PDA

View Full Version : DSLR Scanner Project: Does it work?



Drew Bedo
5-Feb-2017, 06:28
The DSLER Scanner thread is over 60 pages now and that's a lot to read.


Would someone with an understanding of this body of knowledge please post a summery of the current state of this DIY Art?

Peter De Smidt
5-Feb-2017, 13:10
Yes, it does work. While the end result will depend upon your camera, the lens, ....., you can get significantly better results than with an Epson flatbed. With my D600, I get about 3000 spi, and at least a stop more dynamic range than with an Epson scanner. Daniel gets about 4000 dpi with his D810. These results were tested with step wedges and a high resolution chrome on glass target. The downside is that you have to build and test it. We detail the Arduino control system, and anyone can freely download the Arduino sketch that we use, but clearly this is way less turnkey than buying a scanner. In addition, once you have the files from a negative, you then have to process them in stitching software. This takes some time the first time, but after that you use the settings as a template for future stitches, and so there's no problem with detail-less areas. So, there's more time involved with getting a file into Photoshop than there is with most commercial scanners, although this will depend on how many frames you're stitching. So, higher quality but more time.

Things change if you're after fast instead of high quality. If you have a high MP camera, and taking just one picture of the negative is enough for you, for example for web use or small prints, than a system can be very fast and easy to use. Going up to about a 3 frame stitch can be done completely manually, i.e. without stepper motors...., and is pretty easy to do.

Here's a video showing Daniel's system:
https://www.youtube.com/watch?v=PXy7RJwIBAo

Here's an early version of my scanner:
https://www.youtube.com/watch?v=gmRHTausFls

Drew Bedo
7-Feb-2017, 09:43
Thank You for that concise sujmmery of this option.

How many folks are going this rout?

Peter De Smidt
7-Feb-2017, 10:25
I've had about 6-8 people contact me over the years saying they wanted to build a system. I don't know if many of them followed through. A couple of people on the thread did so, and a Google search shows that are more. So, not a lot. The option is really best for people who would otherwise want to buy a professional flatbed or drum scanner, but don't want to for a variety of reasons. The other would be for someone who owns a D810 class dslr, had a Rodagon D or similar, perhaps from doing dupes in the old days, and wants to quickly scan a whole bunch of 35mm slides.

Serge S
7-Feb-2017, 11:15
I started out using a Canon 6d to shoot my 4x5 negatives (mirror lock up + live view + macro lens) with a copy stand when my old scanner died.
I eventually bought an Epson V800.
The results are not as sharp, but they come out decent and it is less trouble for me.

Jeff T
7-Feb-2017, 13:46
Drew,

I've tried this many times when I needed a quick capture of slide films for previews and sharing on social media. For printing and high resolution capture, DSLR scans has many hurdles to overcome. The setup can simple but it can get complicated.

Things you must consider when going this route, a proper light source with a mixing box, a stage to keep the film flat and free of dust, a solid camera mount to reduce vibrations, a decent macro lens without abberations, a workflow for inverting negatives.

I am sure you can invest in a XY linear stage like a CNC router and capture multi samples and stitching them to get a high resolution image, especially for capturing art work.

Its not my best tool but I'd be happy to use a DSLR to scan if that's the only tool I have.

For 35mm negative film my old dusty Kodak 3570 scans at 6Mp and does a better job than a DSLR.



Sent from my iPhone using Tapatalk

Drew Bedo
9-Feb-2017, 07:07
Thank you everyone.

I am now sure that I will stay with my Epson 750.

mksystem
18-Feb-2017, 00:21
I works for me. But now I scan only my old 35mm negatives. I use 60mm macro and D600, also I make couple tests with macrotubes to scan at around 10.000 dpi. I post this small video at main thread.

https://www.youtube.com/watch?v=p9xJ-R3Ku5s
Here one old photo at 4000 dpi (one frame on D600) and at around 10000 dpi (16 frames on D600).
http://koninin.com.ru/panorama.zip

Drew Bedo
19-Feb-2017, 08:03
Thanks; I needed to see that. Looks like it works pretty well. Stiching 9 or so shots of each neg/slide? Wonder why no-one puts together a commercially made set-up . . .with or without the camera.

Must take a couple hundred shots to cover an 8x10 chrome though. Even at one acquisition per 35mm equivalent area of LF film it would be what, 50-80 or so (didn't do any math, just squinted for a moment).

