Drum scans vs Epson 4990

[Zaitz]

I sent some of my negatives to get drum scanned. I wasn't sure what to expect, but I guess I expected these to be quite noticeably better. Maybe I am just seeings things wrong. Mine are scanned on an Epson 4990 with low sharpening set on the scanner. The drum scans are straight as I received them.

First one is 8x10 HP5. So the file sizes are similar.

4990:
http://img97.imageshack.us/img97/5002/compjm.jpg

Drum
http://img580.imageshack.us/img580/909/com1p.jpg

For 4x5 I lowered the size for the drum scans as they were scanned larger.
4990
http://img851.imageshack.us/img851/4948/epson1c.jpg

Drum
http://img16.imageshack.us/img16/3844/drum1d.jpg

The drum scans are darker. Perhaps because I didn't set a black point.

[Zaitz]

4990
http://img684.imageshack.us/img684/8017/epson12.jpg

Drum
http://img718.imageshack.us/img718/1598/drumg.jpg

[Peter De Smidt]

The drum-scanned images look better to me.

[Frank Petronio]

What were you expecting them be like?

It's impossible to really tell from a couple of jpgs but just looking here, the "darker" ones have increased contrast and the edges of things look sharper.

[Zaitz]

The drum-scanned images look better to me.

They do to me too. But I guess not by as large a margin as I thought they would. It seems I am not gaining much by sending them out. I don't think I'd be able to tell in a print. The 8x10 drum scan looks noticeably softer to me.

[Bob McCarthy]

There is not a lot of detail in the images to help make a comparison. find an image with plenty of DOF and crisp detail and you will see a didderence.


Bob

[Kirk Gittings]

You want a real test? Have Lenny Eiger do a drum scan and then compare them. Not all drum scans are as good as they should be.

[Zaitz]

You want a real test? Have Lenny Eiger do a drum scan and then compare them. Not all drum scans are as good as they should be.
That may be the truth.

4990:
http://img14.imageshack.us/img14/3650/kbhhj.jpg

Drum:
http://img200.imageshack.us/img200/307/22221312312.jpg

The drum scan is better. I guess I'm going to have to print these out to see if I can even tell a difference.

[neil poulsen]

These comparisons are quite illustrative for me. The difference I notice is that the drum scans have more detail in the shadows and in the highlights.

The flatbed scans are kind of hazy and appear to have more grain. But, that doesn't mean necessarily that they wouldn't necessarily print well.

[Roger Cole]

That may be the truth.

4990:
http://img14.imageshack.us/img14/3650/kbhhj.jpg

Drum:
http://img200.imageshack.us/img200/307/22221312312.jpg

The drum scan is better. I guess I'm going to have to print these out to see if I can even tell a difference.

Well the drum scan has considerably more contrast. I don't know how they would compare if you equalized that a bit.

[Roger Cole]

I took the liberty of increasing contrast just a bit in the 4990 scan and applying just a bit of unsharp mask. Let's see how this compares.

Here's your drum scan:



http://img200.imageshack.us/img200/307/22221312312.jpg

And here's the 4990 with a bit more contrast and just a touch of unsharp masking:

http://home.comcast.net/~ragnar93/Test1.jpg

Note: I will gladly delete this from my personal web space after some discussion or at the request of the OP, whichever comes first. Image is the property of the original poster.

[Zaitz]

Not a problem and that shows a good comparison. What I am wondering, I guess, is the small difference (I consider it a fairly small difference) in quality between the scanners what one should expect?

[Roger Cole]

Not a problem and that shows a good comparison. What I am wondering, I guess, is the small difference (I consider it a fairly small difference) in quality between the scanners what one should expect?

I think the real lesson is that it's really pretty much impossible to see the difference (given some adjustments in post processing) at these resolutions. Doesn't mean there wouldn't be at higher resolutions or when printed.

And of course I just did a quick and dirty few clicks, and it's obvious I slightly over did it, exceeded the contrast of the drum scan and probably sharpened a tad too much too.

