I don't know about the '5000 but I have had both the '8000 and '9000.
Basically the '9000 is a scaled-up '5000, so quality should be the same.
1. The '9000 costs $1999 list, much less than the '8000 cost, possibly because
Nikon has reused the '8000's chasis design. Another consideration must have
been competition.
2. I reran tests that I contrived to measure the banding "problem" on the
'8000 and could not cause visible banding on the '9000. On the '8000 the bands
were distinct and clear. For all practical puposes it's pointless to use SF
on the '9000. And then, when I was scanning a Kodachrome to check for flare
and wasn't looking for banding, bands showed up in the flare areas, but not in
the image. But this was noticeable only after a huge increase in contrast to
make the flare visible. Out of curiosity I turned on SF and the banding
disappeared from the flare. The bands were with the scan motion. I redid the
scan a few days later and the bands did not appear!
3. The accuracy (as measured by the standard deviation of monochrome scans)
seems to have improved. I'd have to check back with the ones I made with the
'8000. Still, the '9000 still comes nowhere close to using its 16-bit
potential. Interestingly noise didn't appear randomly in the scan area.
Instead some areas clearly had more than others, much like a flatbed scanner.
Where noise appears, it has a grain along the scan motion. If SF is off you'll
occasionally see short, faint whisps of noise at right angles. Monochrome
scans with SF on and off have approximately the same SDs.
4. Shadow noise is noticeably reduced.
5. Most important to me, someone who has most of his work on KR film, the
'9000 represents a huge improvement in reduction of flare. When I saw the
first scan popup on the screen it was obvious that it's a better scanner than
the '8000. Detail that has been indistinct due to a loss of contrast was much
crisper. In addition details are sharper over the entire scan than the
'8000. The amount of blue adjustment for KR is reduced.
Since these opinions are based on samples of one you have to consider them
anecdotal.
Good luck,
Dane
