A few scanner facts

[Paul Schilliger]

Having been asked many times, I thought I would share a few scanner facts from my experience with different scanner types. Feel free to ad your own links and remarks.

Scan samples (http://www.paulschilliger.com/Pages/Scan/Scanner_samples.html)

[Ted Harris]

Paul, some beautiful scans and they do a very good job of showing the results you can get from high-end scanners as opposed to those we seem to focus on here. I would love to see some original 4x5 transparancies of the type you are working with scanned for comparison on a couple of the high end scanners and on those such as the Epson 4870, Microtek i900, etc.



I know that when I am going to print large for resale I still send the scans out to my printer na dhave him do the scanning on his drum scanner before printing.

[JohnnyV]

Below is a page with many comparisons of scans. The reader has to draw their own conclusion.

http://www.terrapinphoto.com/jmdavis/

My scan is on the page - look for my name: John Vitollo

The image was shot on Velvia with an RZ 6x7 and scanned on my Howtek 4000 drumscanner. I have that image hanging on my wall printed with an old Epson 7000 printer at 24 inches wide and can just barely see the grain if I put my nose to the print. I'm probably seeing more of the ink dot than I am of the grain of the film. It's sharp as a tack too.

Also I moderate Yahoo's ScanHi-End list for highend ccd scanners - like the Nikon 9000 and Eversmart and of course PMT drumscanners. I've been using the Nikon 9000 lately and the scans are really nice...just have to use wet mounting as the scanner has a depth of field problem...as sometimes a corner or three is out of focus.

http://groups.yahoo.com/group/ScanHi-End/

[Ted Harris]

Johnny,



Thanks, I ahve looked at this page before and it really doesn't answers the question for us here. The need is for comparisons between the newer/later model flatbed scanners at ~1000 or less (e.g. Epson 4870, 3200, Microtek i900, 1800f) and the 'higher end scanners' for 4x5 and 8x10 transparancies and negatives. All the scans on the page you nted, unless I missed something, are from MF or smaller material.

[Leigh Perry]

How about a scan-around? A suitable transparency is mailed between volunteer scanner owners. Each person does their best scan, publishes the results, and posts the transparency to the next volunteer. All scans end up on a comparison page on www.largeformatphotography.info. This would be more representative than the comparisons Johnny cited because the same image would be scanned each time.

The transparency gets an international holiday, and we get to see it scanned by every scanner known to mankind (or at least to the LF community).

Is anyone interested?

I can put my hand up for the Epson 3200 (no shortage there) and an Imacon Photo (a friend's -- scans are done in two strips).

[Paul Butzi]

Scan-around - count me in. I have a Microtek ArtixScan 1800f.

It would be interesting to see unit to unit variation, too, so I would encourage other 1800f owners to pipe up.

[Henry Ambrose]

I'll play. Epson 1680, Leaf 45.

[Ted Harris]

I'm in with thee Microtek i800.

[Ted Harris]

Whoops,

I'm in with the microtek i900! Can scan with Siverfast Ai, Vuescan and Microtek's software.





Software variations are also importantt

[Michael Mutmansky]

Hey,

I can contribute scans done on an Optronics Colorgetter Falcon drum scanner, both dry mounted and wet mounted.

One of the real issues is the resolution of the scan. At lower resolutions, the 'smear' that many low-end CCD scanners do is not as apparent, and at high resolutions, the files will be very large to deal with. Sharpening will also need to be turned off (for real!, for the Imacon devices out there)

Someone will need to make some determinations about all the possible variables so the scans are as compatible as possible for the greatest comparison benefit. Also, the slide should be sent around with a reference file that indicates the general color balance and contrast that is desired in the scan.

---Michael

[Leigh Perry]

Michael, good points.

For meaningful comparisons we will need to control the variables.

[1] Resolution. We may need to perform a low-resolution and a high-resolution scan, say at 1600ppi and 3200ppi. Some scanners may not be able to deliver the high-resolution scan.

[2] Sharpening. Paul’s illustration of the scans with and without sharpening is great. We would need to agree on a consistent USM scheme, eg Photoshop USM with fixed amount and radius. Some scanners may need more sharpening that others though...?

