SilverFast and Epson scanners

[Andrea Taurisano]

Hi

Many tests on filmscanner.info indicate that SilverFast enables significantly better images than the scanner's original software. If I get it right, SilverFast kind of gives you full control of the RAW file, instead of an already pre-processed file. This seems mainly to make a difference for highlight & shadow details / color rendering.

However, what I'm least satisfied with, in my Epson 4490, is the awful way grain appears, which I understand is a common problem to many Epson scanners. So my question is: will the use of SilverFast make a difference with respect to this, or does the ugly Epson grain depend on hardware? Please drop a line if you have experience or thoughts about this.

 

[Fernando2]

I hate Silverfast; while I had it bundled with some of my scanners, I always preferred Vuescan Pro and unistalled Silverfast whenever possible.
Vuescan gives at least the same degree of "low level" control on the hardware and the image that Silverfast allows, but in my experience is far more stable, logical and easier to use. Not to say way much cheaper.

That said, a software can't do much to overcome hardware limitations.

As I already shown too many times to repeat again, flatbed like your Epson can't resolve fine details, say beyond 25-30 lp/mm.
Scan software often apply huge amount of sharpening to give the idea of sharpness, but this has the side effect of enhancing grain in a bad way, giving the image a bad look overall.

So you may start by disabling any kind of sharpening on the Epson software.
Then, to try and squeeze a bit more quality: set 16bits/channel output and set the scanner resolution to 4800 ppi (yes it's an overkill but helps in postproduction), import the scan in Photoshop, use a Bicubic Smoother resample to 2400 ppi, switch color mode to LAB, apply a "Smart Sharpen" to the "L" channel only, then convert back in RGB (I'd suggest at least AdobeRGB for maximum color gamut, if you shoot color).

This procedure may help getting better image quality from your flatbed, but the real limitation is hardware.

If you're interested, see my (too many!) other scanner threads to read reviews and comparisons of flatbed scanner, film scanners, drum scanners.

Fernando

 

[Andrea Taurisano]

Thanks Fernando. I know that there is a physical limit to what each scanner can resolve (though I'm still not sure how many lp/mm you actually need to resolve film grain sufficiently nicely). So if I understand you correctly, a scan made following the advice you gave above but with original software shouldn't look any better or worse than a scan made with SilverFast or Vuescan Pro. Right? The next step is looking at better scanners..

 

[Fernando2]

I'm still not sure how many lp/mm you actually need to resolve film grain sufficiently nicely

From what I've seen, at least 100 lp/mm, with no aliasing (otherwise you get grain aliasing, a sort of grain aplification). See my threads on drum scanners vs film scanners for examples (Nikon 8000 vs ScanMate 11000, etc).
Grain size (after developing) on modern emulsions varies from 1 to 10 microns, give or take, depending on film type etc.

So if I understand you correctly, a scan made following the advice you gave above but with original software shouldn't look any better or worse than a scan made with SilverFast or Vuescan Pro. Right?

Well, you don't get the very same quality, because Silverfast and Vuescan give you low-level access to the hardware.
For example, some Epson scanners won't allow single-pass multisampling or variable exposure with Epson Scan, while Vuescan enables them, thus enhancing image quality.
But at least with the suggested workflow you squeeze the most from the hw/sw you have.

Fernando

 

[Andrea Taurisano]

From what I've seen, at least 100 lp/mm, with no aliasing (otherwise you get grain aliasing, a sort of grain aplification).

But this means that even dedicated (but consumer) film scanner like the Plustek 8100 / 8200 or Reflecta RPS7200 Professional may not be really resolving grain, since tests show that their actual resolution is somewhere 3200 - 3800 dpi (60 - 80 lp/mm). Not even the Nikon Coolscan 800 ED according to the manufacturer (and your test) resolves 100 lp/mm. Then I'd better stick to my lousy 4490, or am I missing something? I apologize for my ignorance 🙂

 

[Fernando2]

As I said earlier, I'd prefer not to repeat what I already said in these forums again and again; you can find tons of informations and samples with a search.
There's a large, sometimes huge difference in scan quality between a flatbed and a dedicated filmscanner.

Anyway yes, no available film scanner can really resolve grain.
Not even my drum scanner (Scanview ScanMate 11000, 11000 ppi, 180x125 lp/mm actual resolving power) can resolve grain on many film types.
You'd need a good microscope to obtain that.

But it's not important to "resolve grain": what's important, is to render grain gracefully and to resolve a reasonable level of the details actually captured in your photos.

