On drum scanners and flatbeds

[Ben Syverson]

Recently I was mulling the possibility of getting drum scanner. I've seen a few really great working machines go for $1000 - $1500 with mounting station, drums and Mac G3/G4...

So I decided to revisit the scanner comparison page on this site. I selected my scanner, an Epson 4990, and the Howtek 4500.

The sharpened versions of the scans on the site have not been sharpened enough (or properly) for me, so I grabbed the unsharpened images and had my way with them. There's a lot of detail "hiding" in the 4990. Here's what I came up with.

http://bensyverson.com/images/illustrations/lff/scanner-comparison1.jpg
http://bensyverson.com/images/illustrations/lff/scanner-comparison2.jpg

I've done two passes of simple USM on each image.* To me, the Epson images arrive in the same ballpark as the Howtek scans...

Keep in mind that these images are 100% views of 2400 DPI 4x5 scans. If you were to print the entire scan at 40x30 inches (300 DPI, 8X enlargement), these detail crops would only be 1.666 inches wide. To put it in perspective, here is the fabric detail reduced to 72 DPI (life-size section of 40x30" print, if your monitor is 72 DPI):
http://bensyverson.com/images/illustrations/lff/scanner-comparison-actualsize.jpg

I would have to see the actual prints in front of me, but I don't think I would be able to tell the difference between the Epson scan and the Howtek at 300 DPI at a reasonable viewing distance (say more than 12"). Then I started to think: I shoot mostly 8x10, and an 8X print yields an 80 x 60 inch image. :eek:

In other words, with a $200 discontinued scanner, I can make an 80 x 60 print from 8x10 that I can't distinguish from a drum scan, even up close. Cut, print, it's a wrap.

Any thoughts?

- ben

* On the first comparison, it's Epson: {300, 1.7, 0.0} then {200, 0.7, 0.0}, and on the Howtek: {84, 1.7, 3.0} then {112, 0.7, 0.0}.
On the second comparison, it's Epson: {200, 1.5, 0.0} then {164, 0.7, 0.0}, and on the Howtek: {67, 1.7, 3.0} then {91, 0.7, 0.0}.

[Ron Marshall]

[QUOTE=bensyverson;448591]In other words, with a $200 discontinued scanner, I can make an 80 x 60 print from 8x10 that I can't distinguish from a drum scan. Cut, print, it's a wrap.

QUOTE]

As long as you are satisfied, there is no point spending extra money.

Have you had a drumscan made of one of your negs or positives?

[aphexafx]

Perhaps the level of sharpness can be made to appear in the ballpark of the Howtek scans, but the (huge) tradeoff is, of course, noise...

But would this affect your output at 300dpi? I hope you will do some print tests and let us know!

The rope image is going to be better at hiding the noise increase because of the complex, higher frequency textures, but the smooth paper really tells the story, imo.

I hate color-noise personally, it is my enemy, and I think it robs prints of definition even at 300+ dpi, even on matte paper, which goes back to the lack of definition in the file in the first place. Even in the scaled version of your comparison, I can see this effect as a loss of smoothness due to the color-noise related aliasing, and this is exactly what I see on noisy prints to paper. But this is nit-picking. Still...

Having just knocked it, I must reiterate that I am wholly impressed with the level of film scanning that we are now seeing from cheap desktop machines - compared to the last time I worked with desktop scanners for film, we are entering a revolution.

[Peter De Smidt]

Someone better wake Lenny ;)

[Don Hutton]

The operator of those scanners needs to know exactly what he's doing to extract the best out of both of them. I've made a couple of hundred scans with the Epson 4990, with Doug Fisher's mounting station (IMO, an essential accessory with that scanner) and with the Howtek 4500 (the same images - rescanned after I purchased the Howtek). In every single case the Howtek scans are a long way ahead of the scans from the Epson in detail, native sharpness and microcontrast. And it's not only in huge prints that you see the differences - I was working on the attached image yesterday, scanned on the 4500 (shot on 4x5) and had previously scanned it with the 4990 and made a number of prints at different sizes. Even on prints as small as 11x14, there are substantial differences - mostly because the Howtek 4500 extracts all the tiny tonal differences which the Epson 4990 cannot, resulting in increased microcontrast and apparent "texture detail"... On color positive, the drum scan has another huge advantages because Dmax is much, much better than the 4990.

