New article by Leigh Perry: Collaborative scanner comparison

[QT Luong]

A new article
Collaborative large format scanner comparison (http://largeformatphotography.info/scan-comparison/) by Leigh Perry from Sydney, has been posted. Please feel free to leave any constructive comments on this thread. Also, please note that Leigh is always interested in adding results from other scanners, so please contact him if you'd like to contribute a scan.

[Doug Dolde]

I'd say the Tango does the best job but that's not surprising.

[Darin Cozine]

I think the colorgetter pulled out more info in the dark areas though. I wouldnt say there is an overall best, as they have different stregnths and weaknesses. Also note that a flatbed scanner is much smaller and cheaper than a drum scanner (of course!).

All in all a great job, and an excellent reference for the community. Thank you.

[neil poulsen]

Really an excellent comparison. I noticed that, if one has a wheel on their mouse, select one of the options in the drop down and then turn the wheel with the cursor placed over the drop down. This makes is easy to compare one scanner to another.

While there was one image where it lost texture, the Heidelburg stands out the strongest for me. I was a little surprised at the Epson 3200.

[Will Strain]

I love my Epson 3200 for everyday personal work...but now I'm especially glad I get pro jobs professionally scanned or when i need to go big.

Great reference. Thanks for sharing it. I was considering upgrating to the Microtek i900...but based on that comparision, I think I'll save for something bigger later.

[Leigh Perry]

I noticed that, if one has a wheel on their mouse, select one of the options in the drop down and then turn the while with the cursor placed over the drop down. This makes is easy to compare one scanner to another.

PC users can also click on the scanner combo box (dropdown) and use the up and down arrows. That way you can keep the mouse over or away from the image so that you get the sharpened or unsharpened versions.

I plan to improve this dinky presentation, hopefully using checkboxes so that you can select the scanners you are interested in -- that will then populate the combos/dropdown with only the required scanners.

[Leigh Perry]

Forthcoming scans are:
<ul>
<li>Epsons 4870 and 4990
<li>Microtek ArtixScan 1800f
<li>Crosfield drum scanner
<li>Fuji Lanovia C550 flatbed
<li>Creo Eversmart Supreme
<li>Howtek 7500
<li>ICG 380 drum scanner
<li>Canon 9950f
</ul>
Many of these are thanks to the ScanHi-End Yahoo group, so there will be a preponderance of drum scanners and executive flatbeds. It would be nice to fill in the prosumer gap a little more.

[Matthew Cromer]

This is truly great stuff.

How long until the Canon 9950 and Epson 4990 will be added?

[Bruce Watson]

I own a ColorGetter 3 pro, so I'm not exactly an objective observer. That said...

The ColorGetter did well, as expected, especially in shadow detail. I'm surprised that the others had such a difficult time with that however, especially the Tango. I'm also surprised at the Howtek scans - it seems a bit harsh to me for some reason.

I wish such a comparison had been available when I made my choice. I had to do it blind, without any comparisons - just recommendations from people I trusted. I was lucky my referals were good. I'm glad that others can get a good look for themselves.

Good on QT for posting this - 'tis valuable information.

[Harley Goldman]

Excellent! Very interesting to see the side-by-side comparisions. Thanks very much for putting this together! I am looking forward to the new additions.

Harley

[DYMERSKY MICHAEL]

How long until the Microtek ArtixScan 1800F will be added ?

[Matthew Cromer]

For those who don't want to wait on the 4990 comparison,

Quentin Bargate has a pretty impressive comparison linked from the boards at photo i:

Comparison of Howtek Drum scan and Epson 4990 @ 2400 DPI (http://www.photo-i.co.uk/BB/viewtopic.php?t=49)

[Leigh Perry]

The additional scans will occur over the coming weeks and months.

The Epson 4870 is happening in Canada right now. The transparency will come back to me me here in Sydney for an Epson 4990 that I've organised, then off to Cairns for the Crosfield drum scanner and Fuji Lanovia C550 flatbed. The remaining scans are in the UK and USA.

