>
> > manufacture specification says it's 3.6 density or 3.4 etc. I would like
>
> > to publicly announce these specification for dynamic range are totally
>
> > false! I wonder if anyone cares.
>
> > I am doing a scanner dynamic range density test of several different
>
> > manufacturer scanners. Polaroid Sprint Scan, Umax PowerLook 3000, Howtek
>
> > Drum scanners (three models), Imacon Flextight II, and Nikon Super Cool
>
> > scan.
>
> > My testing began with the Kodak standard scanner density step wedge
>
> > chart (positive transparency T-126). It is labeled for density responses
>
> > for each wedge. I validated the target densities by checking the values
>
> > on my X-Rite Status A densitometer. The Kodak values were all within +
>
> > or - .02 density of the printed values on the Kodak Target. I then
>
> > scanned on each scanner the target based on the histogram Dmin value and
>
> > Dmax values set to the step wedge back and white points.
>
> > I then took the US air force 1954 optical quality target and scanned
>
> > it on each scanner with the stated maximum resolution in DPI. This
>
> > target shows with numbered lines down to one part per millimeter optical
>
> > resolution.
>
> > WOW, I'm shocked the various scanner specifications have little to
>
> > no relevance to real optical resolution or color fidelity. I expected
>
> > them to be stretching the truth, but I had no idea. How about a reported
>
> > 3.6 D dynamic range scanner who only can see a density of 2.2 status A
>
> > density. No CCD scanner can see a density greater than 2.7 D. Or a 3400
>
> > dpi CCD scanner that can only resolve optical quality of about 1300 DPI
>
> > for $7,200.
>
> > Does anyone care or notice that this is why professional quality
>
> > highlight and shadows cannot be seen on most CCD scanners ???
>
> > There was one notable stand out CCD scanner and that is the Imacon
>
> > Flextight II (made in Denmark). It has a real optical resolution of 4000
>
> > dpi (advertised for 5600 dpi) and a Dmax of 3.7 (advertised for 4.1).
>
> > Its price is $16,950. In scanners there is no free lunch. With the drum
>
> > scanner prices coming down, a HOWTEK 4500 is now $18.900 at AZTEK
>
> > Imaging. There just isn't any comparable price performance to these drum
>
> > scanners. The scanner saw all of the dynamic range of the test target a
>
> > true 3.6D the same as Ektachrome, FujiChrome and Agfa Rs-200 and the
>
> > optical quality is above 4000 dpi. The Howtek scanners were the only
>
> > ones tested that meet their specifications.
>
> > Another interesting note was that use of the Binuscan software
>
> > (Polaroid, Agfa, Umax) stretched the color response but did nothing to
>
> > improve the dynamic density range from the film.
>
> > I still wonder does anyone really care, because the pretenders like
>
> > HP, Microtek, Nikon, Agfa, el al keep saying their products do one thing
>
> > and they really do another. The public is buying them up with these wild
>
> > claims and these firms flourish. While professionals struggle to
>
> > reproduce film quality digitally with inferior tools, and firms like
>
> > Optonics, Scanview and Howtek struggle to have any understand the BIG
>
> > LIE! None of the pretenders hold a candle to a real Photo multiplier
>
> > (PMT) drum scanner..
>
> >
>
hlee
Mr 645
original is within the limits of the scanner. I currently have a Umax S-6E ,
Agfa Duoscan and an Optronics Color Getter drum scanner. Even though the S6E
is a $99 scanner I can pull very nice scans from many originals. It's a 300
ppi, 24 bit scanner and I have tested it's Dens. range to be about 1.9. When
scanning prints that do not have extreme shadow or highlight detail in them I
can pull scan off of it that you would think can from a higher end machine.
Jon
sho...@ibm.net
I certainly agree about the S6E, which I also have (almost all the photo
scans in my gallery, URL below, are from the S6E). I've had good
results getting shadow detail using MagicScan. Its weakest area in my
experience is extreme highlight detail. A measured photo density of 1.9
seems to fit well within the approximately 100-to-1 maximum contrast
range of color prints. How did you measure the density?
