6400 dpi 4x5 scanning problem, Epson V800 Vuescan

[kaif]

Hi,

Are there any forum users which use Vuescan on a PC to scan 4x5 negatives at 6400dpi on an Epson flatbed?

I'm using an Epson V800, latest Vuescan version (on a Win 10 Pro 64 bit PC with 16GB RAM) - and have no problem choosing the 6400 resolution in the interface, but the image files produced are only 3200 dpi. I do get 6400 dpi files from 6x7 medium format without problem. I found a 5 year old thread to this topic here on the forum, but without any clear-cut solution at the time other than scanning parts of the image separately and then stitching them together - which I would be keen to avoid.

I have increased the 'image memory' size setting in prefs, to more than 9GB, which should be plenty I believe, but without success.

I do realise that many people suggest that scanning above 3200dpi with the Epson flatbeds won't give you additional image quality, and I do respect that oppinion. I would be keen though to scan at the highest optical resolution possible (6400) and then to reduce file size in PS - even if just to compare the results. So if you have encountered the same problem and managed to solve it, or even if you have just from the beginning been happily achieving 6400dpi at 4x5, it would be great to know your settings!

Some people might suggest going to a Mac, and I do have that option, but it would simplify things greatly for me if I could get this to work on Windows.

[Ken Lee]

I do realise that many people suggest that scanning above 3200dpi with the Epson flatbeds won't give you additional image quality

Several tests show a lower number.

You might find this article interesting: Scanning Tips with EPSON and VueScan Software (http://www.kennethleegallery.com/html/scanning/index.php#max). (The link takes you to the section which discusses resolution.)

[Jim Andrada]

Scanning at higher resolutions and downsampling averages the noise in the image so it's not a bad idea. On the other hand I'd be awfully nervous about using the 6400 DPI setting. I just don't believe it's going to give you any better result than 2400 or even 1800. Of course I could be wrong!!!!

Anyhow, maybe the best thing would be to scan the same image at 1800, 2400, 3200, 6400 and see what the difference is. Then you could see whether there's any merit to using the higher settings or whether you get higher resolution at the expense of lower contrast.

[Doug Fisher]

To accomplish what Jim said, you can scan just a small section of the negative at 6400 and should be able to get that to work. Then just compare that same area from your 3200 ppi scan. That should give you an indication whether it is worth your time to continue your efforts to make 6400 work.

Doug

[koraks]

Have you tried it with the Epson scan software? It may not be a definitive solution for you, but it'll at least allow you to try out the workflow and see if it's worthwhile pursuing.

[kaif]

Koraks, just tried this, thank you. Epson scan software doesn't let me select 6400 dpi (well, it does, but then says 'file size too large' and doesn't start the scan.) It does let me select and scan at 4800 dpi, but then it produces a file which Photoshop can't open ('unexpected end-of-file'.)

Having thought about this all, looking at those numbers, this is all likely based on a bit of a school-boy error on my end though:
An uncompressed 48bit RGB TIFF at 3200dpi 4x5 gives me file sizes of 1.2GB, a TIFF using compression is just under 1GB. So at 6400dpi, it would be four times that file size, either just above or just below 4GB... which of course is the file size limit for TIFF files!

With that knowledge: If I set custom resolution on Vuescan to 4800dpi, I do get 4800dpi scanned files which open fine, and are around 2GB.

At these levels, of course I see some grain in shadows long before I see pixels, and it does all become a pixel-peeping exercise at enormous enlargements to spot any differences at all - so I think I'll settle for that and do some processing to see if I can notice any difference at all further down the editing line between 2400, 3200 and 4800dpi scans. (And yes, I'm very much aware that the consensus is that above 2400dpi, I will likely be hard-pushed to spot any real life gains at all - while resource-hogging will be very real.)

From what I've learned during this little exercise, I assume that 8x10 users who scan on Epson flat beds would likely max-out at 2400 dpi for single image scan file sizes? Of course you'd very much benefit from the higher film resolution and less grain using only half the enlargement. So in a way, 8x10 sounds like the ideal way to get maximum detail into film-based images for scanning, at least for those of us who are not blessed with drum scanners!

Ken, thank you for your link. I was interested to read on your website that Epson Scanners work best at multiples of 360dpi. So in high res ranges, would that mean there was likely any noticable advantage in going for, say, 3600 rather than 3200, or 2520 rather than 2400 dpi?

