Flat bed vs DSLR scanning

[McDiesel]

When the true 5000dpi-resolution dedicated film scanners appeared on the market, many claimed this was useless and that the already existing scanners (circa.2400dpi) were good enough, that there wasn't enough detail in the frame anyways. And then the results proved otherwise.

In the very same way that when 6MP digital cameras appeared, many people claimed they were already surpassing 35mm film in quality and being comparable to medium format...

When you scan a frame with a really good film scanner, you're gaining clarity of detail. It doesn't just mean that you're resolving up to the tiniest hair of detail, that's a minor benefit. The MAJOR benefit is that the MTF curve of the whole system is higher for all frequencies. In plain english, the clarity of the whole image will improve, not just on the parts that have tiny details.

Again, this is perfectly illustrated by the portrait/model picture above.

No arguing here. You are right, and @Helge 's corrections to my post are also accurate. I was arguing against the 3000dpi number but didn't go far enough

I will add though, that real world tangible benefits of increasing scanning resolution are log-shaped. Diminishing returns are real and the "good enough" moment happens much quicker than most commenters in a scanning thread will admit. So I understand @Alan Edward Klein point of view quite well.

Also, what bothers me somewhat is that people rarely share anything. We have DSLR lovers and Coolscan lovers argue over theoretical maximums but no real full-sized scans from either side. Not crops please, find a way to post a URL to a giant fucking JPEG in 2022. Why? Because it's easy to see how the 5000dpi (ish) HP5+ scan that I provided far exceeds most practical applications for a 35mm HP5+ negative. We need more of that. If someone feels that 5000dpi isn't enough... well show us. Because when I scan most ISO 400 films using a 24MP camera I get mush in mid/low contrast area, using a 32MP camera - same mush, using 61MP camera - same, enable pixel-shifting - same, try 100MP GFX - same, ask my lab for an X1 scan - same. The only difference is gradually improving grain, which has its own diminishing returns. I am yet to meet the elusive 100lp/mm B&W or color negative in my life.

It's trivial to see why 6MP (!) DSLRs killed film almost instantly. It happened because the practical negative enlargement limits are modest and far more complex than the theoretical lp/mm emulsion rating.

 

[MattKing]

So the sharpening is really adding contrast to the edges (acutance). It's not really adding more resolution. That's fixed by the sensor or film.

And with respect to film, frequently discarded during scanning or digitizing - sharpening doesn't miraculously restore it to a file that doesn't have it in the first place.

 

[MattKing]

Of course, a drum scanner resolves better than a flatbed. So in that case it should be more naturally "sharper".

Apparent sharpness is much more dependent on edge contrast than it is on resolution.
In many cases, higher resolution images actually appear to be less sharp.

 

[Paul Verizzo]

Missing from almost all of these comments is, "What is the output goal?" You don't need 5000dpi drum scanning capability for an image on your screen.

My scanner mindset uses the old "photo quality" printing standard of 250-300 lines per millimeter. For 35mm, that is on the short side.

As a simple example, if you won't print over 8x10 (or 12", I remember some paper decades ago so sized. How logical,) all you need is 2400 dpi on the scan to get 300 lpm on the output. This is far more than adequate for scanning those old family slides and negatives. Even if you scanned one and a year later you said to yourself, "Dang, I want an 11x14 of that, it would still be almost "photo quality."

Two of my Canon scanners are native optical at 4800x9600 dpi. One is 9600dpi both width and length. At 300 lpm, that would give a photo quality print 32" on the short side.

But, wait! As others have noted, there is the matter if INPUT quality. The capabilities of the lens and the film. Each have their own limits. I've forgotten most of the details of my research and experiments a dozen years ago. But I recall that a general bottom line of typical lens, typical modern film, came up with the fact that any digital sensor over 24 megapixels was wasted. (Talking full size image; no cropping, no "binning" of mps' to make a better image.)