Where possible, I ask the volunteer to just sent the film to the next volunteer, but it still takes at least two weeks per scan because of airmail times and operator availability.

[Edwin B.]

I’d say that’s a commendable effort and an interesting comparison. However, a few limitations of such a comparison should be pointed out at least with respect to the drum scanners. It is often claimed that for drum scans the operator is more important than the brand and for the more recent models I think there is a lot of truth to this. The first drum scan I saw was done several years ago by a well regarded pro lab using a Howtek 4500 and the results were less than exceptional with tons of noise and artifacts. I complained and was told that the original was grainy and that it was as good as it gets. Fast forward a few years and I own the same model and couldn’t be happier with the hardware. After a lot of late nights of fiddling around (about six months worth) I rescanned the same original with far better results. In the process I learned a thing or two about scanning.

With respect to sharpness the scanner’s focus is critical and the automatic setting may not be good enough. This may be where the difference between the tango and the others arises. With the Howtek I have to use manual adjustments for critical focus in which case the grain structure is clearly visible especially in the case of a film like Velvia 50. At 4000dpi the Howtek is generally capable of resolving finer detail than most real-world large-format film images contain. If the unsharpened scan is fuzzy at 2000dpi and the original is critically sharp the scanner is probably not the problem.

Other explanations are possible. The scanning resolution is not very meaningful without also controlling for aperture. The aperture really determines the resolution of the image signal (the area of the image that is viewed by the scanner at any given time) and the scanning resolution then determines the rate at which that signal is sampled. Thus, a scan with a relatively large physical aperture and a high scanning resolution results in an over-sampled lower resolution image. On the other hand, a scan with a relatively small physical aperture and low scanning resolution results in under-sampling with aliasing and a noisy image. Both parameters need to be controlled or statements like “scanned at X dpi” can be misleading. I won’t even begin to go into the issues arising from resampling all of the scans from optical resolutions to 2400 dpi.

Shadow detail is more difficult to deal with and controlling for white point and black point is a start but the slope of the analog-digital converter’s gain curve near the black point is also critical. In this respect each of the scanners is probably setup differently at the hardware level and most software packages provide (unfortunately) very limited direct control over the A/D converter.

What this all means is that comparing drum scanners is really an exercise in comparing a nebulous combination of software, hardware settings, and operator decisions and probably doesn’t really indicate a whole lot about the particular instruments’ capabilities. However, these sorts of comparisons are useful to the extent that they represent the quality of results that an average user or customer is likely to realize. It’s just never clear how representative a particular scan is.

[tim atherton]

if you are itnerested I can do a Umax Powlerlook III in Canada - an oldie but still a goodie

[Leigh Perry]

Tim, I've emailed you at your kairosphoto address to coordinate the scan.

[Nghi Hoang]

This is great stuff! Thank you!

[Jeremy_D]

Where you had to interpolate up, what method was used?

[John Hennessy]

Leigh,

This is a great idea. The Epson 3200 results were discouraging to me because I have one. I notice many who use a scanner such as an Epson probably do not use its own program to drive it. I use Silverfast AI as do many others. For some elements of your test, e.g., noise in shadows, Silverfast may make a difference. Also, I would bet that Epson scanners are not subject to much quality control so one may be far worse than the next. At 350 usd, Epson is not going to cull very many out.

So, I am volunteering to be included in your loop to test my 3200 with Silverfast to see how it compares to the other 3200 you have already tested.

[Gary Barnett]

Great, long overdue comparison!! I would love to see other prosumer models included also, such as the medium format Coolscan 9000, given the unusual feature of multiple scans to reduce noise.

I own the Imacon Flextight II, and the noise I see in the dark areas is all too disturbingly familiar...if I'd had this comparison available to me at the time of purchase I think I would have tried for one of the drum scanners instead, especially because my Imacon "loses" ppi as the film size increases, and I also shoot 8x10, a size it does not handle at all... oh well. I've gotten my money's worth out of it in scanning well exposed film!

