Digitizing 120: Opinions regarding DSLR vs flatbed vs film scanner?

[glen_h]

I'm not sure if that's encouraged or not, but there definitely isn't a hard limit.

 

As far as I know, it does whatever reduction that it needs after uploading. It was always extra work to reduce them before.

[glen_h]

For ordinary MF uses, an older scanner should be fine. The V500 or V600 or so.

 

For 35mm, I prefer a dedicated 35mm scanner, which images the 24mm frame width onto a 1D sensor.

(I have a ScanDual III and ScanDual IV, and for negatives when I want them fast, a NexLab F135.)

 

Mostly because I don't have the highest quality 120 negatives (most are from antique cameras),

I got an Epson 3200 for a low price. Works fine for me. If I find a good price, I might buy

a V500 or V600. I don't think I have any that need a V800.

 

For a DSLR, you need a good way to hold the negative/slide, a good light source, an appropriate

lens, and for color negatives, the appropriate conversion.

 

Also, the DSLR has a Bayer array sensor, where the usual scanner uses a 1D sensor, either with separate

R, G, and B rows, or switching the light source. You get real R, G, and B at each pixel.

 

The V500 is 6400dpi, so 14400 pixels across a 2.25in negative. That should be fine for most users.

 

I believe some of the later scanners have two lenses and two sensors, or maybe two lenses and

only one sensor, for more than 6400dpi. The Epson 3200 that I have is only 3200dpi, so 7200

across a 2.25in negative.

 

Many films are rated out to 80 cycles/mm, or about 4000dpi.

 

http://imaging.kodakalaris.com/sites/prod/files/files/products/e4051_portra_160.pdf

 

A 6400 dpi scanner should be good enough. I think used V500 and V600 are somewhat

more reasonably priced than the V700 or V800.

[mag_miksch]

Also, the DSLR has a Bayer array sensor, where the usual scanner uses a 1D sensor, either with separate

R, G, and B rows, or switching the light source. You get real R, G, and B at each pixel.

 

That makes a big difference

 

The V500 is 6400dpi, so 14400 pixels across a 2.25in negative. That should be fine for most users.

 

This flatbeds just lie, the real resolution is 1200-1600 dpi, maybe some bring 2000.

 

The real problem with the flatbeds isnt the resolution but the D-max^^

[glen_h]

 

(snip)

This flatbeds just lie, the real resolution is 1200-1600 dpi, maybe some bring 2000.

 

The real problem with the flatbeds isnt the resolution but the D-max^^

 

It seems that there is ISO 14473 describing how to measure scanner resolution.

 

But yes, there are many things related to resolution that are not covered.

 

I suspect that even if a scanner gets the full resolution, say 6400dpi, that there might

be distortions, such that those 6400 pixels might not match up exactly with 6400

pixels on paper or film. Is that bad?

 

Film MTF curves often go down to 50%, which isn't bad, but is it good enough?

 

As for Dmax, color negatives have a low gamma, so Dmax likely isn't a problem.

 

Ones I know claim 16 bits/color, or about 5 (log10) density units.

If the levels are right, that should be enough for all films.

 

Since the intensity comes out on a linear scale, it doesn't so well match the log10 scale

used for density. So, yes, at large density values, a small number of bits represent the

value, and so the density (intensity) resolution is low.

 

From: http://imaging.kodakalaris.com/sites/prod/files/files/products/e4051_portra_160.pdf

the density goes to about 3 for blue (and Dmin of 1 for blue). That shouldn't be so hard.

 

From: http://imaging.kodakalaris.com/sites/prod/files/files/resources/f4016_TMax_100.pdf

again, the curves only go up to density 3.0. Maybe the film can do more, though.

[Ed_Ingold]

Scanning with a camera has the disadvantage that large negatives are "scanned" with the same number of pixels as small ones. If I use a Sony A7Riii, that is 5304 pixels to cover the narrow side of the negative, regardless of the format. That is more than adequate for 35 mm, exceeding that of a Nikon scanner, but for medium format that number shrinks to 2400 pixels/inch. That's not nearly up to an LS-8000 (> 3600 ppi), but twice that of a consumer grade flatbed. 6400 ppi is a bogus specification. It may be determined according to ISO standards, but in practice, you get only a fraction of that value.

