High Detail Large Format Scanners

grat said:

Just checked, and Vuescan can indeed do 6400 PPI if "Transparency" mode is selected. If "Transparency 8x10" is selected, then you only get 4800 PPI as your max resolution.

Click to expand...

I believe the point is that the scanner optical design itself is a limitation. I have a V850 and a Creo (top-end scanner) and the results are pretty dramatically different on 6x17, 4x5 and 8x10 particularly in the shadows and fine details. I'm archiving negatives so I want as much detail as possible. We all understand that the final print size and output type should be driving much of how we scan either way I assume.

BCM said:

I believe the point is that the scanner optical design itself is a limitation. I have a V850 and a Creo (top-end scanner) and the results are pretty dramatically different on 6x17, 4x5 and 8x10 particularly in the shadows and fine details. I'm archiving negatives so I want as much detail as possible. We all understand that the final print size and output type should be driving much of how we scan either way I assume.

Click to expand...

Considering the price difference, it would be a major shock if the Creo wasn't substantially better.

On the other hand, my Epson cost $900, new, with warranty, and works with modern OS's.

But, no, the point I was responding to was whether Vuescan could utilize both lenses, and the answer is, yes, it can.

Maybe sharing which version of Vuescan you are running would help Donald? Also, I don't think Vuescan uses SANE drivers on Linux so I don't think SANE/Linux distro version would matter but @grat could share that as well...

@Donald Qualls have you tried reseting Vuescan settings?

grat said:

Indeed. I have a 6/12 core Ryzen box with 32gb of memory and blazingly fast SSD-- and it struggles with a 1 gig TIFF.

Click to expand...

Discovered last night that by default, Affinity limits itself to 16Gb of memory usage.

Raising that to 32Gb to match my system helped performance, but it's definitely not "snappy".

brbo said:

Maybe sharing which version of Vuescan you are running would help Donald? Also, I don't think Vuescan uses SANE drivers on Linux so I don't think SANE/Linux distro version would matter but @grat could share that as well...

Click to expand...

Whoops. I was testing vuescan on Windows.

One linux test coming up, if I can make the scanner work with a USB extension-- My linux box is on the other side of the desk.

So VueScan 9.7.95, on Manjaro Linux (and you're right, it doesn't use SANE-- although XSane does "mostly" work with the Epson, it crashed because I "only" had 8G free in /tmp, on a 4800 PPI scan) supports 4800 PPI max for "Transparency 8x10", and 6400 PPI max for "Transparency".

Curiously, 4800 PPI isn't an option on the drop down for "Transparency"-- you have to select "Custom" and enter 4800 manually.

Also, irritatingly, you can't have it auto-select a 4x5 inch frame. "Maximum" gives the full scan area, and "Auto" maxes out to what appears to be 6x9 cm.

Donald Qualls said:

Vuescan 64 for Linux, as of the last time I used it, did not offer resolution above 4800. Have you tried this software in Linux?

Click to expand...

grat said:

Just checked, and Vuescan can indeed do 6400 PPI if "Transparency" mode is selected. If "Transparency 8x10" is selected, then you only get 4800 PPI as your max resolution.

Click to expand...

I run Linux since ever with Vuescan for scanning. I have the V700. (two in fact, one for spare...)

Vuescan offers indeed 6400 dpi when "Transparency" mode is chosen (at least in "Professional" UI, I never user "Basic" and "Standard") but the scanning area doesn't cover 8x10, it's narrower. 8x10 coverage is possible only as "Tranparency 8x10".

example with a 8x10 sheet:

"Transparency":




"Tranparency 8x10":






that said, about resolution of the V700/V750/V800/V800/V850, there are many threads over the years at different places. Often it is said that it does a real ~2200 to 2400 dpi... "real" as "physical" as per hardware, and that other figures are internal firmware extrapolation.
A reference page is this one, here about the V800:l

https://www.filmscanner.info/en/EpsonPerfectionV800Photo.html

of course many people have been doing their own tests.

Just for illustration:
---------------------------

for reasons too long to explain and that may be wrong, I scan BW as "BW negative" but with bits per pixel set as "48 bit GB", and save a "16 bit grey", as RAW in TIF format/container.