I haven't been really happy since the guys at "Flash", the last pro lab in Houston that would do an optical enlargement to printing paper closed ten years ago. In the end, I think they had to pay someone to haul off their two 10 x 10 enlargers for scrap.

While Acer's in the Heights will still develop 8x10 B&W (no chrome) they will only do an ink-jet print from a scan. So each single print can cost, altogether, about as much as a box of HP (again, just a squint estimate).

Leaves me stuck with my own talentless scans on the 750 followed by a few clicks in PS. The whole process is just not as much fun as it used to be.

Peter De Smidt
19-Feb-2017, 09:51
MKsystem, it's awesome that you have a working system! In my experience, though, using a light box like that will cause flare. You might consider masking off the light that doesn't go through the negative. You might also run some tests to see if scanning at 10.000 dpi is of any benefit. I've compared scanning at 1x with a Rodagon D at f/4, 2x with a Mitutoyo M-Plan APO, and 5x with a Nikon 5x Measuring Microscope. With a resolution target, each step up in magnification gives the expected bump in on subject resolution, but with a 35mm frame of Kodak's Technical pan exposed on a heavy tripod with a 105 lens at f/5.6 with mirror lock up, I didn't see an obvious increase in quality moving up past 1x, even at %400 on screen. Using super high magnification will lead to less depth of field, much longer scanning and processing times, more wear on your shutter, and bigger (but not likely better) files. But since our systems are different, you should test for yourself.

Drew, there are commercial products. They are made for universities and museums, and they cost $15K + . Here's one: http://www.gigamacro.com/

mksystem
24-Feb-2017, 02:44
I agree about maximum resolution. I don't see any new details at 10.000 dpi (9 to 16 pics on D600) vs 4.000 dpi (1 pic), I made this mostly for experiment. I have only 35mm negatives and I need to try panorama stitching software.:)

I use AutoPano Giga Pro, it have plugin to import from Gigapano which works well for my mosaics. It take few minutes to stitch one negative. I also tried Microsoft ICE, it works too, but I prefer AutoPano.

BetterSense
24-Feb-2017, 05:45
one of the downsides of the method: you need a good dslr, and if you count to cost of the dslr, it's a pretty expensive scanner. I do have a D70 lying around somewhere but you would have to stitch for sure with a 2MP camera.

mksystem
24-Feb-2017, 07:13
You could get Nikon D3200 for $200-$300 and get more resolution from one negative than D600. 60mm/2,8D for $200-$300.
I also spent $150-$200 for steppers+electronics+LED strip+plywood+wheels+rails etc. Microsoft ICE is free but AutoPano Giga is 200 euro.

It is cheaper than Nikon 9000 but still you need to know how to make this work: engineering, programming etc.

BetterSense
24-Feb-2017, 08:15
I didn't realize D3200 were that cheap.

Doing a whole negative in one shot appeals to me the most.

Are there any mirrorless cameras that have high resolution, or is it only DSLRs that have top-end sensors?

Peter De Smidt
24-Feb-2017, 09:06
Just for the record, I use a template in PTgui to stitch hi-res files. It works very well. I've also briefly use Autopano Pro for other stitching tasks, and it gave great results. I ended up going with PTGui, as Daniel is very knowledgeable about it. When Microsoft Ice works, it's terrific, and it has a .....I"m forgetting the term....mode for stitching where the camera is moved (instead of pivoted), and that's ideal. But when it doesn't work, it really fails. For small stiches, say 3 frames, I do it simply in Photoshop. Load all three, say, files as layers. Turn the top layer invisible. Select the second one. Turn it's blending mode to "difference". Pick the move tool. Now use the arrow buttons to nudge the file into place.

Regarding cameras, some of us have to have a digital camera for other purposes. If I was buying a camera just for scanning, I'd buy a cheaper APS camera, as the smaller frame puts less stress on the lens, as the corners are usually what can go a little softer. For big stitches, a camera that can be controlled by an Arduino is helpful. Nikon is no problem. I expect that Canon wouldn't be hard to get to work, but I'm not sure about other brands. If they can be used tethered, you're probably ok. Until recently, Nikon cameras had a bit of an edge in dynamic range.

Sure there are mirrorless cameras with high resolution. Sony has a bunch.

