What is the best non-commercial scanner on the market right now?

[regular]

It makes sense, indeed. It is all part of the theory of information.

Following Nyquist's idea, the sampling frequency needs to be at least twice the higher frequency you want to capture.

When applied to sound, it means that you need to sample at 44kHz (audio CD) if you want to capture sounds between 0 to 22kHz.

When applied to graphics, the ideal scanner should have a resolution of half the grain size. Without it, there is aliasing : high-frequency spatial details (grain) are confused with lower-frequency details (subject of the photo).

That is an issue close to "moiré" then.

 

[hanskerensky]

Sure is, and to make matters worse this aliasing can be further "worsened" by the graphics hardware you use to see the image. Guess this is the "moiré" part.

Saw several times that when an image looks acceptable at full resolution view it can start to look terrible when zoomed out at a certain rate.

 

[Kamph]

hanskerensky, I believe we once discussed grain aliasing on Flickr as well!
I found that scanning at max dpi with my Plustek 120 didn't yield any significant advantage in real resolution compared to 5300 dpi (no surprise there really), but that the grain aliasing at times indeed was less promient!
Going by the theory that Fernando presented, this might be because the lens is only capable of around 5000 dpi thus the aliasing problem is reduced when scanning at a higher resolution.

 

[Wenge]

just another sample from Pacific Image XE + Kodak gold 200 scan @5000dpi

 

[hanskerensky]

hanskerensky, I believe we once discussed grain aliasing on Flickr as well!
I found that scanning at max dpi with my Plustek 120 didn't yield any significant advantage in real resolution compared to 5300 dpi (no surprise there really), but that the grain aliasing at times indeed was less promient!
Going by the theory that Fernando presented, this might be because the lens is only capable of around 5000 dpi thus the aliasing problem is reduced when scanning at a higher resolution.

That surely would be a solution when scanning 35mm
However when doing 6x6 or even worse 6x9 the scan processing times will go sky high. I prever 2650 ppi there.

The theory of Fernando sounds very good and that would explain the phenomena.

In my case the aliasing can be most notable in the sky part of an image. If it is very bad then, in Photoshop, i select the sky part and apply a mild softening filter. Works good but is of course yet another step to perform.

 

[mfogiel]

Given what people write about the Sony A7(r), I wonder why Sony, or somebody else, does not produce a contraption, which has all the usual scanner amenities ( film holders, advance motor, etc), plus an interface, where you can stick in this camera with a macro lens. Perhaps they are waiting for sensor prices to go even lower to make this palatable.

 

[Scrambler]

Given what people write about the Sony A7(r), I wonder why Sony, or somebody else, does not produce a contraption, which has all the usual scanner amenities ( film holders, advance motor, etc), plus an interface, where you can stick in this camera with a macro lens. Perhaps they are waiting for sensor prices to go even lower to make this palatable.

Something like this?

All the cheap second-hand slide duplicators will now be worth something - they crop on smaller than full-frame cameras.

 

[k__43]

Given what people write about the Sony A7(r), I wonder why Sony, or somebody else, does not produce a contraption, which has all the usual scanner amenities ( film holders, advance motor, etc), plus an interface, where you can stick in this camera with a macro lens. Perhaps they are waiting for sensor prices to go even lower to make this palatable.

YES!

I'm an electronics engineer, if we find a mechanics guy with knowledge in prototyping I'm willing to join an open source hardware group.

 

[hanskerensky]

There are allready quite a lot of people who do/did experiment with these kind of set-ups.
Consideration points are :
Color Negative reversal. Strong blue filter needed to compensate orange film mask.
Lens : Standard camera lenses not really perfect. You will need very good reproduction (or enlarger) lens as source is on a flat 2D plane. Normal camera lens (even macro) designed more for curved 3D plane.

 

[Fernando2]

The theory of Fernando sounds very good and that would explain the phenomena.

I'd love it to be my theory; it's just Shannon's theorem.

By the way: if you scan at less than the maximum hardware resolution, the aliasing get worse.
Because the lens still pushes whatever maximum frequencies it can resolve, but now the sensor samples the signal at even lower frequency!
Which gives even more aliasing.

Ideally, you should always scan at maximum resolution, then resample down in Photoshop, using Bilinear (even better: prefiltering with appropriate Gaussian Blur).

Fernando

 

[hanskerensky]

By the way: if you scan at less than the maximum hardware resolution, the aliasing get worse.
Because the lens still pushes whatever maximum frequencies it can resolve, but now the sensor samples the signal at even lower frequency!
Which gives even more aliasing.

Ideally, you should always scan at maximum resolution, then resample down in Photoshop, using Bilinear (even better: prefiltering with appropriate Gaussian Blur).

Fernando

Fernando, surely will give that i try with different resolution settings.
One question though, shouldn't it be that you scan with a resolution that is higher then the optical resolution ?

For instance my Canon 8800F flatbed scanner has an optical resolution somewhere around 1700ppi. That would be a waist of scanning time if i did a scan at 4800ppi resolution.

 

[Fernando2]

Fernando, surely will give that i try with different resolution settings.
One question though, shouldn't it be that you scan with a resolution that is higher then the optical resolution ?

For instance my Canon 8800F flatbed scanner has an optical resolution somewhere around 1700ppi. That would be a waist of scanning time if i did a scan at 4800ppi resolution.

