Revised Epson 4870 resolution comparisons

[Mike Engles]

Hello



Here, using Wayne Fulton's suggestion, is a revised resolution
comparison.



Resolutions 1200, 2400,3200 and 4800,upsampled to 4800.



http://www.btinternet.com/~mike.engles/mike/compare.tif


and


http://www.btinternet.com/~mike.engles/mike/scompare.tif

The second has sharpening 300,0.8,0.

I have the separate files if anyone is interested.


Mike Engles

 

[Kennedy McEwen]

Here, using Wayne Fulton's suggestion, is a revised resolution
comparison.

Resolutions 1200, 2400,3200 and 4800,upsampled to 4800.

http://www.btinternet.com/~mike.engles/mike/compare.tif

and

http://www.btinternet.com/~mike.engles/mike/scompare.tif

The second has sharpening 300,0.8,0.

The sharpened images showing that the 4800ppi scan reproduces more
detail than the others.

However I am a little concerned that the 4800ppi image shows something
akin to astigmatism - almost like rectangular pixels. This is most
apparent on the rounded characters of the test patterns, but it is also
present on the test patterns themselves, with the 34 lp/mm being barely
resolved in the vertical axis, whilst the 40, 48 and even 56lp/mm
horizontal resolution tests are clearly resolved.

This artefact may well be present on the other images at source since it
is impossible to tell if it has been concealed by the interpolation.

Do you know which direction corresponds to the CCD axis and which to the
stepper motor axis?

 

[Mike Engles]

The sharpened images showing that the 4800ppi scan reproduces more
detail than the others.

However I am a little concerned that the 4800ppi image shows something
akin to astigmatism - almost like rectangular pixels. This is most
apparent on the rounded characters of the test patterns, but it is also
present on the test patterns themselves, with the 34 lp/mm being barely
resolved in the vertical axis, whilst the 40, 48 and even 56lp/mm
horizontal resolution tests are clearly resolved.

This artefact may well be present on the other images at source since it
is impossible to tell if it has been concealed by the interpolation.

Do you know which direction corresponds to the CCD axis and which to the
stepper motor axis?
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Hello

The line of scanning is across the image, so I suppose it is the motor.
I can provide a sample from a different area of the slide.

Mike Engles

 

[-]

Do you know which direction corresponds to the CCD axis and which to the
In my test target scans, I can see that the scanner definitely gives better
resolution in the vertical direction (lengthwise on the scanner bed) as
opposed to the horizontal direction. In other words, a line that is
oriented so that it runs down the length of the scanner bed will appear
sharper than the same line oriented so that it runs horizontally across the
scanner bed.

I actually was surprised at this. My guess would have been that the lines
going horizontally across the scanner bed would have been sharpest since my
understanding is that scanning in this horizontal direction does not rely on
the stepper motor.

Doug (the other Doug)

 

[Mike Engles]

The sharpened images showing that the 4800ppi scan reproduces more
detail than the others.

However I am a little concerned that the 4800ppi image shows something
akin to astigmatism - almost like rectangular pixels. This is most
apparent on the rounded characters of the test patterns, but it is also
present on the test patterns themselves, with the 34 lp/mm being barely
resolved in the vertical axis, whilst the 40, 48 and even 56lp/mm
horizontal resolution tests are clearly resolved.

This artefact may well be present on the other images at source since it
is impossible to tell if it has been concealed by the interpolation.

Do you know which direction corresponds to the CCD axis and which to the
stepper motor axis?
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Hello

Here are two scans of the same portion at 4800DPI.
http://www.btinternet.com/~mike.engles/mike/Lengthcrop.tif
http://www.btinternet.com/~mike.engles/mike/Widthcrop.tif

Maximun Contrast.By length I mean the orientation of the slide on the
length of the scanner bed.

When I looked at the actual target,I noticed that the ruling of the 34
mm lines are not even.

Mike Engles

 

[Kennedy McEwen]

In my test target scans, I can see that the scanner definitely gives better
resolution in the vertical direction (lengthwise on the scanner bed) as
opposed to the horizontal direction. In other words, a line that is
oriented so that it runs down the length of the scanner bed will appear
sharper than the same line oriented so that it runs horizontally across the
scanner bed.

