What are the film scanner options in 2017?

[Sasquatchian]

I'm pretty sure you're going to see less flare (at the micro level, not the huge overall flare veiling a large part of the frame) from the drum scan, as long as your lens is clean and you're autofocus is working correctly. It should also be sharper overall given the way the optical system on the Howtek works and the fact that the lens on the Howtek is only recording a very small area of the film as it passes by the lens and does not have to worry a bit about whether it's sharp at the corners. I've never used the older 4000 scanners from Howtek, only the 4500, 7500 and 8000 and I don't know if the 4000 has the lens on end of the fiber optics cable that focuses the light into the actual scanner lens that the 4500 has but was removed from the 8000. The 8000 definitely benefits from opaque taping outside the film right up to the edges of the film, which will remove any tendency to see flare coming in from the clear drum area. The 4500 only does this to a very minor degree. It might be a good idea to tape off the film on your scanner just in case. I used to use that expensive metallic slide masking mylar tape but when that got harder to get and more expensive, I just started using simple, cheap, black electrical tape, taped to the outside of the C-42 overlay material. Simple, cheap, and very effective. Phil Lippincott once gave me a long convoluted explanation for the flare and claimed that it was caused by Trident using "log" mode to scan with. He was wrong. It wasn't and his software was actually worse in that regard when I did a side by side comparisons between DPL and Trident, much to his chagrin. Having Evan do a custom alignment of the optical path of the scanner did make a small visible improvement, so that was worth it. If we can keep these scanners going another fifteen or twenty years, I'll be happy, or perhaps, dead. Ha!

[Rich14]

I've added two files to the Dropbox folder, https://www.dropbox.com/sh/flir6e4j7...w9q9ljyPa?dl=0

They are Panorama2.tif and Panorama3.tif

Here's a low resolution jpeg of these images:



Panorama2.tif is a 4000 dpi Howtek D4000, oil-mounted drum scan of Pali's Velvia image. It's a ridiculously large file. So large that I couldn't save it with any layers as either a tif or Photoshop file.

So I downsampled to 2000 dpi (Panorama3.tif) and added a monitor (output) sharpening layer and a layer to slightly lift the shadows. You can drag these layers to the larger file (recommended).

These scans have less than 1% of the dust seen in my D800E scan, but I didn't do any spotting, so there may still be some.

The color in the D800E scan is much closer to the original transparency. Although the D800 scan is a little under-saturated. Reds are exaggerated everywhere in the drum scan. Honestly it is very hard to see the actual color in the deep shadows with a loupe on the light table or with a hot light. The shadows are just too dense. The drum scanner is digging detail out there and a combination of the Velvia color curves going haywire in the underexposure and the PMTs operating at full bias are resulting in color distortion.

Detail is significantly better reproduced in the D800E scan. Look at the little white sign on the water's edge below the trees. I may have slightly blown it out in the camera scan, but it is sharper and the tree trunks and branches on the shoreline are far more three-dimensional and detailed. They are flat in the drum scan by comparison. Their color and the color of leaves and other debris on the shoreline in the camera scan are more faithful to the transparency and significantly more life-like.

I will do one last effort - an oil mounted D800E scan. I still haven't figured out what is giving me the flare on the sides of the image. I've got to get that under control. It's only happening with this image.

I'm surprised and delighted that my camera/lens combination is optically outperforming the optical bench in the scanner. Going into this effort, i doubted that would be the case. When I first saw that months ago, it was the biggest surprise of this entire exercise. And I think I can better the performance of the Micro Nikkor 105mm AF 2.8D with one of the APO lenses I have. That will be some time away, though.

This whole effort was only going to compare the "DMax" capability of the D800E vs a Tango scan. I've put a lot of time into this. After I post the oil-mounted D800 scan, I'm going to give this topic a rest for a while. I'll leave all the files in place for a while. Feel free to download them for analysis or comparison.

I'm very eager to see more scans of this image which I'll be mailing back to Pali shortly. Please feel free to comment on any aspect of this whole exercise.

Rich

[Pali K]

Rich, thank you for continuing your efforts with this image. Download and take a look at the Tango image I posted in my previous reply for your comparison as well. What I like about this exercise is that everyone gets to see the output from various methods and can make their own conclusions.

I would contribute by scanning on a Eversmart Supreme and a Imocan if someone handles the dynamics of sending the original around.

