Scanning: Real vs. advertised resolution

[martin_patek_strutsky]

I use a flatbed scanner (Epson 4870) with good results but I am still

confused about what resolution settings I should use. The goal is to

get the best master scans possible with this scanner and adjust for

different print sizes and printer resolutions later when the scan

gets prepared in Photoshop.

 

Several reviews have stated that the Epson 4870 has a real optical

resolution of around 1650dpi and that the published 4800x9600dpi are

nothing but snake oil.

 

Does this mean I could always use 1650dpi as a standard setting and

up sample the pic with Photoshop or other software only if necessary?

This approach would save a lot of scanning time and disc space but

would it have any shortcomings?

[johann_fuller]

Why not do a test yourself and compare the results? Additionaly manufacturing tollerances will mean each unit is different.

[john_shriver]

The sensor element has a horizontal resolution of 4800 dpi. The 9600 dpi is vertical, and that's just how fine the stepper motor steps.

 

The optics of the Epson flatbeds are nowhere near as sharp as their pixel resolution. But, there's a plus to that, the optics serve as an anti-aliasing filter. (Look up grain aliasing.)

 

I would scan at a resolution that is a clean factor of 4800. You probably won't see a real difference between 4800 and 2400, but try for yourself. For many things, 1200 will also be fine.

 

Using resolutions that aren't a factor of 4800 could lead to interpolation artifacts.

 

On the other hand, don't scan at 600. You might be subject to grain aliasing. I wouldn't do lower than 1200, if you need less, downsample in the photo editor.

[martin_patek_strutsky]

John,

 

thanks a lot for the clear answer. For now I will go for the following settings:

 

35mm 4800dpi

MF 2400dpi

4x5 1200dpi

[leonard_evens]

John's answer covered it pretty well, but let me add a few additional comments.

 

The scanning resolution has to be distinguished from resolution in the usual sense of how well the scanned image distinguishes fine detail.

 

The scanning resolution tells you how many samples the scanner collects per unit length. The maximum from the 4870 is 4800 pixels per inch or about 189 pixels per mm. According to digital sampling theory the maximal possible resolution in the other sense obtainable from that is half or about 94.5 line pairs per mm. But no scanner actually delivers that maximum, although some come closer than others. Various people have tested the 4870 by scanning test patterns. The best results I've seen are about 34 lp/mm, far short of the theoretical maximum, but more than adequate enough for 4 x 5 and adequate for most medium format. Given these figures, you probably won't get better results by scanning at 4800 ppi than at 2400 ppi.

[bruce watson]

I don't have a problem with John's answer. I explain the same thing somewhat differently. That is, there is a difference between the ability of the scanner to gather information from the film, and it's ability to generate data for the file. Data and information are not the same thing, so look at them separately.

 

You measure data generation in term of ppi. You measure information gathering in terms of lpmm the scanner can read from the target. Data generation is a given, while you can test information gathering capability with a test target such as a 1951 Air Force Resolution Target (there are several on-line sources for these).

 

As John said, I think Epson decided to use lesser optics to cut down on the cost of the scanner and at the same time to provide an anti-aliasing filter. This trade off effects the maximum resolution the scanner can actually read off the target film. It also effects the number of people who can afford the scanner.I think Epson made a pretty good decision - the scanner gives pretty good bang-for-the-buck as it were.

 

Like Leonard suggested, you might find it interesting to do a test scanning the same filme at both 4800ppi and 2400ppi. My bet it that you'll see little new detail at 4800ppi and can thus save yourself a lot of scan time by scanning at 2400ppi instead. The reason to do the test is, of course, that YMMV.

[bobmichaels]

go read and digest www.scantips.com It'll take you several hours but Wayne Fulton explains this and many other things that would be helpful for you to know.