Interesting - this is completely the opposite of what Mike is
demonstrating, assuming I understand both of your descriptions of length
and width correctly. ;-)

I actually was surprised at this. My guess would have been that the lines
going horizontally across the scanner bed would have been sharpest since my
understanding is that scanning in this horizontal direction does not rely on
the stepper motor.

I wouldn't say I expected one to be better than the other - at the end
of the day the pixels on the CCD are square and the lens should have
circularly symmetric resolution. If Epson try to cut the scan time to
an absolute minimum then they may not have given sufficient time for the
scanner head to have been moved to its new position and settled there
before initiating a CCD exposure, which would reduce resolution in the
direction of the motor travel, but that would imply bad design and,
since they are trying to get optimum resolution on this device, I would
not expect them to cut that particular corner.

 

[Kennedy McEwen]

Here are two scans of the same portion at 4800DPI.
http://www.btinternet.com/~mike.engles/mike/Lengthcrop.tif
http://www.btinternet.com/~mike.engles/mike/Widthcrop.tif

Maximun Contrast.By length I mean the orientation of the slide on the
length of the scanner bed.

When I looked at the actual target,I noticed that the ruling of the 34
mm lines are not even.

I assume, from the look of these images, that they are unsharpened or at
least minimally sharpened. Adding some sharpening in PS brings out the
resolution limits quite clearly and, unfortunately, also the astigmatism
again. :-( Neverthless, the scanner clearly has better resolution
along the CCD axis than along the motor drive axis. For example, on
widthcrop.tif the 56lp/in lines are clearly discernible post sharpening,
whilst they merge into a blur on lengthcrop.tif irrespective of the
amount of sharpening applied. Similarly, on lengthcrop.tif the 68lp/in
are just on the limit of separation with adequate sharpening, whilst
they cannot be discerned at all on lengthcrop.

The fact that the scanner has different resolution in one axis than the
other is not, in itself, an issue. I would be concerned if this was my
scanner and it exhibited as much difference as this. It also seems
strange that it is in the opposite axis to Douglas' nominally identical
scanner though. That would imply that one of the two of you has a
deficient unit, but from this distance it is impossible to say which.

 

[Mike Engles]

I assume, from the look of these images, that they are unsharpened or at
least minimally sharpened. Adding some sharpening in PS brings out the
resolution limits quite clearly and, unfortunately, also the astigmatism
again. :-( Neverthless, the scanner clearly has better resolution
along the CCD axis than along the motor drive axis. For example, on
widthcrop.tif the 56lp/in lines are clearly discernible post sharpening,
whilst they merge into a blur on lengthcrop.tif irrespective of the
amount of sharpening applied. Similarly, on lengthcrop.tif the 68lp/in
are just on the limit of separation with adequate sharpening, whilst
they cannot be discerned at all on lengthcrop.

The fact that the scanner has different resolution in one axis than the
other is not, in itself, an issue. I would be concerned if this was my
scanner and it exhibited as much difference as this. It also seems
strange that it is in the opposite axis to Douglas' nominally identical
scanner though. That would imply that one of the two of you has a
deficient unit, but from this distance it is impossible to say which.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Hello

Thanks for the reply.

I could take it along to Epson, who are not faraway in Hemel Hempstead
UK.

I might just e-mail them with the images, but if they are anything like
Nikon, I don't expect I will get very far.

This is the second scanner I have had.
I bought the first from a different firm.
It arrived with the box pretty battered, so I refused it and cancelled
the order. I got the present one from Amazon.
These devices are precision instruments, but I am not convinced that the
handling by delivery companies is very good.

Mike Engles

Mike Engles

 

[Mike Engles]

Interesting - this is completely the opposite of what Mike is
demonstrating, assuming I understand both of your descriptions of length
and width correctly. ;-)

I wouldn't say I expected one to be better than the other - at the end
of the day the pixels on the CCD are square and the lens should have
circularly symmetric resolution. If Epson try to cut the scan time to
an absolute minimum then they may not have given sufficient time for the
scanner head to have been moved to its new position and settled there
before initiating a CCD exposure, which would reduce resolution in the
direction of the motor travel, but that would imply bad design and,
since they are trying to get optimum resolution on this device, I would
not expect them to cut that particular corner.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Hello

By length I mean the slide was scanned with its long side parallel to
the length of the scanner bed(stepper motor direction?) and width means
the slide was turned through 90 degrees.

