Conference noted::ibmpc-95

  Just read the "Best Buy" review of the HP 5p (I think that's the number) in
  Home Office Computing.  Does anyone have a personal review of the scanner? 
  How does it compare with its big but older brothers 4p and 4cx?

  Thanks.

  	mikeP

    PC Connection is now advertising the MicroTek ScanMaker E3 for $199.
    
    BTW, has anyone else had any experience with the "Inland Technologies"
    scanner that was mentioned a few notes ago?
    
    Burns
    

	RE: .-1 and a few others

Sorry, I only get into the office in DEC once a month or so, and don't read
notes very often. 
    
>    BTW, has anyone else had any experience with the "Inland Technologies"
>    scanner that was mentioned a few notes ago?
    
	I have one. Actually, it's the *second* one; the first one
	died after a few weeks.

	The importer replaced it without too many questions. However,
	the replacement unit makes some rather odd noises, though it
	does "scan" reasonably well.

	Both units have an annoying intermittent hardware malfunction which 
	happens about once in about 20 scans, and which ruins the scan. 
	You can hear the problem, and you know to not expect anything, 
	but just to repeat the operation. Annoying, but not serious.

	The "scantastic" software is a "bundle": that means there are
	some bugs, and the company (I spoke with them in Taiwan) is in
	no position to do anything about any of it. You get it, you got it.
	
	When I first started out scanning, I used some of the "scantastic"
	tools. However, I soon replaced all of them with much better
	software. The "scantastic" tools are still installed on my system, 
	but I haven't used them in months.


Now, on scanning negatives, flatbeds, etc.:

I have done a lot of scanning in the last year (100's of images). I have
learned many lessons from this work. Among them:

	1. The processing software is more important than the device
	   interface. When I started using Photoshop (instead of things
	   like Paint Shop Pro), suddenly my scanning improved by an
	   order of magnitude, using the same hardware and driver.

	1a.Photoshop is able to "invert" a color negative and give
	   decent results in positive.

	   I am doing a set of filters for Photoshop which will allow
	   small editing changes in this regard. I hope to have them for sale
	   in a few weeks. Write to me if interested at donf@netmedia.net.il

	2. There are "negative / slide" adapters available for most
	   flatbeds, but not for the Scan-Tak.

	   I would definitely advise *against* the use of same for
	   any *serious* use, particularly for small negatives (e.g.,
	   35mm).

	   For small negatives, you need to get a real resolution of
	   300dpi or more. Though the flatbed scanners will theoretically 
	   "do" this, everything has to be exactly right to get good results 
	    -- and the results are always inferior to the drum.

	   I routinely get amazingly better results using a drum
	   scanner from an agency I use. I have given up trying to
	   scan anything except prints on flatbed type scanners.

	   Don't get me wrong: I'm not ready to get rid of my flatbed;
	   I have many uses for it. But I'm learning to use the right
	   tool for the right use.

	   For larger negatives and transparencies, e.g., 6x6 cm,
	   4x5", etc., results from an "adapted" flatbed are decent.
	   But the drum will always be better.

	3. It is a very reasonable strategy to have *good* prints made,
	   and scan them on the flatbed. If you have a decent size print,
	   i.e., at least 20x25 (8x10"), you can scan it at, say,
	   85 dpi, reduce the resolution and adjust the sizing in
	   Photoshop, and get good results for display on 800x600
	   displays.

	4. The so-called "true color" dynamic range of "24 bits" (which
	   is actually an 8-bit dynamic range of each of R, G, B) is
	   simply not adequate for really high quality photographic
	   use.

	   I have had to "segment" some of my photos: scan twice, with
	   different driver settings, and then re-assemble the image in
	   Photoshop, in order to get reall shadow and highlight detail
	   in the same image. The range of brightnesses available in
	   a good photographic print is much more than 2**8.

	   This is a pain in the neck. It would be much better to have
	   30 (i.e., 10+10+10) or more (36) bits from the beginning.
	   For my use, anyway.

	5. The flatbed resolution numbers are essentially bullshit, and are
	   mostly useless to you.

		- You do need an optical resolution of 300-ish dpi
		  if you're planning to do much OCR from the scans.
		  More than 600 dpi doesn't add anything. The 300*600
		  optical resolution is more than adequate for all
		  but the most critical applications.

		- The higher resolutions are obtained by SW interpolation.

		  Two simple question for you: if you work is so critical 
		  that you need 2400 dpi, why are you willing to accept
		  SW interpolation? It's not good enough. 

		  On the other hand, if your work isn't critical enough
		  for 2400 dpi (99.99% of everyone else), what for?


Anyway, those are some thoughts. I hope they're useful. See my website at
http://ducque.simplenet.com to seem some of the results of my scanning.

Don Feinberg

    re: 557.222 by TAV02::FEINBERG
    
    Don,
    
    This is a warning to you regarding advertising for your side business
    in a DIGITAL notesfile.  It is strictly prohibited by DIGITAL policy.
    It may be slightly different if you're giving freebies to DIGITAL
    employees.
    
    /Bill (moderator)

    Bill,
    
    in Don's defense, his "sin" was 1 line out of >100. Indeed, I had to
    read the note 3 times before I caught the reference you are upset at.
    If the whole note had been about 'buy this you beaut software filter
    from me' I too would be upset, but a small reference, almost as an
    aside while referring to the main subject. Well... as I said, I had to
    look 3 times before I saw it.
    
    I apologise if you don't see this the same way I do, I mean no offense.
    
    Harry

re .208:  CompUSA in Nashua is now carrying the Spot Scantak at $199.

Burns

        Today I was able to buy the UMAX Vista S6E scanner at CompUSA
        Nashua, with the Digital discount, at $163.64.  This was
        less (by $3) than the Digital discount price of the ScanTak
        scanner.

        Bob

I bought a UMAX S-6E from Computer City two weeks ago for $240.
It was in the return section.

Very nice scanner!

Now if I could only get PhotoDeluxe to work properly under NT 
with my video controller (the support battle continues ....)

Scanner works fine off the adaptec controller too!

Doug. 

	Anyone have any examples of scanned images from these
	scanners on their home pages? I'm very interested in getting
	a moderately priced flatbed scanner for some genealogy work
	I'm doing.

	thanks,

							mike

Which is the better scanner?  Is the software with the Scantak better
as one of the previous notes indicate?  I see noters praise both
scanners.


Gim

The reviews I've seen put UMAX near the top with performance and near
the bottom on price, and so they come highly recommended (the 300dpi UMAX
often compares well against competitors 600dpi units).

The Scantak is a reasonable unit with good software and small footprint
and recent pricedrops to below the $200 mark makes it very attractive.

The scantak come with a propriety controller, which was just one of the
reasons I decided against it (I wanted SCSI).

Reviews on the Scantak seem to be few and far between ...

Doug.

	Bob pointed me at a couple of images scanned by his UMAX.
	I'm impressed. (certainly for $163!) I'll let him post
	the pointers if he chooses. Note that if he does, check
	them out with hi-colour resolution. At 256 colours, they
	were grainy. At 16m colours, they were very nice, albeit
	slightly soft.

	Anyone thing that the UMAX S-12 is worth the ~$200-$300 more?

							mike

        MacUser and MacWorld recently reviewed scanners.  One of
        them really liked the UMax S-6.  The other criticized it
        for not producing very good results with automatic scans
        and gave some Agfa(??) scanner a much higher rating.

        I believe the MicroTek E3 was panned by the one magazine
        that rated it, and its output used as an illustration of
        how poor scanner quality could be.

    >Anyone thing that the UMAX S-12 is worth the ~$200-$300 more?

    The UMAX S-12 is going for $500 these days. Is it worth the money - or
    any money? That depends...

    A bit of a  l o n g  story...

    I bought an S-12 about three weeks ago via mail order. I installed it
    on the external port of my trusty AHA2940 with UMAX's supplied
    pass-thru  terminator attached at the end of the external cable (as the
    scanner was the only external device) and the 2940 was happy to detect
    the scanner. 

    The UMAX supplied Twain software ("VistaScan"), however, does not
    provide an inf file and miniport driver, so whenever Win95 was booted,
    it would (a) detect an unknown device, then (b) figure out it was a
    UMAX scanner, but (c) when it asked for a driver, (d) I had nothing to
    offer it. After seeing this message too many times I eventually  told
    Win95 not to ask.

    If the scanner was installed with the included single-device ISA SCSI 
    card Win95 will never even know there's a scanner attached (so it won't
    complain about lacking a driver) as the UMAX ISA card doesn't use any
    IRQs or DMAs (their driver polls it). But I didn't have any spare ISA
    slots to live with that solution either.

    So far, no show stopper, just an irritation, as the TWAIN driver did 
    in fact see the scanner regardless of which host adapter it was connected 
    to. However, the show stoppers appeared rather quickly:

    - the scanner painted a bright blue "haze" down the entire left side of
    any scan taken from a glossy image. The "glossier", the brighter the
    blue. It didn't make any difference where the object was placed on the
    bed. Interestingly, the haze would fade such that the right two-thirds
    of the scans looked absolutely fabulous! If I only ever wanted to use
    the right two-thirds of every scan, I would have been quite pleased
    with the S-12...

    Looking back on this I'm sure that this was due to a defect in the
    scanner electronics (probably in some gain control or gamma control),
    but when I contacted UMAX "tech support" (being kind there) after a
    three hour wait in the queue (!!!) they were sure it wasn't a defective
    scanner and that my SCSI connection was at fault(!) So they first had
    me remove their terminator from the cable - the techie claimed the 2940
    didn't require an external terminator!

    Arguing with him was fruitless, so to make him happy (and keep him on
    the line) I pulled the terminator and rebooted. Of course the system
    croaked horribly, detecting double of everything installed on the
    internal SCSI bus (4 disks, two tapes, two CDs!) but not detecting the
    scanner on the external bus. While I re-installed the terminator and
    rebooted, he got educated by a supervisor: When he came back on the
    line he said "ya learn something new every day!" 

    (And right then I knew I was in deep doo-doo).

    When that didn't fix the blue haze problem, he had me re-connect the
    scanner via *their* ISA SCSI card, requiring me to remove an ISA IO card
    to make room for it. And when *that* didn't fix the problem, he told me
    "we thought we fixed this problem with the current firmware", and then
    "please put a sample scan on a diskette and mail it and the photo to
    us"...

    Yeah, right...

    - I also discovered that I couldn't operate the scanner using the
    supplied/supported Twain driver when I had a dial-up connection
    established. Didn't make any difference which host adapter I used, or
    what SCSI address the scanner was set for, if I tried to acquire a scan
    with a DUN session active, the system would lock up tight, requiring a
    pushbutton reset from the front panel. After another long wait in
    UMAX's queue, their answers were hardly amusing: first, "Well, then don't 
    use the scanner when you're using DUN", and then from a supervisor: "That's
    a known problem that we're working on a fix for"...

    (sigh)...

    On a tip from the usenet scanner newsgroup, I downloaded a different
    Twain driver from UMAX's web site (MagicScan 3.0) which while not
    *officially* supported for the S-12, does  include support for it in
    the software. Using this package I still got the blue haze, but at
    least I *could* operate the scanner with a DUN session active. Clearly
    the VistaScan package was touching IO addresses that did't belong to
    it...