The 8x19 sheet seen in these screenshots of Vuescan gives the following sizes:
2400 dpi : 808 mb
4800 dpi: 3,2 gb
my photographic stuff is on a workstation with 48gb RAM and 2 Xeon 3,8Ghz quad-core each, so it's comfortable


a 1280px width 100% quality JPG of the 2400 scan after inversion of the negative in ColorPerfect grade 2 virtual printing:





a 100% zoomed crop of the first beast in the center, this here is not a JPG but a lossless PNG chunk of the original RAW:




by the right egde on the whole image there's a black thombstone, 100%, same, lossless PNG :







now, from the 3,2 gb 4800 dpi scan. crops of the same, at 67% instead of 100% just for practical purpose of not providing too big image here:










on the thombstone of Svend Lien there's not so much difference in the resolution of " födt" and "död" ....





now, if I am happy with the narrower scan that can be made as "Transparency" at 6400dpi, which is possible in this case, because I can keep the church and leave part of the sky and of the foreground. Image ratio is of course elongated then.


a web rendering at 1280px reduction in 100% JPG:




a 100% crop of the beast, same, on the original RAW, saved as PNG. heavy file for web ~ 1mb, so you need a good internet speed:






and Svend Lien:

by the way, the sheet there is a Fomapan 100. I use Fomapan or X-rays film with 8x10, otherwise too costly for me, as I can shot quit some.

important about the "Transparency" and "Transparency 8x10": Epson manuel mentions that 8x10 is do be done on the glass bed of the scanner. Epson 8x10 holder is a positioning mask but film lays on the bed. As per the manual (page 40) :

"Place the film in the film area guide (up to 8 × 10 inch film) with the shiny base side facing down. Your images and
any wording on the film should appear backwards on the side that faces up."

there are infinite byzantines discussions about this: emulsion down or up, and is emulsion up, possible Newton rings from the shiny side down....

anyway, the first thing that I learned myself as LF shooter was to cut glass .... Because soon or later one needs to replace a focusing screen, or make your own for a custom camera, and to cut glass to hold sheet flats....

I have 2 settings for 8x10 and 5x7 (18x24, 13x18....) best summarized in pictures:

sheet on the flatbed, hold with a plate of glass, and this for "Transparency 8x10":



sheet sandwiched between 2 plates of glass, the lower one high adjusted with some paper padding (after countless test and try the focus). This for "Transparency" mode. For 4x5 or 120 I sometimes do this instead of using the Epson holders:




in fact I did tune my 4x5 Epson holder for optimal focus. See a piece of thick black paper (from automotive gaskets material) under the native high adjuster:




otherwise for 120 I use Digitalizia rather than Epson holders, with custom high adjustment:

I think it's important to clarify a few important points. First of all, the Epson v850 doesn't actually optically resolve anywhere near 4800DPI. At best it can manage an optical resolution of 2600DPI. These guys have tested it here with resolution test targets: https://www.filmscanner.info/en/EpsonPerfectionV850Pro.html. My guess is that resolution cited by the marketing materials for this scanner refer to the number of steps the motor can move the sensor per inch, but the "high" resolution lens can't actually resolve that much detail.

Second, I think that discussions of how much a given medium resolves is way more complex than it first appears. Whether analog or digital, the storage medium generally out-resolves the lens these days. In pretty much every scanning system, the lens is the limitation, and the best commercial scanners rely on specialized lenses designed for high resolution across a flat field. The reason drum scanners can resolve so much is that they focus a beam of light onto a single point, and detect the light passing through the negative with a (very sensitive) photo-multiplier tube rather than a CCD or CMOD sensor. The limitation here is the minimum size of the spot of light you can create, and how many samples you care to make. What all of these share in common is that their resolving power is best expressed as a function of system that is capturing the image. So the detail in a photographic negative is a function not just of the size of the film or the ISO, but the lens and the stability of the platform, etc.

I found this an interesting read regarding the resolving power of film: https://www.tmax100.com/photo/pdf/film.pdf. The silver halide crystals in black and white film are 0.2-2µm in size, but the actual resolving power of the finest grain film is around 6µm, or 80lp/mm. Black and white film grain is 10-40µm. Dye clouds are 10-25µm. The sensor sites in the A7r IV are 3.76µm. But that doesn't tell the whole story, because as we know with digital cameras, the pixel pitch is not the whole story, and small pixels are actually worse than large ones because they receive less light, and therefore more of the information is noise. When you get into the weeds, the story is complicated, and again the determining factor is more often than not the lens.