If you want to make a decision on scanners based on a single sample point from an unknown operator, do for it, but I'd strongly suggest that you may well end up with something you don't want or didn't need. I believe that the Howtek 4500 example on that collaboration has been discussed many times with the conclusion being that it was a poor example.

I don't think anyone needs to see another drum scanner/EPson flatbed wankathon - if you think they are remotely comparable, get a sample done of the same image by a reputable professional and make up your own mind.

[Bruce Watson]

There is more to a scan that just simple sharpness. Come on, lets see a show of hands -- how many of you have ever made a photograph that's just about sharpness? Anyone? No? I'd like to have one to put up with the one I have that's only about saturation. ;)

Even if sharpness is all you care about, if you start with better sharpness, you don't have to work so hard to get to the end point. IOW, you are sharpening your consumer flatbed scan a good deal more than you are the drum scan. In addition to the scene detail, you are also sharpening any noise (grain clumps). I can see this "coarsening" in the non-print portion of your newspaper example.

If you are scanning negatives in particular, most drum scanners give you some inherent advantages in image quality. Most of them can apply their entire digital range to the exact density range you specify for the film when you set your black and white points in the scanning software. I've never seen this in a consumer flatbed. What this means is smoother tones and tonal transitions in the scan file and of course in the print (which is what counts).

Just due to the way drum scanners light the film they have certain inherent advantages. All the flatbeds I've seen light the entire film at once. This means that light scatter is a problem, particularly in the less dense areas of the film. I'm always impressed at how well a drum scanner scans the shadow detail in B&W, where the flatbeds I've used left a "veil" over the shadow details.

To get the most out of any scanner however, you have to spend some time and climb the learning curves. A good operator can make a good scan from a consumer flatbed, while a bad operator can make a bad scan from a drum scanner.

The bottom line here is always the prints. If you are happy with your prints, no matter how you get there, then good on ya.

[Ben Syverson]

Perhaps the level of sharpness can be made to appear in the ballpark of the Howtek scans, but the (huge) tradeoff is, of course, noise...

There is more noise in the Epson scans after sharpening, but keep in mind that I'm applying the sharpening to 8-bit JPEGs that were compressed for the web.

The odds are that a nice 16-bit scan from the Epson would clean up the majority of that noise. Even if it didn't, there is a lot of really good noise reduction software out there now.

But even with the noise level as it is, I think I'd be hard pressed to tell the difference once it was printed at 300 DPI. The slight dithering of the inkjet spray would probably mask that difference. And the lower resolution of a Lambda / C-print would obscure some noise and fine detail as well.

[Ben Syverson]

I will indeed get a drum scan of one of my images done soon, so I can make a better comparison. Maybe I'll post again after I do.

[Lenny Eiger]

Someone better wake Lenny ;)

Thanks, Peter, I'm here. And I am truly glad the message is delivered with humor. We all need a little these days.

When it comes down to it, I would echo what Don and Bruce said...

I have been testing film for the past three weeks. I don't have full conclusions yet. Many things look good in N light, out in the sun... However, I will say that the 4x5 looks very good. But it isn't an 8x10. The 8x10 has a quality that feels 3-dimensional. I'd rather carry the 4x5... but the 8x10 is clearly superior...

The same is true of drum scanners vs flatbeds (or at least consumer-prosumer level flatbeds, I haven't done my own testing of the top models). The flatbeds leave me flat...

On occasion, things can come together and produce a good result with just about anything. I strive to be able to do it time after time, and to do that I need good equipment, cameras, lenses, scanners, printers and good ink. And I have to know what I am doing with it all... and it keeps changing. A drum scanner takes one variable out of the puzzle of it all, not unlike a great lens... and that's comforting to me.

Lenny

[Ben Syverson]

Lenny, that makes plenty of sense!

[Ed Richards]

2400 is not going to show much difference with a good LF NEGATIVE. I have had very good drum scans made by two different excellent operators to compare with scans from my Canon 9950 running Vuescan.