A note to other Imacon owners: if you are scanning color negs, I once had a tech on the phone with me for about an hour (he was let go, unfortunately) and learned that one should use a NEGATIVE number (-60 was suggested) in the USM window, to reduce film grain and general noise. He also suggested a +152 USM setting for chromes, using a 4 grain limit in both cases.

I haven't found a lot of use for their proprietary noise reduction feature yet...it seems to do as much damage as help in most cases, so I prefer to use selective USM in Photoshop for the final image.

Anyone want to sell me an Optronics? I should mention I am broke, but would gladly trade in my Imacon, which is in perfect condition with spare bulbs still in original packaging. These bulbs never seem to quit!

[QT Luong]

Leigh, thanks for adding the new Epson scans.
I just saw this link
to a
Tango v. Epson comparison (http://clarkvision.com/imagedetail/drum.vs.flatbed-scanners/index.html). The difference between the Tango and the Epson is much less
than in the images you posted. In fact, this difference appears minimal to my eyes ! Yet, your findings seem to indicate that the Epson is clearly inferior to the Tango (as I expected), and even to the Microtek. Comments ?

[Jeremy_D]

You see these things sometimes - people using an unsharp original to evaluate a scanner, or bad technique with a scanner, or whatever.

Having done MANY comparisons of Epsons vs Imacons and Drums, I find the differences easily visible in prints of A4 size even from medium and large format trannies. The Imacons and Drums are close (especially off a 949), the Epsons miles behind in sharpness when you scan an original that is actually sharp. (But 95% of people can't actually shoot sharp anyway!)

People with flatbeds are kidding themselves. It's easy to see it in real prints and unfortunately with the standards of testing floating around the web you just have to do the test for yourself to see it...

[QT Luong]

Jeremy, what makes you think that Clark's comparison is flawed ? If you take a tour around his site, you will
see that he writes a lot about image quality and goes into great details.

[Jeremy_D]

First and foremost, experience.

Scanning an obviously unsharp original is not a great way to test a scanner for one.

I have an Epson 4870 sitting next to an Imacon 949 on my desk, and a good friend owns a crossfield drum scanner. I've scanned over 100,000 transperancies over the last 4 years (professionally). I doubt too many people can say that (and I wish I couldn't as well!!).

I have learned, for instance, that the Imacon at 1600 PPI picks up more detail than the Epson 4870 at 4800PPI even though the files are vastly smaller (same way I know a Canon 1DS MK2 file has more detail than about 90% of people's 35mm trannies but a good shooter with faultless technique with Velvia still gets a much better print above A4 with the Velvia than the Canon, as long as the Velvia is scanned on a Flextight or a Drum).

It's a really easy test to do. Take a really good, ****sharp*** tranny with lots of deep shadows, and scan it on every scanner you can get your hands on. Then, scan it about 50 more times until you really understand each scanner. Profile the scanners if necessary (pref. with a Hutch target).

Then, take the files, do your very best with each one in terms of processing the files, and have each one printed at A4, A3, letter etc (sorry, thats roughly US letter, double US letter, more). Then, look at the prints. And then have a whole bunch of other people look at the prints.

Once you get above very small prints you will see - the Epson prints will have less real detail. Rock will look less three dimensional because the really fine detail has been obliterated by the piss poor optics. Fine blades of grass will not be as distinct. Tonal separation on the Epson prints will be inferior in a way that is hard to put your finger on but is very visible. The Epson prints will lack any real shadow detail and have a kind of clumpy mess look to it.

Incidentally, I have owned a 2450 before the 4870 and I do think they are good tools for the money - great for proof scanning multiple negs at a time etc. But don't kid yourself there up there for scanning high quality landscape trannies unless all you print is 10 by 8s.

Of course if you scanny a blurry, bad tranny then all the scans will be blurry too and the whole thing is moot.

I've read Clark, (and Koren, Reichmann etc) in detail and I'm certainly not anti-digital or even anti-Epson (all my commercial work is done on digital), but I personally shoot 6x12 and 4x5 and scan on the Flextight 949. I do this because I've tested it and at 20 by 40" the differences are night and day. Most prints I sell are 15 by 30" up, so that's what I look at.