Bob, please PM me your address if you are still interested to give this image a go. I'll mail it to you once I have had the change to scan it on my ES Pro and Tango (Wet Mount) again.

Pali

[Sasquatchian]

I downloaded the 4000 ppi drum scan. Quick question: Did you slightly rotate that either in Trident (I'm assuming that's what you're using) or later in Photoshop? I'm asking because your scan is showing telltale signs of a bug in Photoshop's resampling math when it comes to image rotation. I discovered this and traced it to the method of interpolation used in the rotation. You need to use Bicubic Smoother and nothing else and preferably do it with Free Transform where you can conveniently choose your type of interpolation in the options bar. What I'm seeing are the large regular soft cross hatching that is the result of rotating with the wrong interpolation.

Secondly, how has your scanner been profiled and why are the blacks so red? How did you get from Scanner RGB to ProPhoto? Was that in your scanning software or in Ps? What is your current scanner input profile?

Well at least the skies aren't clipped anymore.

[Rich14]

I downloaded the 4000 ppi drum scan. Quick question: Did you slightly rotate that either in Trident (I'm assuming that's what you're using) or later in Photoshop? I'm asking because your scan is showing telltale signs of a bug in Photoshop's resampling math when it comes to image rotation. I discovered this and traced it to the method of interpolation used in the rotation. You need to use Bicubic Smoother and nothing else and preferably do it with Free Transform where you can conveniently choose your type of interpolation in the options bar. What I'm seeing are the large regular soft cross hatching that is the result of rotating with the wrong interpolation.

Secondly, how has your scanner been profiled and why are the blacks so red? How did you get from Scanner RGB to ProPhoto? Was that in your scanning software or in Ps? What is your current scanner input profile?

Well at least the skies aren't clipped anymore.

The scan was done with Silverfast AI 6.

I was very surprised at the excessive red also. I commented on that earlier. The shadows are indeed quite red, but not this exaggerated.

I can't remember which software I used for the ICC profiling. I used both a Kodak IT8 4x5 and a 35mm Velvia from Silverfast. I haven't profiled in a while, but no other scans show any kind of color shift. I don't rely solely on a profile to produce scans. The final scan is specified by the customer to meet one of a number of criteria. Either to look essentially the same as the transparency when viewed by daylight (very common request) or to match a piece of artwork that has been captured on the transparency, or balanced for skin colors or a particular product. The image space is specified by the profile in the scanning, not in PS.

I would have color balanced this one to appearance by daylight, but it's so dense, that I really can't see the color well. As I said, the D800 scan is much closer to the visual appearance of the film. Pali did a dry mount scan which appears on page 3 of this thread. It also has a lot of red.

I rotated the image 90 degrees in Photoshop CS6. No other rotation applied anywhere. I've not been aware of a rotation bug in PS. The only interpolation was downsampling the 2000 dpi image, Panorama3.tif.

I realize that every aspect of the scan is up for scrutiny. But I was interested only in the ability to see into the shadows in these scans, As I mentioned in my first long post after I had finished the D800E scan, I consider this image useless for reproduction, and any scanning attempt is a "rescue effort" that is intended to dig data out of the deep murky shadows at the expense of other image qualities. Without question, if requested to print this image, I would work on the color balance. But a drum scanner is not a magical device that can turn any kind of image into a thing of beauty.

[Rich14]

Correction:

I rotated the image in Photoshop CS2. The scanner software is Power PC software. I run it under Mac OS X 10.6 Snow Leopard. The scanner writes a .tif file which I then usually bring into PS6 running under Mac OS X 10.12.6 Sierra. I have PS CS2 on the Snow Leopard partition and I opened the file there and rotated it and saved it before bringing it over to the Sierra partition and PS CS6. One single 90 degree rotation.

[Rich14]

I've put one more file in the Drop Box folder. A 2000 dpi scan that I've brought into Photoshop CS6 only and rotated 90 degrees. Let me know if anyone sees artifacts of the rotation.

And how come no one noticed the "scan lines" in the previous scans? They're more noticeable in this one.

Fixed now. A recalcitrant power supply lug that apparently vibrated loose from its tab and just needed to be reseated and tightened.

Part of the care and feeding of these beasts is to periodically open them up and reseat every connector and chip in the machine. After all, it's a 27 year old computer. And although it's built like a tank electrical connectors that are not soldered in place don't stay perfectly in contact forever. Even soldered joints can eventually cause problems.