    Completely PO'd with this unit and the incompetent UMAX support, and
    with no particular reason to believe that a replacement S-12 would
    fair any better with any of the problems I was having, I contacted the
    mail order company and got an RA to return the unit for credit towards
    an HP-4C. 

    Yeah, bit the bullet, took the hit, but HP's scanner works beautifully,
    provides both an inf file and miniport driver so it shows up under
    Device Manager, runs perfectly on my 2940, doesn't paint blue crap on
    my scans, uses an active terminator that is alive even if the scanner
    is turned off, comes with calibration software to match scan to screen
    and to printer, *and* came with a MUCH more capable set of utilities
    (including Adobe Photoshop and an awesome OCR package that just blows
    me away with its accuracy). 

    And I suspect if I *do* have any problems with this unit, HP's tech
    support won't consist of a bunch of uneducated morons that think
    terminators will cause gain control problems ;^)

    fwiw: Mega kudos to Rich and Kate at Midwest Computer Works for
    being entirely sympathetic and taking extra special care of me through 
    this mini-crisis. Not only did they take the S-12 back with no
    questions asked, they cross-shipped the 4C on two-day air without any
    extra charges, and it arrived before I'd even gotten the S-12 over to
    UPS for the trip back.

    /dave (a bit poorer - but a very happy HP-4C user)

    ps: UMAX's tech support is apparently famous (or iss that infamous?)
    for  their lack of competence. I pity anyone that has a warranty problem 
    having to call them to deal with it...
    

    btw/fwiw: I neglected to mention that I've been receiving a fairly
    steady stream of similar reports (about the problems I noted in .230)
    from usenet readers. Most are centered around blue or green haze
    problems that vary according to the object being scanned. 
    
    However, to be fair, it is just as clear that there are at least *some*
    S-12's that don't suffer from the haze problem. I have no idea what the
    percentage of problem machines are of the total, just that it's
    obviously not zero.
    
    Sadly, most of the people that have contacted me stuck it out (for
    varying reasons - some because they couldn't get through UMAX's support
    line queue for days on end, some because UMAX tech support scammed them
    with the "send in a copy of the scan" routine) until the merchants that
    they bought their scanners from refused to accept a return for credit. 
    
    So...the message I'd like to convey is: if you get a good/haze-free
    S-12 you'll love the scan quality. But if you get one that suffers from
    any noticeable hazing, SEND IT BACK! Don't mess around or you'll own
    it!
    
    /dave

    How do you scan an image out of a magazine, and get a "decent" replica.
    I called tech support and they stated that magazine images are scanned
    at a very low resolution, and a scanner does a very high resolution, so
    the images will be blochy. I do set the resolution to 72dpi, but the
    image is still very, very grainy!!!
    
    I am not an "image" manipulater by any means, I just like to through a
    photo on the flatbed, and hit scan. Photo's come out great without
    having to adjust anything.
    
    Does any one have any "quick tips" on getting good resolutions for a
    magazine picture??
    
    thanks,
    
    don grangruth 

Well, the first thing is to not automatically pick a high resolution.  For
the majority of scanning, anything over 150DPI is wasted. I have read that it
often helps to pick some odd resolution, such as 143, to avoid problems caused
by the scanning resolution being a multiple of the picture's resolution.

I haven't noticed a problem with magazine pictures with my scanner.

				Steve

    fwiw: Some TWAIN drivers have "de-screening" capability that will
    automagically do what Steve is referring to. You can specify the class
    of target display/output device and the driver will twiddle scan lines
    to avoid the moire patterns that cause that blotchy look...

    Sometimes skewing the page during the scan helps.  Place the page on
    the scanner at a slight angle and try it.  If you still get moire, then
    increase the angle slightly.  You can typically find an angle
    (different for any given page) that will result in less moire or
    perhaps more desirable moire.
    
    Using the de-skew feature of the imaging software will tend to soften
    the image somewhat, but this generally is more positive than negative.
    
    jeb
    

    Smoothing filters will also blend the separate dots together to end up
    as more consistent gray/color scale.
    
    /Bill

I have found that displaying an image at 256 colors that was scanned in
at 16bit color are displayed very grainy ...

I'd also vote for the low res.  Somehow, the preview image always looks
better than the final scan, indicating to me that it's possible, and
probably done with low DsPI.  Post-scan fixups with, say, Paintshop's
soften or blur functions might also help.

	Well, I ended up purchasing the UMAX Vista S-6E the other night.
	I set it up last night and I'm happy. For $163, it's a great
	bargain. One of my friends here in ZK has already gone out
	and purchased one. 

							mike

 You've got to love these notes files.

 Thanks to the information found here I returned
 my S-6e to Computer City and bought a new one at
 CompUSA saving $80 in the process. (If you remember,
 the one bought from Computer City was a return that
 I picked up at a discount).

 I owe someone a beer ...  Thanks!

 The Unit from CompUSA came with newer software. It 
replaced the MultiScan NT scanner software with 
a version of VistaScan that runs on NT. However, 
VistaScan dissapears out from under PhotoDeluxe fairly
regularly.

I'm off to the web site for a software upgrade ...


Doug. 

There wasn't a very big pile of S6E's left a CompUSA yesterday (like 5 or so).
I don't know if they have more in the stockroom/on order or not of course.

Burns

	When I bought mine Tuesday night, the woman behind the counter
	said I was the 3rd Deccie that day to buy one. They are moving
	fast! :)

							mike

    Is there a specific "de-screening" software package, or look for the 
    function in an existing package??
    
    thanks,
    
    don grangruth

    Some TWAIN drivers offer de-screening as a user-selectable function
    (ie: they put "De-Screening" in a menu selection somewhere).
    
    Other TWAIN also offer de-screening, but embed the function as part 
    of the output or target device selection menu. 
    
    (imho, the former case is more user-friendly, the latter I find
    user-antagonistic).
    
    UMAX's VistaScan is an example of the former. HP's DeskScan is an 
    example of the latter...
    
    /dave

	re: a few ago

	If you like, you can see images I scanned with the
	Spot Scantak scanner on my website.  URL is
	http://ducque.simplenet.com

	All of the images which come from negative film
	were scanned on this scanner. The few images which
	were made on Fuji Velvia (identified in the captions)
	were scanned on a high-res drum scanner.

	Just to add to the "fun": I just installed Corel 7.0,
	which has a very interesting scanning front end.
	This software can work in place of the manufacturer's
	driver for "twain" compliant scanners.

	The Corel driver tries to determine characteristics of
	the scanner before use.

	Bottom line: it refuses to work with the Scantak.
	So, twain .NE. twain, I guess.

Don Feinberg
	

I bought a UMAX S6E also.  Best DECDiscount I've ever gotten!

I'm not especially impressed (yet) with the Adobe PhotoDeluxe (which I assume
is an OEM version of PhotoShop or some such).  It does not seem to do a lot of
fairly obvious things like smoothing/interpolating.  But I have not used it
all that much.  I spent most of the weekend making space on my disk, so I did
not get to spend mucht time actually scanning!

What really surprised me though was how well the OCR works.  I don't have much
experience with OCR to compare things with, but this one had somewhere on the
order of one error on every couple lines of text.  Not only that, but it spell
checks the whole thing and shows you the failing words in a different color,
so it is not too hard to fix them.  And this was with the page slightly
skewed.  I tried exactly the same page on Xerox Textbridge (V2, I think and
not the "Pro" version) and the results from Textbridge were nearly worthless.

Be warned, though, that everything takes SKADS of disk space.  I installed
everything I could onto by D: drive, but it still ate up 8 or so megs on C:. 
This is especially a problem since C: is where the printer driver and/or
spooler buffers things you are going to print, and guess what!  Images take a
lot of buffering!

Also be aware that the S6E is quite large physically, and it comes with a
fairly short SCSI cable (probably about 1 meter), so you are somewhat
constrained as to where you can put it.

Burns

    
    They had 12 in stock when I was there last week. .242 Doug mine works
    fine. I even shut it off when I'm not using it and when I turn it on
    the software fines it no problem. 
    
    Scott

    
    Also you might try looking at www.scanshop.com, the photodeluxe is
    $49...
    
    scott

>    They had 12 in stock when I was there last week. .242 Doug mine works
>    fine. I even shut it off when I'm not using it and when I turn it on
>    the software fines it no problem. 

 Are you running NT?
 Are you running if off of an adaptec controller?


 I suspect that, under Windows 95, using the supplied controller, it would
 work as you describe.

 There were no S-6Es on the shelf at lunch today ....

 Doug.

    re: back a few
    
    "PhotoDeluxe" is not the stripped-down Photoshop. That's Photoshop LE,
    which actually provides quite a bit of the full blown functionality of
    its big brother but leaves off things like color calibration and a ton 
    of filters. I've use it, it's more than adequate for everything I do...
    
    On the other hand, PaintShop Pro (shareware - V4.1 for the 32-bit kit)
    is considerably faster than Photoshop and provides most everything
    needed for moderately intense photo-editing. Considering the cost of
    the full Photoshop (~$600 list, I think) to the cost of PaintShop Pro
    ($70),  it's hard to imagine what all that extra $$ gets you other than
    a whole lot of filters that will probably never be used...
    
    /dave

>    "PhotoDeluxe" is not the stripped-down Photoshop. That's Photoshop LE,

Aha, but as .250 implies, it is certainly a low-end product.

Thanks for the reference to PaintShop Pro.  If I find myself with not enough
functionality, I may give that one a try.

Burns

You're right - I was trying to convey that PhotoDeluxe is quite well 
below PhotoShop LE in capabilities. I tried it (briefly) when I 
temporarily owned an S-12. PaintShop Pro runs rings around it...

/dave

    .252:
    
    Another vote for PaintShop Pro.
    
    And, some (many?) PhotoShop filters work in PaintShop Pro as well.  It
    is, in my opinion, a totally copacetic product, and one that I happily
    paid the shareware fee for.

    re .248:  I finally had time to mess around a little more with my UMAX
    scanner and s/w tonight.  I am even MORE impressed with the OCR.  I
    scanned in a newspaper article---one of these front page under the fold
    things that is about 4 columns wide and 3 or 4 inches tall wrapping
    around a bigger font with a highlight sentence and a big headline over
    the top (it's the article in today's Nashua Telegraph about the guys
    with the telescope if anyone around here cares).  Anyway, the OCR got
    it close to perfect and took under a minute.  Not only did it get all
    the columns, different font sizes, italics, etc. but it also only
    missed about 2 words in the whole article (North -> Nortll and tele-
    scopes became tefe- scopes.  Oops...looks like it translated Larry as
    Lanny too.  Still darned good.  It must be optimized for fonts like
    that.
    
    Burns
    

    Hmm.  Just for the heckuvit, I tried it with Textbridge.  TB also got
    most of the words right.  It was terrible with the format, though.
    
    Strange...

    After 24 years in this business I consider myself a fairly jaded
    "computerphile". But I've run into something that actually has me
    excited and I just have to tell someone besides my SO ;^)
    
    I recently acquired the Corel 7 upgrade, and along with the
    unified/spruced up UI's used throughout this tool suite and the
    rewriting for 32-bit speed and function, I am especially pleased
    with the CorelSCAN application. I thought it would be one of those
    throw-aways that Corel has seen fit to toss in the box with previous
    Corel kits, but I couldn't have been more wrong. 
    
    CorelSCAN is a 32-bit TWAIN driver that completely bypasses the TWAIN
    driver supplied with the scanner (in this case an HP-4C, with their
    DeskScan-II v2.4 driver). And it is quite the marvelous little piece of
    software. I've only been using this for a few days but I'm already
    using it full-time in place of the HP driver. 
    