[3] Levels. A consistent levels adjustment regime would also be required, for example adjustment so that the darkest pixel has value 0 and the lightest pixel has value 255, without any clipping whatsoever.

[4] Colour balance. The image should have a nominal grey region that can be used to set the colour.

[5] Bit depth. Where possible scans need to be done in 16 bits before last-minute conversion to 8-bit.

Any better suggestions from the more knowledgeable out there?

[Ted Harris]

Leigh,



I think you covered most of it. I would add one more item and a thought. The item to add is that while we don't all ahave to use the same scanning software and in fact comparisons of software will also be of interest ..... we do need to agree, along with Leigh's comments, to the same scanning settings. To keep things simple I recommend NO image adjustments during scanning. Just do a straight scan with no sharpening, etc.



Now the thought .... are we going to try to put these results on the web? We are going to have some large files if we are scanning 4x5 chromes at 1600 and 3200ppi. Nothing wrong with doing it but it will need someone with lotsa space on their website as this test could easily run to 30 or 40 MB or more since most of the files will be in the 250 to 500 MB range if we are going to make the fukkbkown TIFF files available for folks to use (and I think we should). I am in the process of both redesigning my website and switching providers to a company that gives me unlimited space .. if all works out I could host the page. Finally, we might want to make a CD available.

[Michael Mutmansky]

Leigh,

A few thoughts based on your excellent list, and I'm sure others will chime in with others.

The contrast of the scan (or the built-in gamma, if you want to think of it that way) will have to be set. The easiest way to do this might be to have an image scanned that also includes a step tablet of some type, so the values of several of the steps in the middle can be normalized.

Generally, I don't try to scan to 0 and 255 because these values are unprintable with any texture. I aim for about 8-12 on the low end, and about 245 on the high end in the scan. If you want a full range scan, 0 and 255 is possible, but it is not the typical values for a pictoral scan in my experience (which I admit is limited). For a test, getting 0 and 255 would be possible with a step tablet that has values beyond the density range of the slide, even if the slide is still within the reasonable range for reproduction purposes.

My scanner has native resolutions of 4064, 2032, 1016... dpi. I could give you 3200, but it will be interpolated and I don't know how valid that will be for the test. The protocol for this will have to be decided ahead of time so everyone knows what to do.

My scanner will scan in 12 bit (and write to a 16 bit file). Photoshop actually operates in 15 bit, even though they say it is 16 bit depth. Once again, some of these issues will have to be addressed to maintain consistancy.

It would make sense to record the time for the scan and other variables like that to give some indication of the workflow issues with a particular scanner.

I think it would be very instructive to also determine the TRUE dmax of these scanners. There is a procedure to do this using a kodak test target that may be useful for comparison purposes. A step wedge from Stouffer may also work for this purpose. Also, along with a pictoral comparison, it may make sense to scan a lpm test chip like the one from Stouffer or another source so that TRUE resolving resolution can be determined for each scanner.

Dealing with the issue of resolving resloution and dmax alone would make this set of tests very valueable to many people.

---Michael

[Leigh Perry]

Michael, this is rapidly becoming very scientific! I don't know a lot about the test targets you mention, but I am happy to participate in those tests too. As well as such tests, I would like to see real-world scans such as those that Paul originally provided.

Perhaps the best solution would be to pass around two items to scan -- a real-world transparency such as the one Paul featured, and the sort of target you talk about. Most of us can come of with the first type, but are you able to provide / source the second?

Ted, I had initially envisaged producing only an extended set of scan details such as Paul's, but if we can find somewhere to publish gigabytes of full scan for people to test print, all the better.

[Paul Schilliger]

The scan-around idea proposed by Leigh is excellent! I'm in with the scanners I have of course.

