----- Original Message -----
From: "Tony Sleep" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Monday, September 18, 2000 3:51 AM
Subject: RE: Group Scan site has posted SS4000 scans
> 1. Highlight detail can be clipped in software, but that's the software
> and/or operator. 16bits will *always* get everything, with some headroom,
> and is necessary.
Not correct Tony. If the scanner has a limited Dmin or incorrectly sets the
exposure so that the CCD saturates then it will clip and the software/operator
can't fix that. 16 bit does *NOT* always get everything let alone with some
headroom.
I assume you have not actually read our test protocol... It uses Vuescan in
16 bit mode with wp=0.001%. This was intended to establish the "clipping"
point of the data and reflects what the scanner is actually doing with exposure
barring Vuescan bugs.
There may be Vuescan bugs depending on whether it calculates the exposure
setting or uses the scanners firmware. Ed would have to give us some guidance
on that point.
> Shadows can get clipped even in 16bit. This is a
> hardware limitation - the minimum black point is set unnecessarily high
> in firmware
Software can also affect this unless the bp is set to 0.0% which we have
specified in the test protocol. So our test protocol gets all the black data
the scanner provides -- precisely the information we are seeking. We want to
see what the scanner/firmware does to the black point.
> 2. The Pol4000 is inherently as sharp as the pixel resolution allows. Give
> it a sharp TMax400CN neg, turn A/F on, and the result will be nearly
> indistinguishable from an A4 bromide made with a Schneider Componon-S.
> There is a very minor loss of micro-detail in the scan, but it is
> critically sharp. However you will need to apply just enough USM to see
> this fully, as aliasing (not grain, vanilla) masks sharpness in all CCD
> units to various degrees. More than approximate comparisons here are very
> difficult then, because (a)you can't actually see how sharp the scanner is
> without using USM, and (b)the amount of USM which is optimal varies for
> each scanner, partly as a function of pixel resolution, partly depending
> on the CCD and electronics themselves.
>
Using images and playing with USM (which actually reduces resolution) is
completely uncontrolled since you have no method of determining the images
inherent resolution and no method of quantifying any differences. Also, until
your compare with other scanners it is hard to know if your scanner is the
same, better or worse than the competition. Your tests are anecdotal and high
subjective. For our uses, the way to test the resolution is to use a
resolution slide. The scanners with more resolution scores higher.
Also, USM is a red herring since one can apply it and fool the eye into
thinking there is more resolution than exists. This can be done with any
scanner. The way to avoid this is use a real resolution target since the
presence or absence of USM doesn't affect the resolution score when using a
proper resolution target.
If you have an Polaroid and wish to contribute to the groups knowledge then
please scan a resolution target and post the differences between a 3000 dpi
scan and a 4000 dpi scan. That isn't perfect but it would help us determine
the likelihood of some kind of problem unfairly affecting the Polaroid results.
Also, you are welcome to get in on the groups list and scan the resolution
target (and the other images) so we can have another Polaroid sample in case's
Johnnys unit is faulty. But if you do participate you must read and use the
test protocol. It is designed to remove the skill level of the operator as a
factor and minimize the software impacts.
Rather than critisize, contribute. It just sounds like sour grapes now.
> 3. CCD noise just is not an issue with this scanner, to date it is by far
> the quietest device I have looked at, matched only by the LS2000 using 16x
> multiscanning. I suspect that the black-point clipping masks amputates
> this to an extent, but the fact is that I have never seen noise of any
> significant amount in *any* scan I have gotten from a 4000. Shadow
> compression happens, and you can tweak the gamma to try and linearise this
> and still not be troubled with noise. I do this routinely.
Well noise is clearly an issue on the Halloween slide. So we have to deal
with that somehow.
> Anyway, the blown highlights are not a 4000 problem at all, but can happen
> with Vuescan used in 8 bits. If you have a scanner which is capable than
> >8bits worth of dynamic range, something is going to go out of range.
Hmmm, are you saying that the Polaroid saturates the CCD when it is switched
into 8 bit mode?
If so then this sounds like a flaw with the Polaroid.
A 8 bit scanner should set the zero value to the minimum reading and the
maximum to the maximum reading just like 16 bit mode. The difference is in
the intervening quantization levels not the dynamic range.
Anyway since the test protocol uses 16 bit mode, this point seems irrelevant
doesn't it?
> A noisy CCD reduces dynamic range. It would therefore be unsafe to conclude
> that a scanner which doesn't exhibit those symptoms is superior.
Both a noisy CCD and a quiet CCD have "data" at the very bottom sample values.
The difference is that in one the "data" is noise and the other it is real
data. Vuescan cannot know what is noise and what is data so must treat both
the same. I
Anyway, even if I accept your arguments which I'm not convinced, your point
seems a little strange. You are saying that the quiet scanner is losing
highlights but the user is supposed to consider this superior to the noisy
scanner which preserves the highlights !!! Surely, losing highlights is a bad
thing no matter why it happens. What am I missing?
> I agree! But the alternative is testing of all scanners which is full of
> holes.
>
You haven't shown any holes. In fact, quite the contrary to the extent you've
raised an issue the test protocol has already dealt with it.
Perhaps there *are* holes, but please read the test protocol carefully and
suggest concrete things we can do to fix them. Better to have done it when we
started but still not too late.
> A further issue is interpretation. It is easy to be misled by initial
> appearances, eg excessive saturation looks punchy and wonderful and low
> saturation doesn't, but it may be possible to correct the latter easily
> and the former not at all. Finding out takes detective work, and without
> that, conclusions are just first impressions which are quite mistaken.
> That's why, quite often, the text of my reviews often bears little
> relation to the scan shown, which leads to mystified mail from time to
> time. Not easy, is it? :)
>
I agree.
Anyway, I realize this is awkward for someone such as yourself. You have a
vested interest in preserving control of the "definitive" scanner tests since
you want to do it by subscription and make money from it. A free user test
site that provides the same basic answer would be unnerving. I'm not sure how
to handle that since no-one wants to put you out of business but it really is
quite easy to define suitable comparative tests that show how scanners perform
to a first order approximation. Scanners aren't yet that good that subtle and
sophisticated tests are needed. We're still dealing with grain, noise, shadow
detail problems etc that are quite gross. Saying it can't be done is just
scare mongering IMHO.
Having said that, making final conclusions based on a single data set would be
silly. There is enough people claming the Polaroid is better to at least
require caution. And the recent data provided by the Scan Elite tests add more
info which suggests very high variability in the resolution (at least) of the
Polaroid.
To me, the biggest value of the group site is when all the scanners have
scanned all the data. It will be come quite clear which ones do well and which
don't. And despite all the rhetoric, I'm quite sure the scanners that look
better most of the time *are* better. I'm not interested in some theoretical
claim that scanner A is better than B if after 50 scans it clearly is worse....
So, the results are coming online. I've led the charge with observations of
what I see. If someone sees it differently or has information that suggests a
different interpretation or I've missed something important then put that
forward for discussion. Criticism, however violent, doesn't help anyone.
The test may not be perfect but they are definitely good enough to provide
meaningful results if people actually contribute rather than tear down..
Cheers,
Byron
====================================================================
The filmscanners mailing list is hosted by http://www.halftone.co.uk
To resign, <mailto:[EMAIL PROTECTED]> with UNSUBSCRIBE FILMSCANNERS in the
title, or UNSUBSCRIBE FILMSCANNERS_DIGEST if you are reading the Digest.