There is clearly more detail at 100% on the screen when you compare a 4000 dpi scan drum scan to be best I can do with the 9950, but it is very hard to see it on on prints. I think Bruce has the key - unless your pictures depend on ultimate sharpness (pictures of fields of grass, perhaps) or you make really crummy negatives (get the negative too dense or too flat and consumer flatbeds are much less effective), it is hard to see much difference between a properly run flatbed and a drum scan for black and white negatives.

This is from an aesthetic point of view - if you think a print should be judged with a loupe, then make contact prints or get a scanning back and quit messing with scanning film. (I do not shoot color, so I cannot speak to it.)

[Brian Ellis]

Someone better wake Lenny ;)

LOL! I was thinking the same thing.

[sanking]

Just for fun, here is a recent comparison between my Epson 4990 and an EverSmart Pro. The 4990 scan was done at 4800 spi, the EverSmart Pro at 3175 spi. Both files adjusted to the same resolution in the attachments.

I think this kind of comparison is more informative than dickering around with scans done by other people.

I could fool around with unsharp mask until the cows come home but no way could I get better resolution with the 4990 than with the EverSmart Pro. And the Howtek 4500 is at least as good as the EverSmart Pro in terms of sharpness. You just can not turn pig ears into pearls.

The target, AF 1950, has resolution of 228 CPM, well beyond the potential of both scanners. And this is typical of the difference. I have made the same comparison several times in the past. I do it to keep my head level every time some one claims that Epson flatbeds are as good as drum scanners.


Sandy

[dwhistance]

"I could fool around with unsharp mask until the cows come home but no way could I get better resolution with the 4990 than with the EverSmart Pro. And the Howtek 4500 is at least as good as the EverSmart Pro in terms of sharpness. You just can not turn pig ears into pearls."

Sandy - I couldn't agree more. These threads tend to take on a degree of unreality when it comes to discussions of "Prosumer" scanners. They are great within their limitations but manufacturing price constraints clearly limit the absolute quality they can achieve

David Whistance

[Martin Miksch]

But the flatbeds are much better scanning plates^^

[Ben Syverson]

Sandy, to me it looks like the film is in the wrong plane of focus for the 4990. What holder are you using?

I shrunk the attached 4990 image down to 2400 DPI and compared it with my own raw 2400 DPI scans. My scans are much, much sharper than your USAF 1951.

As for the "unreality" of using prosumer scanners... I get that drum scans are sharper. I'm just trying to figure out how big I would need to print to see a real difference. I plan to print big, so it's not academic.

[sanking]

No, the target is in the right plane, about 1.5 mm over the glass. I have verified this is the best plane of focus several times.

I think if you were to scan my target with your scanner you would see very little difference in results. The results I am getting correspond to about 36 l/mm, or about 1800 dpi, which is about average for the 4990. It is a glass target and you get some reflection around the bars which seems to degrade the image, but that is true for both the Epson and the EverSmart so they are on equal ground in that regard.

It is not my position that you can not make good prints from scans made with Epson flatbed scanners. I just want to show a real objective comparison of an Epson flatbed with a good professional flatbed, done by the same person applying equal effort to get the most out of both scanners.

Sandy





Sandy, to me it looks like the film is in the wrong plane of focus for the 4990. What holder are you using?

I shrunk the attached 4990 image down to 2400 DPI and compared it with my own raw 2400 DPI scans. My scans are much, much sharper than your USAF 1951.

As for the "unreality" of using prosumer scanners... I get that drum scans are sharper. I'm just trying to figure out how big I would need to print to see a real difference. I plan to print big, so it's not academic.

[Ben Syverson]

Crazy... well, it may come down to the difference between scanning film and scanning 1951...

[sanking]

The scan of the target shows pretty much what one would can get on a print in terms of dpi. If one reads the results conservatively (as opposed to my earlier more optimistic reading) the resolution with the 4990 is about 31 l/mm, or 1500 dpi. This would allow a print from a 4X5 negative at 16X20 size with 325 dpi, and most people would find the print critically sharp. That is not bad, but not great either.

Beyond that size you are in the world of genuine fractals.

Sandy





Crazy... well, it may come down to the difference between scanning film and scanning 1951...

[Ben Syverson]

Interesting...