The key I think will be in the printing, especially at large print sizes. From 4x5 I doubt most would see any difference (if carefully post processed to match) at anything smaller than 4x, 16x20, but I admit that's a WAG.

[photobymike]

I did the same test 3 years ago, Drum vs Epson v750... The problem i ran into was consistency in quality of the drum scans.... v750 was not as good but in the long run with many scans to do it was cheaper. I scan anything from 35mm to 4x5 BW, color neg, and positives. Just recently i got my epson v750 to do a 12,800 highest resolution on 35mm film.it is like 200 mega pixels .... cool very cool...... and the Digital Ice is a dream come true. Its just like my scanner has taken the place of my Omega D2 enlarger. In the last three years i have scanned thousands of pictures with my trusty 750. It is given me constancy in quality output on my schedule... late at night. I make prints on an Epson 4800 printer.....

The the bottom line is the print.

Viewing prints from both (same image) I am absolutely sure you could not tell the drum scan from the v750 flat bed scan. (My wallet could feel the difference.) I have tried and challenged other photographers to tell the difference in the prints.... heck they cant tell the difference between my digital and my film pictures.....

http://www.mikepic.com

i love the smell of fixer in the morning

[timparkin]

I've done a couple of side by side comparisons on these and here are a few comments

1) it looks like the drum scanner operator hasn't focussed correctly and they haven't used a large enough aperture for the scan. The noise levels in the image when sharpened to match the 4990 demonstarte this

2) The drum scan has had some shadow/highlight applied or some excessive medium radius sharpening by the looks of it - there is a lot less large scale contrast in the drum scans than the epson scan (particularly in the forest shot)

3) The drum scan will demonstrate it's difference in the handling of highlights. This is born out in the shot of the chandelier which shows a lot better handling of the parts of the lampshade overlapping the left hand bulb.

4) The 4990 shot of the chandelier looks like it has had some strong noise reduction applied? The difference in the noise within the light area of the chandelier goes from hardly any noise to nasty digital noise

Finally, I wouldn't expect a huge amount of difference between a drum scan and the 4990 for 8x10 neg. Yes you will get a better tonality if you choose the right aperture and you should get a more representative gradation if the scanner has been profiled properly. However the 4990 resolves about 1600-2000dpi which is probably as much detail as you have in your image. The only area that would be obvious is in highlight handling and tonality (although the latter presumes the scanner operator knew what they were doing)

[Ari]

There's a noticeable difference in sharpness or acuity; what size enlargements do you typically make?

[Ken Lee]

A more rigorous and profusely documented comparison of scanners appears on this forum already: see http://www.largeformatphotography.info/scan-comparison/.

When making comparisons, it's important to explain the details: size of sample, resolution, degree of enlargement, etc. Otherwise, it's all just speculation. For example, see this page (http://www.kenleegallery.com/html/tech/8x10Detail.html), also a scan from 8x10. There's no ambiguity about sample size.

[Brian Ellis]

The drum-scanned images look better to me.

The second image (drum) on my monitor is so soft in some areas, mainly the fabric in the upper left corner and the wood, that it appears to be out of focus. It looks much worse than the first (4990) to me in those areas. However the large area to the right has slightly more detail with the drum scan than the 4990. I agree that with the second set he drum looks better. However, a lot of the difference appears to be a matter of better contrast, something that could be adjusted easily in post processing.

It's been said here many times that there's nothing magic about just getting a drum scan. Some operators are more talented than others, some know what they're doing and others don't (or don't care enough to take the necessary time to do it right).

And even with a good operator the degree of noticeable difference between a prosumer scanner like the 4990 and a drum scan seems to depend to some extent on the image. In my admittedly limited experience with having drum scans made, the main difference has been better detail in shadow areas. If an image doesn't have important detail in those areas then the differences have been pretty hard to see.

And of course print size is also relevant. The largest prints I make with my 3800 are about 16x20. I suspect (or hope) that with larger prints the differences would be more apparent.

