Beginner at film scanning (4" x 5") - learning resources?

[Doug Peterson]

I'm curious how you're comparing the DT system to drum scanners. I can see the possibility the DT system could capture the same highlight and shadow detail that a PMT scanner could. But what about printing from those captures? Do you think it's possible to make large prints from your system (for example, 40" x 50" from a 6x7 negative), with comparable quality to a print made from a drum scan?

We don't make "comparable" scans - we have higher quality than a drum scan on everything under 8"x10". This sounds like marketing hyperbole, but keep in mind that the last year significant R+D money went into improving drum scanning technology, Apple hadn't switched to Intel yet and Carrot Top was "big".

You'd be welcome to do a test scan the FPS Film Scanning Solution at our office or at the various conferences we bring it to (just came back from MCN and will be at IS&T Archiving in May), or we can send you sample scans. Feel free to shoot me an email so we don't further draw this thread off topic.

[jsiva]

Doug, I have seen one of your kits in person and it is indeed very impressive.  For a passing moment, I thought I'd buy myself a modified Sprinter, put one of them in the back and drive around the country doing repro work as a "Bureau on Wheels".

There also appear to be some great projects here on that system, things people would not have been able to justify from a labour/capital standpoint just a few years ago.  Well done!

[250swb]

Do you have any suggestions for "learning how to scan well"? Should I be asking myself if I want to be photoshopping and printing ginormous scan files on the current computer? Is Epson the only reasonable option short of eBay old drum scanners?

The Epson V700 is an excellent scanner for 4x5 and with a good workflow produces very good results. But the best scanning technique is really the simplest if it is for home use, make sure the black and white points aren't going to clip, and aim for a low contrast flat looking scan. This will have all the tones available and will not have compressed any. But then you do need to use Photoshop to make a likable image from it. So with little effort you can get the best scan, with more effort you can turn it into whatever you like. This is far better than scanning for a finished end result, then deciding you don't like it and having to re-scan again. But you will be looking with colour for a file size in the 2gb region before re-sizing to your final image dimensions and resolution, so your RAM needs to be sufficient. For B&W this comes right down though. You don't do or need fluid mounting with the Epson V700/V750/V800/V850 unless you are scanning larger formats than 4x5 directly from the bed, but the standard 4x5 film holders are OK and work well.

Steve

[8x10 user]

I'm having trouble believing your statement.

(1) Under idea circumstances a 4x5 can hold up to 200 megapixels of information. A photograph 60-80 mp photograph of a 200mp negative leaves a lot on the film
(2) Your system uses a bayer grid
(3) Drum scanners and high end film scanners allow one to adjust a number of settings to the signal before it is sent to the analog digital converter. A 12 stop 14 or 16 bit capture would be ok for high contrast slide films as 12bits equals a 1:4096 contrast ratio and a density range of 3.6 is a contrast ratio of 1:4000. However negative films can have a density range of less then 2. Density values are logarithmic and for everytime there is an 1.0 increase in density value the ratio grows 10x. So a density range of 1.0 has a 1:10 contrast ratio, a 2.0 s 1:100, a 3.0 is 1:1,000 and a 4.0 would be a 1:10,000, contrast ratio. So a density range of 2.0 is 1/40 the contrast ratio of a 12 stop devices range. That is 2.5% of the percentage of light  transmission that would be visible to a 12 stop camera. If we think of the numbers in terms of stops, 6 stops is a 1:64 ratio, and 7 is 128. So if you are using 6-7 stops of your 12-stop 14 or 16 bit capture then you are only using half or your data and after curves and toning adjustments your histogram is going to be skewed, combed, and posterized. As far as I know it is not possible to lower the dynamic range of the camera to allow the bits to have a finer distribution in lower contrast films.
(4) Film grain clouds do not appear in a grid but are rather stochastic in nature. A grain cloud is like a halftone pattern. If you use a sample size that is smaller then the film grain cloud then you will get a bunch of color noise from measuring individual dye particles instead of the combined effect of the particles. The best way to scan film is to set the sample area to match the grain cloud and then to oversample in two directions as the cloud can appear anywhere because it is stochastic. A good drum scanner operator knows this and will us a 8-10 micron aperture for scanning chromes even at 4000 PPI. The results are better then using the same aperture at its native resolution of ~2500PPI

We don't make "comparable" scans - we have higher quality than a drum scan on everything under 8"x10". This sounds like marketing hyperbole, but keep in mind that the last year significant R+D money went into improving drum scanning technology, Apple hadn't switched to Intel yet and Carrot Top was "big".

You'd be welcome to do a test scan the FPS Film Scanning Solution at our office or at the various conferences we bring it to (just came back from MCN and will be at IS&T Archiving in May), or we can send you sample scans. Feel free to shoot me an email so we don't further draw this thread off topic.

