Scanning film: Canon 5D Mark II vs Drum scanner vs Epson V700 with bonus Sony A7r!
This article originally appeared over at Addicted2light in November 2012 and has been kindly updated by author Gianluca Bevacqua. Gianluca’s taken a bit of time to update his initial findings with his updated DSLR “scanner”, a Sony A7r. For the impatient amongst you, you may jump on over now.
Did someone say “scan”?
If you shoot film and you don’t are much into chemicals, or don’t have a basement in which to keep a gigantic 5×7″ enlarger, you’ll soon end up with the need of a way to import those beautiful pictures you’ve taken on the computer. What? Why I didn’t say straight on “you will need a scanner”? After all it’s not 1987 anymore, and scanners are common like toaster ovens.
Well, I didn’t say “a scanner” because this it’s not the only way you can digitize those negatives. Indeed, turned out, even though it’s the first, and often, only way most people will think of, it is the most inefficient and time consuming. And it can loose a lot, i mean A LOT, of the quality of the original slide or negative.
But now there is a much better alternative…
Let’s cut to the chase: I’m proposing here to use a digital camera of high pixel count – full format / dx format doesn’t really matter – mated with a good macro lens to scan the film using multiple shots, like in a panorama. “A good macro lens” that it’s like saying “a macro lens” because, with the possible exception of some russian misassembled lemon, they all range from really good to exceptionally good. And if you have a bellows you can use, instead, an enlarger lens (an Apo-Ronar, for example, will put you back of only 60 / 70 euros).
But what about the quality you say? That it’s what this post is for!
First a brief overview of what I’ll be using for this shoot-out:
Flatbed scanner: Epson V700
The film-holder height has been calibrated. I did not use fluid mounting, but I taped the films to the film-holder and / or used a glass to keep the films flat; so fluid mounting should only make a difference in terms of absence of dust, appearance of the grain clumps and, possibly, slight better tonality
Drum scanner: Dainippon Screen DT-S 1045Ai
The films have been professionally scanned by an external service.
DSLR: Canon Eos 5D Mark II
In addition I will be using a Contax Zeiss 60mm f/2,8 Makro Planar (for medium and large format); Canon Eos 5D Mark II + Nikon Nikkor 35mm f/2 O non-Ai version inverted (for 35mm films). the methodology I used to capture my “scans” using the Canon is described in further detail here: how to scan film using a digital camera.
Given that the Canon 5D Mark II it’s the challenger we will compare it separately against each of the opponents.
So let’s check first how this setup fares compared to the Epson V700, an excellent flatbed per se (still a flatbed though).
The images below are all 100% crops. First the “usual difference” between the output of the two systems: those crops belongs to a 6×4.5cm negative shot with an ultra-sharp Fuji GS 645 on a sturdy tripod and with a soft release. The Epson crop has been sharpened, the Canon one NOT (no kidding).
Yes, the Epson (or any other flatbed scanner, for that matter) looks like an old man who is in desperate need of new glasses…
And now the best possible case (I saw the Epson behave so well only in rare occasions, like once or twice in a blue moon). The crops below were shot on a Hasselblad 500C/M with mirror up and using the standard 80mm f/2.8 Planar, tripod etc.
Yep, you just witnessed the death of flatbed scanners as film-scanners. So buy the cheapest all-in-one or LIDE model you can get, just for bills and invoices, and be done.
But surely a drum scanner–a thing that costs more than many cars, and whose scans can cost 60/200 euros a pop–will put the Canon setup to shame?
These crops were shot on a Linhof Technika 13x18cm with a Symmar-S 240mm f/5,6 lens. The drum image has been sharpened by the photo service, the Canon one is unsharpened:
No, I’ve not made a mistake.
Actually, I made one when I loaded the files in Photoshop. I gave both the same name – putting them in different folders – to make a sort of “blind test”. Well, I saw immediately that there was no contest, even though I made all the tests anyway. Boy oh boy I was up for a surprise…
This surprise came when I was saving the files: I used “Save as…” because I wanted to change back their names to something meaningful, and then I discovered that the file I was absolutely sure was of the Dainippon drum scanner, because obviously superior, was in fact the one shot with the Canon!