 

I've used flatbed film scanners before and after getting an LS-8000. I determine the best resolution as the point at which you no longer see any improvements in detail, plus a little more. Increasing the pixel count only adds to the time and storage space required, not to the quality. I don't place blind faith in ISO standards, which are consensus standards within the industries affected by them. While ISO standards lend consistency to specifications, they are for the manufacturers' benefit, not the consumer. ISO 9000, for example, refers to the quality of the paperwork used to document processes and procedures, and nothing about the quality of the products themselves. "Trust, but verify!"

 

I have yet to try "scanning" using pixel-shifting, which effectively doubles the resolution on a stationary subject, That would bring medium format copies into the realm of grain-sharpness (if the lens is up to it). I've used pixel-shifting for landscapes, and the results are impressive. There is better software for pixel-shifting than as provided by Sony, with a high degree of forgiveness for small variations between frames, like blowing grass or leaves. Motion is a moot issue when copying film.

 

DMax (maximum density) refers to the densest area which can be scanned with useful information, i.e., at the noise level of the sensor. Gamma of the film is of only peripheral importance. Kodachrome, with a DMax exceeding 4.0, is a challenge for any scanner, along with overexposed B&W film. Ektachrome tops out about 3.6, within the scope of most scanners, and color negative film is 3.0 or less.

[rick_jack1]

My two cents. I have scanned 6x7 & 35mm film using a D850 DSLR, Nikon Super Coolscan 9000ED, Epson V850, and Imacon. My purpose was to achieve the best image quality that would allow me to make the largest possible print (limited by grain and the image). A DSLR with the best macro does not cut it. It's OK for putting an image on Facebook, not much more There is no Digital ICE, you will spend hours in Photoshop spotting dust and scratches. My favorite choice was my Nikon 9000ED, with the multi-scan option and digital ICE, It was fine ten years ago and I archived thousands of images at high res. For 24x36" prints it was every bit as good as the Imacon we have at work. A few years ago Nikon chose not to service these anymore and parts became scarce. Requiring Firewire was a pain as well. I sold it while it worked and bought a Epson V850 flatbed. It is OK, it takes quite a but of setup and testing to get the focus right and the results that I'm looking for. It is capable of doing an excellent job. For the price I'd recommend this over a DSLR.

[lukpac]

My two cents. I have scanned 6x7 & 35mm film using a D850 DSLR, Nikon Super Coolscan 9000ED, Epson V850, and Imacon. My purpose was to achieve the best image quality that would allow me to make the largest possible print (limited by grain and the image). A DSLR with the best macro does not cut it. It's OK for putting an image on Facebook, not much more There is no Digital ICE, you will spend hours in Photoshop spotting dust and scratches. My favorite choice was my Nikon 9000ED, with the multi-scan option and digital ICE, It was fine ten years ago and I archived thousands of images at high res. For 24x36" prints it was every bit as good as the Imacon we have at work. A few years ago Nikon chose not to service these anymore and parts became scarce. Requiring Firewire was a pain as well. I sold it while it worked and bought a Epson V850 flatbed. It is OK, it takes quite a but of setup and testing to get the focus right and the results that I'm looking for. It is capable of doing an excellent job. For the price I'd recommend this over a DSLR.

 

Digital ICE is unfortunately not an option for B&W, which is 99.9% of what I'll be scanning.

 

I agree that something like the 9000 would be preferred, but that seems to be beyond my level for money and headaches.

 

There seems to be some disagreement whether something like the V850 would produce better results than a DSLR or not. I'm not sure I want to spend $1000 to find out.

[Ed_Ingold]

Copy fixtures have a diffuse light source, which suppresses minor scratches like those on the back of commercially processed film. Damage to the emulsion side is less common, but can be remedied through the good services of Photoshop or Lightroom, if you're lucky. ICE is no substitute for careful cleaning. ICE also slows the scanning process and adds visible artifacts to low contrast areas like open sky. TANSTAAFL!

 

There aren't many SLR lenses which can take full advantage of a high resolution camera (like the D850). The venerable 55/2.8 Micro-Nikkor is one, although near the bottom of the acceptable list. You can't rely on auto focus. To focus on the grain you need live view with digital magnification. Even at f/8, the DOF is razor thin, so keeping the film flat requires considerable care. That's doubly true with a flatbed, where focus is by trial and error.