[Noah A]

There are so many variables here so it's somewhat hard to make any conclusions from the samples you've posted. Overall I think the drum scans are quite far ahead of the Epson scans. They're sharper and have more contrast without losing detail in the highlights and shadows.

What are the scan resolutions? What aperture was used on the drum scan? What have you posted here? Are they 100% crops? (I doubt it, unless the scans are on the small side.)

Roger's adjustments definitely improve the Epson scan, but he posted it next to a raw drum scan, which could also be improved by some adjustments.

The same thing happened to me when I posted a similar comparison on a different forum. Someone sharpened the heck out of the Epson scan, boosted the contrast and then compared it to the raw drum scan. Clearly both scans can benefit from some post processing, and I'd argue that since it needs less sharpening and less contrast adjustment, the drum scan will likely make a better print in the end. At least it will if you're picky about your work and know what you're doing on the print end of things.

Also, if you're making 11x14in. prints from 8x10, you won't see as much difference between the scans as if you're making 48x60in. prints from 4x5.

Assuming that prints are your final desired output, I would suggest that you optimize both scans and make good prints of each.

[Nathan Potter]

Comparisons between scanners can be difficult. The fact is, what are you comparing? What precisely is the objective of the comparison? A comparison of resolution is common, but that cannot be reasonably decoupled from contrast. The story can only be told accurately by quantitative data take under a set of known conditions.

In some sense you hardly need to do a comparison. The published specs and a few tests will indicate the capabilities of drum vs flatbed.

If you scan a high resolution target of increasing spacial frequencies you'll find one set where the clear space between lines no longer shows a minimum density (256). At this spacial frequency the spot (or aperture) is about the size of the space. I've done this recently for the V750 and found this size to be close to 1 mil (25 um). This size is a result of limitations inherent in the lens and detector. It's consistently different by a bit in the X and Y scan direction. It simply says that each pixel of digital information represents the average color and grayscale within that 25 um diameter area. No detail can be recovered smaller than a nominally 25 um area.

Now in fact the best drum scanners are capable of far smaller apertures, ideally down to less than 5 um diameters coupled with higher Dmax capabilities due to PMT technology.

Of course this assumes that the instruments are adjusted properly and operating properly.

In order to capture detail in the original film accurately we assume that the 25 um aperture is the Nyquist frequency (20 lp/mm), so by D =Nq X (2), we really can only replicate detail at a level of 10 lp/mm (50um features). This would be of course at 100% contrast or so - a very tough standard. Reducing the contrast increases the apparent resolution and is a very reasonable thing to do - say in the 10 to 50% range.

A drum scanner with a 5 um aperture very clearly has capability that greatly exceeds that of an Epson V750.

If you don't see a difference then the detail wasn't in the scanned film to start with or the machines were not adjusted to optimum performance prior to scan.

It's probably best not to obsess over this since Photoshop manipulations will largely corrupt the original film image and transpose it from reality through the artists license to his/her personal vision.

[Roger Cole]

Roger's adjustments definitely improve the Epson scan, but he posted it next to a raw drum scan, which could also be improved by some adjustments.

That is of course quite true. But contrast at least isn't something where more is simply better. I was just trying to make the two match more closely, my point being that the drum looked better in part because it had a more pleasing contrast for that scene, and this can be adjusted as desired with post processing, for either scan.

Of course the sharpness is something else entirely and while you may be able to sharpen this low res version of the Epson scan to match the low res of the drum scan, one can't actually sharpen into revealing detail that it's there and it becomes easy to over sharpen.

In my comments about magnification I missed that these were 8x10 originals. I think that makes the difference even less in practice. Even a 16x20 print is only a 2x magnification. If both were carefully optimized I'd be surprised (but not really shocked) if most people could tell the difference in such a print. OTOH, start with a smaller negative and the differences would be more apparent.

[Noah A]

That is of course quite true. But contrast at least isn't something where more is simply better. I was just trying to make the two match more closely, my point being that the drum looked better in part because it had a more pleasing contrast for that scene, and this can be adjusted as desired with post processing, for either scan.