    In concert with the Corel Color Manager - which has an extensive
    selection of pre-configured templates (it had my ViewSonic 17, my Epson
    Stylus Color in both 360 and 720 modes, and of course the HP-4C) 
    CorelSCAN is producing scans with very accurate color matches, when
    either displayed or printed. And you can use it to acquire scans on
    behalf of any other TWAIN-compliant application (eg: PageMaker 6.5,
    PhotoShop, etc)...
    
    This software has full understanding of the HP-4C features and modes of
    operation, and it also supports a long list of other scanners. Plus it
    has a bundle of filters that can be used automatically  or manually
    tuned, can automagically deskew, and has fantastic moire filtering 
    (aka "descreening) that works a treat on magazine artwork! - no more 
    blotchy scans of magazine covers and photographs!
    
    The HP-4C comes with a capable-but-clunky TWAIN driver. Capable because
    there's adequate function that is "user-controlable" within, but 
    clunky because you have to go to some length to set it up and use it
    (obtaining the best scans is not an intuitive process). 
    
    In contrast, CorelSCAN provides even more functionality, and its scan 
    "wizard" puts all of the settings, functions and filters in a logically
    presented procedure.
    
    Obviously, spending ~$190 on a TWAIN driver - even one that I'd have to
    call the "Bee's Knees" isn't something I'd recommend as a "must" for
    everyone. But if you have some serious requirements and the interface
    provided with your scanner is less than stellar - and you can take
    advantage of those "other application"s provided with the CorelDRAW kit ;^)
    - it may well be worth the money for you.
    
    As a long-time/hard-core CorelDRAW and PhotoPaint user, it certainly
    was for me...
    
    This is just a "fwiw". Obviously I don't have any financial stake in
    Corel, but I sure like their software ;^)
    
    /dave
    

Well, shoot.

Just as I've finally become at peace with DeskScan.  (It's always amazing
how it can show me a preview that looks perfect, then scan something
entirely different.)  Now, it's time to go find something new...

I wonder if the product will be sold separately.  Better yet, I wonder what
the academic pricing on this suite will be.

    does the corel 7 recognize the Mustek 1200sp???
    
    just wondering,
    
    don grangruth

    I'm wondering if it recognizes the Canon IX-4025 scanner?
    
    Jason

The Corel Twain software does a "test" of your driver to see if it can 
control all the functions.  It doesn't like the Plustek scanners, so I just
use its normal interface.  You can still use the Corel scan wizard, which
is quite nice.

				Steve

    re: Directly supported scanners
    
    I wish I had noted the list of scanners that appear to be directly
    supported when I selected the HP-4C from the list (during the
    installation). 
    
    But I took a stroll through Corel's web site (www.corel.com) and found
    a list:
    
    Scanners Directly Supported by Corel 7.0:
    
    Epson ES Series
    Fujitsu ScanPartner 
    Fujitsu M309xx 
    HP ScanJet 
    Kodak PhotoCD
    Panasonic FX-R5308c
    UMAX UC630/1200s
    Microtek SM II and III
    Nikon COOLSCAN
    RICOH IS/FS2 & IBM 2456
    
    Don't know if this list is current or not.
    
    Otherwise, as Steve noted, you can have CorelSCAN use your scanner's
    native interface. I don't know how that actually works as I haven't
    tried it here, but it appears from Steve's note that it does work
    well...
    
    /dave

    Yes, Corel ScanWizard will use your scanner's native interface - you
    just lose the additional tweaks that Corel's interface could provide.
    
    				Steve

    
    I used the OCR software last nite and scanned in the NNYBA Handbook.
    I only got 4 errors at most per page! I found 400dpi the best for
    the handbook I was scanning. lower rates I got more errors. I'd scan 
    the page, drop it on Word and run the spell checker. Bingo I had the 
    handbook in the computer in no time. So much better than a hand
    scanner. 
    
    scott 

	I was in CompUSA the other night. The UMAX S-6E scanners that
	many of us got for $163 (digital price) are now out on the
	floor instead of behind the wall. I know of at least 5 DECcies
	that purchased them. I'm sure there is alot more.

	How much is the CorelScan stuff?

							mike

I think the CorelScan software is part of CorelDRAW only - I don't know
that they sell it separately.

				Steve

    What's the best hand scanner anyone's seen? Best meaning most
    reliable at reproducing the text being scanned.
    the reason I ask, is that both my retired parents do genealogy
    research, have recently purchased a laptop, and have asked me
    about scanners and what to buy, or stay away from. Their questions 
    at first didn't put into context the need for something to be
    portable, so naturally I suggested a bed scanner. Now some bed
    scanners are fairy large, and don't lend themselves to portability.
    Once I realized that they were looking for portability, I suggested
    hand scanners, but didn't have any working knowledge pro/con on
    them. I also suggested that digital cameras are an option depending
    on the level of macro or closeup capability they can find.
    For small pieces of information found in books, simply transcribing
    the information is of course the default option.
    
    Bob

    I just called our local CompUSA and was told the Digital Corporate
    price for the S-6E is $169.41 -- didn't know where $163 came from.
    
    FWIW, I was also told that there will be an advertisement starting 
    Sunday (April 6) for a $70 instant rebate and $30 manufacturer 
    (mail-in) rebate.  Can't qualify for the $70 instant rebate if
    purchased through the corporate discount program, but can send in 
    for $30 manufacturer rebate.
    
    At $140 (+tax), I may just get one.  I don't have a SCSI adapter,
    however.  Any recommendations for an inexpensive one of these? :-)
    
    -Andy

RE: .269

	It comes with an ISA SCSI adaptor. 

	Too bad I missed the mfg's rebate! But I still got a deal.

RE: .268

	Logitech has a large hand scanner that scans full pages. 
	I wouldn't get one of the 4" hand scanners. Stitching
	pages together sucks. The Logitech hung off a parallel port.

							mike

If it's a manufacturer's discount, chances are it's not tied to CompUSA's
2-week sale cycle.  Maybe it's already on.  Look for a rebate coupon next
time you visit the store and see if you qualify.

    
    the discount is ONLY good if you bought the scanner (S-6e) this week.
    It's right on the coupon.... dam... Everytime you think DEC gets a good
    deal.. your screwed...
    
    scott

There are UMAX $20 rebate coupons at ComputerCity. Good for purchases
between Jan 1 and March 31, but you need to mail it by
April 15.


Doug.

RE: Corel software.  As a Digital employee you can order CorelDraw 7 for
$159 (inc S&H).  If anyone is interested, I have posted a copy of the
EPP form outside my office (ZK3-4/X57.)  You can also get the form off
of Corel web page.  Look for EPP.

The UMAX S-6E scanner is on sail with rebates for $149.  One heck of a
deal.  Works EXTREMELY well.  It does with with Corel 5 using supplied
Vista TWAIN interface.  I have had problems with hangs every blue moon,
but still am pretty happy with it.

	- mark

    Interesting turn of events...
    
    Bought the S6E scanner, and a 6x 4-disc CD-ROM drive (NEC) at CompUSA's
    corporate desk.  Prices were $169.99 and $119.99 respectively.  Retail
    (sale) price on each were $179.99 and $129.99 respectively before
    rebates ($30 rebate on scanner; $20 rebate on cdrom).  
    
    When checking out, I was told I would be unable to use the manufacturer's 
    mail-in rebate available for either product because I was purchasing at
    a corporate discount.  I went ahead and purchased at the corporate rate
    anyway as I had been told differently Friday, and then called another
    store (the store I spoke with Friday) to verify this information.
    
    Turns out that the rebates are only good on retail (sale) purchases, 
    not corporate purchases.  So, I retract my earlier note stating the S6E
    could be had for $140.  Still, $150 ain't bad.
    
    -Andy (who's gotta go back now and do some transaction changes...)

>The UMAX S-6E scanner is on sail with rebates for $149.  One heck of a

I guess they must be stuck with a boatload of them!  <yuk yuk>

Burns

    Naw.... they're trying to unload excess inventory, so they can
    make room for a new & improved UMAX model for $149.00 
    The S-6E will become obsolete in the blink of an eye. :-)
    
    Bob

I heard UMAX was helping CompUSA celebrate the opening of new stores 
or some such.

UMAX has been giving a $20 rebate for 4 months now so $30 doesn't seem
like a big corporate change. ComputerCity is still selling these at
$249. They're still on the net at $249.

I won't worry about a replacement until I see these in the online
actions  ....

    Office Depot has some page scanners, not sure the brand, but appear to
    be the standard 300dpi optical resolution 1200dpi interp, small desktop
    unit marketed under many names. In store price is $119. 24-bit color.
    W95 compatible.
    
    Mark

>>The UMAX S-6E scanner is on sail with rebates for $149.  One heck of a
>
>I guess they must be stuck with a boatload of them!  <yuk yuk>


Sail.  Boat.  Sheesh...and you guys thought I was seriously talking about why
they were on sale?  Where is Paul when we need him?

Burns

I assume the ISA adapter for the S-6E has selectable interrupts?
What are interrupts can be used?

Also - can I attached other SCSI devices to the SCSI bus?

Gim

	I don't think the device uses interrupts. It's all driver
	dependant. Probably just uses some I/O.

	The connector off the back of the card is a 25-pin. The
	cable is 25-pin to Centronics. I doubt you could hang
	anything else off this. If you want to "hang" devices off
	of a SCSI adaptor, get a real on. (Or use the slow Trantor
	SCSI adaptor on the PAS-16 sound card, if you have one)

							mike

 Or is the scanner really SCSI?

RE: .284

	A friend is hanging his off an NCR 53C810 SCSI adaptor. I 
	suspect it'll run fine off an Adaptec. It has a Centronics
	connector on the back with an ID selector to select the
	SCSI device name. The only reason I haven't hung it off
	my 53C810 adaptor is a lack of a cable at the moment.

							mike

><<< Note 557.281 by skylab.zko.dec.com::FISHER "Gravity: Not just a good idea.  It's the law!" >>>
>                            -< It's a joke, son... >-
>
>>>The UMAX S-6E scanner is on sail with rebates for $149.  One heck of a
>>
>>I guess they must be stuck with a boatload of them!  <yuk yuk>
>
>
>Sail.  Boat.  Sheesh...and you guys thought I was seriously talking about why
>they were on sale?  Where is Paul when we need him?
>
>Burns

    
    Hey, if you put in notes like this, I don't need to. Though I do
    notice that it takes a Fisher to see the connection with boats.

> Or is the scanner really SCSI?

 Yes it is. 

The controller supplied with the scanner is specific to the scanner only and
has its own proprietary driver.

If you hook the scanner up to a real scsi controller, such as I have mine hooked
up to an Adaptec 1540, a different driver is installed.

I suspect that if you use the controller they supply, you can turn you scanner
on and off without a problem. If you run the scanner off of a real scsi 
controller, then the scanner must be turned on before the controller 
initializes (boottime) if you want to use the scanner. Else the driver fails
to load for lack of seeing the controller.

Doug. 

    >Hey, if you put in notes like this, I don't need to. Though I do
    >notice that it takes a Fisher to see the connection with boats.
         
    Yeah, and it takes a "Fisher-man" to get hooked by his own bait. :-)
    
    
    Bob

Aaaaagh.  You win!