 

The strategic problem is how to replace dedicated film scanners like Nikon, which have gone the way of crurotarsians (Triassic-Jurassic extinction). Flatbeds were a good alternative while DSLRs hovered around 12 MP. Now 50 MP is a reality and affordable 70+ MP is on the horizon. We also have much better lenses now, pursuant to the mirrorless revolution. It doesn't hurt that you can accomplish in 1/4 second what takes 2 minutes on a Coolscan.

[orsetto]

Digital ICE is unfortunately not an option for B&W, which is 99.9% of what I'll be scanning.

 

I agree that something like the 9000 would be preferred, but that seems to be beyond my level for money and headaches.

 

There seems to be some disagreement whether something like the V850 would produce better results than a DSLR or not. I'm not sure I want to spend $1000 to find out.

 

We've all been posting our experience with various options thru 2018, which is nice for future readers looking for an overview of the subject. But lukpac, this is one of those technical questions where you SERIOUSLY need to take a moment, decide what your most probable endgame is for these images, then tell us so we can give targeted suggestions. Otherwise none of this meandering will prove useful to you in the long run. Making grain-sharp large exhibition prints (bigger than 16x16) or publishing a coffee table book will require a dedicated film scanner, Nikon CoolScan minimum. Casual, non-insanely-critical prints, up to 16x16 or perhaps larger, can be made from careful scans via Epson flatbed or DSLR. Ditto for web use. BTW, I wouldn't spend $1000 for a new Epson: pick up a nice used v700 for around $400, or v500/600 new for much less.

 

Your confirmation that these images are almost exclusively B&W helps narrow things down: digitizing B&W is difficult no matter what the technique, OTOH you don't have to deal with the endless horrors of color scanning, and B&W is far more amenable to flatbed scanning if you choose that compromise. Each dedicated film scanner design is optimized for one type of film over others: the Nikons are stellar with Ektachrome and Fujichrome transparencies, bitchy with Kodachrome, a PITA with color negative, and more often than not a disappointment with B&W. The CoolScan LED lightsource emulates a brutal condenser enlarger, which harshly emphasizes grain-dust-scratches, but Digital ICE is useless with B&W. In their heyday, when alternative medium-format "desktop" scanners were available and repairable, the Minolta, Artix and Polaroid MF scanners were often preferred for B&W and Kodachrome due to their softer CFL light sources.

 

Unfortunately, today all but the Nikons are impossible to repair, and the Nikons are only really still viable because they gathered such a cult around them that DIY repair posts eventually arose. Keep in mind even the Nikons are hanging on by a thread at this point: the only "desktop" film scanners still sold new and supported are the ludicrously overpriced Hasselblad FlexTights, which have the same harshness with B&W. Most working pros didn't bother with the "desktop" contraptions at all: they sent their film to service bureaus that scanned entire rolls at a time with $50K Creo, Scitex, or Screen Cezanne hi-res flatbeds (critical work went to Aztek, Howtek, ICG, Heidelberg etc drum scanners).

 

A few posts back I mentioned the long-discontinued Polaroid Dust & Scratch Removal utility. This remains the best software-based cleanup tool for film scans: not as good as hardware ICE, but comes remarkably close. For B&W or Kodachrome, there's no contest: the free Polaroid software works effectively with those, while hardware ICE does not. The huge drawback with this utility is it hasn't been updated since version 1.0 back in 2002, so it has compatibility issues with modern Mac/Windows 64-bit OSes and PhotoShops. But if you have a large number of B&W or Kodachrome film scans to clean, its well worth getting an old 32-bit laptop just to run the Polaroid tool (if using an old scanner, you're probably gonna run it off an old PC anyway, so no problem). The Polaroid software installs as both a standalone utility that you can open files into directly, and/or as a plug-in filter for PhotoShop. The default settings do a pretty good job, and you can customize them. Some additional info, B&W samples, and fairly recent download links can be found in threads like here and here.. As with any "free" downloads, make sure you only click the links from a highly protected computer (Apple Mac, or a PC running hyperactive virus scanner like Norton). The KR-provided download link is probably safe, but you never know.

Edited December 5, 2018 by orsetto

[glen_h]

 

(snip)

It doesn't hurt that you can accomplish in 1/4 second what takes 2 minutes on a Coolscan.

