Should I Drum Scan, X1 Flextight Scan, or use the Epson V850 w/Aztek Wet Mount Kit?

[Corran]

Since Day 1 I have said the Epson was a perfectly reasonable scanner for 4x5, but had its limitations. And as the format got smaller from there, its defects become a bit more apparent for obvious reasons. There's also more options out there for 35mm / 120. (Edited to add: the 4x enlargement mentioned by RD below is exactly where I have considered the limit of an excellent Epson scan to be.)

Many of us have used Epson scanners and have a lot of experience with them. I used to help teach students how to scan film and the photo lab was equipped with a half-dozen of them. They are also sometimes available to use at local libraries, free. Nice!

They still have their limitations and if one needs something a bit better, options exist, even if some are more troublesome than others due to aging hardware/software. I upgraded when I was in the middle of preparing a big gallery show and was making many large prints and had the budget to buy a better scanner. When printing 30 to 40 inches on the long side I found the results to be overall better than I had gotten before, in various metrics.

I'm not a test chart guy. I just don't care. I use the gear and I make photographs. Way back in 2013 though, I was bored and did a little comparison. At the time, I had a Microtek M1 scanner, which is pretty similar to an Epson but actually has autofocus and scans in a "tray" so no glass in-between the scanner lens and film. A pretty good scanner. I also had a shiny new Nikon D800E. I setup a shot and used the D800E to test this shot with studio strobes. Made a few exposures on various chromes, settled on the Provia 100F transparency. I scanned the film on that Microtek M1 and made some comparisons, which I'll leave aside for the moment.

Recently I was looking through some older film and came upon that transparency, and figured I would do a new scan on my Cezanne. I pulled up my old raw tiff files and Nikon raw image and sized it all up to comparable resolutions and then did some basic editing to match things up color-wise, at least as close as I felt like getting without serious work.

This is by NO means a scientific test. I also think chromes are the most challenging thing to scan well, and b&w would be a whole other kettle of fish.

Link

One other comment. Leaving completely aside the issue of performance, there's other reasons to prefer scanners. For one, I can scan 12 4x5 negatives at once on my scanner, or my 8x20 film on the platen. This is not insignificant from a workflow perspective. Even if someone came out with a new film scanner that blew out everything before now, I'd still probably stick with the Cezanne merely for this fact alone.

[Peter De Smidt]

Good stuff, Bryan. Thanks for posting!

[interneg]

not sure what exactly you mean by "pixel resolution" and I'm sure someone can jump in and deliver a sermon on the difference between dpi and ppi but basically someone with a 4x5 inch negative asking for a 2000 ppi scan means that after opening the scan in photoshop and going to image/image size the file dimensions are 4 by 5 inches with a resolution of 2000 ppi.

Or depending on whether it's full frame or not, approx 10,0000 pixels on the long side.

DPI = the dots per inch an output device delivers in whatever screening pattern - stochastic or otherwise. PPI input resolution equates to a particular DPI dot output, but that can vary depending on screening patterns, dot gain etc, etc. There is a tendency to confuse the two throughout the industry and (for example) Canon describe the print mode on their printers as '600dpi' in the driver, when they seem to mean 600ppi... Clear as mud, eh?


Regarding the pixel resolution, yes, it's much as you described - so with Sperdynamite's Panasonic S1R, which delivers 11168px on the short side of the sensor in the pixel shift mode, the native resolution with 4x5 filling the frame as far as possible is theoretically 2972ppi, 5x7 should be about 2233ppi, 8x10 should deliver 1396ppi. But it's not necessarily test chart resolution that matters, it's total optical performance that I'm interested in - so HP5+ in 120 would do fine, if the camera was set up as it would be used for 4x5, 5x7, 8x10, as it's only a chunk of the image at about 50% that's needed for useful comparisons - preferably covering smooth grey mid tones as the visible granularity of the film has a bell-curve distribution. In other words, the most visible granularity essentially denotes where the mid-tone of a particular film is located - which may be a distortion from notional 'reality'. I have a matched set of 1500, 1600, 2040ppi (and I think 2400 and 1200ppi, will need to check) scans I have made on a Hasselblad X5 from HP5+ from the same negs and I'd like to see how they visually relate in terms of granularity to a camera scan - ie the hypothesis I want to test is how overall resolution of the granularity and general sharpness compares - it seems to be a good quick 'image content' type of test in my experience - and one one that quickly separates the low performance optical systems (Epson) from higher performing ones.

