Computer "'horsepower" for scanning?

[Peter Mellis]

I am planning on scanning MF film (with a Canon 9000f) - 645, 6X6, 6X7 - and am curious as to whether or not my current photo dedicated computer is up to the task. My goal is to be able to do contact style sheets and some scans for printing;based on my reading, I'm assuming that these would be done at around 2400 dpi, which seems to be the effective true output of the current crop of flatbeds. I don't figure on printing any larger than some reasonable multiple of the film size @ 2400 dpi; if I want something larger/better, I'll have a drum scan done. Won't be doing 35mm, so that's not an issue. That said, the computer is a Dell Dimension 8400, Pentium 4 3.2ghz, 3.25 gb of ram, running XP. Speed aside (I'll live with whatever that is/isn't), is this machine up to the task? How about handling the files in Lightroom?

[Mark D Segal]

I think you should be fine. You have all the gear, so just try it. I'd question the need for 2400 PPI to make contact sheets of medium format photos, but that's another topic. You won't need more than something in the range of 240~360 OUTPUT PPI to make a fine print, so depending on how large you intend to print each image, you may not need 2400PPI input resolution to handle that. For example a 6*6 negative is about 2.4 inches per dimension. Multiply 2.4 inches by 2400 PPI, and you have a total of 5760 pixels to play with. Divide that by 240 PPI, and you'll see you can make a 24" print, or divide by 360 PPI and you can make a 16 inch print. However, there is an argument that scanning at full optical resolution and downsizing produces a better quality image than dimensioning your PPI to more accurately reflect the size of the intended output. I've tested this principle numerous times and never been convinced, but either way works well.

[Christoph C. Feldhaim]

I have some performance  problems scanning with my Nikon LS 9000 and Silverfast.
Its a 64 bit Windows 7 machine with 8 gigs ram and once the scan starts the PC is hardly usable.
It works, but seem to use up incredible resources.

[Mark D Segal]

I have some performance  problems scanning with my Nikon LS 9000 and Silverfast.
Its a 64 bit Windows 7 machine with 8 gigs ram and once the scan starts the PC is hardly usable.
It works, but seem to use up incredible resources.

Christoph, that's interesting and strange. Just out of curiosity I turned on my activity monitor and scanned a slide in my Nikon 5000 SuperCoolscan using SilverFast Ai6 Studio. The activity monitor shows a trivial use of processor resources. Not the same set-up as yours. Mine is a 2.66 GHz 12-core + hyperthreading (24 virtual cores), 24 GB RAM MacPro, but the fact that it only consumed such a small fraction of two or three cores at any one moment suggests simply that this computer is heavily over-dimensioned relative to what is needed for running this scanner and software. I also had Acrobat, Bridge, Photoshop and Firefox open at the same time. So I would think your computer should really do better than what you are observing. But this is a difficult area. Much depends on your processor design, capacity and speed, and how the operating system is allocating instructions to the processors. I've been told by people who know better than me about this stuff that on PC systems the binding constraint is usually the front bus capacity, but I'm no expert in that area. I was reasonably confident in my suggestion above to Peter, because before I bought this computer I was on Windows XP using a Dell 690 Workstation - 32 bit, 4 core 4GB RAM, and felt no computer keep-back whatsoever using the same scanners with it.

[Christoph C. Feldhaim]

Yes - most likely the frontside bus.
Another annoying thing is, that the extraction of the infrared channel with with silverfast on the LS-9000 on windows doesn't work.
This is a software problem not yet fixed by Lasersoft and together with the performance annoyance might lead to me buying a used dedicated Mac for scanning only ... 

[Bart_van_der_Wolf]

Yes - most likely the frontside bus.
Another annoying thing is, that the extraction of the infrared channel with with silverfast on the LS-9000 on windows doesn't work.
This is a software problem not yet fixed by Lasersoft and together with the performance annoyance might lead to me buying a used dedicated Mac for scanning only ... 