Fomapan 100 rates itself at 110 lpm. Remember, this is in best conditions, no lens! The short side of a 35mm frame is 24mm, ergo, 2,640 dpi. Scanning at higher dpi accomplishes nothing. Fomapan 400 is rated at 90 lpm. I couldn't find any other film manufacturer that succinctly state MTF as lpm.

Now, I admit that I may not understand something. And that is, is a "dot" the same as a line pair? My thinking is no, that it would take two dots/pixels to make a line pair. So, very worst case, 5280 dots/pixels per 24mm short part of the frame would capture everything.

Said capability by a flatbed scanner snoozing.

 

[Paul Verizzo]

Esos dos centavos tienen muchísimo valor...

I was pretty sure I got the drift....mi Espanol es muy malo y poquito!...but I did the translation thing; very funny. Thank you!

 

[grat]

Prove me wrong.....literally.....but it seems to me that Epson scanners have a focusing problem. You gotta do "this" to get best results. I've never seen anything about Canon scanners needing focusing.

Actually, I agree 100%. The focus is non-adjustable, done at the factory, isn't always at exactly 3mm (for the second lens), and seems vulnerable to moving the scanner around without the transport locks engaged. For whatever reason, I seem to have a "good" model, as does Alan. Nick Carver, a professional photographer, is OK with his V750 for the majority of his work-- he does occasionally have special projects drum scanned.

I would say they're limited by a lack of adjustable focus. The film trays help, especially the V8xx style, as they have 5 height adjustment positions (the v7xx trays have two).

It's also worth understanding that there are two separate lenses-- one 6400 DPI calibrated for 3mm above the glass, and only works for the 7" middle strip of the scanner bed, and one 4800 DPI calibrated for the glass bed, and covers most of the bed is useful. It's also worth remembering to place the film emulsion side down so that there's nothing except the glass bed between the imaging unit and the film, is also worth doing.

I've never said the Epson is perfect. It could be much, much better, and it is a bit fiddly, and takes some work to get the best out of it.

But it's not nearly as bad as claimed by a few bad reviews.

It does, however, have a warranty, unlike the no-longer-made Canon negative scanners.

 

[Adrian Bacon]

You'll need to be very very careful (i.e. play a lot of time with height adjustment and/or glass holders) to get real 2700dpi out of the Epson V850, and even then, it would be "smeary" resolution, with little clarity.

A 36mpx digital camera like the Sony A7R with a good macro lens can extract a good 7360 x 4912 image. This represents over 5100dpi.

No. 4912px over the 56mm of a 120 frame width is barely 2200 dpi if doing single capture scan. You can do multi capture if you want more resolution, but that introduces a lot of other variables and introduces way more time to get to a usable image. 2200 dpi is very easily within what a v850 can do.

 

[Adrian Bacon]

Because, as I've said above and as you can find out in the internet if you do the research, your scanner doesn't have more than about 2700 real, true DPI. That's why your "6400dpi" scan won't look much better than "the 3200ppi" version.

Now, run the very same frame (assuming it's a sharp frame) through a really high quality dedicated scanner like a top Nikon scanner, and you'll see all the detail you've been missing.

Just a simple illustrative example -- although you'll find plenty on the web.

Epson V700 -- tested to reach about 2400dpi maximum. The V850 doesn't improve too much over this figure, even if using glass carriers and adjusting height.

View attachment 309427

No research on the internet is required. I have an Epson v850 Pro (which is not a v700) and a glass etched 1951 USAF test chart and have tested it extensively. If my memory serves correctly, I even have some threads here on Photrio with some of the gathered performance data.

I also have a dedicated 45 MP digital camera with a very high resolution macro lens and process and scan many hundreds of rolls of film every month, so I’m speaking from experience and not speculating about the performance of the two platforms.

 

[Adrian Bacon]

Hm... I wouldn't put it this way. There are several sharpening techniques. Your statement describes "smart sharpening/upsampling" in Photoshp. But the good ol unsharp mask is basically just contrast adjustment. It's quite literally takes what's already there and makes it more visible without adding anything at all.