I've yet to see any evidence of the claimed ability to resolve granularity realistically at the resolutions claimed by the Epson defenders - but if you combine two 1200ppi scans & interpolate them to 2400ppi by fairly crude means, you end up with an optical signature very similar to the Epson in terms of the rather poor visible sharpness and mushily vague granularity. Funnily, enough that seems to be essentially what the Epson is doing - combining two 1200ppi sensor lines per colour that are offset from each other by 1/2px and trying to interpolate them together.


It all comes back to a threshold not of resolution, but of acceptable sharpness of the total image, granularity and all - all of which may contradict the old-as-the-hills amateur tests of high contrast resolution alone, because it's possible for two optical systems to exist where one has higher resolution and poor MTF across the resolution range & the other system very high MTF performance yet lower absolute resolution, but because of where the threshold of visually perceived resolution falls, the second system will always perceptually outperform the first. And that's the data point where the inexperienced get visibly tripped up by their confidence outperforming their experience - because they have no experience of the differences of the systems in practical use, they dismiss the second system on the basis of an intuitive (prejudicial) reliance on poorly informed reading of data sheets and a cognitive bias towards testing that is poorly set up (or a wilful misreading of the purposes of the test objects for the same reasoning) because simplistic resolution tests are easily digested compared to the complex and highly mathematical testing that has led to an understanding of the relationship between MTF, RMS Grain, latitude on the ability of a film (or other imaging system) to record visually useful and ( more importantly) visually pleasing information - Psychophysics if you want a one-word answer.

It's the same reason why a car with a big engine, bad handling and a poor power to weight ratio is going to be outperformed in both real-world and racetrack environments by a smaller, nimbler car with a more efficient engine.

[rdenney]

Okay, folks. Let's not make the lifeguard blow the whistle and require everyone to exit the pool.

Now that I've done my moderator thing, let me return to my photographer thing.

I have an Epson V750 scanner. I have digitized slides on a Canon 5D/5DII using an Illumitran, and also using a Nikon bellows, Micro-Nikkor, and slide duplicator. I have scanned negatives in my Nikon 9000ED scanner (and in many film scanners of that type before that one, including a Nikon 8000 and a Minolta). I have constructed an apparatus for scanning 4x5 negatives in tiles using a DSLR camera back when Peter and others were exploring it. I have not yet tried using that apparatus with a Pentax 645z and 120mm macro lens, but I do own a Pentax and have done macro work with it. Based on all that, here are conclusions, in no particular order, based on my own personal experience. With a few opinions mixed in, of course.

1. The Epson does not generate more detail, dynamic range, smoothness, or noiselessness at 6400 than it does at 2400, even with many adjustments made for position.

2. I find scans from the Epson look very good when enlarged up to about four times. By "very good", I mean that they meet my requirements for sustaining the illusion of endless detail in the print. In other words, I want to be able to look at the print from any distance and never be taken out of the impression that the only reason I don't see more detail is that I'm not closer. That is, of course, an illusion, but it is conspicuous by its absence. So: I can make 16x20 prints from 4x5 negatives scanned in an Epson that meet my requirements. Transparencies are more challenging--the Epson is only able to capture all of what is in a Velvia scan because of the inherent flare in the system. At that level of enlargement, gradation seems good enough to me. I participated in the print exchange a few years ago, and compared my offering--4x5 Fuji film scanned in the Epson and printed at 11x14 Exhibition Fiber Glossy in an Epson 3800--to other prints in the exchange and did not find mine inferior in any technical dimension, including tonality and gradation.

3. Digitized slides made in the Illumitran have too much contrast for the Canon sensor. When I engage the Contrast Control Unit, the effect is too strong and the image posterizes when I try to restore contrast. I have thought about using a neutral-density gel over the CCU light source to moderate its effect, but never got to it. The copy lens on the Illumitran was...okay.