[kaif]

...apologies Ken, of course your website does mention you don't recommend ever going significantly beyond 2400dpi with the Epsons! That might suggest 2160 (6x360) or 2520 (7x360) rather than 2400, going by your recommendation of 360-multiples - did you notice any real life differences at all, going for multiples of 360 at these ranges?

[Ken Lee]

...apologies Ken, of course your website does mention you don't recommend ever going significantly beyond 2400dpi with the Epsons! That might suggest 2160 (6x360) or 2520 (7x360) rather than 2400, going by your recommendation of 360-multiples - did you notice any real life differences at all, going for multiples of 360 at these ranges?

The widely circulated notion is that Epson printers work in multiples of 360 dpi. I never read that about their scanners and would be surprised if that were the case. I suspect that resolution is limited by the quality of taking lens and the ability of the stepper motor to move the scanner in tiny regular increments. After all, this form of scanning is really an exercise in precision stitching. Resolution is often higher in one direction than the other :rolleyes:

If you look at Nathan Potter's graph (http://www.kennethleegallery.com/images/scanning/Epson750MTFNathanPotter.jpg) you can see that the trend line is fairly smooth. Any slight bumps in the numbers with respect to a 360 dot interval (or any interval for that matter) are probably determined by the stepper motor and the software which does the stitching, and are probably below our ability to measure. I have never tried.

I think the core problem here is that we are struggling to squeeze the last few % out of our equipment, either because it's an adventure to try, or because we are reluctant to upgrade due to cost and inconvenience. We can see a minor improvement with b&w film by scanning the Green channel only (on Epson flatbeds) but if we simply jump to a larger format - or use a better scanner - we will obtain a greater increase in image quality, with far less effort.

[Pere Casals]

I do realise that many people suggest that scanning above 3200dpi with the Epson flatbeds won't give you additional image quality


One thing is a shot in the lab, another thing it is a shot in the field. I would be nice to know what % of LF shots have more than 2300 "real dpi equivalent" in film.

So it depends on the particular shot and film, if shot has some little blur from shake, lens, focus, diffraction etc a higher dpi can be worthless. Also it depends on film, Adox CMS 20 or TMX are sharp films, but it also depends on image microcontrast in textures and silhouette contrast of subjects. TMX resolves 50Lp/mm in low contrast conditions, but 200Lp/mm in high contrast (1:1000) patterns, in the first case a LF lens can be better than film.

What I say is that depending on the resolving power the negative has it may be useful or not going to 6400 dpi with V800. In general LF negatives are not resolving a lot of Lp/mm, compared to smaller formats, but this depends on your shooting style. Just inspect the negative with a x30 loupe and compare with your scans.

What's about V800 recommended dpi, horizontal bars and vert bars of a USAF 1951 target have different limits, horizontal bars do not improve beyond 3000 dpi, while vertical bars do improve slightly when scanned at 6400 dpi (from 2300 to 2600 optical), see table C2 here:

http://archivehistory.jeksite.org/chapters/appendixc.htm


It should be pointed that 4x5 sheets are scanned with the high res lens of the V800, it scans until 5.9" wide. For 8x10 negatives IMHO it is not worth going beyond 4800dpi, as low res lens is used.

I practice 4800dpi scanning may be a lot.

[Peter De Smidt]

How a scanner interacts with grain can be an important issue, whether or not more scene detail can be extracted from the film. My flatbed isn't an Epson, but scanning at higher resolutions with my scanner leads to a better grain rendition, even though I doubt my large format camera/lens system can achieve the corresponding level of detail on film. So for instance, you might scan a high resolution test target. It gives you a max dpi for your scanner. You might be tempted to scan at that setting, but while that might give you the smallest file size that'll lead to the maximum number of bars resolved on the target, it may not give you the best scan for other reasons, such as grain size...

[Pere Casals]

How a scanner interacts with grain can be an important issue, whether or not more scene detail can be extracted from the film. My flatbed isn't an Epson, but scanning at higher resolutions with my scanner leads to a better grain rendition, even though I doubt my large format camera/lens system can achieve the corresponding level of detail on film. So for instance, you might scan a high resolution test target. It gives you a max dpi for your scanner. You might be tempted to scan at that setting, but while that might give you the smallest file size that'll lead to the maximum number of bars resolved on the target, it may not give you the best scan for other reasons, such as grain size...