 

To me, this is the real advantage of using a DSLR instead of an actual scanner.

 

(Note that the word scanner comes from moving either the sensor or object, to scan across

the image. I suspect that the word applies even when using 2D sensors, but isn't quite as obvious.)

 

Not so long after I got my D70s (as in new), I had a few hundred slides that I wanted scanned.

I had some scanner that didn't work all that well, and was very slow.

 

I got a black plastic tube that would fit over the end of my AI 55/2.8, such that the end was the

right distance to position a slide. I also have an extension ring that is just about the right length

to image a slide on the crop sensor at the right size. Even more, I didn't need the highest quality

for these pictures. Screen image or 4x6 prints was enough. (Specifically, they are pictures from

my college dorm years. Things like water balloon fights with the neighbor dorm.)

 

I never felt like doing others that way, though.

 

I later got a Minolta ScanDual III, and not so much later than that, a ScanDual IV.

(Conveniently they use the same negative and slide carriers, as those are sometimes

hard to find.). I have an XP machine to run the software for those.

 

But it takes a long time to get a few hundred slides through the ScanDual scanners,

especially at the higher resolution modes.

 

For the DSLR, it does take more than 0.25s, by the time you get the slide in position,

maybe check the focus to at least be close, and make the shot. But a lot faster than

the ScanDual.

[Ed_Ingold]

For the DSLR, it does take more than 0.25s, by the time you get the slide in position,

maybe check the focus to at least be close, and make the shot. But a lot faster than

the ScanDual.

True, but even with external time (which is required for other scanners too), I can finish at least 5 boxes of 36 slides in an hour, using a Nikon ES-1 or ES-2 copying attachment. I can process 3 or more tolls of 120 color negatives in an hour, including inversion of RAW files using Silverfast HDR. If I don't like the results, I can return to the RAW file and redo the conversion, without a need to re-scan.

-

Dust has not been a problem, given perfunctory cleaning prior to each scan, and they're not scratched. My negatives were cut and stored in archival pages within an hour or so of processing, which also renders them dead flat.

 

I've use Polaroid Dust Removal software, but find it unreliable, and resulting in noticeable artifacts.

[rodeo_joe1]

Also, the DSLR has a Bayer array sensor, where the usual scanner uses a 1D sensor, either with separate

R, G, and B rows, or switching the light source. You get real R, G, and B at each pixel.

 

"That makes a big difference"

 

- No, it doesn't.

The de-Bayering, or de-mosaicing, process is widely misunderstood and believed to reduce resolution. That's not the case in practice, and a Bayer patterned sensor can resolve right up to the theoretical spatial frequency limit of its photosite spacing.

 

Plus, the generally higher quality of a dedicated macro lens means that a DSLR/MILC copy can easily scrape all the detail from a piece of 35mm film that it's capable of holding.

 

There are plenty of devices that hold film in front of a suitable macro lens as a rigid assembly with the camera. Using such a device is simplicity itself. The light source need be no more complicated than the camera's popup flash pointed at a white surface in front of the camera - copier paper, white card, whitewashed wall, whatever.

 

The reversal from negative to positive is just as easy or problematic, regardless of the method used for digitising.

 

Having compared dedicated 35mm film-scanner output with that from a modest 24 megapixel DSLR copy, I can categorically say that the DSLR provides results almost indistinguishable from a good dedicated scanner, but in a fraction of the time.

 

If you already have access to a digital camera with an even higher resolution, then the choice of digitising method is pretty much a no-brainer.

Edited December 5, 2018 by rodeo_joe|1

[orsetto]

I've use Polaroid Dust Removal software, but find it unreliable, and resulting in noticeable artifacts.

 

As with everything else, "it depends" or "YMMV". I've had the much-hyped hardware ICE in my Nikon CS8000 do comically bad things to scans the Polaroid software handles with aplomb (and vice versa). For color scans, I'll usually go with ICE. As long as my patience holds out, anyway: "two minute" scans on a CS8000 are an optimistic dream once you engage "fine" mode and ICE (the CS9000 is much faster, since it doesn't require fine mode to avoid banding). Hardware ICE (in film scanners) doesn't work at all with B&W: for that its the Polaroid tool or manual spotting. And if you don't have hardware ICE in your scanner at all (like my SprintScan 4000+), the Polaroid tool is handy. Used carefully, it can save a lot of time vs manual spotting alone. It takes awhile to learn when to change settings like "adaptive" mode, just as it takes awhile to intuit what mischief ICE is likely to do with a given scan. I like having both cleaning tools available, since my environment is near impossible to rid of dust contamination (in the ten seconds it takes me to load the scanner tray, my pristine sleeved-at-birth film strips inevitably attract dust).

