Scan of Grain Texture at 11'000ppi

[dokko]

I recently did a comparison of Kodak T-Max 100 vs Delta 100 developed in Adox FX-39II, Adox XT-3, and Spur HRX. Since the discussion involved scanning questions, it was suggested that I'd start a new thread in the scanning forum. The original discussion with the detailed info about the film development can be found here:

Comparison of Kodak T-Max 100 vs Delta 100, in Adox FX-39II, Adox XT-3, and Spur HRX

I thought this might be interesting for some, so I created a separate thread from the discussion at https://www.photrio.com/forum/threads/similarity-between-fx-39-and-ilfosol-3.198293/ I ran a film/developer comparison recently because I wanted to check two new magazines for my Hasselblad and...

www.photrio.com

I scanned the films at 11'000ppi on my custom built scanner. to make things easier to see, I've created a animated gif at 200% scale since many people will watch this on a high ppi screen (open the link below in a new window for larger display size). no digital sharpening has been applied:

https://live.staticflickr.com/65535/53276525073_0971bd39ab_o.gif

individual jpegs can be found here (go to download → original size for the unscaled unsharpened version):

Film Comparison

Explore this photo album by dokko scan on Flickr!

www.flickr.com

I've also created a layered .tif file with all the samples in one file:

Dropbox

www.dropbox.com

 

[MattKing]

Thanks for starting the separate thread. When I made the suggestion, it was oriented more toward the particulars of your scanning approach than the example results.

 

[_T_]

I am very interested in your scanning setup and procedure. 11,000ppi is an impressive resolution.

Can you give us a rundown of what you’ve done to achieve this and how you have measured the performance?

 

[dokko]

my scanner is a proprietary custom build machine that I've spent more time and money on than I would like to admit to myself, so unfortunately I can't share specific details.
But I'm happy to discuss different scanning methods in general and share my findings about the various challenges.

there's really nothing groundbreaking in the general principle of my scanning method itself though, it's just a light source, a lens and a sensor, like nearly every other scanner (apart from drum scanners, which by now are 20-30 years old).

the problem arises once we want to get more than 5000ppi in resolution, everything gets very difficult very fast. it needs extremely careful optimisation and very expensive components. for example, only the components of the optical system of my setup cost about 40'000 EUR.

as a comparison, the Imacon/Hasselblad scanners, that sold for up to 25'000EUR until 5 years ago, uses a lens worth around 1000EUR.

here is a comparison of a 35mm negative (Adox CMS 20 II) scanned on an Imacon 848 at 8000ppi and on my dokko scanner at 11000ppi, scaled to the same size for comparison. first the overview of the full frame with the crops that follow marked in red:

my scanner at 11'000ppi:

and the Imacon at 8000ppi:

(as a side note, the distortion of the Imacon scan comes from problems in its line scan mechanism).

as I mentioned in the other thread, my scanner is able to scan up to 40'000ppi. However, such an extreme resolution is usually not worth the effort since is very time-consuming and doesn't always result in a better looking image. I found that the sweet spot for extremly large prints is around 10'000-20'000 PPI for 35mm and 7000-14'000 PPI for medium format and large format film.

the main reason why I started this endeavour was because I was frustrated by the digital scanning artefacts of the existing scanner models and wanted something which kept the organic nature of analog film in a hybrid workflow, even on high enlargements.

 

[Alan Johnson]

The Imacon scan looks like it could be improved with some sharpening.
Is this not usually done with this brand of scanner?

 

[dokko]

The Imacon scan looks like it could be improved with some sharpening.
Is this not usually done with this brand of scanner?

The Imacon scanner always adds digital sharpening, which can't be turned off. There has been quite a some discussions in forums that said that you have to set the software to -120 (the minimum) to disable it, but a simple test comparing sharpening se to 0 and -120 shows that -120 actually adds a software blur to smooth out grain (it looses image detail which can't be recovered later). the above 8000ppi sample has been scanned with sharpening set at 0, which adds a certain default digital sharpening by the scanner software. of course we can add more, but it will start to look like an image with a lot sharpening applied, which I personally dislike.