+1

i was about to disagree with @MattKing on this. Sharpening at its most basic is nothing more than adding contrast along the edges in the image. Good algorithms give you control over edge detection, and crap algorithms apply settings that generally look good for many images, but aren’t very ideal for scans of film.

 

[brbo]

2200 dpi is very easily within what a v850 can do.

Yes. Do you see anyone disputing that V850 can do 2200dpi?

Question is, can you show it resolving 6400dpi?

Some people here obviously think Epson V850 can resolve 6400dpi. Others know it can't.

 

[McDiesel]

Missing from almost all of these comments is, "What is the output goal?" You don't need 5000dpi drum scanning capability for an image on your screen.

I think the goal of scanning should always be the same: to be able to throw the negative away after scanning. Reproduction device doesn't matter because they come and go. A scan simply needs to capture all usable information from a negative, to be reproduced on suitable media today or in the future.

Now... the definition of usable is probably subjective, but if I were to formalize it I'd probably start with the resolution of a human eye.

How much detail can a healthy 20 year old with normal vision actually see in an image without moving their head around? Well, the Internet suggests that our angle of view is up to about 150 degrees!! But it should not be a surprise that only about 60 degrees in the middle are actually utilized for seeing (as opposed to sensing), the rest is peripheral - we can't even read outside of that range, and the optimal resolution is only in the middle area of that. So much for that "corner to corner sharpness" requirement hehe.

So within those 60 degrees the "normal" vision allows us to see 30 cycles per degree, that's 30x3x60=5,400. Yes, these are extremely rough approximations but they match perfectly to my experience. I downsample my 35mm scans to 5000 pixels and medium format scans to 5000-7000 pixels on the wide side. I just never found any use for additional pixels. Anything extra is just pixel peeping, i.e. for viewing at a distance so close that you can't see the entire image.

And for that reason I don't need the same scanning resolution for medium format as I use for 35mm. The bigger negative offers finer grain and smoother tonality, which is another advantage of extra resolution (not just detail) but 5000dpi MF scanning would be grossly excessive for my needs.

So perhaps it's not a coincidence that the megapixel race in the digital world exhausted itself around 24-30MP. It doesn't seem that anything above that is valued by anyone for anything other than cropping.

 

[Adrian Bacon]

Yes. Do you see anyone disputing that V850 can do 2200dpi?

Question is, can you show it resolving 6400dpi?

Some people here obviously think Epson V850 can resolve 6400dpi. Others know it can't.

My issue isn’t that, but that many poo poo the v850 and play down it’s performance, when based on my own experience of actually owning and regularly using one, if used optimally, it tends to perform quite a bit better than what many on the internet give it credit for. 6400dpi? No, but for black and white material, quite a bit better than it’s usually given credit for, again, if used optimally to get the best out of it. Combine that with some intelligent application of sharpening to boost the contrast of the lower contrast finer details and it gives quite good results.

 

[brbo]

Epson V850 is given credit of up to 2800-3000dpi. Can you, please, show us how much better it is than that? I understand you have a good resolution target and a V850, many here don't have both or either so I'm sure we would all appreciate it.

 

[Alan Edward Klein]

+1

i was about to disagree with @MattKing on this. Sharpening at its most basic is nothing more than adding contrast along the edges in the image. Good algorithms give you control over edge detection, and crap algorithms apply settings that generally look good for many images, but aren’t very ideal for scans of film.

Adrian, I usually scan flat with my V850 without sharpening and sharpen in Lightroom V6 purchased, not the CC variety. Would I be better off sharpening with the Epson during the scan? Do they do a better job?

Which software does the best sharpening of film scans?

 

[flavio81]

As a simple example, if you won't print over 8x10 (or 12", I remember some paper decades ago so sized. How logical,) all you need is 2400 dpi on the scan to get 300 lpm on the output. This is far more than adequate for scanning those old family slides and negatives.