Burns

    re .276, UMAX S6E
    
    I was not told this at checkout - this is interesting.  I was, however
    given a compusa substitution rebate form, as it seems that they did not
    have the originals available - does anyone have one or a copy? (at
    least the address of the org handling the rebate?)  Even if this is the
    case, I am going to try to send in the rebate anyway (does it say on
    the rebate form that corporate prices are not eligible for the rebate?)
    
    
    Regarding rebates not being honored on corporate discounted items:
    
    I don't see how this could be an issue with the scanner manufacturer,
    compusa paid the same price for each scanner regardless of who buys it
    or for how much.   BTW - I paid 163 and change, plus MA tax...
    
    -Ken

>    Naw.... they're trying to unload excess inventory, so they can
>    make room for a new & improved UMAX model for $149.00 
>    The S-6E will become obsolete in the blink of an eye. :-)

Looking at Umax's web page I don't see any mention of the S-6E on their
scanner page, but the page for the Astra 600S says it is the "The award
winning S6e upgrade at a new price point!"

Jeff

    Computer City is advertising the Astra 600S at $149.
    
    			Steve

    -.290 the rebate says you have to send the UPC from the box and the
    sales slip. ALSO you have to buy it between 4/6-4/12.. Got me there...
    
    I'll just take back the S-6e and go to Computer City and get there's...
    ;v)
    
    I'm under 30 days...
    
    scott

Hmmm ...

Didn't see that on their web site before ...

the 600s is listed at 8.5" x 14"  (extra 3 inches)

and a 25pin to 25pin scsi connection ...

    I called two local Computer City stores here in the Atlanta area and
    was befuddled to find that none of these stores carry the Umax Astra
    600s, at least not yet.
    
    I then called the Computer City on Daniel Webster Hwy in Nashua (found
    their number through www.switchboard.com) and they, too, said they do
    not carry it.
    
    Steve, can you tell me where you got your info?
    
    -Andy

    
    I got the Computer City flyer in the mail yesterday and it shows the
    Astra 600. I don't remember it being $149. I'll look tonite.

It was in the flyer I got.  I might be misremembering the price.

				Steve

    The Comp C. flyer I just looked at had the Astra 300 shown (not the 600, 
    at least I did not see it), listed for $199. This is the parallel port 
    version. 
    
    The Nashua store doesn't have the 600 nor does it show up in the Comp C. 
    computerized database.
    
    Antony.

That may have been it.  Sorry for the confusion - I didn't have the flyer
in front of me.

				Steve

Before to many folks feel like they got snookered ....

	The Astra 600 is a 30bit 8.5" x 14" scanner
	the Astra 300 is a 24bit 8.5" x 14" scanner

The cheapest price I found on the net for these scanners

	Astra 600 w/PhotoDeluxe,OCR,drivers     $262
	Astra 300 w/PhotoDeluxe,OCR,Drivers     $222


   Doug.

CompUSA is willing to help those people who bought their scanner BEFORE
the dates for the rebate.  

Bring in your receipt and explain that you need a new receipt for the
required date range.  They will get you a new receipt, which is what
they end up thinking is a return and then you bought it again!

This saves them from having people bring the machine back in and buying
a new one, just to get the refund!

They just did it for me, so I get the $30!

They did play the game that the scanner was no longer $163 for DEC
employees and said it was now $179 - $30 = $149.  In my case I paid $163
- $30 = $133 for the scanner.  I plan on bringing this up to them for a
friend who is buying one.  The deal that DEC has is 12% over and I have
the receipt to prove the price.  Of course they may say the price went
up...

	They were VERY good about it and no hassle.
	- mark

http://www.dpi-scanner-authority.com/144int/umax/vists6e.html

>	They were VERY good about it and no hassle.

	Based on personal, notes, and usenet experiences, the hassle factor
	at CompUSA varies by store, manager, phase of the moon, etc. It might
	be of help if you post which store this happened at and the name of
	the manager if one approved this transaction.

Larry

RE: .303

	Mark is in ZKO, so I suspect it was the Nashua store.

>	Mark is in ZKO, so I suspect it was the Nashua store.

	I'm in ZKO too, but since I live in Belmont, MA, I frequently find
	myself in the Woburn and Brighton stores. CompUSA is notorious for
	advertising products without adequate store stock, so store hopping
	is required if you want it NOW.

Larry

    
    I just picked up a Mustek Plug and Scan 600 II SP at Office Depot.  It
    is a parallel port 300 DPI flatbed color scanner($199).  It works great
    except for one thing that is driving me crazy.  Every time that I
    reboot my machine, I have to re-install the scanner software to get the
    scanner and the printer to work again.  They will both then work fine
    until the next reboot when I have to repeat the ordeal.  I have tried
    to reach Mustek tech support but gave up after waiting 30 minutes on
    hold.  Anybody have any ideas?
    
    John
    

re   <<< Note 557.306 by CSC32::J_MANNING >>>
                -< mustek scanner >-

do you have it turned on before you boot?

what is the interface?
what op sys?

What is the big difference between the scanners for under 200 and the 
scanners for 400+?  

When I first thought about buying a scanner I figured I would have to spend 
in the $500 range (ie HP flatbed versions).

I know, as with all things PC'ish, the question of "what are you going to use 
it for".  My answer would be "scanning 4x6 glossy photos" "right now".  I'd 
like to be able to edit the scanned results (mostly cropping and maybe some
adding of text) and end up with a printable reasonable picture at the end.

Brian J.

    re .307
    
    I found the problem.  It was the parallel port driver for Seagate
    Backup.  I am using an internal tape drive so I just renamed the
    driver. 
    
    John
    

The major differences are the quality and features of the software and
color accuracy - though the latter can be mitigated in software.  The price
of the low-end scanners has fallen through the floor in the past few months.

					Steve

>>The major differences are the quality and features of the software and
>>color accuracy - though the latter can be mitigated in software.  The price

The software they bundle with it (like half the products out there include
Netscape on them!) or the drivers that control it?  

Are there scanner packages that don't include the SCSI interface cards (I've
already got a SCSI card)?  It might not be worth them packaging it up that
way (limited consumer base).

bjm

RE: .308

	The under $200 scanner would do you fine. Spend the additional
	~$200 on fancy software like Corel.

	One thing to ask yourself, do I have a system with enough memory
	and resources to do mega-dpi scanning? Even a 4x6 photo at
	1200dpi is going to take ALOT of memory for very little visible
	payback. 

	Myself, for the needs I have for a scanner, the $163 Umax will
	suit me fine. 

							mike

I mean the scanner control and related software.  The more expensive models
allow you to fine-tune the scanning more, include more sophisticated graphics
and OCR programs, etc.  But I have a cheapo (well, it was $260 last year and
is now $149) parallel-port Plustek OpticPro 4800 and it works fine for me.

				Steve

    Latest on the S6E for those still wanting one: CompUSA told me
    definitively yesterday that UMAX will NOT be shipping them any more of
    this product. Instead, they will be receiving the 300 mentioned earlier
    in this string (parallel port only) from UMAX.
    
    Antony.

Caution:  If you are scanning color photos, you might want to get the next step
up in scanners from rock bottom, one that has 30 bits rather than 24 bits of
color.  You might thing 2**24 would be plenty of colors, and it would be if they
were distributed in the right places, but they are not.  You tend to see
"iso-color" lines on things that should be smooth gradations (like faces).

Burns

    re: .315 by skylab.zko.dec.com::FISHER
    
    >> You might thing 2**24 would be plenty of colors, and it would be if
    >> they were distributed in the right places, but they are not.  You tend
    >> to see "iso-color" lines on things that should be smooth gradations
    >> (like faces).
    
    Burns, that doesn't sound right.  Are you sure you are not seeing a
    limitation in your own video environment?
    
    /Bill

24 bits is plenty for all but graphic professionals creating images for
magazines or other high-resolution print media.  Consider that in the video
world, 24-bit is considered "true color" and is all that inkjet printers are
capable of.

				Steve

Well .315 was the kind of info I was looking for but there appears to be a 
few questions to its accuracy :-)

re: lots of memory

The system is a P133 sitting on a MB with 512 cache and 32M (70ns).  Gee just
a month ago I wouldn't have know what the hell that ment!

bjm

    
    I have a P133 with 512K PB cache and 32 MB 60 NS EDO memory running
    Windows 95 and I can tell you that for high DPI scanning/editing, it is
    SLOW!
    
    John
    

    It sounds like most of you components are okay, but I'd replace or
    upgrade that last one, you know, Windows 95.
    
    jeb :-)
    

re: W95 

Ya I know - I was debating WNT but opted to hold off until I got this up 
and running with W95.

RE: 32M

It was hinted that anything more than 24M on W95 doesn't buy you much, is this
true or false?

bjm (shutting down for the weekend)

    re: .321 by hndymn.zko.dec.com::MCCARTHY
    
    >> wife said WNT - I said not yet...
    
    Now we know who is the brains at home.  ;-)
    
    >> RE: 32M
    
    >> It was hinted that anything more than 24M on W95 doesn't buy you much,
    >> is this true or false?
    
    IMO, 32MB Win95 is still better than 24MB Win95.  64MB NT is even
    better.  ;-)  Or 128MB...  ;-)
    
    Have a good weekend, everybody!  Cheers,
    
    /Bill
    

I have no practical experience with 24-bit vs. 30-bit or "true color"...

But considering that 24-bit only gives you 8 bits of intensity (i.e.,
8-bit greyscale, only 8 bits per color, etc.), I can see where it might be
a trifle limiting, compared to 30-bits.

Even in color, given a fairly uniform color where the intensity varies
slightly, like sky blue or flesh tones, you really only get effectively
some 8 bits of shading in that color.

Being able to print all that color variation is one thing, but I think
better resolution is needed at the input device, because one typically
tweaks the colors, which throws a bunch of your resolution away.

re amount of memory for scanning:

as mentioned, it can take a lot.
a friend has 100MB on a Mac that he used mostly for image processing.
my daughter, a grad student, found it worthwhile to go from 24 to 40MB
for her grayscale image work, even spending her own money.

ask yourself three questions:
what fraction of your work will be image processing?
how impatient are you?  
would you rather have the speed or a different new toy?

of course, people will have different opinions; 
they have different preferences.

RE: Note 557.304 by AXEL::FOLEY

.	Mark is in ZKO, so I suspect it was the Nashua store.

Yep, this is the ZKO store.

Computer City is willing to match the $179 price, but not the
additional manufacturer reabte of $30.

	- mark

RE: Note 557.315 by skylab.zko.dec.com::FISHER

.Caution:  If you are scanning color photos, you might want to get the next step
.up in scanners from rock bottom, one that has 30 bits rather than 24 bits of
.color.  You might thing 2**24 would be plenty of colors, and it would be if they
.were distributed in the right places, but they are not.  You tend to see
."iso-color" lines on things that should be smooth gradations (like faces).


I have had no problems...  I have scanned in pictures and they look great.
I would suggest you are getting a limitation of video.  Do you have
real color turned on, 256 colors, or 65535 colors?  It is important to have
1mil or more colors to see an accurate picture.

	- mark

    
    RE: the 30 bit vs. 24 bit scanner debate....
    
    For what it's worth;
    
    Most software can't process all 30 bits (and sometimes 36 bits) that
    some scanners can represent.
    
    Most printing presses have a far more limited range than a color
    monitor or a 24 bit scanner.
    