In addition to the factors mentioned above, I would add that a good target slide should include some dark brown and dark purple-red motives, as these are the colors that CCD scanners have more difficulties to "see" and to map than PMT scanners. It should also be tack sharp and include some very fine textures such as stone grain. It should have deep detailed shadows and highlights. Maybe this will require more than one slide. I also believe that there is no need to have the full scans online, but crops of some chosen parts and full resolution crops. The best would be that they would be treated by the same person from the raw scans so that apples would be compared to apples. We should choose the native resolution of each scanner closest to a real world scan of a 4x5" transparency, between 2000-2500 or 3200 dpi for example.

It would have been nice to have an image made for that purpose and and that would include elements such as mentioned in the different posts, grey card, stone, fabrics of different colors, purples, highlights and deep shadows etc. If someone is inspired!...

[Lars Åke Vinberg]

Count me in, Nikon LS-8000, Microtek 1800f.

Might be a good idea to scan two 35mm slides - one calibration target, and one with subject matter.

[Leigh Perry]

We are currently working out the details of the scanning test by email, but anyone else who can contribute, please register interest here.

Thanks.

[Ben_4657]

Hi folks,

I've got an Epson 4870 and results are very nice for the price but the scanner is very slow (45 min with Ice for a 6x17) and the sharpness is below my Nikon Ls-8000 (i'm happy, given the price of this latter).

For the 6x17, I scan the 2 parts of the 6x17 in 6x9 without auto exposure and without having to turn them (I use the function to adjust the strip and see on the thumbnails if they have some overlapping) then I stich them together with Photoshop (first part in opacity 100% and the other in 50%) like the focus in a telemeter camera. The stich is perfect (you only loose a few mm because of the overlapping and the fact that the Nikon scans on 88mm for the 6x9)

Recently I bought a Microtek i900 (on the paper it's got everything : multipass scan, glassless holders,ICE). It was a mistake (the ICE only works with prints, the sharpness is on Epson's level but not Nikon - the glassless holder do not allow 6x17 and the multipass scan exhibits some banding (the shadows are not better than with the Epson on a single pass). I keep it for damaged prints (rare for me).

So I made a quantum leap a few weeks ago when I got the opportunity to buy a demo Imacon 848 (around 11000 $) I know it's a lot but the results are nowhere near the two other ones (it lacks the ICE but the Flextouch soft makes a good job) in terms of sharpness, dynamic and speed (it's a Porsche Turbo). I keep the Nikon when I need the Ice function.

This was my 2 cents experience.

[Danny Burk]

Count me in too. I have a Howtek 4500 drum scanner running with Trident (Mac) software.

Regards, Danny www.dannyburk.com

[Leyland Smith]

I can contribute the Epson 3200 if needed. I think someone has already volunteered one. I believe this scanner is good for posting files to website and for viewing on computer monitor maybe, but I don't think the quality will hold up for more that 16x20 print from a 4x5 transparency even at resolutions in 2400-3200dpi range. Native scans are very soft.

[Brian Stein]

This is a great idea.

Given the effort that will be expended by the many with scanners I wonder about trying to maximise this.

Suggestions:

1. I think the 'real world' and the target scan is a good idea. I like the idea of being able to determine non-manufacturers Dmaxs (and actual resolution too)

2. I think the 'as it comes out' part is useful, but many folks will also want to know how much a few simple manipulations (eg the unsharp masking) will help. Perhaps both?

3. I would encourage multiple versions of the same scanner to be used. We all know there is item-item variation, but almost all photogrpahy reports are samples of n=1. It would be nice to change that.

4. Many flames are generated by the 'I cant see any difference...' vs 'Its blindingly obvious...' argument of ... vs ... Could we incorporate an experiment once the scans are done but prior to their general release?

For example this might involve sending the various scans on CD unidentified as to their origin to folks who express interest and asking for a ranking. Alternatively a bunch of scans could be sent which include duplicates and folks be ask if they can pick the difference. We wouldnt need that many people to generate a rank ordering and the reproducibility of judgement.

If there is any interest in this last suggestion, not having a scanner yet I would be happy to help out with distribution etc

[Leigh Perry]

Brian, thanks for the suggestions. We are currently discussing the scanner protocol in a dedicated Yahoo group: http://groups.yahoo.com/group/ScanAround/

Anyone who wants to volunteer their scanning services or contribute in some way is welcome to join the Yahoo group.