The Plustek 8100 is a popular entry-level 35mm film scanner. It produces significantly higher resolution scans than any accessibly priced flatbed scanner I know of (thinking specifically of the Epson V600/V800 flatbeds recommended so often on the internet), but it is limited to film formats smaller than 35mm. This page details an effort to relax this limitation.

The punchline is that a modified Plustek 8100 can be used to image a 6×7 film frame by stitching together 4 sweeps, together taking around 6 minutes per frame at the highest resolution setting. If we are to believe filmscanner.info's claim that it captures 3800 ppi, we end up with almost 84 effective megapixels or a 33"×28" print at 300 dpi. It also allows inclusion of borders in plain 35mm scans.

Two changes are involved:

• Modify the driver to sweep the scan head longer than usual ("taller" in terms of the image plane).
• Replace a part on the film carriage with one that avoids the scan head's field of view.

Modification procedure

My guess is most of this post applies identically to similar-looking Plustek scanners (7200, 7400, 7600i, 8200i) but I can't test this. Smarter people than me have said this so I guess I will say it too: I'm not responsible if anything bad happens to you from following these procedures.

Software

The usual interfaces to this scanner only produce captures of a ~36×24mm region. Up until a few months ago this device was limited to use with expensive proprietary software (SilverFast, VueScan). No longer, thanks to a recent update to the sane-genesys backend (Link to gitlab, cf sane-backends 1.0.31 NEWS). With this software, all that is needed to make the 8100 perform longer sweeps is to change some values in the source. The necessary changes are in a github repository here:

Link to github.

Only the genesys backend needs to be compiled.

Hardware

The film carriage must be modified to enlarge the scanning stage. As it is, the scanner's view is obscured by the carriage for part of the elongated sweep. Here a picture of the inside of the device and what needs to be removed:

It isn't that difficult to get the part off, but I was very conscious of keeping the scan head clean during my process.

Here is a render of the replacements and their thingiverse link:

Link to thingiverse.

For fun, here is a picture of the measurements I took to help sketch this in FreeCAD:

Here is how the parts look printed and installed on the carriage:

.

How to scan

I jury-rigged a really crummy film tray out of some report binders and a hacked-up Starbucks card. It succeeds in keeping the film decently flat but it doesn't consistently get the negative on the scanner's focal plane as you'll see below...

A single 6×7 frame must be run 4 times, once for each quadrant, then stitched together. The command I use for scanning is:

./scanimage -v --mode Color --resolution 3600 -y 40 -t 10 -o quadrant_1.tiff

Notes:

• For previewing replace 3600 with 600. You can also run 7200ppi scans.
• In mm, -y controls the length of the scan and -t controls how far out the head goes before it starts scanning.
• Each 40mm run takes about 1m20s.
• Make sure the sum of -t and -y does not exceed 60mm to avoid damaging the scan head actuator.
• The --mode Color option is necessary for sane to talk to the device properly.

Here is what the raw scan files look like in GIMP during coarse alignment:

And after merging, inverting and playing with levels, here is the result:

You can see there's a few problems: I was lazy and just did the merge by hand instead of with Hughin. Aside from the obvious dust and scratches everywhere, you can tell the negative wasn't on the scanners focal plane in the bottom left run since the grain is noticeably blurrier there. But all this stuff can be fixed with more practice and a better scan tray!

Possible extensions

Here is a list of extensions I think would be valuable.

• To make this hack useful someone really has to design and print a 6x7 film tray that keeps the negative flat and in-focus. Up to my measuring abilities the in-focus plane is 3.5mm above the bottom of the carriage rails.
• Hypothetically one could also design an automated feeder like the Cameradactyl Mongoose (Link to Youtube) to automatically scan the length of an entire film roll.
• Larger formats are possible in theory but they would require destructive modification and lots of patience to work with. The effort might be worthwhile since then you could digitize a 4×5 negative into a ludicrous 288 effective megapixels. This translates to a 4'×5' print at 300dpi. The cost of getting a single transparency drum-scanned at this resolution is a significant fraction of the price of an entire 8100 scanner.