So, in summary, the reason I think Clark's test is flawed is because I have, many times, done the same test and reached a vastly different conclusion - I'm 99.99% certain Clark is wrong on this one. I don't really care about all the other bunkum on signal-to-noise versus spatial etc, I just care what the prints look like. I'm very output centric because all I can sell it the output!

Maybe Clark has a freakishly bad drum or a freakishly good Epson somehow? I don't know, but having owned and extensibvely used several of the scanners in question, I know Clark's results are not the norm.

[QT Luong]

How does your 949 compares to known drum scanners (such as the ones in the test) on 4x5 ?
What about 5x7 ? I am always a bit worried by the decreasing resolution of the Imacons as film
format increases. Would you be willing to scan Leigh's tranny on it ? If so, please contact him.

[Jeremy_D]

In my experience, it compares very favourably. With 6 by 12s I've never felt a need for anything more than the 949 and I haven't met an operator who can consistently deliver me a better file than I can with the 949. The 949 is a big improvement over the earlier Precisions/Photos etc that are in that test.

I don't shoot 5 by 7 and I'm not sure there is a holdr for 5 by 7 for the Imacons in any case, sorry.

Very occasionally if I have call for a VERY large print of a 4 by 5 then I go for a drum scan, but that is the only time and its not very often I run out of pixels. The Imacon is so clean and sharp that even if you end up with 200PPI or so, you can still get a very nice print.

I'm happy to scan Leighs tranny...although I think the 'wack the levels up' shadow test at the end isn't very good.

[Keith S. Walklet]

Having done my own tests with the Epson vs. the Nikon 8000 and TANGO, I look forward to receiving Leigh's tranny. Regardless, I am also really impressed with the functionality of the website he has devised to compare the results. So much so that I can imagine only one way to improve upon it. If there were two columns of windows so that one could have two different scanners results side by side, it would be an excellent method to compare the results. Awesome effort regardless!

[QT Luong]

Keith, I've thought about that, but it's easy enough to open a second browser window.

[Leigh Perry]

Sorry for the delay in responding. I'm just back from holidays.

I'm aware of the comparison by RN Clark, but I don't really have any insight into his contradictory results. Inevitably the operator has some bearing on the scans, although perhaps more so with the high-end scanners than the flatbeds. Because our comparison has been performed by different operators under the control of a range of software, objective conclusions cannot really be drawn. I must admin that the results I have collated do fit with my own experiences and expectations a little more than RN Clark's, but I am by no means an authority on scanning -- I'm as much an interested observer in all this as the rest of you, and a sizeable proportion of my motivation in coordinating the study is the self-interested research of my next scanner purchase :-)

I do see a consistency in our results across the three Epson models -- the scans show an incremental rather than quantum increase in resolution from each model to its successor. But I also look forward to Keith’s version of the Epson 4870 scan, since his experiences seem to affirm that with operator expertise the Epson can approach drum scans.

I have now received a Howtek 7500 scan that I’ll process tomorrow night. In the next few days are ICG 350i and ICG 380 drum scanners.

If I get time over the coming weeks, I'll also look at alternative display options that might make comparison easier.

Cheers.

[Keith S. Walklet]

Another browser window! Duh!

With QT's suggestion, Leigh, I wouldn't worry too much about a second column.

[Leonard Evens]

I don't see any images under Mozilla Firefox. The page did work properly in a previous version. I do have javascript enabled.

[Alan Corker]

The scan target itself is flawed it seems to be shot under artificial light and therefore has a very constricted colour-space or tonal range thereby favouring the cheaper scanners.

I use on a daily basis a Heidelberg Topaz that wipes the floor with my own Epson 4990 in every respect (as you would expect at that price, however the point is a relative quality comparison, not a price comparison). I only usually use my 4990 for 1:1 reflective document archiving which is all it really does well.

Your results only show how good each unit is at scanning that particular individual target, and should not be used to form an overall and authoritative or definitive relative comparison.