The problem is, that their scanner lens only resolves about 6000ppi for 35mm to start with.
A normal film scanned with their 8000ppi setting looks a bit sharper than the 6300ppi setting, but if you examine the files in detail, it's mainly due to a digital pixel pattern they use to make the grain look sharper rather than true film resolution.
with the Adox CMS20 II having nearly no visible grain at 6000ppi, that approach falls short.

one can start to see the Adox CMS20 II grain at 20'000ppi though

 

[Alan Johnson]

@dokko, you mention:

"the problem arises once we want to get more than 5000ppi in resolution, everything gets very difficult very fast. it needs extremely careful optimisation and very expensive components"

Is this the reason why scanner manufacturers appear slow to provide scanners of higher true resolution? Thanks.

 

[chriscrawfordphoto]

The Imacon scan just looks like its out of focus or has a crappy lens. My Nikon 8000ED scanner gave sharper results than that, and its supposed to be inferior to the Imacons.

 

[dokko]

@dokko, you mention:

"the problem arises once we want to get more than 5000ppi in resolution, everything gets very difficult very fast. it needs extremely careful optimisation and very expensive components"

Is this the reason why scanner manufacturers appear slow to provide scanners of higher true resolution? Thanks.

I don't know the reason for the lack of a new, dedicated high-end film scanner in the last 20 years.. my guess is that the market is too small to warrant the R&D, manufacturing and support.

but yes, one reason could well be that it needs very expensive parts and tight tolerances to get to 10'000ppi and above, which means it's pretty much impossible to sell one for an affordable price.

there is a copy stand scanning solution from PhaseOne, which sells around 100'000EUR, and it doesn't reach 10'000ppi (to be fair it's designed more for versatility in mind than highest resolution)

 

[dokko]

The Imacon scan just looks like its out of focus or has a crappy lens. My Nikon 8000ED scanner gave sharper results than that, and its supposed to be inferior to the Imacons.

the lens in the Nikon 8000ED is actually significantly sharper for 35mm and 120 formats than the lens in the Imacon scanners.

the main reason for this is that the Nikon has been optimized for one single format (120), so the lens is highly optimized and actually outperforms the 4000ppi rating of the sensor quite a bit.
but since for 35mm the scanner simply crops the sensor, you still only get 4000ppi even on the smaller format.

The Imacon was built to scan from 13x18cm or 5x7" film down to 35mm using optical enlargement, but since it doesn't have floating element it performs quite poorly at 8000ppi, resulting in scans which have lower resolution than the advertised number.
This is actually rather easy to explain since the lens has a maximum aperture of F5.6, which means diffraction kicks in even if it were a perfect lens (which it isn't) and we only get around 6000ppi.

also note that the Imacon scan of the example above has been upscaled to match the size of the 11'000ppi scan, which of course shows the softness more than if it would be displayed at smaller size. but it only seems logical that an image of around 6000ppi looks softer than one with 11'000ppi.

 

[dokko]

by the way, I scanned that negative several times on the Imacon to make sure it wasn't a focus problem. Comparing the results made me realize that is also has a serious problems with distortions since the line scan mechanism is not fully uniform. here the image scanned twice with the same settings:

for comparison, here also the same image scanned at 6300ppi.

 

[warden]

the lens in the Nikon 8000ED is actually significantly sharper for 35mm and 120 formats than the lens in the Imacon scanners.

the main reason for this is that the Nikon has been optimized for one single format (120), so the lens is highly optimized and actually outperforms the 4000ppi rating of the sensor quite a bit.
but since for 35mm the scanner simply crops the sensor, you still only get 4000ppi even on the smaller format.