IMHO grain depiction can be very important for 35mm with all films, and for MF with cubic film like TX and HP5, but IMHO that interaction has little sense for LF. You have a Cezanne, IIRC, 4x5 is scanned at 2000 dpi as a maximum with the Cezanne, and sure you have no problem with that. It would be interesting to test that with the Cezanne: if there is a difference because grain depiction with normal scanning (2000dpi) compared with stitching strips (8000dpi). This may also depend on film, perhaps depending on tabular vs cubic...

Also it should be pointed that (in the past) some Nikon LS users were complaining about too much scanner resolution that was aliased with grain...

With the Cezanne "pixel size" can be adjusted because the zoom lens, it should be a good platform to see grain vs pixel aliasing...

[Peter De Smidt]

I've done those tests.... Grain can matter with LF. A friend produces large prints in a tradional darkroom from 8x10. With some films; development, and printing grain is clearly visible.

[Pere Casals]

I've done those tests.... Grain can matter with LF. A friend produces large prints in a tradional darkroom from 8x10. With some films; development, and printing grain is clearly visible.

Of course you can see TXP/HP5 grain in a x8 enlargement at reading distance, but x8 from 8x10 this is a 2m print, normally 2m prints are not seen at 1 foot, but from 1m far at least, so grain structure cannot be seen, just tonality, for "normal" viewing distance. A x5 enlargement (1m from 8x10) from TMX/D100 won't show grain at reading distance.

IMHO if one wants a good grain depiction, better to use an enlarger.

[Peter De Smidt]

Have you ever seen a photo show in a museum (or a gallery) were people don't get as close as they can to the print? I've seen my friend's large prints in a number of shows. In each case I looked at the print both from a few steps away and up close. Looking around the museum, I wasn't alone in that practice.

Back to scanning. Make tests at a number of different resolutions. Decide on the largest size you could conceivably print. Then make some test prints at size, or, at least, make 8x10 portions of images such that they correspond to how big that area would be if the whole print was made. Put the prints up in good light, and judge for yourself. Even better: Do a double blind test and have a bunch of people look at the prints.... Ok. That's a fair amount of work. But it's much less than having to go back and scan, spot, and process your files more than once because you later realized that you could've gotten a better file scanning in a different way.

[Pere Casals]

Have you ever seen a photo show in a museum (or a gallery) were people don't get as close as they can to the print? I've seen my friend's large prints in a number of shows. In each case I looked at the print both from a few steps away and up close. Looking around the museum, I wasn't alone in that practice.

Back to scanning. Make tests at a number of different resolutions. Decide on the largest size you could conceivably print. Then make some test prints at size, or, at least, make 8x10 portions of images such that they correspond to how big that area would be if the whole print was made. Put the prints up in good light, and judge for yourself. Even better: Do a double blind test and have a bunch of people look at the prints.... Ok. That's a fair amount of work. But it's much less than having to go back and scan, spot, and process your files more than once because you later realized that you could've gotten a better file scanning in a different way.

For sure it is impressive to see perfect close up detail in a 2m print... it is not essential, Rembrandt paints show strokes seen from close and this is not a problem. But it is true that depicting very fine textures in a monster print from very close distance it can be catching, still at that close distance the picture is not seen as a whole, composition and message is absent in a close inspection, but inspecting close texturized detail can also be a resource for expression.

A limitation I find is dealing with very large files, a 8x10 top qualtity scan that shapes well cubic grain (the big grains at least) may take some 7Gb, in a 16 bit file. JPG compression may be avoided. Then the file has to be sized to the output device requirement, one pixel for each dot, to be sure about the result. It is true that today's printers have very optimized algorithms for resizing, but we may want to sharpen the grains to an optimum after resizing to the destination device.

The exhibition I saw taking extreme care for grain depiction was Salgado's Genesis. There were the pre 2007 shots, made with 645 TXP 220 format (now ony TX 120 is available) directly enlarged, showing darkroom skills of the french "tireur". Then there were also the post 2007 digital shots comming from a Canon DSLR, with DXO film pack artificial grain, LVT printed on Delta 100 sheets, and enlarged in darkroom. I spend several hours comparing "silver grain" to "digital grain on silver" there.

Grain is a very delicate and challenging resource... IMHO, in general, it is overlooked in LF. Often we hear about it as a defect.

[chassis]

To me grain is analogous to brush strokes in a painting. The paint needs to get into the canvas or panel or cave wall somehow. In silver based photography, grain, grain clusters and dye clouds form the image. So to me grain may or may not be a defect. It depends on what the photographer's creative goal is.

