Scanner comparison: Epson V750 Pro added

[Don Hutton]

Matt

I've seen plenty of very sub-par drum scans...

[sanking]

Point that Michael Mutmansky made earlier.

Let's strip away the restraints and let the big boys do what they can do. And let's have a look at the raw scan , before someone presents it through the prism of a print on paper.

We could do that within the parameters of the scanner comparison on LF Forum, if we don't limit sample resolution.



Sandy King

Another thought is that this pretty much IS a flatbed vs. drum scanner comparison going on and so each flatbed should be used to its full potential. I really want to see the best that each flatbed can do and I am not personally interested in fair practices used between them. Mutli-pass and super sampling included.

[Paul Kierstead]

Yeah, but you just know it would turn into the kind of pixel peeping that digital is plagued by, where the end goal gets totally lost. Mind you, it would still be interesting.

[Michael Mutmansky]

The Falcon scans are mine, and from a time when I had just gotten the scanner. Since then, I've learned to get sharper scans out it it through a little defocusing (I presume the drum is a little out of spec, or the focus assembly needs a little adjustment). I won't comment on the other scans specifically, since they aren't mine.

As I recall, since this wasn't scanned at the maximum setting, I had to decide whether to scan at a larger aperture and/or defocus a little. I's been too long to recell exactly what I did, but I think I defocused a little to provide a smoother scan with less grain in the final scan. That compromises the sharpness a little in favor of a smoothness in the image. For B&W scans, I prefer to see the grain mostly, so I favor sharpness (grain imaging) over smoothness. Otherwise, I wouldn't be shooting TXT all the time...

However, the other points (about clipping and other serious artifacts in the shadow zones) that seem to show up on some of the scans are mostly a case of taking care when setting the B&W points on the scan to ensure that absolutely no clipping is occurring. I suppose on some of the drums it may be an issue of a lower DMAX capability, but if the scan were set correctly, the black areas should still ultimately be black, and they don't appear to be that way on many of the scans.

The M1 IS a very nice scanner. I have one, and have printed images from 4x5 film up to about 80" long with it. They hold up pretty darn well considering the reproduction ratio. I think it will beat the V750 in most respects, especially considering it is much easier to wet mount on it (and wet mounting produces MUCH better scans).

But to me only the Optronics Colorgetter Falcon looks impressive. The Tango
and Howtek 7500 on the other hand look rubbish. The Tango does get
*something* from the dark areas but I am not sure I would want that clipped
mess in my scan.
From the cheap Flatbeds the Artixscan M1 looks very nice to me. Smooth tonality
down to the deepest shadows and very consistent (obviously not as sharp).

Can anyone with drum scan experience explain how to interpret these examples?
The Howtek 7500 even has artifacts that look like stuck hotpixels on a CCD.



Joseph (jb7): I will write an explanation and post it tomorrow. Its very easy.

[Michael Mutmansky]

I've had large portions of several 4x5's scanned at 8000 SPI on the Aztek drum scanner. The files were on the order of 2+ gig. The resultant images were enlarged 75x to 100x and the darn things looked great. I was originally planning to shoot the subjects with an 8x10 (which would have produced a MUCH better end result), but access to shoot the subjects became an issue, so I was 'stuck' with the 4x5 images I had originally shot.

Regardless, the scans were fantastic, and they completely held all of the information in the film, as they were resolving the grain. You really can't get much better than that.

The files ended up being in the 8+ gig range once I had layers going. Thank goodness for the 16 gig capability of the Mac Pros.

My point is that as soon as you start to crop, 2400 is a bit meager for scans that need a decent size final result (say, 16x20 or so). As it is, we know that the difference between the consumer flatbeds and a drum scan or high end flatbed is really visible at 4x, and is visible but not meaningful at 3x. Whether it's meaningful at 4x is up to debate.

In defense of 2400 spi, if that is the typical required resolution then the ability to do higher resolution is not better (in the larger sense of the scanner as a whole), just higher; it could only be better if it served some purpose. The rare outlier where much higher resolution is required should not be the determining criteria. In any case it is extremely clear from the scans that those who perform better at 4800 also perform better -- and significantly so -- at 2400. If this were not the case then their "better" resolution at 4800 would be rather suspect, since things such as resolution should taper off by a curve, not a wall.

[harrykauf]

Thanks for the explanation, Michael.

[Paul Kierstead]

I've had large portions of several 4x5's scanned at 8000 SPI on the Aztek drum scanner. The files were on the order of 2+ gig. The resultant images were enlarged 75x to 100x and the darn things looked great.

I think you'll agree: that is not representative of normal use. Or even a deviation or two from it; that is an outlier. It is an interesting one (and I've love to see the monster print), but is rare.

My point is that as soon as you start to crop, 2400 is a bit meager for scans that need a decent size final result (say, 16x20 or so).

Assuming 300 dpi output, 2400 spi will let a 2"x2.5" neg print at 16"x20". That is a 1/4 (using 1/4 of the original neg) in 4x5. If your cropping much more then that, you probably should have used a different lens or MF, because you're likely to get some issues outside of scanning ones.

As it is, we know that the difference between the consumer flatbeds and a drum scan or high end flatbed is really visible at 4x, and is visible but not meaningful at 3x. Whether it's meaningful at 4x is up to debate.