I even double checked the EXIF data, because I could not believe my eyes.
The amazing thing is that I did NOT used the lens at 1:1, or at 2:1 or 3:1 magnification like I do on 35mm film. So, in exceptional cases of extremely sharp negatives–say ultra sharp lenses and microfilm like films–I would be able to pull out even higher resolution!
And, just to clear any doubt you may have, here the two crops above after a good dose of sharpening:
To put things in perspective: those crops are from a 660MB greyscale file; seeing it like this on a monitor it’s like peeping at a print of 5.2 x 3.7 meters…meters.
At 240dpi I could still print it as large as 2.3 x 1,65 meters!
As I mentioned above, by increasing the reproduction ratio you can extract even more detail; see for yourself. All the following 100% unsharpened crops came from a 6x6cm negative shot like the ones before with a Hasselblad 500C/M, mirror up, standard 80mm f/2,8 Planar, tripod, etc.
The final crop has been resized at the 50% (at 3:1 there is more grain than detail, so keeping a gigantic file is pointless). And please ignore the tonality; this is a shot from a color negative, and I’m struggling a bit to find a suitable curve:
I showed here only few examples, but I tested this thoroughly with many images, colors and black and white, slides and negatives, and I consistently found the same results.
Even the tonalities of the films were much better preserved with the Canon than with the scanners. And, as an added bonus, including the picture borders or importing into the computer some odd format shots – 6×17, for example – it’s a breeze.
Summing up: forget about scanners.
Yes, if you have 3,000 euros laying around and you need to scan 100 shots a week by all means buy a Coolscan or an Imacon – but in this case, and for the sake of your own sanity consider going digital and lose film!
Instead, if you need just to scan your best shots, you may follow my advice and use the system described above.
Updates as of August 2016
I figured out my “scanning” procedure (documented here), while I was still shooting with a Canon 5D Mark II (21Mp).
This was in 2012 and since the arrival of much higher megapixel cameras my workflow has changed a bit, though not that much. Please keep in mind that a 36MP file prints at over 1 meter with ease, and that taking multiple shots of a piece of film as per my original procedure is more for the need to oversample the film than for the actual presence of so much detail on the film itself.*
Basically I now shoot with a Sony A7r and my process has been updated as follows:
For medium format, I tend to take just 1 shot for the bulk of my frames. To do so I had to add another metal lens hood as a spacer. The new hood fits snugly inside the old one and for added security they are also taped together.
For the very special pictures, I now take between 2 and 4 shots, depending on how detailed they are to begin with.
I could maybe extract a smidgen more detail taking more “sectional” images, but the results are something that would be hardly visible in a very large print, if at all, so why bother?
Besides, if you read the article linked at the foot of this page you’ll see that most of that added detail would be grain / dye clouds.
For 35mm film, even just 1 shot is normally way more than enough to extract all the detail, even accounting for oversampling. If I really want to be sure (super detailed films like Velvia, that I don’t shoot anymore anyway, or Adox CMS 20), I take 2 shots.
That’s it, all the rest has stayed the same. For a short time I substituted flashes for the light table (using an LED flashlight to focus), but honestly there wasn’t any/enough difference in terms of sharpness to justify the added complexity.
On the software side, I use Photoshop CS6 or Lightroom 6 indifferently to stitch the files together should I need to. I tried Autopano, and while it is a tad better in the trickiest situations, these don’t occur often enough for me to justify the expense and the disruption of my Lightroom-centric workflow.
*You can find wildly different values on the internet for the resolution of a piece of film. In my experience, even a medium format slide shot with perfect technique doesn’t have much more than 40Mp, at most. Should you be interested, you can see how I reached my conclusions, with 8.500ppi scans to back them up, here:
Contribute to EMULSIVEEMULSIVE NEEDS YOU. The driving force behind EMULSIVE is knowledge transfer, specifically engendering more of it in the film photography community. You can help by contributing your thoughts, work and ideas.
Help drive an open, collaborative community - all you need do is drop us a line and we'll work something out.