Edited December 6, 2018 by orsetto

[glen_h]

"That makes a big difference"

 

- No, it doesn't.

The de-Bayering, or de-mosaicing, process is widely misunderstood and believed to reduce resolution. That's not the case in practice, and a Bayer patterned sensor can resolve right up to the theoretical spatial frequency limit of its photosite spacing.

 

(snip)

 

You can't create information that isn't there. There are many tradeoffs between signal/noise

and resolution, and the de-mosaicing is one of those. So, yes, under some conditions you

can get full resolution, with some cost in S/N.

 

With non-linear deconvolution, you can increase the resolution of lower resolution, but high

S/N images. Looks at some of the early results from Hubble telescope to see what can be done.

 

Someone else mentioned Dmax. High density regions will have low signal, and so low S/N.

Those will complicated extracting resolution from de-mosaicing.

[Jochen_S]

The real problem with the flatbeds isnt the resolution but the D-max^^

Aye! - That 's a big issue. - I tried rescue scanning a too dark slide on a Heidelberg Topaz flatbed they had at work, no chance!

the DSLR has a Bayer array sensor

That shouldn't matter much. - At very best a 18MP Monochrom equals a 24MP Bayer, but I wouldn't spend on Visoflex and vintage macro (or enlarging?) lenses to cobble on it, to get rid of Bayer arrays, when higher resolution MILCs & native DSLRs are at hand.

 

My 2ct on the entry question: Sod the idea of scanning the heck out of a 120 neg yourself. - Sorry but going that lane will no way ever pay off.

The OP has a couple of inherited negs. How few of those can be totally stunning and awesome plus end covering walls? Sorry for punching below the line, but real estate seems in short supply, especially if you spend your fortunes on photo gear...

My bet: There is a 24 pics exhibition to salvage from that heirloom. It might contain 3 "mega stunners" that could justify desperate drum scanning efforts. But: I'd doubt anybody to own more than one wall dedicateable to holding such a picture.

Math: Please divide #price of dedicated MF film scanner through #price of drum scans. I expect the result to be "x >7" Now lean back and ponder "How much will I have to sink in film & processing, to shoot 4 mega stunners worth(!) their own walls?" (Sorry about the "mega stunner" term. - How would you call the creme of the elite among keepers? - I have no clue how rare mega stunners are. I expect a decent keeper every 2 rolls if you are a great shooter. So I dare to guess 100 rolls per mega stunner? <- I am still "optimistic"!)

What are 400 rolls + processing in your €/$s? Please look up that answer. I am at 4K€ + DIY work.

I am not the richest kid; while I'd be able to shoot MF cameras I have once in a blue moon I couldn't guarantee reaching a point where I'll need 4 drum scans made, during the rest of my life.

Next level of questions: How would you(!) shoot your MF? - My grandpa carried an Agfa Box & hand held it most of the time. - MF? Yes / Need for drum scans or resolution beyond least spectacular DSLR? No

I walked a TLR loaded with 400, more frequently 1600 ISO, handheld 1/250sec. That might be more solid results than Tri X 35mm press work but still: No need for drum scans! / Unlikely to need / exceed DSLR scans. I'd expect my Monochrom to punch in the same league results wise, so if I am into happysnapping to look & see and not in a special MF mood I should take that digital camera instead.

If you are likely to use your MF differently, at lower ISO, always from a tripod.... Wouldn't make buying a used H3 39 more sense than getting a film kit + scanner + film & processing? And how long will it take you to go through 4-600 rolls?

While "MF on tripod" had or has it's space; I don't exactly understand what is the issue with accepting: "Whatever I'll do; I won't get more than (if at all) 2000dpi scanned out of my negs, with somewhat affordable gear."? So if a lot of megapixels is desired, its time to break out a 5x7"... - Why should anything have changed during the last century of photography? - Progress only added alternatives.