[Doug Peterson]

I'd be glad to address your points in fullness. Feel free to email me at dep@digitaltransitions.com so as to not further hijack this thread. Or you can send me some film for us to scan so you can compare end-results rather than theory. The real world testing of the final version we've done has been far more valuable to us than any of the research and theory we did while designing it.

One quick clarification though, for 4x5 film at very high resolution we'd recommend a two-shot stitch, yielding around 140mp (840mb TIFF) of capture, which we've found to extract more useful detail from the emulsion than a Tango drum scanner. Since our system tightly controls the Y and Z axis, and the lens has essentially zero distortion and even sharpness across the field a two shot stitch is arbitrarily easy to accomplish (scan+stitch) at a rate at least 10 times faster than scanning it in a drum scanner, before you even account for the wet-mount time.

For 8x10 one can do a four shot stitch if they feel they need more than a 840mb file. Simply do a two-shot stitch on one side of the film, and rotate the holder 180 degrees for the 3rd and 4th shot. Time to scan and stitch will be about 4 minutes with a fast computer. Again the holder and stage are manufactured to extremely tight tolerances for flatness so this is much more straightforward than you might think if you've done this only with ad-hoc systems. This results in around a 260mp file (12,840mb TIFF).

In more ordinary usage like 120 strip film or 35mm, or even most users needs for 4x5 or 8x10 the results from a single shot are extraordinary.

Scanning technology stopped being improved a decade ago - are you really surprised we've so clearly exceeded it?

I'm having trouble believing your statement.

(1) Under idea circumstances a 4x5 can hold up to 200 megapixels of information. A photograph 60-80 mp photograph of a 200mp negative leaves a lot on the film
(2) Your system uses a bayer grid
(3) Drum scanners and high end film scanners allow one to adjust a number of settings to the signal...
(4) Film grain clouds do not appear in a grid but are rather stochastic in nature. ...

[EricWHiss]

Doug,
I'm still curious to learn more about how the CH version of C1 does color conversion of color negatives (or does it at all?).   

Having used the Aptus 12 (80mp back) with very high quality lenses on my own negatives and also with the multishot and micro step backs and some high end scanners, I highly doubt your claim that the your system can produce better quality than a drum scan on anything up to an 8"x10".  I have to call you out on that - very doubtful.  I've 'scanned' thousands of slides and negatives with both the Aptus 12 and CF528 multi-shot back.  It seems clear to me that the multishot back achieve better results even for black and white negatives and this is more comparable to the scanners since they also do true color non interpolation.  However, I'm sure the digital capture process is much faster at least on capture compared to scanning - even the multishot backs over a light table are hugely faster than scanning - but the color conversion and dust / scratch spotting are a significant part of the total job. With wet mounting for scanning you can skip a lot of that work.   

What CH C1 now needs is a sensor and lighting set up that can also do multi spectrum capture so that you can really get the most information.  Also with the IR you could do the digital ICE stuff that a lot of scanners can do which saves time in processing images.

[Doug Peterson]

The CH version has a dedicated "repro negative mode" specifically for handling color negatives. We should have video tutorials out later this quarter. In the meanwhile I can give you a remote demo if you'd like.

As to the quality of scan, again - glad to have anyone send me a negative they have a reference scan for. We have done this for multiple people with several high end scanners and a tango drum scanner operated by a long term vet of drum scanning. Proof is in the pudding. Send us your pudding and we will provide the proof . Or you can email me and I'll send you the files from previous tests (I only suggest doing your own test with us so there is no doubt about the quality of the scanner file).

[EricWHiss]

Doug,
I tried to sign up for the trial of the CH version of C1…. but didn't get a download link.  Was very excited until I saw the price -  $4000?  Woah, I guess it might work for museums but not for me.   

Using the 80mp back for 4x5 film you can only realize about 2000 pixels/inch and 8x10 you can only get 1000 pixels/inch …. but hey scanners can do 4 times that or even more.  So you really want to say that it's going to beat a drum scan?  A drum scan of 8x10 could have easily 16 times the data of an 80mp back - and the 80mp backs are interpolated color - you really see the difference at the grain level.    Wasn't there a guy posting here on LuLa scans from 8x10's - he was in the scanning business doing a bit of advertising. You might want to go have a look at those and also check Tim Parkin's comparison shots of drum scanned 8x10.   If I were selling that rig, I'd pump the work flow not image quality. I still use my multishot back for digital capture even though I have the 80mp back too because I think its better.