To really talk about the resolving power of an imaging system, it's somewhat pointless to discuss the size of the fundamental units that record the image, or how many of them they are. As with any information recording system, it's a matter of information theory. What is the point in the system beyond which any additional information recorded is just noise?

pwadoc said:

I think it's important to clarify a few important points. First of all, the Epson v850 doesn't actually optically resolve anywhere near 4800DPI. At best it can manage an optical resolution of 2600DPI. These guys have tested it here with resolution test targets: https://www.filmscanner.info/en/EpsonPerfectionV850Pro.html. My guess is that resolution cited by the marketing materials for this scanner refer to the number of steps the motor can move the sensor per inch, but the "high" resolution lens can't actually resolve that much detail.

Click to expand...

They were using a demo unit (most likely), with the glass target sitting on the bed. Maybe. To be honest, they didn't really describe their testing methodology. There's no way to know if they had it in focus or not, because they only provide crops. They certainly weren't testing the high resolution lens, because it's set at a fixed focus 3mm above the bed (or should be-- moving the scanner with the transport locks disengaged can result in the calibration heading south with the birds).

They also claim that the new holders add TWO layers of glass: "But now Epson added two more glass panels in the film holders."-- My film holders have one layer of glass, between the light source and the film-- how that extra layer of glass between the light source and the negative is supposed to make it harder to scan the negative, I'm not sure.

They also commented that their negatives were hanging about 0.4mm below the ANR glass-- so you have to wonder if they compensated by raising the transparency holders one notch.

Do you have an Epson scanner? I do. I get much better results than filmscanner's tests suggest are possible. So either I have a magic scanner, or they had a lousy one, or they didn't know what they were doing. When a reviewer's tests come nowhere near matching the reality of the unit sitting on my desk, I have difficulty with the review.

pwadoc said:

The silver halide crystals in black and white film are 0.2-2µm in size, but the actual resolving power of the finest grain film is around 6µm, or 80lp/mm.

Click to expand...

Fuji rates their 400H at 100lp/mm resolution, and Kodak rates T-Max 100 from 60 to 200, depending on contrast ratio of the negative. Acros II has similar ratings. God knows what Adox CMS II 20 can achieve.

grat said:

Do you have an Epson scanner? I do. I get much better results than filmscanner's tests suggest are possible. So either I have a magic scanner, or they had a lousy one, or they didn't know what they were doing. When a reviewer's tests come nowhere near matching the reality of the unit sitting on my desk, I have difficulty with the review.

Click to expand...

Have you performed your tests in a reproducible way with a resolution test target, and can you share those results? Hell, I'd even go for a scan with a v850 that achieves a higher resolution than a Coolscan 9000, since that would go a way to proving the resolution claims.

I owned and used a v850 for years, and I've compared the results to other film scanners and to tests I've done with resolution targets with DSLR setups. The results aligned pretty much exactly with the findings in that article.

grat said:

Fuji rates their 400H at 100lp/mm resolution, and Kodak rates T-Max 100 from 60 to 200, depending on contrast ratio of the negative. Acros II has similar ratings. God knows what Adox CMS II 20 can achieve.

Click to expand...

This is the MTF of Fuji 400H, the scale at the bottom is logarithmic and they don't include as many divisions as the Kodak data sheets, but it looks like it falls below the 30% contrast mark at about 60-70lp/mm. I recommend reading the article I linked, it has a lot of information including a formula for deriving the resolution you can achieve in a camera system given the MTF chart of a given emulsion.

[edit] I want to expand here on why I think the published resolution of the V850 is bogus, and why most claims of scanner resolution tend to be as well. I'm pretty sure even Creo are inflating their numbers are a bit, though they are at least honest enough to state a 10,000DPI resolution, but only a 5,000DPI _optical_ resolution in their marketing materials. This is the distinction Epson fails to make, and represents the number of steps the motors can move the sensor within an inch rather than the resolving power of the lens. The very best lenses you can realistically buy, lenses that cost many thousands of dollars, top out at a resolution of about 140lp/mm. But that's for regular lenses, the lenses for scanners are very specialized as I mentioned. They are macro lenses generally, but optimized for a flat field and for a certain magnification. Most of the lenses I've pulled out of scanners and tested are optimized for 1:1. This is pretty damn demanding set of requirements for a lens, and these lenses generally cannot get anywhere near the resolving power that the best general lenses achieve in the center of the frame. In my experience, with a very good 1:1 macro reproduction lens, 90lp/mm is extremely good performance. That's about 4500DPI. My Sony 80mm Macro manages about 92lp/mm at the very center at f5.6, which is the sharpest aperture for the corners, and in the corners it falls to 80lp/mm. It's widely regarded as one of the best 1:1 macro lenses you can buy. The lens in the V850 would have to be better than that, yet still economic to stick in a $1,000 scanner. Most of the reason the Coolscan 9000 is so damned expensive is because it has one of the best 1:1 flat field macro lenses ever made inside it, though it's limited to a single, rather wide aperture.