Why not scan a Kodak IT8 5x4" transparency target and produce an ICC profile for each unit under test, then graphically overlay each IT8 profile onto the Adobe 1998, or some such, common reference colour space? I can't think of a quicker more accurate way to quantify the capability of a scanner and quickly identify it's weaknesses in terms of colour reproduction. For testing resolution, just a simple LPMM chart should answer any question you have?

Very best regards.

[Paddy Quinn]

"Why not scan a Kodak IT8 5x4" transparency target and produce an ICC profile for each unit under test, then graphically overlay each IT8 profile onto the Adobe 1998, or some such, common reference colour space? I can't think of a quicker more accurate way to quantify the capability of a scanner and quickly identify it's weaknesses in terms of colour reproduction. For testing resolution, just a simple LPMM chart should answer any question you have?"

Why? because most of us don't photograph charts and targets, rather "real" things. This is the same argument made by the old lens testing/lpm/USAF target crowd. The results were often good in theory but not always so in the real world.

"Your results only show how good each unit is at scanning that particular individual target, and should not be used to form an overall and authoritative or definitive relative comparison."

nor does it claim to

(I think the test image here was shot in daylight by the way?)

[Ed K.]

Great stuff. I don't think that the Epson 4990 test results are

very far off, if at all. It's a good scanner for proofs and small prints

and that's about it. My results, even with Silverfast AI yield

similar results to those shown - soft, artifacts and close, but no

dice kind of output. For those of us who have to get drum scans done

at a lab, it would be very revealing to send the same slide around

to various labs to compare the results.

[Doug Dolde]

A Seybold scanner report worth reading addresses the high end scanners.

http://www.seyboldreports.com/SRPS/subs/3001/html/pixelperfect2.html

[Steven]

The Microtek i800 and the Epson V700 are out. Are there any plans to try and add them to the list? They both look to be promising (for the price, or course).

[Leigh Perry]

I'll include any scanners that I can get access to. If someone has either, just let me know.

[Aaron Rocky]

Crosfield gives the best shadow detail IMO.

[Ron Marshall]

The 4990 is fine for a three times enlargement from b/w or color neg, it doesn't do as well with transparencies.

[DrPablo]

I'd been having a hard time getting shadow detail from positive transparencies to scan well using Lasersoft Ai and my Microtek i800. But I've discovered that if you go in and manually increase the gamma to 2.2 or so and scan in 48-bit HDR there will be plenty of shadow detail. You can then use curves in PS to restore the contrast.

[chip_reuben]

First, let me say this source of drum scanner comparisons is EXCELLENT! With that said, I have a few comments and suggestions:

1) I would like to see the Aztec Premier added to the list
2) I would have preferred that in the shadow-brightening examples, that the curve not have been so exaggerated in order to get an example that is closer to or more likely to be a real-world example of shadow brightening.
3) As a writer, I must criticize the conclusions section, which is inadequate. It is ok to allow the reader/viewer to come to their own conclusions, but it is, then, the writer's responsibility to articulate in greater detail here what factors are believed to be of importance in making the evaluation. It is these factors that should have been thought through while designing the experiment.

[John Harvey]

I find the Optronics scanns to be very soft when the sharpening is removed. Did the operator hav the scanner correctly focused?

[John Kelly]

The tests appear to neglect the influence of negative carriers and focus. This is critical with Epsons flatbeds.

I'll simply note that the old Epson 3200 is quite obviously sharper, printed to 11X14 , scanning with a focus-adjusting 3rd party negative carrier rather than the original. http://www.betterscanning.com/ (the non-focusing carriers are better than OEM as well, but not as good as the focusing version)

Focus is critical within a range of less than 1mm. In other words, once one finds a good focus adjustment, if one drifts from the midpoint by more than 0.5mm on is likely to see a loss of sharpness on relatively close inspection. With my particular 3200 the optimum film plane appears to be about 1.5mm from the glass, as contrasted to the general folklore calling for 1mm. 0.5mm difference is visually significant.