Don't ask me why if I have both 80mp backs and micro step backs why I still shoot film - I guess I like film. :-) 

[Doug Peterson]

Eric, I understand now why you doubted my statement - I think you missed post #24 above. We are not limited to those PPIs. The system provides tightly controlled X or X/Flip-Y stitching.

So for 4x5" our practical limit is around 4500ppi. For 8x10 our practical limit is around 2300ppi. In theory we could go much higher but the practicality is strained.

We can do the practical limits numbers at a workflow rate much, much faster than drum scanning.

When scanning smaller formats like 120, or when scanning 4x5 to a 480mb TIFF a single capture works very well and of course is even faster.

I said "We don't make 'comparable' scans - we have higher quality than a drum scan on everything under 8x10" - I stand by this 100%.

[ErikKaffehr]

Hi,

I have done some experiments using duping techniques and I have little doubt that very good results are possible, under optimal conditions. Also using X-Y stitching almost any resolution is achieveable.

Regarding DR, I would say that Velvia with its density range of around 10000:1 is a challenge for CCD scanners and probably also CCD based cameras. Drum scanners are photomultiplier based, and I guess they can handle higher density ranges.

Negative film has large, but non linear, dynamic range. That DR is handled in a compressed density range and may therefore be less problematic.

Best regards
Erik

Eric, I understand now why you doubted my statement - I think you missed post #24 above. We are not limited to those PPIs. The system provides tightly controlled X or X/Flip-Y stitching.

So for 4x5" our practical limit is around 4500ppi. For 8x10 our practical limit is around 2300ppi. In theory we could go much higher but the practicality is strained.

We can do the practical limits numbers at a workflow rate much, much faster than drum scanning.

When scanning smaller formats like 120, or when scanning 4x5 to a 480mb TIFF a single capture works very well and of course is even faster.

I said "We don't make 'comparable' scans - we have higher quality than a drum scan on everything under 8x10" - I stand by this 100%.

[EricWHiss]

Doug,
I guess stitching would help, but I am still in doubt about your claims.  The other Erik has already pointed out that some drum scanners use other technology that can be superior.   I also see with my own work that the 528 mulitshot produces a superior file with about 2 stops DR advantage over the 80mp dalsa sensor and more detail as well.   You addressed the detail somewhat with stitching and I presume you could also take multiple exposures too which could be all automated.       Photomultiplier scanners may be best but the good ccd scanners have the linear sensors with 80db SNR (about 12 db more than the 80mp dalsa) and they also make a non-interpolated file.   

So many forum comparisons are made only on luminous detail, but that's only one facet of IQ.  I'm not saying your set up isn't great, but I'm just saying if I were selling it, I'd not make that claim about IQ, but instead sell the workflow advantage.   Probably the post process is faster with a good wet mounted drum scan - but then most of the work is archival and they don't have to do post on every shot right away.

[8x10 user]

The highest d-max I've seen from velvia is 3.9 -.3 for the base is 3.6.  

I did my own tests with 35mm kodachrome using a 8000 DPI drum scanner and the "best" CCD scanner made. The CCD scanner was almost as sharp as the drum scanner and had better shadow performance. The R&D behind the eversmart Supreme was huge. They spent millions perfecting the design. The CCD is actively cooled by a peliter cooler, infared is bounced away from the sensor using a cold mirror, it automatically stitches and does multiexposure, and the ADC is true 16bits.

Recently I digitized some black and white negatives using a homemade copystand and a Sinar 54H in 4-shot mode (@ iso 25). The problem was that the negative only used maybe a third of the histogram. I dont see how the result would be any different with a newer phase one back.

FYI I've owned 3 Drum Scanners and the Eversmart Supreme so I've had a lot of first hand experience.  

[8x10 user]

I said "We don't make 'comparable' scans - we have higher quality than a drum scan on everything under 8x10" - I stand by this 100%.

Care to wager one of your "scanners" on that statement?

[Doug Peterson]

Care to wager one of your "scanners" on that statement?

The only problem is when I win all I get is one of your drum scanners .

Seriously though if you'd ever like to do a test or article we'd be very glad to provide our facilities. I think we'd both enjoy the process (I have the several times I've already done such tests with clients).

[8x10 user]

The only problem is when I win all I get is one of your drum scanners .

Yeah.. I can only imagine what the price tag might be on your system. It has to be close to the sticker price of a drum scanner when they were sold new. Your system looks great for chromes and high volume work. I'm sure its worth every penny if you are to digitize thousands of slides. Now you just need to make a version with an automatic slide feeder attachment. 

[8x10 user]

I had a PM in regards to learning about drum scanner use and I thought I should make the post here for others to see.

Mounting:
There are good mounting videos on Azteks website. I slightly prefer Luminas fluid for scanning. Azteks drum cleaner and film cleaner are nice.