True, but next to your improved Epson version, the drum scan looks a little flat. Frankly I think the drum scan still looks better. But to be fair it would be best to optimize both scans...

Really, the main problem with the comparison is that we know nothing about the original scan resolutions, etc. Nor do we know about the OP's intended print size, which is important since as you say, for small prints from big negs the differences between the scans aren't as critical.

[Zaitz]

The Epson scans are at 2000dpi. The drum scans at 4000 except 8x10. The drum scans are down rezzed to match size in the 4x5 shots. Crop is about 50%.

[SergeyT]

The Epson scans are at 2000dpi. The drum scans at 4000 except 8x10. The drum scans are down rezzed to match size in the 4x5 shots. Crop is about 50%.

There isn't much to discuss.

* No point to compare scans made at different res and more so based on some down sampled crops.
* A negative film does not put a stress on scanners density range nor DMax; read : easily fits into the working range of a pretty much any scanner.
* 2000dpi is not beyond the resolution limit of an Epson and usually not sufficient for making quality prints larger then 16x20 (from 4x5).

To make a real test, consider scanning a piece of a properly exposed 4x5 Velvia shot in daylight at no less than F22, perfectly focused, with both scanners at 4000 dpi, no sharpening applied and then compare 100% crops.

SergeyT.

[Kuzano]

I've heard it said that the real issue on the quality of drum scans is not the hardware, but the operator. However, I don't think that answers the question. Rather, I think it begs the question, would a truly expert operator be able to get as near a quality scan from a flatbed scanner as from his drum scanner.

So the real issue and question to be answered, is "How good are you with your flatbed?" compared to a knowledgable drum scanner operator.

Perhaps, it's only marginally a hardware issue, or even not at all.

Bottom line, who is the best drum scanner operator that anyone on this forum knows about, and what would I have to do to become as proficient as that person with a flatbed. At that point remaining quality scan issues would be hardware differences.

I've seen these questions bounced around repeatedly and the real solution to me seems to be...."forget digital". Proceed with a wet lab and chemical process.

Or, perhaps, Simply... forget digital!!!

Another solution would be to convert to digital capture from the beginning. It occurs to me that the weak link in the entire work flow, is an after capture (film) transfer to digital (scanning). Solve the problem. Take scanners out of the equation.

No one seems to be offering a consistent scanner solution.... Well OK, that's not quite right. The real solution may simply be to lower your standards substantially. Lower standards might make all scanner output look just fine?

[Zaitz]

I've heard it said that the real issue on the quality of drum scans is not the hardware, but the operator. However, I don't think that answers the question. Rather, I think it begs the question, would a truly expert operator be able to get as near a quality scan from a flatbed scanner as from his drum scanner.

So the real issue and question to be answered, is "How good are you with your flatbed?" compared to a knowledgable drum scanner operator.

Perhaps, it's only marginally a hardware issue, or even not at all.

Bottom line, who is the best drum scanner operator that anyone on this forum knows about, and what would I have to do to become as proficient as that person with a flatbed. At that point remaining quality scan issues would be hardware differences.

I've seen these questions bounced around repeatedly and the real solution to me seems to be...."forget digital". Proceed with a wet lab and chemical process.

Or, perhaps, Simply... forget digital!!!

Another solution would be to convert to digital capture from the beginning. It occurs to me that the weak link in the entire work flow, is an after capture (film) transfer to digital (scanning). Solve the problem. Take scanners out of the equation.

No one seems to be offering a consistent scanner solution.... Well OK, that's not quite right. The real solution may simply be to lower your standards substantially. Lower standards might make all scanner output look just fine?

That'd certainly work. :D

[Kuzano]

That'd certainly work. :D

Well, it did come to me as a last resort....

Primarily because that's what I found that worked for me when I was dating the ladies!!!:D

In fact, I was somewhat surprised at how far I had to lower my standards!

