Canon FS4000US noise

  • Thread starter Thread starter SJS
  • Start date Start date
S

SJS

Hi,

I am using an FS4000US (new purchase) to scan slides and negs and I am
getting much more noise than I expected. This shows up as coloured
pixels in a dark area with a fairly even balance of each colour.

I am using Vuescan to drive the scanner and have scanned a black slide.
The slide has matt black tape on both sides. I set Vuescan to image
mode with exposure lock at 1 and no colour correction.

I expected the pixels in the raw file to be 0 or pretty close but find
that each of the colours can range from 0 to about 9 (24 bit raw file)
or from 0 to about 240 (48 bit raw file).

Can anybody tell me if this seems reasonable for this scanner. I can't
see the value in 14 bit colour resolution with such a high noise floor.

Thanks,

Steven
 
I have been using this scanner since it first came out but I admit I have
never scanned a uniform black image to measure individual pixel values. This
is an interesting idea but there are many places in the system where the
variation in pixel values might arise. Have you compared the pixel values of
a scan made with the Canon software? Have you tried it with other
scanner/software combinations?

The reason I ask about the comparison is that, while I have never scanned a
black slide and measured pixel values I have made many hundreds of 4000dpi
scans with this now ancient machine. Unless preparing images for CMYK
printing I don;t think there is a real-world value, as opposed to
technically measurable value, to 48 bit scanning: no monitor or inkjet
printer can display that color range and their drivers will arbitrarily
truncate the color information anyway. Do you really need the >100mb scan of
every image that results from 48 bit/4000dpi scanning? I have several CDs
full--that is exactly 6 images per CD.

I question whether Vuescan actually achieves the level of control over this
scanner that the software interface implies. I have used Vuescan through
many iterations, including version 8, and have not found it very usable, for
practical purposes, on this particular scanner. I get better images using
the Canon software. Although the software interface of Vuescan tells you it
is performing a function at a certain level of precision how do you know
this is true or even if the scanner is capable of performing this function
at this degree of precision? Vuescan is not specifically written for the
Canon scanner. You may want to download the demo version of Silverfast for
comparison as well. I am not sure the latter is worth the $250 price tag for
a scanner that now sells for $500 and is ancient in terms of its technical
design.

However, since the Canon takes so long to scan each image anyway time may
be better spent scanning real images and seeing if you are satisfied with
the results. If you are not satisfied, using Vuescan, try the Canon
software. I wish I could justify getting the new Minolta 5400 but, slow as
the old Canon is, it actually makes very, very good scans.
 
I have been using this scanner since it first came out but I admit I have
never scanned a uniform black image to measure individual pixel values.

I have to admit that scanning a black slide does sound silly (almost as
silly as scanning a blank slide holder, which I have also tried) but my
scans do seem a bit noisy. I have read that the FS4000 is noisy but I
don't have any readings from other FS4000's. My scanner is new so I
could ask to have it checked under warranty. I have tried things like
SCSI instead of USB (meant to reduce noise but probably an urban myth)
and also keeping the scanner remote from the PC and monitor but have
noticed no difference.
Have you compared the pixel values of a scan made with the Canon software?
Have you tried it with other scanner/software combinations?

I get slightly worse noise using Filmget but even with its exposure
locked and options disabled I suspect it is doing more processing than
Vuescan so perhaps the test is less reliable.

Today I tested using a Silverfast demo. All colour values for the black
slide are 43, 43, 43 (+- 1) but I can't see how to lock the exposure so
the test is not a valid comparison.
I question whether Vuescan actually achieves the level of control over this
scanner that the software interface implies. I have used Vuescan through
many iterations, including version 8, and have not found it very usable, for
practical purposes, on this particular scanner. I get better images using
the Canon software.

I hope that the raw file from Vuescan (during scan) is really raw.
Perhaps if Ed reads this he could comment.

Some of my scans are better when using Filmget rather than Vuescan. I
notice that some white areas are tainted with red using Vuescan but not
with Filmget or Silverfast. I hope I can overcome this as I think
Vuescan is much nicer to operate. Silverfast seems good but the cost is
prohibitive for me.

If anybody with an FS4000 and Vuescan could scan a black slide (image
mode, exposure lock 1, no colour balance) and tell me what colour values
they get in the preview screen it would be most appreciated.

Thanks,

Steven
 
Back
Top