Aperture selection:

The size of the average grain or dye clump vary with film type, and developing. As a drum scan operator you are able to change the aperture to optimize the scan for your media as well as your personal tastes. A larger aperture will produce a smoother scan with less film grain, if your aperture selection is on the large side for your film type then you will also have a bit less detail and sharpness. If too small of an aperture is use then the image will have more noise and grain without extra details.

With color film types a huge amount of color noise (versus luminance noise, aka film grain) appears in the image when. This is known is grain aliasing and it is the result of measuring individual dye particles rather then the grain clump as a whole. Each grain clump is like a pixel that is made by a halftone of many dye particles. It is possible to digitally "descreen" these files like you would a half tone pattern but it is recommend that your resolution is 10x the grain size for descreening so this is not really practical or possible for our purposes.

Some software suites come preloaded with reference values for different film types. These are not exact and you can still adjust for your personal tastes. Some software can be a bit tricky and instead of giving you exact aperture measurements they give you a the choice between the 5 or 6 options that are closest to your native resolution. In these cases you want to remember to use a smaller selection for chrome and fine grain black and white films. A common problem with comparison scans between a tango and other scanners is that too large (default) of an aperture might be selected... One of my scanners is a Primescan.

3 Micron - Used for the very finest of black and white films, Technical Pan, ATP, CMS II, electron based scientific photography films,
high resolution surveillance film.
6 Micron - Good for other unusually fine grained films, kodachrome, technical pans, ect.
8-12 micron - High quality E-6 films, very fine grained black and white films Tmax, Delta, Acros
12-17 micron - medium speed black and white, color negative films
17 micron+ unusually grainy films.

One should do their own testing with their exact film types... Scan a small high detail area of the film at the resolution of the final scan or your scanners max. Select 4-5 aperture settings and compare the results from each scan. Remember the figure for later. This is the smallest aperture that you should use. It is recommend that you never use an aperture that is smaller then your scanning resolution. Setting a resolution that is higher then your aperture selection increases the number of details captured due to the stochastic nature of film... Those clumps of grain appear at random and not in a perfect grid.  Converting your aperture to PPI can be done by dividing 25400 (microns in an inch) by your aperture.

3 Microns = 8466 (most say 8000)
6 Microns = 4233 (most say 4000)
8 Micron = 3175 PPI
10 Micron = 2540 PPI
12 microns = 2116 PPI
14 Microns = 1814 PPI

End Points / curves:

When selecting the endpoints of your scan you want to make sure that there is no clipping so select a white and dark value that is just outside of your films range. I prefer to use curves that make the image appear slightly flat (low contrast). Contrast and endpoints can be enhanced in post using non destructive editing practices.

Calibration:

Color slide types can be ICC profiled to give you a very accurate starting point. There are some great ICC calibration guides on Don Hutchinson website. You want to make sure that your endpoints/curves are wide and stay the same for every scan. I prefer to use the Max DR method where you increase contrast slightly before running the image through the color profiling software. Profile targets should come with your scanner and are also available from Hutch color, Lasersoft, and Wolf Faust. Wolf (coloraid.de) makes profiles for the more film types then the others. Hutchcolor makes very nice targets, and lasersofts are pretty much identical to old kodak targets. If you want to be super pro you can use a color reference monitor that has been calibrated to ~130 candles, 5000k, and a gamma of 2.2. Put this in a room with 5000k lighting, neutral walls, and a 5000k adjustable brightness viewer to make comparisons between the slide, computer colors, and actual prints.

Color negative film:

Your software should include tools for correcting color negative film. These film types cannot be profiled and need custom correction each time. Correction can be done in post but you want to make sure to optimize your end points and curves so that you dont have any clipping and the majority of your bits are within the scanners range. For color negatives if it best to do this for each color channel if possible.

[ErikKaffehr]

Hi Doug,

If you have a transparency with density of 4.0 and say a good amount of sky, the light coming from the transparency can bounce back from both equipments and walls. Intensity doesn't help with that.

Another consideration I may have is flatness of film and lens aperture used. Glassless mounts are not keeping film flat.

With drum scan it is easy, as film is wet mounted on drum.

Best regards
Erik

Only sort of. It's not that the room needs to be very dark, it's that the ratio of the transmissible light to the room light must be very high. We use a very bright transmissive light to keep shutter speeds fast and allow workable dim-but-not-dark working environment. After all you can only scan hundreds of frames per hour if you can see well enough to retrieve/prep/load/unload/return the film! We also strongly gate the light to keep the operator position from being bombarded with bright light.

If naming a DIY solution with a standard grade light box as your transmissive source then I'd definitely suggest working in near complete darkness.