Jump to content


Photo

4k scans for Super 8


  • Please log in to reply
10 replies to this topic

#1 Brian Drysdale

Brian Drysdale
  • Basic Members
  • PipPipPipPip
  • 4858 posts
  • Cinematographer

Posted 23 February 2017 - 09:38 AM

The possible advantages of higher resolution scans (this article goes into 5k scans at the end).

 

https://dl.dropboxus...onse - 2012.pdf

 

Note that 4k scanning for S8 is being offered in the near future by Spectra Film and Video, an established Laboratory in North Hollywood.


  • 0

#2 Michael Lehnert

Michael Lehnert
  • Sustaining Members
  • 1085 posts
  • Other
  • London, UK / Basel, CH

Posted 23 February 2017 - 10:01 AM

Excellent paper, and very good pictorial evidence that is very helpful.

 

For more than 20 years now, people have been saying that the then contemporary highest resolution achievable "is totally sufficient" or is already "overkill" for Super 8… with early VTR, then telecine transfer to Betacam video, then first progressive HD scans, 2K, 4K… man, just you all recall the shitstorm in this forum when Santo put his "Santo's Method" forward: people went nuts… "720p for Super 8? Nonsense!!"

 

I just think of all the people who transferred their S8 "home movies" in the 1990s onto S-VHS, and then threw the rolls away because "S-VHS is superior to Super 8 already" – German video amateur and pro magazines were filled with adverts and letters to the editor on this subject.

 

We are still going to have this discussion on the worthy transferability and resolving power equivalence from one analog format into another digital medium when 8K and 16K will be the regarded as the digital resolution of the day.

 

2K for S8 should be default for projects who are not willingly out to use the format for "degraded low-res aesthetics". Kodak going for 2K with their new in-house lab eco-system is a good indicator to support this practice.


  • 0

#3 Tyler Purcell

Tyler Purcell
  • Sustaining Members
  • 2977 posts
  • Other
  • Los Angeles

Posted 23 February 2017 - 12:51 PM

Twas a good read for sure, but I do think the results were dubious.

First off, nobody is watching a still frame of your movie, they are watching a moving image. So resolution means very little if the frame is constantly jumping around from lack of decent registration.

Second, if an image was scanned at 4k and then scaled to reproduce 1080p, it will look entirely different then a non-scaled image from a true 2k scanner for instance. Where the glass and imager are designed to capture a full-frame image at 2k.

Third, the resolution chart was from 16mm scans. We all know that modern 16 can resolve 2.5k and it's been tested time and time again. The chart was crisper, but I could still see the same lines at 1080p as I could at 3k. So you aren't gaining more resolution.

Since super 8 is HALF the size of 16mm, you'd just make the assumption there is HALF the data within it. Since 4k is way overkill for 16mm and test after test shows this, that would mean 2k would be over-kill for super 8.

I mean you can scan super 8 at 4k all you want, but all you'll end up with is a file that looks more noisy and can't be compressed well because the MPEG compression systems we use today for most distribution can't deal with the film noise very well. It just turns the noise into mud and it really looks awful.
  • 0

#4 Nick Collingwood

Nick Collingwood
  • Basic Members
  • PipPip
  • 27 posts
  • Other
  • New York, NY

Posted 24 February 2017 - 06:27 PM

Tyler, I'm admittedly an amateur in comparison to your experience level and not trying to be inflammatory but I'm truly perplexed by your posts on this forum. You seem to hate Super 8 for all that it is. And want it to stay in its place and do nothing more than crappy home movies and only 16mm and up are really legitimate formats for true filmmakers.

 

If you seriously can look at the 720p vs 1080p vs 2k crops of the old man's face and tell me there's not a sharpness/resolution/image quality difference, then I don't know what will convince you... And seeing the same lines has nothing to do with it. For a digital comparison, I can see the take a photo at 2mp of a mug and a photo at 12mp and while I can still see the same mug in both photos, there's an obvious difference in resolution.

 

I can say from my own Super 8 scans that I noticed a dramatic improvement in image quality when I first got a 2k scan vs all of the 1080p telecines I had done before that. The difference was quite stark. Seeing my Tri-X scanned at 2k looked much more similar in grain/sharpness to projecting it on my Elmo ST1200 than the 1080p telecines.

 

With that said, I don't see as big of a jump between 2k vs 4k but there's no reason not to other than cost. I do all my scans at 2k overscan currently as I like the ability to see the full image captured. You can also apply some image stabilization in post to help with the frame jitter but not everyone detests it as much as you might.


Edited by Nick Collingwood, 24 February 2017 - 06:29 PM.

  • 0

#5 Tyler Purcell

Tyler Purcell
  • Sustaining Members
  • 2977 posts
  • Other
  • Los Angeles

Posted 24 February 2017 - 06:48 PM

I agree with you, there is absolutely a difference, but I don't think this test really exemplifies that difference.

Again, if you digitally degrade an image to prove a point, you aren't proving the point scientifically, you are manipulating the image to prove the point you wish to make and nothing else.

Also remember that "scanning" and "telecine" are two completely different processes and a scanner will generally deliver a better over-all image in both color accuracy and crispness due to it stopping for every frame in a pin registered method.

My point above is simply about resolution, no more, no less. Film can resolve greater crispness/sharpness using a higher resolution scanner, but is the added information beneficial or detrimental? I personally think it's detrimental because all you're getting is noise and grain, which is something you can't just process out and make look good. The noise and grain makes encoding and viewing very complicated. It's far better to scan at the max resolution (based on the line chart) your format can handle in my opinion, then have all that noise to deal with.

Since we know that vision 3 50ISO 35mm 4 perf negative resolves a little bit north of 5k, it's very easy to do the math on the narrow gauge formats based on that data. Super 16 is around 2k, though over-scanning at 2.5k is smart for cropping to 1.85:1 reasons. Which makes super 8... half of that?