Actual details depend on camera, lens, shooting technique, shooting conditions, subject, etc.
For example, if your shooting style is about using a top-line camera with top-class lenses, tripod, low sensitivity film, optimal focus and aperture, static subjects with lots of details, then of course you really need a very good scanner to get the most out of your pictures.
But if you like street shooting at waist level, with lens on hyperfocal and look for interesting situations on moving subjects, your scanning requirements may be different.

So it's better to understand one's needs, and buy accordingly, than to struggle aimlessy for the "very best". I'm often caught in this naughty spiral, by the way. 🙂

Fernando

 

[Andrea Taurisano]

Far from me the thought of upgrading my scanner for the sake of upgrading. And I don't want to make anyone repeat what's already said. I actually did check your tests, Fernando. I do see and understand the difference in resolution between drum scanners, dedicated film scanners and flatbed scanners when this is measured with a resolution chart. Charts show a striking difference, yes. But what that means for my negatives and the particularly the rendering of grain, is much less obvious to me, without trying those scanners myself.. But I guess it's simply me who find that step difficult.

 

[Fernando2]

Indeed it's difficult to anticipate the different "grain rendition" given different films, different scanners, different sharpening settings etc.

In general, flatbeds have the less intrusive grain; and with accurate sharpening (maybe masking the sky and other uniform areas and/or with "high-pass sharpening") may give you enough details for small enlargements.

Plustek and Nikon filmscanners give a reasonable compromise between OK grain rendition and details extraction.

Minolta filmscanners are usually the best in details capturing at a given resolution, but have the most intrusive grain renditions, because of grain aliasing (too low sensor sampling frequency vs. lens resolving power).

Of course drum scanners are the best for both aspects, but...! 🙂

Fernando

 

[wallace]

Hi

Many tests on filmscanner.info indicate that SilverFast enables significantly better images than the scanner's original software.

Mr. Wagner, who runs filmscanner.info, wants to sell software (Siverfast).
Personally, I am very satisfied with vuesan and would give it a try.


Thomas

 

[Andrea Taurisano]

Fernando, this is the best I can get out of my 4490 and a 35mm negative using the advice you gave me. It is better than I had got before from the same negative. Sure I will never know how good it COULD be with better scanners, but it is good enough for print of my trade books and I guess for prints up to A4, perhaps more. The only way to know is try and print I guess.. (the output file is about 4500 x 7000 px, for what it's worth)

100% crop (I guess it is the grain of Tri-X we're getting a hint of, although the scanner has too low res to resolve it)

 

[Fernando2]

I tend to agree with Wallace. 😉

Andrea, this is an example (not the best, but better than nothing) of different grain rendition between a Nikon filmscanner (8000ED) and a drum scanner.
The Nikon scan was done at 4000 ppi and then upsampled to match the 11000 ppi scan from the drum.
Film is Provia100F.

If you only had the Nikon, you may think the "grain" you see are the actualy crystals/dye clouds. So you may conclude that you are "resolving grain".

Then you see the drum scan, and the grain is much finer.
So what you see on the Nikon scan is not actually "grain". It's grain aggregates, because grain is not fully resolved by the scanner.

Actually not even the drum scan has the basic grain structure resolved, but it's still better/finer/less intrusive.

Another example: Minolta 5400 filmscanner vs. Epson V700 (with optional adjustable holder and reversed wet mounting, so it's the best possible performance for this flatbed).
In this case, I simulated the need for a 8x enlargement (20x30 cm), so I resampled both scans to have the right size (kind of a "real world" comparison).
You see the "grain" is nonexistent on the V700 scan (also thanks to wet mounting), but details are better on the 5400.

100% crops from the unresampled images:

 

[Fernando2]

Fernando, this is the best I can get out of my 4490 and a 35mm negative using the advice you gave me. It is better than I had got before from the same negative.

If you only shoot BW, you don't need LAB-based sharpening.
But you could still benefit a lot from: a) masking uniform areas, b) using Smart Sharpen instead of Unsharp Mask, c) disabling Epson Scan sharpening d) using Vuescan if it gives you single-pass multisampling (you can try the free demo in Advanced mode to see if it's the case).

Sure I will never know how good it COULD be with better scanners, but it is good enough for print of my trade books and I guess for prints up to A4, perhaps more

As I said, it depends on the subject. For this subject, I agree, it's not a picture which lives or dies depending on the resolved details. Some motion blur, slight focus inaccuracies etc. do more damages to sharpness than scanner does, in this kind of photography.

Fernando