The Imacon was built to scan from 13x18cm or 5x7" film down to 35mm using optical enlargement, but since it doesn't have floating element it performs quite poorly at 8000ppi, resulting in scans which have lower resolution than the advertised number.
This is actually rather easy to explain since the lens has a maximum aperture of F5.6, which means diffraction kicks in even if it were a perfect lens (which it isn't) and we only get around 6000ppi.

also note that the Imacon scan of the example above has been upscaled to match the size of the 11'000ppi scan, which of course shows the softness more than if it would be displayed at smaller size. but it only seems logical that an image of around 6000ppi looks softer than one with 11'000ppi.

I just took a look at your web site and now I’m wondering if anyone has ever purchased a scan from you that required 108,000 pixels on the long dimension. What might that be for? Just curious as I can’t imagine. I’m also wondering what would happen if I tried to do something simple, like a curves adjustment in Photoshop for instance, with a 55 GB image.

 

[warden]

by the way, I scanned that negative several times on the Imacon to make sure it wasn't a focus problem. Comparing the results made me realize that is also has a serious problems with distortions since the line scan mechanism is not fully uniform.

Is there any chance that the film sagged? I don’t know the Imacon system but that difference between the two images is odd. Are the two images the same overall pixel dimensions or is one considerably larger in only one direction?

 

[dokko]

I just took a look at your web site and now I’m wondering if anyone has ever purchased a scan from you that required 108,000 pixels on the long dimension. What might that be for? Just curious as I can’t imagine. I’m also wondering what would happen if I tried to do something simple, like a curves adjustment in Photoshop for instance, with a 55 GB image.

yes, I actually made a 45GB scan of a 8x10" slide recently. It was printed 25meter wide for an exhibition in Jeu de Paume in Paris.
It was 120'000pixels on the long side, but cropped vertically.

spotting and retouching was very time consuming since the slide had a lot of micro scratches from an earlier drum scan. I had to do this in parts because even a beefy workstation with 128GB or RAM started to swap after a few strokes.
curve adjustments on the full size image worked quite well, but I had to paint quite a few masks, which I had to do on a lower resolution image, then upscale it, copy it over to the full resolution image... and then press save very quickly and pray

 

[dokko]

Is there any chance that the film sagged? I don’t know the Imacon system but that difference between the two images is odd. Are the two images the same overall pixel dimensions or is one considerably larger in only one direction?

no the total pixel size of the two scans is the same. the distortions repeat in a wave pattern, ie some areas get stretched and other compressed. you can see this on the figure on the left: the foot and eye stay nearly in the same position, the space in between and above warps.

I've repeated this test many times with different types of film as I used to scan with Imacon scanners for years. I never managed to get scans which matched when overlayed (this was actually another reason why I started to build my own scanner).

 

[warden]

…. but I had to paint quite a few masks, which I had to do on a lower resolution image, then upscale it, copy it over to the full resolution image... and then press save very quickly and pray

Holy cow! I thought it would have to be a picture larger than my house, and in fact it is.

 

[Steven Lee]

@dokko why can't you share the most interesting thing?! the details of your home made scanner, of course. All other scanning methods have been discussed to death here. We need something new!

 

[NortheastPhotographic]

Why not post a picture of your scanning set up? It would be interesting to see.

 

[snusmumriken]

This is all awe-inspiring. So the "graininess" one can see on your enlargements of very small areas: is that the graininess inherent in the film, or are we still looking at aliasing patterns, even with such a high pixel count?

 

[dokko]

@dokko why can't you share the most interesting thing?! the details of your home made scanner, of course.

As mentioned it's a proprietary design that I've spent tons of time and money optimizing. When I started I didn't think that this would become a business at some point and did it purely because I wanted to get organic looking scans of my own pictures, but at some point I realized it offers potential for scan quality that isn't available anywhere else.