    Most people can't see all the colors that can be represented by 24
    bits. Most people cannot see all 255 levels of gray that can be
    represented by 8 bits (and Postscript level 2 now gives us 4095 shades
    of gray (and white of course))
    
    Most scanner optics in the home market (under $3000 today) can't
    reliably reproduce all the colors that they must handle. Sure there are
    2**24 different values available, but the optics can't really see that
    many different colors (just like what I said about most humans not
    being able to see millions of colors.) adding 2 extra bits per color is
    cheaply done. Adding optics that can actually use those two extra bits
    is not cheap. In other words, if the optics are the same the fact that
    they added two bits more to the hardware buys you nothing. Even if the
    optics are different, in the <$3000 market, I bet the new optics were
    cheaper to the manufacturer and they make little difference to the
    number of discreet colors the scanner can see.
    
    To the person getting banding in flesh tones from a 24 bit scanner; Is
    your monitor is displaying 8 bits? Is your software displaying 8 bits?
    Are you saving the scans as GIFs? There is no reason to get banding
    with 24 bits *unless* the scanner you are using is not coming anywhere
    near seeing all 16 million colors that can be represented by 24 bits.
    In that case, even though it uses 24 bits its representing far fewer
    colors. I've seen a lot of work done on 24 bit scanners and reproduced
    in magazines and 4 color brochures, with great results, and banding in
    the flesh tones (or any other area) is not an inherent flaw of 24 bit
    color. You may have a hardware problem with the scanner.
    
    Personally, I subscribe to the theory that 24 bits is enough to
    represent all the colors that the human eye can distinguish and then
    some...I just don't believe that $150 is enough to represent all those
    colors (although it should be enough not to give you banding.)
    
    Mario

    agreed Mario.
    
    Another thing (basically Mario said it anyway) is 'you get what you pay
    for' - you buy a cheap scanner, you get "cheap" output via cheap
    software. The good news is that by throwing some decent software (I
    think I saw 'Corel' mentioned?) can go some way to reversing the
    'cheapness' in the hardware - note 'some way', not all the way.
    
    My first scanner was a (relatively) cheap sheet feed scanner. But even
    now, it still does a pretty good job. I mainly stay with 300dpi maximum
    - it's rare that I need more than that, unless I'm doing some sort of
    'enlargement' - and 95% of the time, I'm scaling back to 256 colours
    for web graphics anyway :') (at 100dpi or so)
    
    For professional graphics (which I'm not into yet - but you never
    know), flat bed is a better way to go, and (at least in Oz) be prepared
    to pay nearly $1000 for that professional touch - although you can get
    pretty good flat-beds for half that here.
    
    Scanners owned by friends are becoming a real pain in the backside...
    "Oh Harry, you own a scanner? can you help me get mine going?" 2 hours
    later... ;') I guess it's one way to gain experience in a wide range of
    hardware and software :')
    
    H

Ya the wife's got the brains.  Actually, if my son's "learning games" stopped
working that would be the death of things too!  

Some day I might try to get the thing setup to duel boot - but I'm in no
rush. 

re: memory
Ya more is always better.  When I was first stocking this machine I figured
2 16M sticks was enough for now.  The MB will handle up to 512 so maybe 
next time I'm out I'll buy 2 32M sticks - ya right... I am sticking with the 
standard memory (no EDO) for now.

re: performance
I guess next time I borrow the system here to scan in a picture I'll see what
its got under the hood.  I have not put a stop watch on it because its a three
pass scanner and I thought it would be unfair.

re: 24 vs 30 vs 36
One of the notes mentioned "enlarging".  Since I used to do B&W photograpy and
loved the ability to crop/enlarge etc in the darkroom this is one thing I 
would like to be able to do with the scanned image.

I assume (there I go again) the higher the DPI value the "slower" the film 
(ie I will be able to enlarge more with a higher DPI scanned image without 
starting to "see the grain").

Will having an image scanned in on a 30bit scanner make a difference when it
comes to enlargments?  One would think Yes, but those marketing people get
everyone confused!

thanks

Brian J.

I run an AMD p100 with 48mg of 60ns memory.
It is a dual boot system with W95 and WNT.
Attached is a UMAX S6E.

24 vs 30 vs 36 bit color:

	All 30 and 36 bit color allows you to do is generate much larger
	graphics files. 24 bit color, at least from the UMAX, is very accurate
	and blends well on a 28dpi 17" monitor.

W95 vs WNT:

	Performance is better on W95 by a wide margin, at least for 
	PhotoDeluxe V1 and PhotoShop V4. I don't know about Corel.


CPU vs Memory:

	If you want speed, by LOTS of memory, or scan at lesser resolutions.
	My vote would be for 128mb of memory for color 8x11 portraits. Else the
	system spends it time in virtual (disk bound) memory.

Doug.

re .323:

>But considering that 24-bit only gives you 8 bits of intensity (i.e.,
>8-bit greyscale, only 8 bits per color, etc.), I can see where it might be
>a trifle limiting, compared to 30-bits.
>
>Even in color, given a fairly uniform color where the intensity varies
>slightly, like sky blue or flesh tones, you really only get effectively
>some 8 bits of shading in that color.

This was essentially my point way back.  Millions and millions is a lot of
colors, but it depends how they are distributed.  With 24 bits you can get 256
shades of any of the primary colors, or or white, magenta, or cyan.  Any other
color (like the unsaturated blue of the sky or the pinkish-brown of caucasian
skin) you get many fewer shades.

Having said that, even though it is true, it might be that it is not as visible
as I thought and that the banding I am seeing is due to some other issue.  YMMV.
I don't think the physical facts I describe above are very controversial, though.

Burns

    
    Re: .331
    
    First, I think you're confusing input and output.
    
    My point was that even if you purchase a very high quality drum scanner
    and scan at 48 bits (16 bits per color (RGB)) you still down sample in
    your software to 24 bits. There are very few software products capable
    of handling more than 24 bits. Most hardware is not up to the task of
    manipulating 30, 36, or 48 bit color images. Even if you had the
    software and hardware capable of doing all these things you still could
    not see the results and when you went to print you could not reproduce
    that many colors. (I'm not real knowledgable on monitors, but I rather
    doubt that there are any monitors that can accuratly repoduce 30 bit or
    greater color. I know that you can add bits forever, but I doubt that
    you could see or display the difference between FFFFFFFF and FFFFFFFE.)
    
    The input may be 30 or more bits, the outpu is 24 or fewer bits.
    
    Second I think you're confusing color/darkness with dynamic range
    
    In your example you say there are very small changes in the color of a
    blue sky. I agree with you there. There are no more that 256 different
    colors of blue that your eye can distinguish though so 8 bits is
    plenty. It is not very likely either that the sky you are looking at is
    only blue (especially if you're looking at the real sky and not the one
    on your monitor that you just scanned into your system.) So if there
    are other colors mixed with the blue, you get the 256 shades of blue
    times 256 shades of the other color. It's really no where near that
    simple though.
    
    What I was trying to point out is that banding is not due to the fact
    that you have a 24 bit scanner. 24 Bits is plenty for color work. If
    your banding is caused by the scanner then there is a problem in the
    scanner, the scanner is either broken or the optics are not a high
    enough quality to accurately read all 256 levels in each color channel. 
    
    Most scanners do not have optics that can accurately reproduce all 256
    levels in one 8 bit channel...how many or how few your scanner
    reproduces accurately will determine how much or how little color
    banding you will get.
    
    The ability to distinguish all 256 levels in a given channel is the
    scanner's dynamic range. In a 256 color model intensity changes by .39%
    per level. That means that the first level of black that a scanner
    should be able to see is .39% gray. That's not very much gray. Most
    scanners cannot pick this up and instead they interpret it as white.
    The next level is .78% gray...with most low end scanners this again
    becomes white. The next level is 1.17% gray - at this level you start
    to get more scanners that can actually read the gray. If accurately
    read this level is the first one the scanner sees. So, your scanner has
    read 4 shades of gray (I'm including white or 0%) and output the
    following bits;
    
    0%     = 0000 0000 -
    .39%   = 0000 0000  > Many levels represented by the same number
    .78%   = 0000 0000 - 
    1.17%  = 0000 0011
    
    In a 1024 level color model (or 30 bit color (10 bits per channel)) you
    have a rate of change of .09% per level of gray that can be
    represented. In this case if the optics are the same quality, here is
    what happens to the numbers;
    
    0%     = 00 0000 0000 -
    .09%   = 00 0000 0000  |
    .18%   = 00 0000 0000  |
    .27%   = 00 0000 0000  |
    .36%   = 00 0000 0000  |
    .45%   = 00 0000 0000   > Many more levels represented by the same number
    .54%   = 00 0000 0000  |
    .63%   = 00 0000 0000  |
    .72%   = 00 0000 0000  |
    .81%   = 00 0000 0000  |
    .99%   = 00 0000 0000  |
    1.08%  = 00 0000 0000 -
    1.17%  = 00 0000 1100 <-- More bits represent the same level.
    			      You have more numbers but you can't see
    			      the levels to accurately represent them
    
    Can you see how having 24, or 30, or 36 bits makes no difference if the
    optics are not capable of reading all 256 (or 1024) levels of any color
    channel? Can you see how much more accurate the optics must be to take
    advantage of the additional 2 bits per channel? This is what makes
    inexpensive scanners reproduce less faithfully - not the fact that they
    represent color in fewer bits, but the fact that they cannot accurately
    distinguish all the levels that they can represent with those bits. The
    rest is all marketing. I've seen way too many scanner manufacturers add
    2 bits per channel to last years model and not change the optics and
    claim they now give 30 bit color. The truth is they use more bits to
    represent the same limited number of colors and they would have done
    the buyers a lot more good adding some better optics but there's no
    industry buzz word they could use in advertising for that.
    
    A scanner with 36 bits that has optics that deliver poor performance
    will not produce images that are any better than a 24 bit scanner. It's
    not the number of bits per pixel that's the issue. I assure you that 24
    bit color is plenty enough to reproduce a blue sky with all its subtle
    color changes. The issue is how much of that 24 bits is really being
    used. Most inexpensive scanners cannot use the full range of light to
    dark in each color. That's why banding occurs. It is not an inherent
    flaw of the 24 bit color model.
    
    Here's a simple test, take a look at a good scanning target, one that
    has both color and grayscale bars. Look at the gray bar, in most cases
    it is broken down into 5% increments from white to black. In some cases
    it is broken into 1% incremnents. Even in the 5% case though it is hard
    to see the difference with your eye between 95% and 100% black. In the
    1% targets, most of the top end looks black. The scanner sees basically
    the same thing. Adding bits will not help the scanner distinguish more
    levels, adding better optics will. 
    