The advantage that a flatbed has over a dedicated slide scanner is that it can make multiple passes over the bed with the lens focussed at different areas. This is likely how the V850 manages to maintain its 2600DPI at larger formats, because otherwise the resolution would fall at larger sizes the way it does with the Coolscan 9000 and Minolta Damage Multiscan Pro. This is because those scanners scan larger sizes by moving the lens further away from the target. The Coolscan, very impressively, manages to retain something like 90% of the 35mm resolution, whereas the Minolta drops from over 4000DPI to something like 3600DPI for medium format.

When assessing these scanners you need to think about the compromises necessary to make the optics of the thing work within the constraints of reality of budgets. That's why I'm tempted to believe the Creo's claims of 5000DPI, because a 98lp/mm flat field macro lens is a crazy piece of engineering and it should cost a shitload of money.

grat said:

Just checked, and Vuescan can indeed do 6400 PPI if "Transparency" mode is selected. If "Transparency 8x10" is selected, then you only get 4800 PPI as your max resolution.

Click to expand...

What OS are you using? This may be a limitation for Linux, wherein Vuescan is having to use its own reverse engineered driver for the scanner instead of one provided by Epson.

To answer the OP's original question more succinctly, since the best lenses (35mm) you can buy resolve around 140lp/mm, and that ends up being a ~2µm pixel size. (in practice it doesn't actually work like this, but let's just go with the naive model for the sake of argument here). The highest resolution full frame consumer digital cameras today have a pixel pitch between 3-4µm. Smaller sensors, including cell phone cameras, offer an even smaller pixel pitch, but as you can probably guess that doesn't actually offer much improvement in resolution for reasons we've already discussed. For most of the lenses you are going to be realistically be able to buy there's no point to decreasing the size of the photo sites unless the average lens performance drastically improves. Bigger sensor sizes promise higher quality, but lenses that have to cover a larger area tend to have a lower ability to resolve detail, so there are diminishing returns there.

The real future is likely in computational photography. Systems that take multiple samples and composite them, potentially with multiple sensor/lens pairings. At that point the resolving power of the system is less a factor of what the lens can offer and more about what the algorithm running the process can derive.

So I predict we will see ever-increasing megapixel values, but that imaging power won't be used to increase resolution, but rather to offer additional features. For instance, once you have way more, smaller photo sites than any lens can ever resolve, you can use all of those extra sensors to detect, say, the direction of the incoming light. One of the big tradeoffs with light field cameras was that they traded resolution for the ability to adjust the depth of field post image capture. If you have a 600MP sensor in a 35mm size, you'll never build a lens good enough to actually resolve that much detail, but you can certainly capture a bunch of other information, such as the directionality of the light, that can make things interesting. Combine that with lidar, multiple lens at different focal ranges, multiple image sensors, etc. Imagine all of the stuff that modern cell phone cameras do, plus some additional fun stuff that we haven't thought of yet, combined with a professional, high resolution interchangeable lens camera system. That's the next decade of digital camera development.

Donald Qualls said:

What OS are you using? This may be a limitation for Linux, wherein Vuescan is having to use its own reverse engineered driver for the scanner instead of one provided by Epson.

Click to expand...

Manjaro Linux, which is a rolling version of Arch. It's probably as up-to-date as it can be. But since it works identically under Linux and Windows, I'm assuming it's just how VueScan is written-- and honestly, it's not a limitation, it matches the scanner's specs-- if you're scanning at the bed height, it's limited to 4800 PPI, and if you're scanning at the 3mm transparency height, it's 6400 PPI.