[Zaitz]

Well, it did come to me as a last resort....

Primarily because that's what I found that worked for me when I was dating the ladies!!!:D

In fact, I was somewhat surprised at how far I had to lower my standards!
:D

A bit disappointed. After editing the drum scans in the same way as my Epson scans they just look worse. The 8x10 is not as sharp, as seen in the original post. And these 4x5 scans after being edited looked bad.

Epson up-rezzed:
http://img687.imageshack.us/img687/7751/compareuf.jpg

Drum:
http://img835.imageshack.us/img835/6797/unled1vbn.jpg

Noise reduction doesn't help much. Noiseninja makes a mess of it and CS5 noise reduction just obliterates any detail. Duplicating the layer, applying CS5 noise reduction and then reducing opacity to ~50% does an ok job but it's still worse, imo.

I can send 4-5 negatives to Cheap Drum Scanning in the UK (if they'd ship my negatives back) and it'd be cheaper than anything in the US. I know price should not be what I go by but they also look professional and appear to do a great job. $200+ for a full rez scan is out of the question for me.

[Noah A]

:D

A bit disappointed. After editing the drum scans in the same way as my Epson scans they just look worse.

Is this the same drum scan you posted in your first post? With respect, it looks horrible. Keep in mind two things. First, the drum scan shouldn't be edited in the exact same way as the epson. A good drum scan needs only a tiny bit of sharpening, if any at all. I do sharpen my color scans for huge prints but only a tiny bit of USM does the trick.

Second, a good drum scan will resolve the grain better than an Epson. This may mean that the scan looks slightly grainier, but the grain should be sharp and accurately defined.

This edited example looks like a drum scan with too small of a scanning aperture. I suggest you ask the scan operator which aperture they used. If they don't know, then don't go back to that lab for scans. Do you know which scanner they're using?

I'm still a bit confused because the original drum scan you posted was far better. I didn't want to post it without your permission, but I did a small curves adjustment and a very light USM and got a much better result than the last edited drum scan.

How large do you intend to print your photographs?

[Zaitz]

Is this the same drum scan you posted in your first post? With respect, it looks horrible. Keep in mind two things. First, the drum scan shouldn't be edited in the exact same way as the epson. A good drum scan needs only a tiny bit of sharpening, if any at all. I do sharpen my color scans for huge prints but only a tiny bit of USM does the trick.

Second, a good drum scan will resolve the grain better than an Epson. This may mean that the scan looks slightly grainier, but the grain should be sharp and accurately defined.

This edited example looks like a drum scan with too small of a scanning aperture. I suggest you ask the scan operator which aperture they used. If they don't know, then don't go back to that lab for scans. Do you know which scanner they're using?

I'm still a bit confused because the original drum scan you posted was far better. I didn't want to post it without your permission, but I did a small curves adjustment and a very light USM and got a much better result than the last edited drum scan.

How large do you intend to print your photographs?
You are right I think on the sharpening. I did selective sharpening on the image too though. Perhaps it's just the much more visible grain in the unsharpened parts? Post your image, I've no problem with that.

I am not doing this as a comparison for others on a drum scan vs an Epson. I am just wondering if the results should be much more different. Looking at this example, perhaps I am expecting too much...again.
http://static.timparkin.co.uk/static/scanner_comparison_1/index.psp?code=lichen

I was interested in printing these very large, perhaps 40x50. May be only going to 30x40. I think I'll be sticking with my scans as well.

[D. Bryant]

I think that makes the difference even less in practice. Even a 16x20 print is only a 2x magnification. If both were carefully optimized I'd be surprised (but not really shocked) if most people could tell the difference in such a print.

I have to disagree. A properly made drum scan will easily out shine one made on the Epson 4990 (or V700/V750). The differences are readily apparent in a print, even at sizes less than 2x, at least that has been my experience comparing scans made with a Howtek and 4990. Optimizing a scan from a 4990 just isn't the equivalent to a well done drum scan, even at identical SPI.