The only way to actually test this is to take still pictures of a resultion chart with super 8 film (so you take out the registration and gate issues the cameras have), then scan those still images to see where the resolution stops. On the above sample, the test was bogus because it was old stock... I've seen tests with modern stock, but not on super 8, just 4 perf 35mm.
  • 0

#6 Robert Houllahan

Robert Houllahan
  • Sustaining Members
  • 1561 posts
  • Industry Rep
  • Providence R.I.

Posted 25 February 2017 - 09:52 AM

Nyquist theories sampling means that you really want to sample at 2 times the target resolution to achieve the best results.

 

I think 2K is a great resolution for 8mm film and when that 2K is scanned in the scanner at 4K it results in a better scan with more finely resolved grain.

 

Also most of these scanners are Bayer mask 5K sensors and use some sharpening. I think the general rule of thumb is that a Bayer sensor really resolves 3/4 of the overall sensor resolution.


  • 0

#7 David Cunningham

David Cunningham
  • Basic Members
  • PipPipPipPip
  • 1029 posts
  • Cinematographer

Posted 25 February 2017 - 03:43 PM

I defy anyone to convince me that the 5K sampled Super8 that I finish at 1080p is not sharper than an actual 1080p scan. The difference is night and day for me. I'm ready for 10k LaserGraphics Director Super16! Bring it on! I might finally have a Super16 image that looks as crisp as my Super16 timed prints. Projected Super16 by far exceeds and 1080p image I've ever seen. Even with 50D there is considerable grain, but it's actually quite pleasing. I wish I had the money to shoot 65mm and make and project 70mm prints. Maybe after I hit the lottery. :)
  • 0

#8 David Cunningham

David Cunningham
  • Basic Members
  • PipPipPipPip
  • 1029 posts
  • Cinematographer

Posted 25 February 2017 - 03:47 PM

Have any of you seen the 8K scanned wizzard of oz? The 1080p Blu Ray looks like it could have been shot yesterday. Thanks to modern technology, the three BW negatives could be combined post scan with absolute precision like was never possible in the 30s. We can see the wizzard or oz today better than anyone saw it back then.... and I bet a 4K presentation would be that much better.
  • 0

#9 Anthony Schilling

Anthony Schilling
  • Basic Members
  • PipPipPipPip
  • 1030 posts
  • Camera Operator
  • Portland, OR

Posted 25 February 2017 - 04:31 PM

The only thing in this world that goes down in price over time is technology. As a DIY shooter (like almost all S8 shooters) the question isn't really about resolution, but do the costs outweigh the benefits? For me to upgrade my equipment from 1080P to 2K, the answer is no. I'll wait for the tech to get cheaper, but by then i'm sure 8K will be all the rage. I'm content with a 1080P S8 image and more focused on post scan processing, because that's where your image really counts. 


  • 0

#10 Jeremy Cavanagh

Jeremy Cavanagh
  • Sustaining Members
  • 126 posts
  • Other

Posted 26 February 2017 - 09:30 AM

I think 2K is a great resolution for 8mm film and when that 2K is scanned in the scanner at 4K it results in a better scan with more finely resolved grain.
 
Also most of these scanners are Bayer mask 5K sensors and use some sharpening. I think the general rule of thumb is that a Bayer sensor really resolves 3/4 of the overall sensor resolution.

Robert, I'm building my own DIY scanner for 8 mm and I'm intending to use a mono machine vision camera to do RGB frames and then combine them in software as I don't like the loss in resolution from a bayer sensor (as brilliant a technology as it is). I've been warned off this approach and told that just getting a larger bayer based sensor will get over the loss of resolution however I'm not convinced about this and I was wondering what your opinion was (BTW, I'm an engineer so not afraid of technicalities.......no, a real engineer, not someone who has spent their professional life behind a keyboard.......).
  • 0

#11 Robert Houllahan

Robert Houllahan
  • Sustaining Members
  • 1561 posts
  • Industry Rep
  • Providence R.I.

Posted 26 February 2017 - 11:13 AM

Personally I think a true RGB scan is the best as it is a full image record of each film color channel and it also avoids the interpolated color of a Bayer mask camera.

 

That said I think it depends on the monochrome camera you are using and how well the data from the camera is mapped into a DPX or TIFF frame and what kind of R,G,B illumination you use.

 

The 5K camera in the Scan Station is a $10K camera and the 5K CCD in the color  Xena is about $12K and runs at 2fps. Those Bayer mask cameras will outperform a lower cost cmos monochrome camera in terms of noise and DR so YMMV.

 

I would look for a Imperx or Allied monochrome CCD CameraLink or Gig-E camera on eBay you can pickup a very high end mono machine vision camera for a fraction of the new retail price.


  • 0


ZoomCrane

FJS International, LLC

rebotnix Technologies

Aerial Filmworks

Tai Audio

Quantum Music Works

Glidecam

Abel Cine

Broadcast Solutions Inc

CineTape

Gamma Ray Digital Inc

Willys Widgets

Visual Products

Ritter Battery

Paralinx LLC

CineLab

Zylight

Technodolly

The Slider

Rig Wheels Passport

Pro 8mm

Metropolis Post

Rig Wheels Passport

Pro 8mm

Abel Cine

Willys Widgets

Technodolly

Tai Audio

ZoomCrane

Ritter Battery

Glidecam

Zylight

Aerial Filmworks

Visual Products

Quantum Music Works

Paralinx LLC

The Slider

CineLab

FJS International, LLC

CineTape

Gamma Ray Digital Inc

Metropolis Post

Broadcast Solutions Inc

rebotnix Technologies