Regarding scanning, Peter and Pere are you talking about imaging individual grains, or grain clusters/clumps? My understanding is that individual grains are at least an order of magnitude smaller than the resolutions being discussed in this thread. My understanding is likely not 100% clear on this.

[Peter De Smidt]

By "grain" I mean photographic grain, the kind that's visible in prints.

How grain is presented in the final print is very important to many of us, and that's why we take a lot of care with format, film, developing system....and print size choices.

[Pere Casals]

Me also I was thinking in clusters/clumps. To me grain structure of a film is significative footprint, this is if there is more grain in the darks greys (like TXP) or in the lighter greys (like HP5), but this aesthetic effect is way milder in LF than in smaller formats.

[IanBarber]

Me also I was thinking in clusters/clumps. To me grain structure of a film is significative footprint, this is if there is more grain in the darks greys (like TXP) or in the lighter greys (like HP5), but this aesthetic effect is way milder in LF than in smaller formats.

What is your workflow for scanning 4x5 Pere to get the most out of the scan and how do you do your selective sharpening as not to destroy the grain

[kaif]

Just a slightly different thought re maximum scan resolution: Of course, for a specific project which will get printed relatively soon, it makes sense to work out how large you are going to print - and then work out a sensible scanning resolution based on that. If you later want to work on bigger prints or crops, you could re-scan as necessary.

But then, at least at C-41, Fuji for example gives an average live of 10 to 20 years for their negatives without any loss of quality, if I remember this rightly. So I imagine there is at least some argument for scanning at least colour negatives at the highest resolution you can achieve when you first archive them. (Which is really how I came to ask this question in the first place...) Of course that's somewhat different for b/w, where you'd hope your negatives have a much better live span.

[Pere Casals]

Just a slightly different thought re maximum scan resolution: Of course, for a specific project which will get printed relatively soon, it makes sense to work out how large you are going to print - and then work out a sensible scanning resolution based on that. If you later want to work on bigger prints or crops, you could re-scan as necessary.

But then, at least at C-41, Fuji for example gives an average live of 10 to 20 years for their negatives without any loss of quality, if I remember this rightly. So I imagine there is at least some argument for scanning at least colour negatives at the highest resolution you can achieve when you first archive them. (Which is really how I came to ask this question in the first place...) Of course that's somewhat different for b/w, where you'd hope your negatives have a much better live span.


Important color images can be preserved with color separation technique, this is making 3 contact copies on BW copy film, each copy made with a different basic light color, I guess that depending on if it is a positive or negative image CMY or RGB can be used.

In fact Hollywood color movies are even today stored in this way, including movies shot digitally. Kodak 2237 and 2238 films are used for this: "This black & white recorder film is intended for making archival separations from color digital masters", Kodak says.

For color sheets that technique is also possible with regular fine grained film. Another way is refrigeration (40°F (4.4°C) and 27% RH, for example), that slows chemical changes, and I guess that concurrent inert gas storage may be also good.

[Pere Casals]

What is your workflow for scanning 4x5 Pere to get the most out of the scan and how do you do your selective sharpening as not to destroy the grain

Hello Ian,

I'm not concerned about grain with LF shots, to me it is irrelevant for the up to the 1m prints. Presently I'm mostly using TMX for 45 and HP5 for 810 and Xtol, so I'm not able to see grains at all. IMHO a 300ppi printer is not able to print the grain, nor the human eye would see it.

If you print digitally from 45 TXP, phushed N+2 with Rodinal, enlarged to 2m, and viewed from 1 foot... then you may want to depict grain structure. But printing well all grain shapes from a 45 sheet this requires a very big file, really.

A workflow to conserve (more or less) grain structure would be that:

> Scan at max resolving power, 16 bits for BW. No jpg, always TIFF.

> Make a crop from interesting areas, mostly those that are smooth, as areas with microcontrast hide the grain.

> Make tests reducing image size of the crops and inspect at what reduction there is a perceptible degradation of the grain shapes.

> Also with the reduced crop, use unsharp masking (or other sharpening tool) and test what radius and % will give the best results for grain structure.

> Take a reduced image of the full image to work in it with convenience to make get prototype image working in tonality, etc; record operations.

> Reduce the full raw scan file to the size ratio you found you had no perceptible degradation of grain structure.

> Apply edition (tonality, etc) from prototype image.

> Apply best sharpening settings you found for grain structure.

You have the TIFF output file !!!



But... what happens if you want to print that ????? something has to be done different !!!