[Sasquatchian]

I believe there were also some electrical improvements to the later 4000 scanners, so if you have an early 4000, that could be an issue as well. Never any kind of fan of Silverfast, but depending on what you're using as a scanner input profile, that could indeed be causing a loss of shadow detail, particularly if you didn't use an opaque patch for your 0,0,0 in the target. If you don't do that, the input profile can clip the very bottom end of the shadows. So the target matter, the software used to generate the target and the scanning software too - all matter in getting the most out of the scanner. Even if you think you're not using an input profile, you have to be somewhere even if it's not apparent. The other thing to consider, and I have no way of knowing the history of your scanner, is the condition of the pmt's and whether they need replacing or not.

If you "just" rotated ninety degrees exactly, that will not cause an issue with the furrowing but if you're rotating even a fraction of a degree, either in Ps or SF, you could be bitten. After I get back from Monterey in the middle of the month I'll put together some samples for Chris Cox at Adobe to give him something meaningful to re-code, but you have to be able to show him what to look for and how to reliably repeat the problem. We used to see that cross hatching back in the mid 90's on certain scans we used to get from Howtek's back then, and then, when I started working with Trident and giving John at Colorbyte feedback, we pretty much attributed the effect to a bug in the bicubic interpolation routines within Trident which was avoided by scanning only at hardware optical resolutions.

[Rich14]

I have used every version of Polaris and Trident software Howtek made available. I never want to see them again. Why would you "assume" that's what I'm using, never having communicated before? And why do you assume I don't know anything about scanning?

You're describing problems I left behind twenty years ago. How strange to have that software brought up now. Ancient history.

Yes, I am familiar with the mangling that Polaris/Trident did to image data. Specifically the rotation bug. And that software's endearing trait of throwing away data when downsampling, either to an odd value or an exact divisor of the original.

The only use I had for Trident was as control software to physically run the scanner to get a file into the computer, keeping that software's image editing controls as far away from the file as possible, until I could get it into Photoshop. Howtek made great hardware. But the software they made available was truly painful through the early and mid 90s and not much better after that.

I thought you were referring to a problem in Photoshop. Do you see any artifacts in the latest 2000 dpi image that I rotated in PS CS6?

Silverfast is the best software I have ever used with Howtek machines. I have no experience with it otherwise.

I am quite familiar with Don Hutchinson's advice to apply opaque tape to the IT8 black target to avoid crushing blacks with the resulting ICC profile. In fact, my scans of my targets have his term "stretched" in their title to indicate that they were made with the opaque tape in place.

Silverfast references a ProPhoto ICM file, I installed for that purpose, via Apple ColorSync during the scanning process to output the scanfile in ProPhoto color space.

My intent here was not to get involved with critique of my scanners or their calibration. I am very familiar with their operation and construction, their mechanics, electronics and optics. Drum scanners are becoming extinct. The effort to keep them alive becomes more difficult with each passing month. And each of us who owns these machines knows that it is not a matter of "if," but "when" we will have powered them up for the last time because some major or minor part has finally failed and there are simply no replacements anywhere, anymore. Has anyone tried to get a replacement drum in recent memory? Aztek got out of the drum scanner business for many reasons, but being unable to source drums, anywhere, was right up there.

I agreed to post scans from my D800E for an honest comparison to other drum scanner operators' results with Pali's dark 4x5 so we all would have a basis for comparing a DSLR scan to a drum scan. It will be very interesting to see those scans as they become available.

Thanks for your observations. If you have some experiences with "Film scanner options in 2017" (to me options means alternatives) please chime in. But please try to spare us the name-dropping and references to irrelevant and obsolete technology.

[Rich14]

To all,

I'll have my D800 oil mounted scan up here tomorrow. Just want to tell you it is hard to oil mount to a flat surface (my light table). Real hard.

I've been a drum scanner operator for a long time, and can practically oil mount in my sleep on the curved surface of a drum. A flat surface fights you all the way. I just don't feel that the film is down tight on the surface the way securing it around the drum circumference feels. On a drum you know that film is down.

I taped one end, used a squeegee to smooth it down and taped the other end. Couldn't really tell if the oil flowed out smoothly between the film and glass. Couldn't see the leading edge of the oil film ahead of the blade like I can on a drum.

With the transparency right down on the light source, it's impossible to see any bubbles. On a drum and mounting station up above the light table, bubbles and other imperfections light up like little light bulbs. Not so right down on the deck.

Not fun.