4. Digitized slides made with the Nikon bellows are better than with the Illumitran, and vastly more convenient, but contrast management is still a challenge. I think these are good for enlargements up to about 8x without losing too much tonality (or enlarging the flaws in the original slides too much). Good for a blurb book, not good enough for larger prints. But that would be true even if I was enlarging them under an enlarger onto Cibachrome. I look at my small-format 11x14 or 16x20 prints from my earlier years and sometimes cringe.

5. The Canon sensors in my 5D and 5DII are not good for this work. They lack the dynamic range to capture everything in a color transparency, though they are fine for negatives. But reversing a color negative is a challenge for me, and it takes a lot of time to get it right. When I started making photos with the Pentax 645z, I realized how much better sensors could be. It's not just that the 645 sensor is bigger--I can crop the Pentax image down to a 24x36 capture area and compare it to a Canon image of the same subject, and even at monitor resolution, the Pentax image has more subtle tonality and better color, and much better dynamic range and shadow reach. The Pentax is to the Canon what Kodachrome 25 is to Ektachrome 200 sitting next to each other on a light table, in terms of color and tonality.

6. Digitized 35mm slides are not for making large prints, but then neither were the original slides. The illusion of endless detail cannot be sustained.

7. Pentax 645z images are good for making large prints--certainly larger prints than I can make from Epson scans of 4x5 while maintaining better tonality, gradation, and sense of detail.

8. Rollfilm scans from my Nikon film scanner can be enlarged to at least a factor of 12 without losing the sense of endless detail. I have been able to scan two ends of 6x12 negatives and stitch them with alignment to the pixel level. My largest such composite was an image of a stained-glass window. The purpose of the image was to be able to print the photo onto transparent film to simulate the look of the original window at 1.5x life size--a print that would have been five feet by eight feet. I made three 6x12 exposures in my Sinar view camera, using lens and back rise/fall to avoid any misalignment between the images, and then separately scanned each end of each frame (the Nikon only supports up to 6x9). That's six images that I then merged in Photoshop. Printed at the target size (at least a test portion printed on paper), the illusion of endless detail was sustained even on close inspection. Lens was a 210mm Sinaron (Rodenstock Sironar-N). That was about a 12x enlargement, and the Nikon scan held up. And because it was a photograph of a stained-glass window, contrast was enormous and colors saturated, and I used Kodak Ektar negative film to control it. In the end, the full-size print was never made, which made me sad because I really wanted to see it.

9. I have routinely made excellent 16x20 prints from 6x7 film scanned using the Nikon--a smaller challenge. 6x7 negatives scanned in the Nikon provide better image detail than Epson scans of 4x5 when viewed at the pixel level. The Nikon scans also provide greater tonality and gradation. But both make 16x20 prints that are adequate for me. I don't compare them to chemical prints--that's like comparing oils to watercolors--but both are capable of whatever artistic expression I'm capable of.

10. I've never worked with scans made using a laser scanner. Frankly, I don't want to--it might force me to raise my standards higher than I can afford to sustain. I said as much to one well-known lab owner who offered to make one for me for free. The images comparing the Howtek to the Epson earlier in this thread were clear to me: The Howtek image showed better color and gradation, and fewer sharpening artifacts. (I was reminded of a video some guy posted on YouTube comparing his Nikon to a Pentax 645z--even at video resolution the Pentax images showed much better color and tonality, but he couldn't get his head out of the pixels and thus drew the wrong conclusion.)

11. I suspect my 645z with the 120 macro lens could make a one-shot digital image from a 4x5 negative with sufficient resolution to meet my requirements for sustaining a sense of endless detail in a 16x20 print. Technique would have to be perfect, but I think I could make it work. And the Pentax sensor has much greater dynamic range and I think could do it even with transparencies (but probably not Velvia). But I would no more get all there is to get out of a 4x5 negative than I would using the Epson to scan it.

12. It is very reasonable to suggest that the Epson is good enough for making moderate prints from large-format negatives when budget considerations are included in the analysis. It is also reasonable to assert that other technologies are better at this or that dimension, though at higher cost, space, or difficulty. This is not a religious topic. Those who state that they can only accept the very best that is possible will need to bring their checkbook. But if they have their checkbook in hand and have actually done it, those haven't who still insist they are wasting their money have no standing. If someone was paying me to make a much larger print than 16x20 from one of my 4x5 negatives, I would send it out for laser-scanning.