Well, instead using the "good image size" that has no perceptible loss, first you have to consider how many pixels you are to print. 150ppi are more than good, 300ppi is excellence.


> If you print for example 40x50" at 300ppi you have to resize your file to 12000x15000, stay at that image size, one pixel for each printer dot, (16 bits, no jpg compression: use tiff).

> Then after aesthetic edition (tonality, etc) you may sharpen with best radius and % to have nice grain, find best setting with a crop of an smooth area, for convenience.

> Then you will see what printer will be able show from silver.


Note, with photoshop, for all image reductions, in the Image Size dialog set (combo box at bottom) "Bicubic, ideal for reductions".

This is my normal way of editing scans for grain depiction... I don't know if it is the best way.

Regards

[Jim Andrada]

I recently scanned color negatives taken in 1973. The results were every bit as good as the prints I made at the time. I couldn't see any obvious sign of deterioration. Of course it wasn't Fujifilm either.

By the way. I think the negs were processed as C-41. Which had just replaced the earlier process (C-22???) around 1972.

[Peter De Smidt]

Ctein found increases in quality sending his Epson printer more than a file at 360dpi. If I remember correctly, he saw an perceptual increase in the quality of prints up to about 480dpi, all on glossy paper, of course. I usually send a 720dpi file to my Epson from my 8x10 scans.

[Pere Casals]

Well, there is a mess around: PPI vs DPI...

Speaking about printing and graphic arts jergon, one can say that 1 PPI resolves the equivalent of 4 to 10 DPI. The Durst Theta 76 specs claimed that at 254 PPI detail is as good as a 1200 DPI inkjet.



"DPI - Dots Per Inch. This term has multiple definitions. For printers, it is usually the number of physical dots that can be printed on an inch of paper. Confusion arises when trying to relate printer dots to image pixels. Most printers require a matrix of dots to recreate one image pixel. That’s why printer specifications give resolution as two values, such as 4800 x 2400 DPI for Canon printers. Such printers have a resolution that photos are printed at, which is sometimes called the Native resolution or Photo resolution, given in PPI. For Canon it’s 600 PPI and for Epson it’s 720PPI. When printing images from these printers, the native PPI values are the only numbers that are meaningful."




I usually send a 720dpi file to my Epson from my 8x10 scans.

It is a good choice, 720 dpi may resolve something like 150 or 300 continous tone PPI, a good print.





Ctein found increases in quality sending his Epson printer more than a file at 360dpi. If I remember correctly, he saw an perceptual increase in the quality of prints up to about 480dpi, all on glossy paper, of course.



It is well accepted that human eye see nothing beyond 6 Lp/mm, this should be some 12 pixels per mm, so 300 PPI. (Some would call it DPI, but this is confusing). So we are observing 0.08mm pixelization, pixels are sized like the cross section of fine human hair.


In reality paper can deliver aprox 1:100 contrast, but not 1:1000 contrast. But worse than that, most pictorial textures do deliver much less than 1:100 contrast, perhaps 1:2 or 1:3. In those conditions 150 PPI may be not worse than 300 PPI.

With eye it happens something like than with film. TMX resolves 50 lp/mm at 1:1.6 contrast, but 200 Lp/mm (Kodak says) with 1:1000 contrast. So eye also resolves more with 1:1000 contrast, but paper is 1:100.

7Lp/mm is an absolute maximum for very, very good sight that few pople has, concurrently with very contrasty patterns that are not common in pictorial images.

[Jan Becket]

I scan 4X5 at 6400 on my V700 (in Scan Science oil on a Better Scanning holder), but then choose 2 as the TIFF file size reduction in Vuescan Output. The resulting file is 3200 dpi. I could probably be doing exactly the same thing at 4800 and see no difference, so 6400 is probably overkill.

[Jan Becket]

Several tests show a lower number.

You might find this article interesting: Scanning Tips with EPSON and VueScan Software (http://www.kennethleegallery.com/html/scanning/index.php#max). (The link takes you to the section which discusses resolution.)

Excellent article, Ken.

[Ken Sinclair]

Earlier this year I decided to invest in an Epson 870, using View Scan
As yet I have not scanned any of my 'older' 35mm or 120 films, since the majority of my scans are from 4x5 inch negatives for printing onto Pictorico for the 'archaic' non-silver processes. I have not yet, (anyway) had the need for such high resolution scans from any of my 'pile of negatives' that have all been processed in Pyrocat HD.

Ken