My sense is that the better scan is the width. There is less colour
halo, viewing at 200%

Mike Engles

 

[Mike Engles]

I assume, from the look of these images, that they are unsharpened or at
least minimally sharpened. Adding some sharpening in PS brings out the
resolution limits quite clearly and, unfortunately, also the astigmatism
again. :-( Neverthless, the scanner clearly has better resolution
along the CCD axis than along the motor drive axis. For example, on
widthcrop.tif the 56lp/in lines are clearly discernible post sharpening,
whilst they merge into a blur on lengthcrop.tif irrespective of the
amount of sharpening applied. Similarly, on lengthcrop.tif the 68lp/in
are just on the limit of separation with adequate sharpening, whilst
they cannot be discerned at all on lengthcrop.

The fact that the scanner has different resolution in one axis than the
other is not, in itself, an issue. I would be concerned if this was my
scanner and it exhibited as much difference as this. It also seems
strange that it is in the opposite axis to Douglas' nominally identical
scanner though. That would imply that one of the two of you has a
deficient unit, but from this distance it is impossible to say which.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Hello

Yes that is my sense, the width is better. Less colour halo.

Perhaps this is considerd 'within tolerance'. After all it is basically
a consumer scanner cosying £300UKP.

I might send the images to Epson. Can't imagine they would want to do
anything about it.

I have made a PSD file with the images so that it is easier to compare.

http://www.btinternet.com/~mike.engles/mike/layered.zip

Mike Engles

 

[Ralf R. Radermacher]

Similarly, on lengthcrop.tif the 68lp/in
are just on the limit of separation with adequate sharpening, whilst
they cannot be discerned at all on lengthcrop.

Now, isn't this exactly the direction for which they're even
advertising a resolution of 9600 dpi....?

Ralf

 

[-]

By length I mean the slide was scanned with its long side parallel to

the length of the scanner bed(stepper motor direction?) and width means
the slide was turned through 90 degrees.

I guess it is less important how the slide itself was oriented as opposed to
how the lines were oriented for the scan (we don't know if you shot the
slide in portrait or landscape orientation with your camera). Pick a set of
test lines, think back to how you scanned the slide and thus how the
test_image was oriented (which may be different from the slide holder's
orientation) and tell us whether the lines that appear horizontal in your
image were actually oriented horizontally across your flatbed scanner glass
during the scanner glass. Then you can put your answer in terms of "a line
oriented horizontally across the scanner's bed scans sharper/less sharp than
the same line oriented vertically on the scanner glass.

Doug

 

[Mike Engles]

I guess it is less important how the slide itself was oriented as opposed to
how the lines were oriented for the scan (we don't know if you shot the
slide in portrait or landscape orientation with your camera). Pick a set of
test lines, think back to how you scanned the slide and thus how the
test_image was oriented (which may be different from the slide holder's
orientation) and tell us whether the lines that appear horizontal in your
image were actually oriented horizontally across your flatbed scanner glass
during the scanner glass. Then you can put your answer in terms of "a line
oriented horizontally across the scanner's bed scans sharper/less sharp than
the same line oriented vertically on the scanner glass.

Doug

Hello

The original image was landscape.
The image Croplength was scanned with the slide in landscape.
The image Cropwidth was scanned at 90 degrees to the above.
It is this scan that to my eyes looks better.

Mike Engles

 

[Bart van der Wolf]

SNIP

Interesting - this is completely the opposite of what Mike is
demonstrating, assuming I understand both of your descriptions of length
and width correctly. ;-)

Perhaps it helps to embrace the terminology the ISO uses:
fast scan direction: the scan direction corresponding to the direction of
the alignment of the addressable photoelements in a linear array image
sensor.

slow scan direction: the direction in which the scanner moves the
photoelements (perpendicular to the lines of active photoelements in a
linear array image sensor.)

Bart

 

[Kennedy McEwen]

Now, isn't this exactly the direction for which they're even
advertising a resolution of 9600 dpi....?

Yes, but as I have been campaigning about for years, they are quoting
resolution and meaning sampling density. The two are not the same -
except in the world of scanner specifications!