The problem is that we are talking a lot about resolution, but i think my main point, which is about "clarity", hasn't been taken into account. And this is something I've already wrote in the past here.

To measure the fidelity of an optical system you have the MTF, the modulation transfer function. The combination of lens MTF + film MTF + scanner MTF = system MTF, and will show not just how much "resolution" you get at the tiniest detail, but also how much contrast you get at all kinds of detail, be it narrow or coarse.

A system with a better MTF curve, that is, a curve where most of the frequencies (measured in cycles per millimeter) are transferred with the greatest contrast (idealy 100% which would mean no loss of contrast).

What the really good scanners have, is a very good MTF curve, that is, a curve that stays very high for most of the frequencies. This is where the Epson flatbeds fail -- they smear detail even at lower frequencies. That is, the smear of detail happens even on coarse details. This was readily evident here, in the image at the right:

You can try to counteract this MTF loss by applying sharpening but this will also increase noise and add more artifacts, which give a peculiar, grungy look to the image. As any information engineer knows, "garbage in -> garbage out", you can't fully compensate for bad quality data.

So, what does this mean? Yes, for a small 8x10" you only need 2400dpi on your scan (with a 35mm frame). Yet the very same print, done with a better scanner, like a dedicated scanner, would give a print with much better "clarity" or "fidelity".

This is my point. It's true that the Epson flatbeds are good enough for many purposes. But a better scanner will give better results EVEN in downsized images, and EVEN when the print is not up to the big scale where the maximum resolution is used.

 

[Adrian Bacon]

Epson V850 is given credit of up to 2800-3000dpi. Can you, please, show us how much better it is than that? I understand you have a good resolution target and a V850, many here don't have both or either so I'm sure we would all appreciate it.

There are other members on this same thread that are claiming less than that. I suppose I can go generate performance data and post it if I haven’t already done so in another thread.

at any rate, my original point was that when digitizing 120 film with a digital camera using a single capture, 40-60MP cameras are just now approaching what you can do with the Epson in terms of resolution. The only real advantage over the Epson is that it’s orders of magnitude faster.

 

[Adrian Bacon]

Adrian, I usually scan flat with my V850 without sharpening and sharpen in Lightroom V6 purchased, not the CC variety. Would I be better off sharpening with the Epson during the scan? Do they do a better job?

Which software does the best sharpening of film scans?

I pull it into photoshop, make multiple layers and apply high pass filters to each layer optimized for what I want sharpened, then set the blend mode for each layer to Overlay, then create a layer mask for each layer and mask it out except for what I want the layer to apply to.

 

[Alan Edward Klein]

Thanks. Unfortunately, I don't have photoshop only Elements. Do you think Lightroom's sharpening is better than epson's scan sharpening?

 

[grat]

Thanks. Unfortunately, I don't have photoshop only Elements. Do you think Lightroom's sharpening is better than epson's scan sharpening?

Epson Scan has built-in sharpening. SilverFast has built-in sharpening. The Epson scanner by itself, does not appear to have any level of user-configurable sharpening. If it's doing sharpening internally, we have no way of knowing, or controlling.

Personally, I don't use SilverFast's unsharp tool. I do all my sharpening / tweaking in Affinity Photo, which goes on sale for half off multiple times a year. Even at full price, $55 USD isn't a bad price to pay for the feature set.

 

[McDiesel]

Personally, I don't use SilverFast's unsharp tool. I do all my sharpening / tweaking in Affinity Photo, which goes on sale for half off multiple times a year. Even at full price, $55 USD isn't a bad price to pay for the feature set.

+1 for Affinity Photo. It succeeds where GIMP failed - it beats the Photoshop UI while being completely different, yet simple to learn.

 

[Adrian Bacon]

Thanks. Unfortunately, I don't have photoshop only Elements. Do you think Lightroom's sharpening is better than epson's scan sharpening?