    I will say though that once you add better optics the values I wrote in
    above change drastically. There are more levels for the scanner sample
    and this increases the scanners dynamic range. You may see the
    following happen and you read a few extra levels;
    
    0%     = 00 0000 0000 
    .09%   = 00 0000 0000  
    .18%   = 00 0000 0000  
    .27%   = 00 0000 0000  
    .36%   = 00 0000 0100
    .45%   = 00 0000 0100
    .54%   = 00 0000 0100
    .63%   = 00 0000 0100  
    .72%   = 00 0000 1000  
    .81%   = 00 0000 1000  
    .99%   = 00 0000 1000  
    1.08%  = 00 0000 1000
    1.17%  = 00 0000 1100
    
    You can see that with better optics you can represent more levels 
    of change...but before you say "That's what I tried to tell the fool"
    here's what the same optics would output on a 24 bit scanner;
    
    0%     = 0000 0000
    .39%   = 0000 0001
    .78%   = 0000 0010
    1.17%  = 0000 0011
    
    Each scanner represents 4 unique levels. There is an advantage to the
    30 bit scanner...it may interpret a change sooner than the 24 bit
    scanner. This advantage can lead to the 30 bit scanner having a
    slightly greater dynamic range. But the requirement that the optics are
    better is still there if you expect to get better scans. Just adding
    bits will not give poor optics much greater dynamic range. Dynamic
    range is only slightly, if at all, increased by number of bits used to
    represent each channel.
    
    Mario

I don't think we need to continue this too much since I basically don't disagree
with you except on a couple points.

>    First, I think you're confusing input and output.

No, I think I know which is which.  As I said, I can't (without further
experiment) prove whether the banding I see is from the scanner or from the
monitor.

Another point:  I don't disagree with you that less expensive scanners probably
don't have the intensity resolution to distinguish all 8 bits per channel.  I
was, however, assuming that a 30-bit scanner would be able to distinguish more,
if not the full 10 bits per channel.  If not, then certainly 30-bits is just
hype.  (But they wouldn't do that, would they?  :-)

>    In your example you say there are very small changes in the color of a
>    blue sky. I agree with you there. There are no more that 256 different
>    colors of blue that your eye can distinguish though so 8 bits is
>    plenty. It is not very likely either that the sky you are looking at is
>    only blue (especially if you're looking at the real sky and not the one
>    on your monitor that you just scanned into your system.) So if there
>    are other colors mixed with the blue, you get the 256 shades of blue
>    times 256 shades of the other color. It's really no where near that
>    simple though.

Either I disagree with you here, or the non-simplicities are what I am talking
about.  Let's say we have a color which is represented by blue=255 + green=63 (a
sort of very blue aqua).  My point is that in order to maintain the same hue
(formally, as in the HLS model) but reduce the lightness you must reduce the
blue and green so the proportions remain the same.  You can't say there are 256
shades of this hue, because if you change blue without changing green you don't
have the same hue.  In this particular case, the next lower intensity is going
to be green=62, blue=251, then green=61, blue=247.  So there are only 64 shades
of this particular hue.  If you added a bit of red (and made the color less
saturated) you would have even fewer intensities of the same hue/saturation.

This is all theoretical, of course.  I agree that in real life:

1) The limitations of the output device(s) you are going to use the results of a
scan on may be more significant than the limitations of a 24-bit scanner.

2) A 24 (or 30 or 36)-bit scanner may really have fewer than that many bits of
actual distinguishable colors.

Burns

I'll share some of my experiences, just to provide a bit more empirical
data.

I have a Microtek Scanmaker IIHR connected to a DEC ST450 with 32mb of
ram (and 100mb of swap space). I have a 17" monitor running at 1280x1024
-- AND 16 COLORS (very important factor). With only 1mb of video ram I
can't run true color at that resolution. But, I want the extra screen
real estate that the 1280x1024 gives me. 

So. When I scan in a color photo it looks like crap on the monitor
because of the 16 color limitation. If I print that lousy-looking image
to my HP870 at 600dpi AND use the expensive high-gloss paper it looks
perfect on paper. I've defied people to distinguish the results from a
photo shop print, and they can't. So the lousy appearance has nothing to
do with the actual image with which you're working. Too many people don't
recognize this effect for what it is.

If you're working with an important image, do your scanning, do your
processing with PhotoShop or Picture Publisher or whatever, then write
the file to a zip drive and take it to a full service photo lab and have
them either give you a neg or a slide or a print... and you'll be
impressed.

And, by the way, if I scan in a 5x7 or 8x10 color photo at 600dpi, each
step that I do goes something like this: pick an image effect - say,
gamma correction or stretch detail or remove pattern, whatever - click on
"Apply" and go get a cup of tea. The other night I went and took a bath
while the system was processing one color image at 600dpi. We're talking
about 300mb+ image files here. (My drives have 11ms average access time.)

I've had occasions when I've scanned in just 4 square inches of an image at
1200dpi, put on my coat and drove into the village for the Sunday papers,
came back and it was still working.

My next system will have four to six mb of VRAM and 64 to 128mb of ram.

Art

    >Either I disagree with you here, or the non-simplicities are what I am
    >talking about.  
    
    I think it's the non simplicities that we are differing over.
    
    >Let's say we have a color which is represented by blue=255 + green=63
    >(a sort of very blue aqua).  My point is that in order to maintain the
    >same hue (formally, as in the HLS model) but reduce the lightness you
    >must reduce the blue and green so the proportions remain the same.  You
    >can't say there are 256 shades of this hue, because if you change blue
    >without changing green you don't have the same hue.  In this particular
    >case, the next lower intensity is going to be green=62, blue=251, then
    >green=61, blue=247.  So there are only 64 shades of this particular
    >hue.
    
    You're right if the world were digital...but it's not. In the real
    world red, green, and blue overlap. We can see color in the range of
    light from about 400nanometers to about 700nanometers. Below 400 are
    the ultraviolets and above 700 are the infrareds. In this visible
    spectrum of light there are no hard boundaries. That's how we attept to
    define light in any digital color model though.  
    
    Generally, red is considered to be above 580nm
    Green is from 580nm to 490nm
    and Blue is below 490nm but...
    
    Blue transitions to green from 480 to 510
    Yellow and orange (transitions from green to red) occur between 550nm and
    630nm
    
    Remember, the blue green transitions, the yellows, and the oranges, are
    not mixes of red, green, and blue...they are indiviiidual wavelenths of
    light that when seen by the eye are precieved as these colors. 
    
    I have no idea anymore why I started with that, but since I typed it
    I'll leave it...it is good information and it might help to understand
    the following;
    
    You say that starting with 255 blue and 63 green (a shade of cyan which
    by the way cannot be reproduced on a printing press using CMYK inks) we
    have only 64 steps available if we vary just L in an HSL model. That's
    not quite true though. We have only 64 steps if we vary just G in an
    RGB model, but we have lots more steps in HSL. Here are the real values;
    
    H=225, S=100, L=100 is the same as R=0, G=63, B=255 
    (As a side note, because of the differences in color models (and
    scanners scan RGB, not HSL) these values for HSL are the same for B=255
    and G=62 or 63, or 64, or 65) 
    
    Watch what happens as L is decreased (your example is not correct)
    
    H=225, S=100, L=99 is the same as R=0, G=63, B=253
    H=225, S=100, L=98 is the same as R=0, G=62, B=250
    H=225, S=100, L=97 is the same as R=0, G=62, B=248
    H=225, S=100, L=96 is the same as R=0, G=61, B=245
    H=225, S=100, L=95 is the same as R=0, G=60, B=243
    H=225, S=100, L=94 is the same as R=0, G=60, B=240
    H=225, S=100, L=93 is the same as R=0, G=59, B=238
    H=225, S=100, L=92 is the same as R=0, G=58, B=235
    H=225, S=100, L=91 is the same as R=0, G=58, B=232
    H=225, S=100, L=90 is the same as R=0, G=57, B=230
    .
    .
    .
    H=225, S=100, L=2  is the same as R=0, G=1,  B=5
    H=225, S=100, L=1  is the same as R=0, G=0,  B=2
    H=225, S=100, L=0  is the same as R=0, G=0,  B=0
    
    By adjusting L you have 100 different luminosity (brightness) levels of
    cyan to chose from. This is a bit more realistic actually than the
    number of cyans available with RGB. I do have to say though that after
    L reaches about 14 it all looks black to me.
    
    There are only 3,600,000 diffent values available in an HSL model.
    These 3.6 million values have to represent all 16,777,216 colors
    available in the RGB model. Some overlap has to occur when color is
    converted from RGB to HSL. 
    
    Scanners do not scan in HSL though. And even when using a 30 bit
    scanner, the HSL model still has only 3.6 million values to represent
    the billions of color values available in a 30 bit RGB color model.
    
    I thought it odd that you'd feel that 30 bits is better than 24 bits,
    and then choose the HSL model (which I personally think is just as
    accurate though) to make your point? it has fewer values than even 24
    bit RGB color. 
    
    Mario
    PS I really enjoy this topic in this conference. Although I use a
    Macintosh there is not very much discussion like this in the Mac
    conference. I am not trying to "win a disagreement", I'm enjoying
    discussing the topic. Thanks.

    re .several:
    
    (Talking of B/W for simplicity) - the human eye can cover a contrast
    ratio of about 1:100, and detect about 0.1% differences, resulting
    about 460 levels of gray... i.e. ~9 bits (using a suitable non-linear
    scaling).
    
    There's also the problem that there's no such thing as absolute black
    and absolute white (not even in transparencies, let alone in opaque
    pictures). If the scanner isn't perfectly calibrated, you might leave
    values unused at either or both ends of the scale.
    
    On the other hand, most print media won't achieve that 1:100 contrast
    ratio either...
    
    I don't know much about scanners (which obviously grants me full
    authority on the subject :-) but I find it hard to believe that the
    optics is the limiting factor for gray scale resolution rather than the
    electronics (the CCD or whatever sensors used, A/D-converters) and
    software (miscalibration etc.).
    
    
    

    re: .336
    
    I was interpreting the use of the word "optics" to mean the CCD array -
    as there's not much glass involved within a scanner...
    
    /dave

I don't know much about scanners either (so I guess I'm an authority
too! :-) but want to throw in a bit more to the discussion.

Sorry, I don't quite follow Mario's point about better optics.  It
implies that the optics determine the binary output.  Is this really
true?  I'd have thought that the analog-to-digital conversion takes
place somewhere in the electronics, after the optics (which presumably
includes the sensor) have turned light into an analog voltage.

If the optics are analog and have an analog output, then Mario's
tables of light-to-bits look odd.  If the analog/digital conversion
is monotonic, then you should get a continuous increase down to the
LSB in the binary code.

Admittedly, poor optics (not worthy of an XX-bit scanner) may result
in inadequate accuracy or linearity; the light level that causes a
particular code change may be skewed.  But all output codes would be
represented, won't they?

If yes, then a 30-bit scanner does really give you 30 useful bits, but
perhaps not 30 bits of accuracy.

If not, then all they're doing is adding bits hard-wired to 0, as
Mario implies ... which would indeed be deceptive marketing of the
worst kind.

If the problem with poor optics is one of noise, then that's a
different limiting factor, which indeed makes the lower bits useless.


Regarding needing more bits than you can print, or display on your
monitor, I say yes!  If you only scanned perfect images, then it might
not matter.  But I hear that tweaking scanned images is the norm. 
Add a little brightness here, some gamma correction there, and you
effectively throw away a bit or so in each color.  In a less ideal
case where the original image isn't so hot, such tweaking may drop
three or four bits per color.

It doesn't matter than your eye can't see all those bits in the raw
scanner output.  The fact that you typically throw away bits, means
you need more to start with.

So ideally, I think one needs scanners with lots of resolution (not
necessarily accuracy, just useful bits and monotonic codes vs. light),
and software that can handle those bits until you get the scanned
image tweaked and distilled to the point you want it.

That is pretty much what happens in audio recording, for example. 
Even though the final CD may be only 16-bits/channel, the master tape
will be recorded at 20-bits, and the processing to get to that point
has even more.