[Clay Turtle]

Noted that no HP scanners appear in your list . . . having one I feel that their missing aspect is justified. Purchased scanner software did not function to begin with requiring down of software to be reinstalled . . . with much agro from multiple attempts, finally resolved via Windows support in India somewhere.
Problems: scanning with larger (high resolution?) files seems to default Windos to reboot!
Question: Shooting both 35 mmm & 4x5 formats, it seems ludicrous as capacity of 35 mm scan surpasses file handling ability? Considerations transition into 5x7 format doesn't seem viable? Scanning this mornig of 4x5 slide with TMA adapter with setting of 600% increase in output size @ 200 dpi. & This is the Third time around as system crashes about half way through scanning! So whose running Mac's & do you have these types of problems with adding an HP scanner to a Mac system?

[Andrew O'Neill]

Neat. I had no idea that scanners would render colour so differently...just like different types of film. It was interesting to see Fuji's scanner and how it was slightly "green".

[Faustin Baron]

I have owned two Epson 4990 scanners and on both scanners I have had an issue with the glass. The glass on the scanner always looked like it needed cleaning; but underneath, not on top where you would place your negative. I had it serviced through Epson and the glass was replaced but the new glass was foggy in different places. Epson eventually replace the scanner but the new one has the same foggy look to it and it has become worse over time.

[Harley Goldman]

With regard to the Epson glass, you will have to clean it from time to time. You can google to find instructions how to disassemble the glass. It is quite easy as long as you follow the instructions. I suspect the plastic inside is outgassing and causing the glass to fog.

[Frances Melhop]

hi All ...
I am new to this forum and entirely new to scanning things myself...but i do have 20 years worth of images on all formats of film, transparency ,neg, and polaroid which are in 5 different storage locations around the world and therefore very difficult to access ....

i need to rationalise my library but not knowing what the end use of each image is i need to make the scans at very high quality ......as I am doing all my own postproduction now i need access to my picture s for compositing etc... I dont use stock libraries and prefer to go out and shoot it myself at my own expense if a job needs a certain shot inside the main image....

at this point i would like to thank Leigh Perry and collaborators for the amazing scanner comparison...

I am thinking about a high end flatbed scanner ... (drum scanner is not for me ...spatially... portabilitywise ....and cost wise...unfortunately ... and i have enough to do without learning to be the best scanner operator...)

i found the scanner specs quite confusing.... i think i should be looking for a 16 bit scanner but why do the canon specs say this ?

"The CanoScan 9900F Color Image Scanner from Canon (www.canon.com) features 3200x6400 dpi along with 48-bit depth for over 281 trillion possible colors."

?

my preference in the comparison was the creo eversmart supreme...and i havent been able to find out where this lies in the price market yet ...but it seems to be way above the other flatbeds in quality and the scan had an atmosphere that was as close to filmic as i guess is possible ...

[Keith S. Walklet]

Frances,

The 48-bit refers to 16-bits per channel (Red, Green, Blue) in which the scanner is recording information.

The 3200x6400 refers to manufacturer's claimed resolution (samples per inch) for the scanner, with the smaller number referring to the highest optical resolution, (limited by the number of sensors per inch on the scanning wand (array).

The larger number refers to the claimed sampling interval of the steppor motor that moves the wand in minute increments the length of the image that is being scanned. So, they are claiming that the scanner samples the image data 6400 times each inch that the wand moves.

When evaluating the claimed resolution, it is the smaller number that is of greater importance. If you search the threads here, you'll find quite a few discussions about 1. this subject, 2. a lot of debate about the merit of manufacturers' claims 3. the difference between the various scanning technologies.

[ryanmfritz]

Im a noob that needs an updated scanner (within budget) that is at least 11x17 or larger.

Can anyone offer recommendation for a new scanner for my small business of scanning artwork mainly, and photos and reproducing them in various sizes on a large format 44" HP Designjet Printer. I would appreciate any advice. Budget is secondary, I want a quality product first...

Thanks
Ryan Fritz
www.FargoKids.com