Tim Parkin - Landscape Photographer

The picture above is the last version of my Easdale picture I took on the first week I had my large format camera and one of my favourites from the recent Gower trip. The reason I’m showing both old and new now is that I’ve only just got around to scanning all of my images and here’s why…

So, straight after my trip to the Gower, I decided I couldn’t wait any longer to get my own scans and after a lot of research, splashed out on an Epson v750. The alternative was to pay a little extra for a second hand drum scanner but the deciding factor was that for the majority of scanning I want to do, I don’t need the extra resolution and DMax of the drum scanner and I definitely can do without the extra time taken to drum mount etc. So the answer was a domestic flat bed for the majority of my scanning with the possible option of buying a second hand Howtek, Creo/Eversmart or ICG later.

First of all, I needed to work out how best to get good detail out of the v750 which means … Testing!..

Well the first thing that needed testing was the optimum height at which to scan. The v750 is factory configured to about 3mm height above the glass but supposedly the actual focus point can be +/- 1 in some instances. I bought a betterscanning newton glass mount which allows you to tape a transparency to the underside of it and ensure a consistent scan height. I scanned up and down in 0.2mm increments and found my perfect height was 3.2mm. The actual resolution difference wasn’t a great deal, possibly an extra 100-200 dpi..

Once this was complete, I tried scanning at different dpi to find out where the ‘break point’ was i.e. the point where extra dpi on the slider doesn’t result in any increase in effective resolution. I tried quite a few resolutions and did see a jump around the 1600-2400 mark. If we say that the maximum resolution is typically a multiple of the native resolution, then the frankly ludicrous 28800 dpi should scale down to either 3200 or 1600. I tried scans at 1600, 2400 and 3200 and tried scaling the smaller ones up and could not tell a huge amount of difference. In fact, in some cases I saw more resolution out of the 1600 than the 2400.. I thought this was an anomoly but tried a few more times and it seems like the the 2400 and 3200 does some form of interpolation or double passing which smears fine detail. The 1600 dpi scan definitely holds a little more detail although it is undoubtedly slightly noisier. That said, if I reduce the noise with a good filtering program, the results are no worse than the 2400, 3200 scan.

I also tried the double pass scanning which also smeared resolution and produced some strange patterning in the shadow detail.

My final test was to see if 48 bit scanning made any difference and to my surprise it definitely did. The results were slightly sharper and less noisy so it appears that something wierd is going on in the downsampling algorithm in Silverfast.

Anyway, by my real world tests, the best resolution to scan on ther v750 is 1600 dpi with no sharpening and at 48 bit.

The output scans can take a reasonable amount of sharpening too and I always apply photokit capture sharpener which has a special setting for 4×5 transparencies. This just brings the sharpness up and avoids most haloing – but look out for sharp, high contrast edges where you can use expert capture sharpener and erase the sharpening on them.

To give you an idea of the results I asked Joe Cornish to scan a transparency that I was working on using his Imacon and the results from the Epson were acceptable alongside it but with some reasonably large problems.

The images to the right and at the bottom are the Epson versions.. The major issues are a lack of colour fidelity, colour problems in the shadows and haloing around high contrast edges. All of these things are hardly noticeable until pointed out or shown side by side. I’m happy with the compromise at the moment and will still be saving up for a drum scanner at some point to make the most out of my ‘best’ transparencies..

I should also add that I can get more detail out with the v750 than I can see with a Schneider 10x loupe.

Addendum :

I’ve taken slices from the Rannoch moor photograph and uploaded them to Flickr. I’ve uploaded two photographs, one sharpened with Photokit Sharpener and one not. The order of the slices are shown is ..

1) 1600 dpi – but I’ve tweaked the alignment of the red and blue layers to get rid of chromatic abberation
2) 1600 dpi – straight out of scanner
3) 1800 dpi
4) 2400 dpi
5) 3200 dpi

The CA fix works quite well in getting rid of the red and blue smearing around highlights.. I’m going to have to play more with this. I’ve also been chatting with Julian Barkway who confirms my visual inspection that the 1600dpi scans are definitely sharper.

http://www.flickr.com/photos/timparkin/2525239268 – unsharpened

http://www.flickr.com/photos/timparkin/2524416493
– sharpened

Click on All sizes at the top of the photo and then click on the ‘original’ to see the full size 100% scan.

I’ve also figured out a little more about the haloing that occurs, especially the large scale haloing as seen on the Gower worms head shot (at the top of the island). This appears to be caused by the direction of scan. Because the light spans the whole width of the scanner, when scanning a pixel in the middle of the picture, light from either side of the scanner will be causing flare in the lens which will cause a smearing across the picture. I tried changing the direction of the mount and lo and behold the smearing moved to either side of the island (in this case that was more acceptable).