Not that it means anything, considering the source ... but I could
swear I read in a magazine in the last few months, a warning similar
to Burns' in .315:  be cautious of the limitations of 24-bit color
scanners if you want good accuracy; since you only get 8-bits in
monochrome (or in each color), and sometimes that's not enough.

    
    By optics in a scanner I was refering to the CCD array, the light
    source, the glass, and the mirrors at the least. I also tend to
    consider the A/D converter part of the optics although I don't think
    anything I said has to tie the A/D converter to the optics. Basically,
    I think of most of the scanner except for the parts that move the CCD
    array, the power supply, and the interface as optics. 
    
    To clear up what I was saying, I did not mean to imply that 30 bit
    scanners do not use all 30 bits or that the bits are some how hardcoded
    to certain values. The point I was trying to make is that the CCD array
    (notice how I backspaced over optics and put in a better word) in
    inexpensive/home scanners is not capable of resolving 4096 levels of
    reflected light. So the CCD outputs the same value for several
    different levels of light that are close to one another. The CCD limits
    the number of discrete values produced because the CCD is limited. 
    
    With better optics you can more accurately distinguish more values of
    reflected light. There are many professional scanners costing in the
    100s of thousands that provide 36 and 48 bit resolution. My second
    point was that even in these scanners (and this is a point I feel keeps
    getting lost) the output is 24 bits. They take the best 24 bits and
    provide an image. Your software is capable of manipulating 24 bits so
    the scanner does the pre-processing to provide you those 24 bits. When
    you do your image manipulation in Photoshop, you are using 24 bits.
    When you drop off a bit here or add a bit there through contrast and
    brighness changes or you tweak a few bits through gamma correction you
    are doing those manipulations to a 24 bit image - regardless of the
    number of bits your scanner read, Photoshop supports only 24 bits for
    the RGB color model. 
    
    >If yes, then a 30-bit scanner does really give you 30 useful bits, but
    >perhaps not 30 bits of accuracy.
    
    That's what I was trying to say...I just ended up typing more words to
    say it. 8)
    
    >That is pretty much what happens in audio recording, for example. 
    >Even though the final CD may be only 16-bits/channel, the master tape
    >will be recorded at 20-bits, and the processing to get to that point
    >has even more.
    
    And in the pre-press industry this is also true. I don't know of any
    drum scanners that provide fewer than 30 bits of resolution, most
    provide 36 and some provide 48. With a drum scanner the operator (who
    is very much like the audio engineer sitting in front of that console)
    can adjust the full 30, 36, or 48 bit image before it is down sampled.
    Once the operator has adusted the image to their liking, the software
    takes over and downsamples the image to 24 bits providing the best 24
    bits it can. This resulting 24 bit image can then be manipulated in
    standard applications. (The exception to this is in drum scanners that
    can convert to CMYK and provide 32 bits (8 bits for 4 channels) but I
    know of no scanners in the price range we were talking about that can
    convert to CMYK - so with RGB output scanners you get 24 bit output.
    
    We were discussing the home market though and my comments about optics
    (read CCD/mirror/glass/light) were meant to show that adding more bits
    to poor optics will not create better scans, and in most cases it will
    not even give you more data, just more bits to represent the same data.
    
    Mario

    
>Sorry, I don't quite follow Mario's point about better optics.  It
>implies that the optics determine the binary output.  Is this really
>true?  I'd have thought that the analog-to-digital conversion takes
>place somewhere in the electronics, after the optics (which presumably
>includes the sensor) have turned light into an analog voltage.
    
    The optics do determine the binary output. They provide the voltage that
    is converted by the A/D converter. If the optics can't distinguish tiny
    changes in reflected light they pass on the same voltage for multiple
    intensities of reflected light. This is how .39% and .78% can end up
    with the same binary value...the optics do a poor job of reading the
    reflected light and reproduce the same voltage for two different levels
    of reflected light.
    
    >If the optics are analog and have an analog output, then Mario's
    >tables of light-to-bits look odd.  If the analog/digital conversion
    >is monotonic, then you should get a continuous increase down to the
    >LSB in the binary code.
    
    Theoretically you should see resolution to the LSB but in low cost or
    low end scanners this doen't always happen. The CCD reads light and
    outputs voltage. If the CCD is not capable of accurately reading the
    light then the voltage it outputs is not a smooth constant ramp but
    more of a stair-step. This is why, even though A/D converters are very
    accurate, we end up with inaccurate results...the A/D converter is not
    at fault, the CCD is at fault. CCD's are the most expensive single part
    of the scanner and CCDs (optics) the place were corners are cut to
    bring scanner costs under $200.
    
    Mario

    
>    With better optics you can more accurately distinguish more values of
>    reflected light. 


A major scanning problem, in the photo work I do, is that the CCDs
in a reflective (i.e., flatbed) scanner are progressively worse and
worse at the low (towards black) end in terms of their ability
to recognize shade differences.

I have done several scans by scanning twice, using radically
different driver setups, then manipulating and combining the images
in Photoshop. Even so, the low end color resolution s**ks.

From the same photographic sources, I have been able to get very good
24-bit *drum* scans -- because the drum scanners are much better at the 
"low end". But I can't afford one.

My compromise is: I have tried 30-bit flatbeds. While they still suffer
from the same "low end" problem as any CCD, the low end resolution is
*considerably* better than from 24-bits. When you're talking shadow
detail, the visual difference is very considerable...

My 2 cents.

Don Feinberg

    ok, so we have clarified some definitions... certainly, the (flatbed)
    scanner glass, any mirrors, lenses, whatever might live in such a beast
    can be called optics, but I certainly wouldn't call the A/D converter
    optics (and the CCD is, at most, an optoelectronic device [or is it
    electro-optic?]).
    
    Anyway, the CCD is still an analog device (unless we start counting
    individual electrons), so I still don't buy the fact that it would have
    any significant quantization effect (in this case, taliking about
    8/10/12 bit A/D converters).
    
    I've read somewhere (regarding CCDs in video cameras) that a CCD well
    accumulates somewhere in tens of electrons during the exposure (1/60
    sec) for a totally "black" pixel, and tens (or even hundreds) of
    thousands for the maximum brightness it can handle. Assuming (yes I
    know wone should never assume anything) the scanners are anywhere close
    to having the light source and exposure time optimized to use the full
    dynamic range of the CCD, it would produce tens of thousands "discrete"
    levels of voltage.
    
    The full well charge is obviously dependent on the surface area of the
    pixel on the CCD - the area is probably somewhat larger for typical
    flatbed scanner CCDs than video camera CCDs.
                                                              

>    PS I really enjoy this topic in this conference. Although I use a
>    Macintosh there is not very much discussion like this in the Mac
>    conference. I am not trying to "win a disagreement", I'm enjoying
>    discussing the topic. Thanks.

I agree.  This is fun...no winning or loosing involved, and I am learning a lot
in the process.

>    I thought it odd that you'd feel that 30 bits is better than 24 bits,
>    and then choose the HSL model (which I personally think is just as
>    accurate though) to make your point? it has fewer values than even 24
>    bit RGB color. 

Hmmm.  I chose the HLS system because it models more closely the descriptions
that we use in casual conversation and what typically changes in the real world.
In other words a face that has light shining from a particular angle will tend
to vary in HLS lightness as the reflection angle changes (possibly some in
saturation, but I think not unless there is some over or underload).  But in
your example, isn't the "36,000000 colors" in the HLS system based on an
arbitrary integer scale.  In other words, in the real world if you choose to
scale lightness from 1 to 100, there may really be a 76.54 value somewhere.

I also wonder if the different values that you see in the HLS table you showed
are really based on integer quantization.  However, what I did completely (and
incorrectly) ignore in .333 is the fact (as you say) that we have an RGB device
here and that for a lightness which is actually 1/256 less, the scanner
opto-electronics (how's that?) will round each value of R, G, and B to a value
it can represent.  That means that it would, in fact, come out with a different
RGB value even for a 1/256 lightness decrease even though the hue would be
slightly off.  (This is assuming that the scanner were able to perfectly resolve
256 values each for R, G, and B).

I am starting to be happier and happier that I only spent for a 24-bit scanner
rather than holding out for a 30-bit one, however.

Burns

RE: Note 557.314 by TLE::BOOTH

.    Latest on the S6E for those still wanting one: CompUSA told me
.    definitively yesterday that UMAX will NOT be shipping them any more of
.    this product. Instead, they will be receiving the 300 mentioned earlier
.    in this string (parallel port only) from UMAX.

Does anyone have any more information about the ASTRA 300 that is replacing the
S6E?  It looks the same physically and in the specs.  My guess is they
chose a cheaper route of parallel port for xfer of data.  My guess is that
this will cause a slower xfer of data to computer.

	- mark

    While we're defining things, could you summarize RGB, HLS and CMYK?
    
    I thought they were all just different ways of looking at the same
    data, like statistics.  A swatch of blue will have a unique value for
    each system, but will be consistent.  
    
    I thought the only reason for different systems was for the end user,
    which might find one system or another easier to use, depending on what
    he was trying to do with it.
    
    For example, if I am trying to change a specific color, RGB might be
    easier.  If I want a particular color to be brighter or darker, HLS
    would be better.  And tweaking with CMYK would be easier for someone
    doing commercial four-color printing.
    
    Shows what I know, eh?
    
    jeb
    

Re: .291 (and later notes saying that CompUSA does not carry the Astra 600S):

I saw the Astra 600S at a Mac store (Computer Town) today.  It supports Macs
and PCs with SCSI cards just like the S-6E, but the price was something like
$269.  Since Computer City's price on the S-6E was $279 before CompUSA's one
week $163 sale, this brings new meaning to the advertising claim

>    "The award winning S6e upgrade at a new price point!"

|              <<< Note 557.345 by PCBUOA::BAYJ "Jim, Portables" >>>
|
|    While we're defining things, could you summarize RGB, HLS and CMYK?
    
    Your understanding is pretty close to what happens.
    
    RGB is Red, Green, Blue.  These are the primary additive colors, and
    are what you find on your monitor and TV screen, and they match pretty
    closely how the eye works (and how color photographic film works). 
    This is usually the mode most paint and retouch programs work in,
    because the source material is usually scanned in in RGB.
    
    CMYK is Cyan, Magenta, Yellow and Black.  The first three are the
    primary subtractive colors.  These are the colors of ink you usually
    print with commercially for full color (unless you do special printing
    with specially matched inks) and are the three dye colors in
    photographic color prints and slides.  In theory, printing these three
    together will give you black.  Since printing inks and dyes are rarely
    perfect, most high quality prints also add Black (K, since B is already
    used for Blue).
    
    HLS is one of a number of other ways of specifying colors.  As you
    surmise, there are situations where it's easier to change colors when
    using this model.  It's also fairly similar to the way NTSC Color
    Television signals are sent (the system used in the U.S., Canada and
    Japan).
    
    There are any number of books that give more detailed explanations.
    

    re -.1
    
    They tried to swap that model on me and I refused.  While the specs are
    similar, it is a parallel scanner versus a SCSI.  This means slower
    overall performance and unclear impact if you are trying to share the
    port with a printer.  Most notable is that if you go the parallel
    route, you have no support under NT.  Umax scanners do have drivers for
    their SCSI based scannes under NT.
    