--
Bob
http://www.jps.net/shomler/gallery.htm
Tony Sleep
wrote a whole lot of erudite stuff:
James,
Personally I can see no explanation for many manufacturers' quoted Density
ranges except what will theoretically fit into the available bit depth.
Have a look at http://www.cix.co.uk/~tsphoto/tech/filmscan/compare.htm - a
selection of scanned grey scale step wedges from the Kodak Q60E-3, done
via various film scanners.
Bob Gann
Tony,
I didn't see the message from James, but
I couldn't agree with you more. Particularly when it comes to
reflective based scanners.
Often the quoted density is beyond what you can reasonably create on a
reflective
original, so you cannot test it effectively.
I've just published a book that discusses this and many other technical
and non-technical
aspects of scanners. The book is titled "Desktop Scanners: Image
Quality Evaluation" and
you can see/preview it at
http://www.phptr.com/ptrbooks/ptr_0130809047.html. In fact, the
book is the 4th edition in a series and discusses scanner technology,
etc.
By the way, most of the "typical specs" for scanners are not very
relavent any more (i.e. ppi,
bit depth, etc.). They have been abused way beyond the limit.
Bob Gann
JDeee
>> ranges except what will theoretically fit into the available bit depth.
>> Have a look at http://www.cix.co.uk/~tsphoto/tech/filmscan/compare.htm - a
>> selection of scanned grey scale step wedges from the Kodak Q60E-3, done
>> via various film scanners.
I'll look at the site - I agree that many products are quoted with the
theoretical maximums, which bare little resemblance to the real
capabilities of the scanners. Without a uniform standard for the
measurement of a scanner's capabilities, their specifications are
always going to be reduced to marketing "one-liners" - as an engineer
I'd rather see the full spec published (to an agreed standard), but it
wouldn't make much sense to most scanner buyers...
In the professional graphics market scanners are often best judged
qualitatively, used with their full suite of applications/drivers.
Published evaluations frequently omit the accompanying software apps,
which can affect both the achievable image quality and overall
productivity... frustrating for the developers who can't affect a
"ms/line" scan speed or an optical resolution, but who can double the
productivity of a system, enhance raw image data etc!
>I didn't see the message from James, but
>I couldn't agree with you more. Particularly when it comes to
>reflective based scanners.
>Often the quoted density is beyond what you can reasonably create on a
>reflective
>original, so you cannot test it effectively.
My main contention was that there isn't a conspiracy amongst the
scanner manufacturers to mislead potential customers, as suggested in
an earlier post. There were also a few dubious perpetuations of the
PMT/CCD debate, and some barely disguised advertising...
[snip]
Presuming you can agree how to measure a Dmax value, the important
Dmax and Drange specifications are frequently distorted in a "Density"
one-liner spec of little relevance - even the most basic flatbed
scanners reach or exceed the maximum density of a reflective
photographic original. As Tony pointed out it's difficult to interpret
a manufacturer's density spec even when scanning trannies. The
solution?... a standard specification for image capture devices? a
series of standardised test targets to allow image quality to be
assessed subjectively?
Another example - Using one figure only, answer: How may scans per
hour can your scanner do? <g> In a recent review of mid-range
flatbeds, the "scans per hour" figures (quoted without explanations)
requested from all of the manufacturers were based on totally
different image sizes (varying by a factor of four!), correction and
separation procedures, with or without post-processing software etc.
>By the way, most of the "typical specs" for scanners are not very
>relavent any more (i.e. ppi,
>bit depth, etc.). They have been abused way beyond the limit.
I agree! There's no point in sampling a small colour range from a poor
optical design to 48 bits (but you'd still call the scanner 48 bit!),
you can precisely state the optical resolution of a scanner based on
the number of CCD elements, imaging area etc but the results could
still be poorly focussed and the effective resolution much lower...
Suggestions for a standardised imaging device spec. welcome!
Cheers