Dunno. I always turn all sharpening off until I’m outputting, then do sharpening that is optimized for the output resolution. This is something that many get wrong. Sharpening should be the last thing you do, once you’re at your intended output resolution. Doing sharpening at any other stage is largely negated as soon as you resize the image to the output, especially if scaling down.

 

[250swb]

Dunno. I always turn all sharpening off until I’m outputting, then do sharpening that is optimized for the output resolution. This is something that many get wrong. Sharpening should be the last thing you do, once you’re at your intended output resolution. Doing sharpening at any other stage is largely negated as soon as you resize the image to the output, especially if scaling down.

I agree, sharpening should be the last thing to be done, and generally speaking never ever using something like Silverfast or Epson scan, Lightroom and Photoshop have far bigger brains when it comes to sharpening algorithms. There is no harm in just having a look to see what will happen to a photo with sharpening applied, but if you shoot enough photos and do enough scanning and post processing you end up kind of knowing how it will turn out without having to check.

 

[Alan Edward Klein]

I don't sharpen during the scan but I do sharpen very heavily after I get the scanned image into Lightroom. Then I add some Masking slider to eliminate the "noise" created in the blue sky caused by the sharpening.

 

[Helge]

Missing from almost all of these comments is, "What is the output goal?" You don't need 5000dpi drum scanning capability for an image on your screen.

My scanner mindset uses the old "photo quality" printing standard of 250-300 lines per millimeter. For 35mm, that is on the short side.

As a simple example, if you won't print over 8x10 (or 12", I remember some paper decades ago so sized. How logical,) all you need is 2400 dpi on the scan to get 300 lpm on the output. This is far more than adequate for scanning those old family slides and negatives. Even if you scanned one and a year later you said to yourself, "Dang, I want an 11x14 of that, it would still be almost "photo quality."

Two of my Canon scanners are native optical at 4800x9600 dpi. One is 9600dpi both width and length. At 300 lpm, that would give a photo quality print 32" on the short side.

But, wait! As others have noted, there is the matter if INPUT quality. The capabilities of the lens and the film. Each have their own limits. I've forgotten most of the details of my research and experiments a dozen years ago. But I recall that a general bottom line of typical lens, typical modern film, came up with the fact that any digital sensor over 24 megapixels was wasted. (Talking full size image; no cropping, no "binning" of mps' to make a better image.)

Fomapan 100 rates itself at 110 lpm. Remember, this is in best conditions, no lens! The short side of a 35mm frame is 24mm, ergo, 2,640 dpi. Scanning at higher dpi accomplishes nothing. Fomapan 400 is rated at 90 lpm. I couldn't find any other film manufacturer that succinctly state MTF as lpm.

Now, I admit that I may not understand something. And that is, is a "dot" the same as a line pair? My thinking is no, that it would take two dots/pixels to make a line pair. So, very worst case, 5280 dots/pixels per 24mm short part of the frame would capture everything.

Said capability by a flatbed scanner snoozing.

No ergo.
It’s line pairs.
You need to at the very least multiply the lines pairs per millimeter rating by two, if you are thinking of comparing to pixels.
In reality you need to multiply by three or more, because the real world is not aligned with the pixel array or consists of square waves.

For a bunch of reasons (that are really easy and quick to research or deduce yourself) for optimal results (even at quite small reproduction ratios), you’ll want to sample an analog medium to its full maximum potential resolution whether the lens, microphones or pickups potential resolution warrents it or not.

One of them being that low resolution interaction with the substrates physical high frequency structure (that BTW almost never caps the maximum frequency range of the medium), will send artifacts down the frequency range, that will be quite visible/audiable at smaller reproduction ratios.

 

[McDiesel]

for optimal results (even at quite small reproduction ratios), you’ll want to sample an analog medium to its full maximum potential resolution

I am conflicted. On one hand, this advice needs to be pinned somewhere. The "good enough to share on the web" thinking needs to go away.

On the other hand, "maximum potential resolution" is such a rabbit hole... quickly leading to microscopes, $10K process lenses, and other exotics that incriminate that solid advice above as unrealistic or even insane.