Re: filmscanners: OT: Native intelligence

[rafeb <rafeb@channel1.com>]

At 11:51 PM 7/25/01 -0700, Art wrote:

>Getting back to scanners, why is it there is so much discussion of
>"banding, banding, banding"... is it that manufacturers think we "want"
>banding in our scans?  Of maybe it has to do with problems is design
>(gee, could that be engineers who made errors?... no, couldn't be)...
>
>Or maybe, there is lack of precision in the components?  Changes in
>dimensions due to temperature?, or changes of electronic component
>characteristics?  Or, yes, some might be software programming defects as
>well.


Actually, it's been a couple of days since there was any 
mention at all of banding.

I'm looking, at this instant, at a datasheet for a Toshiba 
CCD (TCD-2901D) freely available on the internet.  There is 
a spec for non-uniformity across the array:  TYP: 15%, MAX: 20%.

So, in a way, it's amazing these CCD scanners work at all.
What makes them work is white-point and black-point compensation. 
The base measurement has to be done as often as possible - typically 
just before each scan.  The corrections are applied on a 
pixel-by-pixel basis, for every pixel in the resulting image.

Of course, the calibration presumes that the light source is 
also constant over the course of a scan.  If that assumption 
fails, all bets are off.  This is why scanners with cold-
cathode tubes often have annoyingly long lamp-warmup times.

And of course, there are 2nd-and 3rd-order effects, some of 
which have been mentioned earlier in this thread.  Eg., line-frequency 
noise, poor grounding (to explain the periodic banding in the 
8000 ED) and maybe a host of mechanical issues as well.

The yellow/brown streaking (not quite banding) that I saw occasionally 
on my SprintScan and Microtek scanners may also be due to poor 
sensitivity in the blue channel.  Again, the datasheets tell 
the story -- the blue channel has about 1/2 the sensitivity of 
the red and green channels.


>My point, very simply is that your assumption that mechanical engineers,
>by nature of their diplomas automatically make they more capable of
>understanding or implementing design is only partially true, and if they
>were all so good at it, we'd like in a world where things held up a lot
>better than they do.


The difference between Austin and you, or me and you, is that 
we *are* engineers.  So, rather than talk generalities, we look 
for root causes.  I've cited one case (just above) where one can 
learn a good deal just by looking at data sheets for typical 
CCD arrays.


rafe b.