Indeed, one should not overdo; if a scanner has an actual resolving power of N, sampling at 1.3*N is all what's needed to prevent aliasing.

Usually flatbeds have optical resolutions way lower than sensor resolution; so they almost never show aliasing, except if set at very low resolutions.

Professional scanners usually have optical and sensor resolution well matched.
And drum scanners have adjustable optical resolution (aperture control) so the operator can always avoid aliasing by matching aperture and sampling resolution.

Fernando

 

[Fernando2]

From a couple of messages I received, I see I was not clear about the issue of scanning resolution.

Let's say we test a scanner with a resolution chart (I use the FSR-1T from Danes Picta), and see that at best it resolves 50 lp/mm.

So at what resolution we should sample, to avoid aliasing for sure?

First we have the sampling theory, which says the minimum sampling resolution must be 2x the maximum signal frequency.
Then we add a safe margin, let's say 25%.

So we and up at 2.5x50 samples/mm = 125 sampled points/mm = 3200 ppi (3175 to be precise).

Sampling at more than 3200 ppi for that particular scanner would be useless.

Sampling at less than 3200 ppi could introduce aliasing, most notably grain aliasing because grain is the smallest "detail" in a film image.

Since the lens has a fixed resolving power (unlike in drum scanners), scanning at low resolutions with CCD scanners is much worse in the (grain) aliasing department, because you are sampling at less than the Nyquist frequency, and are sure to get signal aliasing.

Fernando

 

[hanskerensky]

Thanks for sharing that knowledge Fernando !

Well, i have little hope for my Plustek Opticfilm 120 then. With an optical resolution of about 3650ppi i would need another sensor to scan with 2.5 times that value :-(

Sure will do some test scans with different resolutions on one of my more notorious aliasing photos

 

[Kamph]

hanskerensky, 3650 dpi? When tested mine was about 4800 dpi when scanned at 5300 dpi, and just about 5000 when scanned at 11k.
Some of your photos on you flickr stream does show significant amounts of grain, but then again most of them are scanned at 2650 dpi. I always scan at 5300 dpi, and i haven't seen as excesive grain as some of your scans exhibit. It'll be exciting to see wheter you will notice a diffrence scanning at a higher dpi!

 

[Fernando2]

With an optical resolution of about 3650ppi i would need another sensor to scan with 2.5 times that value :-(

No, because that 3650 ppi figure of yours already includes the 2x sampling ratio: it translates to 72 cycles/mm, which including the margin would need 4600 ppi to be scanned without any aliasing.

That said, seems a bad figure to me.
Looks like your unit has the focus plane a bit off vs. the film plane.
You could try shimming the film gradually to see if you gain sharpness.

Fernando

 

[James1]

Following Fernando's comments on this thread, I decided to rescan a few 35mm negatives at 10,600 ppi that I had previously scanned at 5,300 ppi. My usual workflow is to scan any negative as a RAW colour image in Vuescan, extract the green channel if B&W, then invert using ColorPerfect running under Photoline.

My test was in no way scientific but I am certain that the scans at 10,600 ppi are smoother and have more detail.

I pre-filtered by adding a little Gaussian Blur (1.5/100% - no idea if this is not enough / too much but it didn't really alter the image much) then downsampled using the bilinear algorithim to 7000 ppi.

There was very little difference to the image at 10,600 ppi, and the file was of course considerably smaller. Downsampling in the same way to 5,300 ppi also produced a good image which was probably better than a direct scan at 5,300 ppi.

I am happy with my OP120 - I seem to have a good example. I am sure that the differences I noticed are real, so I will more than likely scan at full resolution, and downsize. It didn't take that much longer.

Only thing I need to now get is a better computer as my current set-up doesn't have enough power to cope with the monster-size files that I'd get when scanning medium format in the same way....

 

[hanskerensky]

hanskerensky, 3650 dpi? When tested mine was about 4800 dpi when scanned at 5300 dpi, and just about 5000 when scanned at 11k.
Some of your photos on you flickr stream does show significant amounts of grain, but then again most of them are scanned at 2650 dpi. I always scan at 5300 dpi, and i haven't seen as excesive grain as some of your scans exhibit. It'll be exciting to see wheter you will notice a diffrence scanning at a higher dpi!

Ok, in the mean time made some testscans. 1325, 2650 and 5300ppi. Results can be seen here :
http://www.flickr.com/photos/29504544@N08/14219791246/sizes/o/

The results start me wonder if that "grain" effect is really caused by Aliasing as the 5300ppi scan looks the worst of all.
Maybe something in the Fujifilm Reala 100 filmlayer which drives my OP120 nuts ?

 

[Fernando2]

Really intriguing.
What if you try 10600, following James1's findings?

BTW the "grain" is so sharp, it's strange you measured such a low actual resolution.

Ok, in the mean time made some testscans. 1325, 2650 and 5300ppi. Results can be seen here :
http://www.flickr.com/photos/29504544@N08/14219791246/sizes/o/

The results start me wonder if that "grain" effect is really caused by Aliasing as the 5300ppi scan looks the worst of all.
Maybe something in the Fujifilm Reala 100 filmlayer which drives my OP120 nuts ?

 

[Kamph]

That sure is a lot of grain! I have never seen anything quite like that on my P120 unless my negatives were really underexposed, but then again I only shoot Kodak negative film, so could potentially be a problem with Fuji's emulsions and the light source in the OP120.