Peter De Smidt
13-Aug-2017, 13:11
For the best quality with a dslr scan, one should maximize the information captured by each of the channels available with a specific capture system. If it's a monochrome sensor, then it's one channel, but if it's an R,G,B sensor, then it applies to all three channels. To do this, you want to Expose To The Right (ETTR), which means to give each channel as much exposure as will almost cause highlight clipping, assuming that no area of the negative should be clipped. This will maximize tonal separation in each channel while minimizing noise. While this is obvious for color film, it also applies to bw negatives, especially if a staining developer is used, although the green channel will be the most important, as most image detail with digital captures depend more on the green channel than on the others, although doing this is not as critical as it is with color images.

This isn't as easy to accomplish as you might think. In the first place, your camera's histogram and display image are based on a jpg that's made according to your camera settings, and that's the case whether you are shooting in jpg or raw. These jpgs often pump up color and contrast to give a pleasing jpg image, but that's not what we're after here. We want to optimize the capture. So if you're using your camera to determine exposure, go into it's picture control settings, and set the controls for the lowest saturation, contrast, sharpness settings as you can. These only apply to the jpg and histogram. Since you're shooting raw they won't be cooked into the captured file. It's ideal if you can save that as a preset. Doing this will often show that you need to give a stop or more exposure to truly ETTR. Underexposing by a stop is a huge deal when dealing with a slide with a very high density range.

If you're shooting tethered, then there's also a gotchya to avoid. Make sure to set the raw processor to do as little processing as possible. Namely, set the input profile to "neutral" or "linear". If you use something like Capture One's Auto, it will significantly brighten underexposed images. That's great if you want an underexposed image to look good, but it's better to not have an underexposed image, and you want your raw processor to show you that your image is underexposed if that's the case. This applies not only to scans but to optimizing studio lighting if you're shooting tethered.

If done properly, these files will appear too bright. The best way to fix this is in raw processing, being very careful not to go too far.

Another point to emphasize is that no stray light should come out of your light source. The only light visible from your source should be coming through your negative, with as little of the clear base of the film showing as you can. If you don't mask the negative appropriately, you will exacerbate flare, and this will cause obvious artifacts, or, worse, a not immediately obvious loss tonal separation. It would be like traditional printing with a slightly too bright safelight.

The implications would seem to be that you want a light source that RGB levels are controllable, and that's what Daniel and I did, but that approach does have some practical problems, as it can lead to banding due to the interaction of pulse width modulation and shutter speed. Could that be overcome? Sure, but it's not straight forward. LEDs have come a long way in the years since Daniel and I built our current machines. In particular, there's now a whole breed of high color quality LED lights intended for shooting video. These are likely much higher quality than the RGB strips in our current scanners. Daniel will be testing an affordable Dracast unit over the next couple of months. Hopefully, it'll also allow us to increase the shutter speed to minimize vibration from shutter actuation. These are most likely to occur with shutter speeds from about 1/8th second to 2 seconds.

Happy Scanning!

williaty
13-Aug-2017, 17:16
Peter, can you link to your write up about the Arduino control system and the motor drive you're using, as well as to your light source? I've read the several threads that are started here and there on LFPF, but you guys have made a lot of references to "oh, I explained that in the other conversation we were having" without actually providing a direct link to concise enough information to allow someone to replicate what you've done.

Peter De Smidt
13-Aug-2017, 17:22
https://github.com/nSomnius/Scanduino-by-ReallySmall

https://www.flickr.com/photos/35334802@N04/

Drew Bedo
16-Aug-2017, 12:18
I have hqad an Epson 750 for ten or so years. From what I can understand from reading this thread, I will keep it and mot try to do it with any sort of DIY scanner set up.

I scan 4x5 and 8x10 negs and transparencies with the 750 and it seems to be working for me.

Thanks to everyone who contributed the good info on this thread.

BetterSense
16-Aug-2017, 21:58
The thing is that 4x5 and 8x10 is the easiest thing to scan, so you can get good results from flatbeds, at sufficient resolution, and without grain aliasing, with large sheet film. The capability of flatbeds breaks down for 6x9 and smaller, and what good smaller-format scanners there were, are no longer in production or supported, so there is every reason to be interested in DSLR scanning for smaller formats, at least as much as for larger ones..