@grat I meant what I was seeing (no offer of higher than 4800) was a limitation for Linux. I use Kubuntu 20.04 at present, which is surely not as up to date as Manjaro, but also is Debian based (Arch is an independent, neither Debian nor Fedora based, says Wikipedia). I'll try to remember to fire it up again when I'm home (haven't used it in a while, my darkroom has been back-burnered by life) and double check.

Donald Qualls said:

@grat I meant what I was seeing (no offer of higher than 4800) was a limitation for Linux. I use Kubuntu 20.04 at present, which is surely not as up to date as Manjaro, but also is Debian based (Arch is an independent, neither Debian nor Fedora based, says Wikipedia). I'll try to remember to fire it up again when I'm home (haven't used it in a while, my darkroom has been back-burnered by life) and double check.

Click to expand...

Same behavior on Windows, though.

grat said:

Same behavior on Windows, though.

Click to expand...

Right, so it might be something specific to Ubuntu flavors or Debian based Linux. Or I might have managed to miss it. Or it might depend on some other setting I've left on "Auto" (Vuescan has some depth to the menus). I need to clear off my scanner anyway, I've got three rolls of 127 that need scanning (and I've got an adapter to fit one strip in a 120 slot). Long weekend coming up...

grat said:

Considering the price difference, it would be a major shock if the Creo wasn't substantially better.

On the other hand, my Epson cost $900, new, with warranty, and works with modern OS's.

But, no, the point I was responding to was whether Vuescan could utilize both lenses, and the answer is, yes, it can.

Click to expand...

I had tried Vuescan a while back on my V850 and it didn't allow me to scan with the higher resolution lens using 8x10 film on the bed. Was I missing an option? Is there a newer version? I still use my 850 for digital contact prints and that would save time.

Correct. The higher resolution lens is set to focus approximately 3mm above the glass bed, and only covers an area of 5.9" x 9.7" inches.

On the other hand, 4800 PPI for an 8x10 negative is good for 16x blowups-- 128" x 160". If you need something bigger than that, consider professional scanning.

momus said:

Note: There may not be a lot of VISIBLE difference between a very good scan and a drum scan, at least for most practical purposes. You have to wonder if the difference would even be seen on a large print? What we see on a monitor isn't much help either, because that just throws yet another variable into the equation. My computer monitor is not other people's monitor.

This is probably why 8x10 shooters prefer to make contact prints. If one does that, it's done directly from the negative, w/o any other steps in between. Every step outside of that introduces issues that are not present on the negative. My experience is that scanning also does something to the film grain on the scan. Usually it makes things more grainy. On, on a wet print, a lot of that will disappear. I doubt it would disappear on an inkjet print, since the dither of the printer introduces it's own stuff.

Click to expand...

I had a fluid mounted scan on a Epson 750, as perfect as I could, looked really good. Had the same negative drum scanned. I wish I could afford a drum scanner.. Noticeable difference on print as well.

Alright, firstly sorry for my absence in this thread. I've been super super super busy. I just got back from Georgia for Thanksgiving. There was next to no service or internet wherever I went. Finally back to home where there is internet sometimes instead of never (Mississippians are smartester and more cultured than Georgians).
What I'm getting from this thread is that with a 6400 dpi scanner you can successfully scan all of the detail from film, just that focusing is the issue with the different scanners?

6400dpi should be plenty to get everything from all but very high resolution film and lenses (shot with good technique and in good conditions).
Scanners that can do true 6000+ dpi are very rare. Consumer flatbed scanners are nowhere near that.

Focusing is a problem. Carefully establishing the plane of focus in a flatbed scanner can maybe bring you about 10-15% better results than straight "out of the box". With a high resolution scanner proper focusing is MUCH more important. Read this review to get a feel of how poorly some high resolution scanners do when film is not in focus (either because film holders can't hold film flat or because they lack proper autofocusing) even compared to a desktop flatbed like Epson V850.

6400 for a 4x5 would give you about 6GB tiff file for color 48 bits.

Darth Musturd said:

know of any good current scanners that can scan large format well at all

Click to expand...

I used a Sony Alpha with 90 mm macro lens to scan some old half-plate glass negatives. It gives me 60 MP which can be increased to around 200 MP with Pixel Shift. If I need more resolution I can always take multiple exposures and stitch them in post.
What are you going to do with a 2 Gigapixel digital image?