13. I don't use large format solely for greater detail--I generally meet my requirements for endless detail in a 16x20 print using 6x7 or medium-format digital. I mostly use large format for its image-management capabilities.

Again, my results are my results. You can apply different standards than I do, or measure image quality in different ways than I do, but you can't change my results.

Rick "what I think I know" Denney

[sanking]

Again, my results are my results. You can apply different standards than I do, or measure image quality in different ways than I do, but you can't change my results.

Rick "what I think I know" Denney

Rick,

Well said. I take my hat off to you for having the patience to write this up.

But what is a laser scanner? I never heard of that?

Sandy

[Pere Casals]

I'd like to see how they visually relate in terms of granularity to a camera scan - ie the hypothesis I want to test is how overall resolution of the granularity and general sharpness compares - it seems to be a good quick 'image content' type of test in my experience - and one one that quickly separates the low performance optical systems (Epson) from higher performing ones.

This is easily adjusted with Photoshop edition:

Look, in this side by side https://petapixel.com/2017/05/01/160...s-500-scanner/ the Epson grain looks a bit softer, a few clicks in Ps and you have exactly the same in the X5 and Epson images.

There is no doubt... a perfect match. The Epson image looks even better... see the letters on the lens... (the crop is seen like in a x12 enlargement in the monitor, so the Epson made a good job in great enlargements)

[Peter De Smidt]

Great summary, Rick. I too hadn't heard of laser scanning before.

[interneg]

It is very reasonable to suggest that the Epson is good enough for making moderate prints from large-format negatives when budget considerations are included in the analysis.

While some CMOS based imaging systems are coming down in price, an Epson isn't a bad option if you aren't making big prints and have fairly relaxed aim points (qualitatively speaking).

The problems start however when:

...he couldn't get his head out of the pixels and thus drew the wrong conclusion

And that is the fundamental problem - it's not the number of pixels, it's how well they're used in the visually important range.



P.S. I think what you meant by 'laser scanning' dates from the era when a drum scan was used to output to a laser film writer for halftone separations - initially direct from transparency on scanner to laser - Rudolf Hell's major innovation was to realise a computer for image manipulation could be put betwixt scanner and output device.

[Pere Casals]

1. The Epson does not generate more detail, dynamic range, smoothness, or noiselessness at 6400 than it does at 2400, even with many adjustments made for position.

2. I find scans from the Epson look very good when enlarged up to about four times.

Let me say my view.

Tecnically, at 6400dpi you get a 25% higher resolving power than at 2400, diminishing returns, but for sure there is a performace increase. Of course to notice it the LF shot has to be sharp enough, and the sheet has to be well in focus.


As we all know the V700 is quite sensitive to the film curling, if negative falls 1 single mm you loss half of the resolving power in the x direction:



Instead the V700 is less sensitive to a film displacement in the up direction.

New ANR glass holders with adjustable height ensure flatness and allow perfect focus.


IMO if flatness/focus is ensures then x8 enlargements are pretty good, if the film goes 1mm down in the scanning then x5 enlargements are still good.

See that in the graph resolved microns change from 7 to 12 if the film falls down 1mm from the ideal position. Film position has 1mm tolerance from 2.5mm to 3.5mm, so it can only fall 0.5mm, beyond those margins we have problems.


So IMO, being able to enlarge x4 or x8 from an Epson is about focus/flatness accuracy.

[sanking]

Per,

Some great comments at the end of that article, which I believe is about three years old.

Quote

"I have a $20 solution that will equal the quality of the Flextight scanner. Put a negative on a lightbox and shoot it with any high quality camera and a good macro lens. I've compare shots done this way to drum scans and there's no quality difference. I now 'scan' transparencies this way and the results are wonderful."


I realize now how much time we are wasting with scanning. Damn, I have a 108 mp Xiamoi cell phone that will probably give results that are just as good as my Howtek scanner?

But what can we expect? We live in a world where neither the emperors nor the peasants wear clothes, and no one notices.

Sandy