 

If you want pixels for cheap: Shoot digital. If you are into the film thing: Get a darkroom. Home-scanning is basically just an option to let the online world know about what you are doing on film. Or good enough for a very conservative approach to film format selection.

[mag_miksch]

The de-Bayering, or de-mosaicing, process is widely misunderstood and believed to reduce resolution. That's not the case in practice, and a Bayer patterned sensor can resolve right up to the theoretical spatial frequency limit of its photosite spacing.

it doesnt reduce resolution in the sense of pixelcount, but the Bayers cannot show fine details.

I shot a lot with 15MP Merrills and comparing the pics to 24MP Bayerpics at 100% I see in the Foveons the finest details where in Bayerpics I just see quark.

[mag_miksch]

So if a lot of megapixels is desired, its time to break out a 5x7".

I shot a lot 5x7 slides and scanning them with my Canon 9950f brought nice results indeed

 

PS: Using vuescan the Canon supports multi- pass and long- pass

Edited December 6, 2018 by mag_miksch

[glen_h]

I think I mostly agree with the above.

 

Well, when I scan with whatever scanner I have to use, I pretty much always use the highest resolution.

(Not including any interpolation modes that might be offered.)

With the work and time to run the scanner, I might as well get what I can get.

 

At the same time, I mostly don't worry that the scanner isn't giving that absolute best resolution I could ever hope for.

 

Diffraction, imperfect optics, and inexact focus will reduce the resolution.

Most often, close enough for me.

 

My digital cameras are most often set for the highest resolution and quality of JPEG that they produce.

Sometimes I do RAW or TIFF, but not all that often.

[mag_miksch]

my conclusion of all this: shoot Merrills^^

[rodeo_joe1]

but the Bayers cannot show fine details.

 

- Well, the 100 lppmm that I see from my 24 megapixel Bayer-patterned Nikon D7200 seems well resolved enough to me. Not 'quark' at all, whatever that means.

 

Considering you'd be very lucky to get 80 lppmm clearly visible on any film, and only then in the centre quarter of the frame; 24 megapixels is more than enough to digitise a 35mm frame, and to equal the result of scanning medium format on a flatbed.

 

This isn't a Foveon versus Bayer thread. It's about the pros and cons of conventional scanning versus digital camera copying. But I'd dearly like to see the evidence supporting the supposed superiority of Foveon sensors. If it was that blindingly obvious, surely Sigma would've made better inroads into the camera market? Which they patently haven't.

[mag_miksch]

But I'd dearly like to see the evidence supporting the supposed superiority of Foveon sensors

 

[mag_miksch]

100% crop

[carbon_dragon]

So what full frame non-foveon sensor are you intending to compare this to? Digital sensors are VERY good these days, especially at low ISO on a tripod with a really good lens with good focus technique. I'm not criticizing the Foveon, I'm just saying they are pretty much all great. And especially if it doesn't have a anti-moire filter.

 

The Foveon's methodology is certainly interesting and different, but it's not quite clear what MP the Foveon's equate to.

[rodeo_joe1]

So what full frame non-foveon sensor are you intending to compare this to?

 

- Exactly!

An isolated image isn't a comparison. What's needed are side-by-side images of the same scene, taken under the same conditions, preferably using the same lens and with comparable Foveon and Bayer cameras.

[lukpac]

But lukpac, this is one of those technical questions where you SERIOUSLY need to take a moment, decide what your most probable endgame is for these images, then tell us so we can give targeted suggestions. Otherwise none of this meandering will prove useful to you in the long run. Making grain-sharp large exhibition prints (bigger than 16x16) or publishing a coffee table book will require a dedicated film scanner, Nikon CoolScan minimum. Casual, non-insanely-critical prints, up to 16x16 or perhaps larger, can be made from careful scans via Epson flatbed or DSLR. Ditto for web use. BTW, I wouldn't spend $1000 for a new Epson: pick up a nice used v700 for around $400, or v500/600 new for much less.

 

At this point, my goal is, while arguably vague, to get the best captures reasonably possible with the caveat that I can't really justify spending thousands of dollars, which basically leaves out dedicated film scanners. So the question is: current DSLR setup vs a flatbed. There's been a lot of discussion about one or the other, but not much directly comparing the two. If a flatbed would give results about the same or perhaps worse than my DSLR, I'm not about to spend the money on one. However, if a flatbed is likely to produce better results, I may consider one.