I object to tossing "high end flat bed" and "drum scan" into the same bin. The samples would tend to strongly differentiate them. And if we all agreed that it was really visible at 4x, there would be about 1/10 the messages on the topic

The problem with "let the dogs loose" is that it will, indeed, answer "the question" of which scanner can scan with the absolute most resolution and DMAX. The problem is that "the question" isn't terribly relevant or useful in most cases. What I like about the current comparison is that it answers a much more relevant question w.r.t. to common maximum print sizes (basically what you could have printed on something like a Epson 9800). Actually, I also like that they didn't wet mount etc. the flatbeds to get "the best" out of them; I'd like to see what they can do under workaday conditions. We all know (or should know) that most of those samples could be improved if need be.

One of the things I like about LF shooters is that they tend to be pragmatic in their approach to achieving output (well, outside of the whole huge camera thing).

I am staying entirely in the realm of LF here. I use a Nikon 5000 for 35mm and wouldn't dream of using my epson scanner for that. I previously used a Nikon 8000 for MF before I gave it up.

[Michael Mutmansky]

Paul,

I was trying to represent that there are other purposes to scanning other than the traditional full-frame 4x enlargement that most people are thinking of when this debate comes up. I know that my examples are not typical, but they are very representative of things that are being done wither smaller portions of film on the scanners.

I print at 360 PPI so I think in those terms. Regardless, it's painfully obvious at 6.6X (for 360 PPI) or even more so at 8X (for 300 PPI) that the performance of the consumer scanners has been left in the dust by the better scanners. Anyone who does not see the differences at that point are not looking with a genuinely critical eye.

Whether the difference in performance is meaningful or of value is wholly the decision of the photographer and what then demand in their imagery.

So while 2400 is just above the performance of all the consumer flatbeds, it is well below the performance of the 'professional' products. In my mind, this stacks the deck in favor of the consumer products when the differences are really considerably greater that is percieved in these comparisons.

There really won't be much of a debate if the comparison were at a higher resolution. There also wouldn't be much of a debate if resolution targets and step tablets were employes as well, because that kind of scientific approach eliminates most of the variability. ;-)

---Michael

[Paul Kierstead]

There really won't be much of a debate if the comparison were at a higher resolution.

Actually, my whole point was that there is a debate; that debate about whether higher resolution is all that meaningful. Clearly it sometimes is, but sending out a neg for a drum scan once a year is vastly different then buying a scanner for everyday use. I don't see how a "which scanner is best regardless of whether you need its capabilities or not" debate is useful. The other part of the debate is about how those capabilities translate into prints. As to a debate as to which is sharper at a very very fine level, anyone with eyes can tell that the drum scan is sharper. The question is where does that start to matter.

As I illustrated, 2400 spi will let you crop to 1/4 of a 4x5 neg and still achieve your 16x20 output at 300 dpi. 360 dpi would not be much different. I think the crop thing is a distraction; unless you severely crop, it doesn't make much difference, and if you are severely cropping then we aren't really talking LF anymore.

[Lenny Eiger]

Actually, my whole point was that there is a debate; that debate about whether higher resolution is all that meaningful. Clearly it sometimes is, but sending out a neg for a drum scan once a year is vastly different then buying a scanner for everyday use.

The "meaningful" semantic just means that it is a personal value. We all have our own ideas of what we want a print to look like. I think everyone who wants to understand where the lines are ought to send a piece of film to a serious scanner operator, not a lab or NancyScans, or scans from India. Pay the $100-$200 to get a great scan, then they can see for themselves.

There are endless conversations about resolution. Resolution is only part of the picture. We aren't shooting test targets for our images. One is not going to get the answer without trying it, or at least looking at someone's work, in person, who has (and it would help if this person was an excellent printer in the same style).

I don't see how a "which scanner is best regardless of whether you need its capabilities or not" debate is useful. The other part of the debate is about how those capabilities translate into prints.

Ok, so you want to know where the base line is. Sounds reasonable. But can you tell me when a Sironar S lens is worth it? How many of us have a lens like this, regardless of whether we need it every time for "everyday use"? How many of us want a Leica lens for smaller formats, or a lens like a Planar 80, or some of the Mamiya 7 lenses? For that matter, can you tell me when a view camera is "necessary". For some reason, we will all go out and get a beautifully made, exquisite, inspirational and expensive piece of equipment for a camera and lenses, and yet want to buy a, at best, mediocre scanner and "get it work". I do appreciate that finances comes into it... especially after getting that fancy camera. Some of have Ebony's and some Tachikara's but we all seem to get absolutely the best one we can afford.

The game has changed. If one wants to print digitally, which IMO has great benefits, then there is that one more piece of equipment required to make it all work. It is, by nature, part of the capture step. It's as important as film choice, developer choice, and the camera and lenses that one chooses. Any weak link in any part of this chain will have an effect.

Personally, I am far more concerned with the sensitivity of the scanning mechanism than I am with absolute sharpness. (Altho' the 750 is way below my idea what what minimal sharpness is.) I am currently running tests to see how much midtone separation I can get from different films and developers. I usually print over 720 for my own work as I can see the difference in black and white...with b&w inks. And it "means" something to me.

Lenny