Hi Christoph,

It's cheaper to buy a copy of VueScan. VueScan and the LS-9000 work very well together, including IR dust and scratches removal!

Cheers,
Bart

[Bart_van_der_Wolf]

However, there is an argument that scanning at full optical resolution and downsizing produces a better quality image than dimensioning your PPI to more accurately reflect the size of the intended output. I've tested this principle numerous times and never been convinced, but either way works well.

Hi Mark,

Grain-aliasing is real phenomenon, although it depends also a bit on the grain structure and scanning hardware (diffuse or collimated lighting) how easy it shows.

People also underestimate the loss of resolution, even if scanning is followed by downsampling. I have an example of the latter, combined with noise on a deprecated and no longer maintained page here: http://www.xs4all.nl/~bvdwolf/main/foto/scan/se5400/se5400-5.htm
It also shows that even 4000 PPI is not enough to extract all detail.

Cheers,
Bart

[Christoph C. Feldhaim]

Hi Christoph,

It's cheaper to buy a copy of VueScan. VueScan and the LS-9000 work very well together, including IR dust and scratches removal!

Cheers,
Bart

Does VueScan also allow the extraction and storage of the IR data for late dust and scratch processing?
Silverfast ICE works great during scanning, but it can't extract the data and store it in the tiff.
Can VS do that (on Windows)?

[Mark D Segal]

Yes - most likely the frontside bus.
Another annoying thing is, that the extraction of the infrared channel with with silverfast on the LS-9000 on windows doesn't work.
This is a software problem not yet fixed by Lasersoft and together with the performance annoyance might lead to me buying a used dedicated Mac for scanning only ... 

Well, it's actually not a software problem that needs fixing - in the sense of broken code in the application. I've asked them about this in the past and the unfortunate reality is that infrared capability on these Nikon scanners (5000/9000) was not provided for at all. I seem to recall a technical reason being mentioned, the nature of which I honestly don't remember - it was quite some time ago. I believe it's still on their radar screen even though the Nikon scanners are sadly discontinued, because there is a large interested user base.

[Mark D Segal]

Hi Mark,

Grain-aliasing is real phenomenon, although it depends also a bit on the grain structure and scanning hardware (diffuse or collimated lighting) how easy it shows.

People also underestimate the loss of resolution, even if scanning is followed by downsampling. I have an example of the latter, combined with noise on a deprecated and no longer maintained page here: http://www.xs4all.nl/~bvdwolf/main/foto/scan/se5400/se5400-5.htm
It also shows that even 4000 PPI is not enough to extract all detail.

Cheers,
Bart

Bart, this is a very interesting page - whether "dated" or not, I hope you intend to leave it up on your website, because it shows interesting things about the evolution of scanner hardware and software treatments at the same time. The difference between the Nikon 2000 and 4000 is remarkable. The fact that the Minolta 5400 beats both of them for rendition of fine detail is the main reason why I bought that scanner back in 2002. The only problem is how damn slow it is with all its gymnastics before the scan, and the scanning itself. But the scan quality is very good. Then a year ago November just before they disappeared completely (at normal retail price) I bought a Nikon SC5000ED, which works much faster and is at least as good as the Minolta for resolving of fine detail. But its LED illumination is, I believe, rather non-diffused, because it shows both image detail and grain detail quite stunningly.

I can also confirm your observation of some colour shift when activating dust and scratch removal with the aid of the infrared channel. (I do this with iSRD in SilverFast, as it's much more controllable than the infrared set-up in Vuescan, and it can be previewed as part of that user controllability; as for ICE - it's either on or off - there is nothing for the user to control.) The amount of hue shift I get seems to be considerably less than what your examples show. It's quite subtle in fact.

As for dealing with the grain, I generally prefer to do grain management and sharpening in the post-scan phase of image editing, because we have more specialized, reversible and controllable applications for handling these functions than anything in either SilverFast or Vuescan.