    The real replacement for the S6e is the Astra 600s which looks to be
    the exact same scanner as the s6e except that it now has 30bit
    scanning.  The going price for the 600s seems to be between 230 and 260
    though it doesn't look like anybody has it in stock yet.
    
    I suppose if you want a decent scanner at a really good price, aren't
    worried about any potential issues on your LPT port (or want to just
    add another one), and plan on running under Win95 ONLY, then the 300p
    might be a good deal.  
    
    So far CompUSA has had no real acceptable alternatives except to offer
    to refund my money for the scanner (or give me the 300p).
    
    Rich
          

re .347, explanation of RGB, HLS and CMYK color systems:

first, i don't disagree with anything in .347.

this is a little more about CMYK. it may be of interest to folks
with color printers.

actual inks are not very pure in color.  if you plot intensity on
the y axis and color (frequency or wave length) on the x axis,
you would like to see sudden changes in intensity, similar to a
square wave, lots of desired color, close to zero of other colors.

in fact, the slopes tend to be moderate.  it has been a long time
since i looked at spectrograms of ink, but about 1/3 of the area
under the curve was at the unwanted colors.  you can get inks
with somewhat better behavior, but at a steeply increasing cost.
printers use black ink to get better looking results.
but they also use it to get more economical results.

black ink costs less than colored ink. it takes a lot of colored ink
to make an approximately black blob.  it takes extra heat to dry
the extra ink, extra energy to keep the solvent from going into
the air, extra postage to mail the heavier result, etc.

the trick is to reduce the amount of C, M, and Y, and replace them
with K. if all three are present, the color that is weakest is reduced
to almost zero.  going all the way to zero yields the discontinuities
that have been discussed in other replies.

Naugebauer (sp?)  worked out the equations to convert between the 
RGB and CMYK systems. (pre WWII I think.) they are gruesome cubic equations.
the coefficients come from the spectrographic data.
to go from RGB to CMYK, there are three equations in four unknowns.
that provides the flexibility to drive one value to almost zero.

I found an interesting article describing the NTSC television system, including,
for the first time I had ever seen, a coherent explanation of that weird-looking
"CIE chromaticity diagram" - the chart that shows the visible color "gamut"
that looks like a bulging triangle.  What was of particular note was the
statement that if you picked three points on the chart, you could represent
any color within the triangle formed by those points with different combinations
of the point colors.  Red, blue and green are the most obvious choices - the
triangle for those colors encompasses most, BUT NOT ALL, of the visible
colors.  Because of the "bulges", some colors (like some bluish-greens) are 
outside the triangle and can't be represented by any mix of red, blue and green.
The diagram also suggests why television (and monitors) have more trouble
with some colors than others, because the triangle isn't equilateral.

					Steve

    
    >557.341
    >A major scanning problem, in the photo work I do, is that the CCDs
    >in a reflective (i.e., flatbed) scanner are progressively worse and
    >worse at the low (towards black) end in terms of their ability
    >to recognize shade differences.
    >
    >From the same photographic sources, I have been able to get very good
    >24-bit *drum* scans -- because the drum scanners are much better at the
    >"low end". But I can't afford one.
    >
    
    This is what I've been trying to point out, and you've experienced it
    first hand. 24 Bits from an expensive drum scanner is enough to
    properly represent the full range of color. The problem with the 24
    bits from the flatbed is that the flatbed is not as accurate as the
    drum scanner...in your case (and most cases) the low end drops off too
    soon. Typically, the high end washes out too soon too.
    
    >557.342
    >Anyway, the CCD is still an analog device (unless we start counting
    >individual electrons), so I still don't buy the fact that it would have
    >any significant quantization effect (in this case, taliking about
    >8/10/12 bit A/D converters).
    
    I'm not sure I understand this, but I think you mean (by reading the
    full note) that a CCD should be able to easily represent all color
    values. That would be true if everything were perfect. As shown in the
    experience above though, everything is not perfect.
    
    >557.343
    >Hmmm.  I chose the HLS system because it models more closely the
    >descriptions that we use in casual conversation and what typically
    >changes in the real world. In other words a face that has light shining
    >from a particular angle will tend to vary in HLS lightness as the
    >reflection angle changes (possibly some in saturation, but I think not
    >unless there is some over or underload).  But in your example, isn't
    >the "36,000000 colors" in the HLS system based on an arbitrary integer
    >scale.  In other words, in the real world if you choose to scale
    >lightness from 1 to 100, there may really be a 76.54 value somewhere.
    
    I did say that I liked the HSL model, I think more in terms of HSL--If
    I want a darker red I drop the value of L, brighter I increase L...in
    CMYK subtracting the opposite color drives me nuts, and in RGB
    increasing everything drives me nuts. 
    
    BTW, I said (I think I said anyway, I didn't actually re-read my note)
    that there were 3,600,000 not 36,000,000 colors that could be
    represented by HSL. The increment of the value L in integer jumps is
    the only value change I've ever seen. My table came from Photoshop's
    color picker, I varied L and wrote down the change for each value. I
    suppose there could be a way to represent L in to a higher granularity,
    but I've never seen it done and it presents some problems. Varying L in
    0.1% increments gives 36,000,000 values we'd require 26 bits to
    represent that value. Since most color programs have only 24 bits
    available to represent color it makes it easier to represent L (and for
    that matter S or saturation) in full intergers. If as in your example L
    were allowed to change by .01% you'd have a color model that required
    29 bits to represent all the values you could generate. Again, this is
    a major change to the software. I imagine that once L could be varied
    in smaller than integer increments we'd want S to vary the same way and
    that would really complicate things. 
    
    I asked my question because I was getting the feeling that you believed
    that if more colors could be represented then there were more colors
    available to display. and HSL did not follow that thinking. HSL was
    designed to work more closely to the way our eyes work and that doesn't
    require lots of bits. 8)
    
    >I am starting to be happier and happier that I only spent for a 24-bit
    >scanner rather than holding out for a 30-bit one, however.
    
    Cool, my work is almost done! 8)
    
    Seriously though, I'm not saying that 24 bit scanners are always better
    (nor am I saying they are always worse.) I was trying to point out
    though that in equipment that costs under $200 the extra 2 bits per
    channel are typically going to give no significant change in the
    quality of the scan. As the optics in scanners improve and the price
    decreases I can see this changing, but today there seems to be more
    hype than performance in the 30 bit home scanner market.
    
    I have an excellent description of the color models too, I've sent mail
    off to the author asking if I might have permission to reproduce his
    work. I hope he says yes, I the person that asked will find it a nice
    addition to the excellent information that was already posted.
    
    Mario

RE: Note 557.346 by NPSS::NEWTON

.I saw the Astra 600S at a Mac store (Computer Town) today.  It supports Macs
.and PCs with SCSI cards just like the S-6E, but the price was something like
.$269.  Since Computer City's price on the S-6E was $279 before CompUSA's one
.week $163 sale, this brings new meaning to the advertising claim

CompUSA was willing to give me the Astra 600S for $219 rather than the
$279 they normally charge, as I had one of the S6E back ordered.  Of
course I need SCSI for a MAC, they understood why the 300P was useless.

For the fun of it, I asked them what the price would be for the DEC
discount.  They guy looked shocked and said, I doubt it will be better
than the special we are giving you, but checked anyway (he is a nice
guy).  The shock on his face (he checked the numbers twice) told me the
discount was better.  The end result:  $209 for the Astra 600S.  A
little cheaper than $219.

The 600S is 9 seconds slower in scanning an 8.5x11 at 300 dpi color.  It
has a larger platen (8.5x14 max vs. S6E's 8.5x11).  It is 30 bit vs 24
bit.  It does NOT come with a SCSI card, where the S6E does.  It has a
mini SCSI connector, where the S6E had a DB25 and large 50 centronics
style.

Those are the major differences I can remember.  The price at $209 is
not bad.  The extra 6 bits of definition really make no difference for
consumers as we do not have printers to utilize it.

But...  If you need a SCSI scanner that works well, the price is not bad
and it may be the best you can get in that range.  I am happy with my
S6E that I currently have.

	- mark

    Well, we just picked up a UMAX 600S for our group from the local
    (Nashua) COMPUsa. The Digital price was $204.55, from the shelf price
    of $249.99.
    
    It does come with a SCSI card. We installed it in a dual processor
    Prioris server, 96MB memory, WNT 4.0. We can't get it to work right.
    We're using the Adobe PhotoDeluxe that also came with it and followed
    the directions regarding selecting the TWAIN driver, etc. The scanning
    head moves and it seems like the system is doing something but we never
    get the image preview screen that allows us to set resolution, etc.
    While this is going on, PD.EXE is using 50% of the CPU. 
    
    Any ideas would be welcomed.

    Re. .356:
    
>    Well, we just picked up a UMAX 600S 
>    It does come with a SCSI card. 
    
    But .354 bought one and says it does NOT come with a SCSI card. Did they 
    recently change the way this scanner is packaged ? It's a much better deal 
    if that price includes the interface card.
    
    Antony.

PhotoDeluxe is a bothersome piece of software.on NT. It's a little better
on W95.

The scanner itself should work great. But you might try a copy
of Photoshop or Corel Draw instead of PD.

Doug.

    re .357
    >>>>...It's a much better deal
    >>>>if that price includes the interface card.
    Yeah, but the SCSI card is probably not a real SCSI card; if it's
    the same flavor as the one that comes with the S6E scanner all it's
    good for is driving the scanner (and occupying a slot).

Not sure if its related but someone posted in the scanners newsgroup having to
set the speed down to "10" for the ID that had a printer attached to it.

I'm not sure about the generic cards and if they have a SCSI setup utility
at boot time but if they do you may want to try that.

bjm

    I've gotta admit that this whole SCSI discussion is over my head but
    I'll pass along your information to our SCSI alpha geek. I'll post the
    results.
    
    BTW: I too was surprised to see that it came with a SCSI card,
    especially after what I had read in this string, but sure enough, there
    it was!
    
    BTWBTW: Last Friday I had inadvertently purchased the Mac version of this 
    scanner (it's not well marked) for the aforementioned $204.55. I
    immediately brought it back for an exchange but they didn't have the PC
    version in stock. They had them yesterday and when we brought it up to
    the business sales desk, the salesperson said our Digital price was
    $210. Still not bad but not as good as Friday's price so we mentioned
    the $204.55 price from last Friday which he then gave us. We talked a
    bit about the prices and he said that on hardware and software, it's a
    flat 12% over their cost but on peripherals, it's supposed to be
    halfway in between their cost and the marked price. He ended up giving
    us the 12% over cost.
    
    But don't get me started on the security guard on the way out...

>    I too was surprised to see that it came with a SCSI card,
>    especially after what I had read in this string, but sure enough, there
>    it was!
    
    Can you tell from looking at the box that it contains an interface
    card? I guess there must be 2 versions of this thing out there, maybe
    the later ones ship with a card.
    
    Antony.

    Yup, it says right on the box that it contains a SCSI card.

Could the SCSI card be the diff between the Mac and the PC version?

Burns

The MAC version doesn't have a scsi card but has the MAC software.

If they are like the S6-E, they may have a MAC/PC combo package that
contains it all.

    Has anyone looked at/reviewed the new Umax Astra 1200S?  It appears to
    be a newer model of the S-12.  It is slightly smaller in height 5.2in
    vs 4.1), lighter (12.8 lb vs 17), scans up to 8 1/2 x 14, and has true 
    30 bit color vs 24.  The price appears to be the same.