[walter23]

Cool. I have a 750 and I never scan much beyond 1200 DPI, and print 8x10s mostly, so it's more than adequate from a resolution standpoint. Pulling out shadows from dark areas of transparencies on the other hand - that's the annoying part.

[sog1927]

Just for fun, here is a recent comparison between my Epson 4990 and an EverSmart Pro. The 4990 scan was done at 4800 spi, the EverSmart Pro at 3175 spi. Both files adjusted to the same resolution in the attachments.
Sandy

Which one's which ?:rolleyes:

[Don Hutton]

Which one's which ?:rolleyes:Amen. I think Sandy's post wins the brevity prize forever on this forum!

[sanking]

As we say in Spanish, !Ja, Ja, Ja, Ja . . ., muy gracioso !



Which one's which ?:rolleyes:

[Ron Marshall]

The scan of the target shows pretty much what one would can get on a print in terms of dpi. If one reads the results conservatively (as opposed to my earlier more optimistic reading) the resolution with the 4990 is about 31 l/mm, or 1500 dpi. This would allow a print from a 4X5 negative at 16X20 size with 325 dpi, and most people would find the print critically sharp. That is not bad, but not great either.

Beyond that size you are in the world of genuine fractals.

Sandy

That agrees with what I have observed.

With a few different b/w images, I've printed from both a drum scan and 4990 scan. At 11x14 I cannot discern a difference. At 16x20 some of the finer details are a bit soft in the 4990 print. Depending on the image, that amount of softening may or may not be a concern.

[Keith S. Walklet]

Ron and Sandy, I want to thank you for your contributions to recent threads on the comparisons between the flatbeds and the drum scanners. This thread and the recent one started by Ron regarding estimating resolution loss in scanning are terrific.

[Ben Syverson]

More to follow as soon as I get my macro lens...

[aphexafx]

Ron and Sandy, I want to thank you for your contributions to recent threads on the comparisons between the flatbeds and the drum scanners. This thread and the recent one started by Ron regarding estimating resolution loss in scanning are terrific.

Agreed. Keith, your study of, and willingness to share, Epson flatbed scanning results is equally terrific.

[Keith S. Walklet]

I was fortunate enough to receive Leigh's collaborative scanner transparency a couple of days ago and am in the process of putting my machines through their paces to generate some updated and hopefully, meaningful comparisons. I particularly enjoyed this thread for its low-key and common sense discussion. Having looked back at the first image set that was posted, I thought this most recent comparison might be relevant.

I was running a series of scans boosting the SPI setting each time to compare the downsampled data results. This scan, done at 12800spi and downsampled to 2400dpi is not realistic for my workflow, as the file sizes are unwieldy. I'd need to make six different scans of a 4x5 at that resolution to piece together. But, I didn't expect the results to be improved beyond 6400spi, maybe 9600spi since that was supposed to be the limit of the stepper motor. But there was an improvement even with interpolation, in this case, or so it seems. So I thought I would see how it measured up to a TANGO scan of the same image at 2400spi.

I sharpened both the first time at 400-1.0 and then ran a second pass on the Epson V-750 at 150-1.0. The attached jpeg is the result. The Epson scan was done with Silverfast, 48-bit HDR wetmounted directly to the platen with a 3/16" piece of plex.

[Findingmyway4ever]

That agrees with what I have observed.

With a few different b/w images, I've printed from both a drum scan and 4990 scan. At 11x14 I cannot discern a difference. At 16x20 some of the finer details are a bit soft in the 4990 print. Depending on the image, that amount of softening may or may not be a concern.

Do you have a couple of prints of an image that you do not use commercially but has good tonality/excellent exposure that you can send a couple of examples in even 8X10 size, one from the Epson scan, and one from the drum scan? I'd like to see if I can see the differences, black/white, color, anything you wish to send. I will pay shipping both ways. BTW, you have scans from at least a Howtek/Aztek level drum scanner, correct?

[Keith S. Walklet]

I haven't generated the prints from this most recent set of tests yet, but will soon.

I have samples from an Epson 4870, V-750, Nikon 8000, and TANGO. The comparisons of Leigh's transparency are limited to the 4870, V-750 and TANGO.

PM shipping details to me.