 

[dokko]

This is all awe-inspiring. So the "graininess" one can see on your enlargements of very small areas: is that the graininess inherent in the film, or are we still looking at aliasing patterns, even with such a high pixel count?

that's an interesting question. what I've found is that the appearance of the grain in films depends on a lot of factors...

for example, even on fully analog prints, grain will look different on different print sizes, different enlargers (condenser vs diffusor box) and to some extent even on different papers.

or for film scans, it will look different when viewed on an older 96ppi screen or a newer retina screens, or when printed on a lightjet or on a inkjet.

with scans, the tricky bit is to differentiate which details are coming from the film itself, and which are from the scanning method, ie noise, aliasing, synthetic pixel pattern.

as an example:
I scanned with an Epson V850 for a long time, which has quite soft optics, so scanning higher than 2400ppi didn't give me any more detail of the image, but it introduces a small synthetic pixel pattern that could be mistaken for image detail because it looked less soft.
similarly, drum scanners add quite a bit of noise that people often mistake for grain texture.
and the Imacon/Hasselblad scanners are very good at assembling the pixels so that the grain looks sharp and detailed, but it's quite edgy and not really the original shape of the grain.


in the examples of the T-Max 100 and Delta 100 scans above, if you open the images in a software that allows you to view them at 100%, you will see the true grain structure with practically zero noise or synthetic digital pixel pattern at this specific resolution.
there will always be some aliasing in any digitally sampled image (or sound for that matter), but if the frequency is high enough it's not going to be visible (that's the concept of oversampling).

I'll see that I can scan a detail in 20'000 and 40'000ppi next time I've set up the system for that resolution (takes quite a bit of time to adjust). at 40'000ppi the grain in many films (specially grainy color negative) starts to become less defined and more like clouds, so while we could say that is the "true" shape of grain, it's not how we usually perceive it (those who have seen images of films under a microscope will know what I mean).

 

[Alan Edward Klein]

For whatever it's worth, here's a comparison I did with Tmax 100 between an Epson V850 and Howtek 8000 drum. They compare favorably, considering.

Howtek 8000 Drum vs. Epson V850 flatbed scanners

Here's a comparison between Howtek HiResolve 8000 drum scanner and Epson V850 flatbed. Negative is mine, Tmax 100, processed normal by North Coast Photo dip and dunk with Clayton F76+ developer (similar to Kodak D76). Taken on a Chamonix 45H-1, Schneider Symmar APO 150mm with B+W orange...

www.largeformatphotography.info

 

[Steven Lee]

As mentioned it's a proprietary design that I've spent tons of time and money optimizing. When I started I didn't think that this would become a business at some point and did it purely because I wanted to get organic looking scans of my own pictures, but at some point I realized it offers potential for scan quality that isn't available anywhere else.

Even better! Will it be possible to purchase your scanner at some point?

 

[dokko]

Even better! Will it be possible to purchase your scanner at some point?

that's unfortunately rather unlikely...

just the bare parts needed to build it cost around 50'000 euros. If I would factor in R&D, assembly, polishing up the software, warranty, training and support and still be able to pay rent, I'd probably have to charge a fortune for it :/

besides, it would turn into a manufacturing/sales business, which I'm not very good at. I enjoy the process of making really nice scans of good images much more.

btw, I just read through your comparison of the X5/IQ3/A7RIV at:

Comparing Imacon X5, Creo iQ3 and Sony A7R IV scans

This thread is derived from this one, where I was comparing the ease of inverting color using camera and Imacon scans. Some of you have asked to see the full-sized crops to evaluate the available resolution. WARNING: I am not a resolution aficionado. Back in my digital days I never felt the...

www.photrio.com

shipping over the atlantic is always a hassle, but if you're interested feel welcome to send me your negatives and I'll run it through my dokko scanner.

 

[alanrockwood]

As a side issue, or maybe it's really a central issue, the results in this thread should permanently put to rest the idea that once you get to 4000 ppi you have extracted all of the information the film can give.