Jump to content

Best Super 8 footage you have ever seen (I bet)


Friedemann Wachsmuth

Recommended Posts

Make sure to watch this in highest quality and fullscreen:
https://youtu.be/3Nh9BTMWj9M


Or better download this version and watch it offline:

https://drive.google.com/file/d/0B9PJ3Kp3qOe4UklQZ0NrM1dNaVk/view?usp=sharing

If anybody interested, I can also share an ungraded ProRes 4444 snippet.

All the tripod shots where shot with the Optivaron (recognisable by the slight vignette in top right, since I did not take the lens hood off). All handheld shots (mostly in the middle) were taken with the Fujinon lens.

The incredible folks from http://www.gammaraydigital.com/did the scanning, once in 2K and once in 4K. Unfortunately, my Mac Pro (late 2008) is a bit too old to handle 4K video, so I had to grade the 2K version. It still looks amazing though, make sure to download the 2K version for full quality!

The clip was stabilized (tracking marker put on the frameline) since the scanner software used currently does not register based on the +2 sprocket hole, but on the sprocket hole next to the visible frame. You can see how bad the sprockets in this cartridge were if you watch the weaving sprocket hole on the left. But watch how image and frameline are like welded together: Absolutely no jitter, weaving and a sharpnes I have never seen before with any other Super 8 camera. This is due to the registration pin -- so thanks to Denmark for finally making a S8 camera that has this professional feature!


If you don't have a Logmar yet, order one now. I tell you, the achievable quality is abso**(obscenity removed)**inlutely addictive. :)

  • Upvote 1
Link to comment
Share on other sites

  • Premium Member

I wonder how much of the achievable quality is down to the camera, though. The overwhelming majority of the technology is in the lens and the filmstock. As is inevitable with super-8, the registration, looking at that sprocket hole dancing about, is poor.

 

P

  • Upvote 1
Link to comment
Share on other sites

I wonder how much of the achievable quality is down to the camera, though. The overwhelming majority of the technology is in the lens and the filmstock. As is inevitable with super-8, the registration, looking at that sprocket hole dancing about, is poor.

 

The registration in this version is misleading, and you shouldn't go by the sprocket hole in this case. Here's the deal: Kodak's perforations are not perfect for Super 8 - their placement relative to the edge of the frame varies from perf to perf. It's a 5-perf cycle, resulting in a kind of sawtooth pattern. In a scanner like the ScanStation, which uses the perfs to digitally register the frames, the picture zig-zags back and forth while the perf is held perfectly still. You *don't* see that in the video above, because Friedemann has further stabilized it in software. That's likely the result of the slight rocking of the image that you see from time to time in this example.

 

It's because of this post-scan stabilization that you see the perf wiggling back and forth. The scanner actually holds it perfectly steady, and the image moves instead.

 

This perf placement problem is specific to Super 8 - you don't see this problem on 16, 8mm, 35, etc. Kodak, for their part, say it's within SMPTE specs, and basically don't seem interested in tightening up the tolerances. That leaves it up to the scanner manufacturers to deal with the problem.

 

It's worth noting that this is not a problem with most Super 8 cameras, only the ones that produce exceptionally steady images like the Logmar. That's why nobody has ever really complained about this before, because it's only in recent years that the camera and scanner quality for Super 8 has gotten to the point where this underlying issue is exposed.

 

It'll get worked out, so that post-scan stabilization isn't necessary, and when it does, we'll post some examples. Lasergraphics is working on a fix for this, which we hope to have soon. We've been working closely with them, to generate test footage to help come up with a workable solution.

 

-perry

Edited by Perry Paolantonio
Link to comment
Share on other sites

  • Premium Member

I can see the slight rotations that happen very smoothly but are the signs of software stabilization. The only reason I can even notice it is because the whole frame is visible. If it was cropped down to just the image you'd never know.

 

Great looking footage, I would still guess Super 8 if I had to but it is really close to regular 16mm.

Link to comment
Share on other sites

In this example the pin-registration of the camera isn't being exploited, other than to ensure the frame is stationary during exposure (the image sharper than it otherwise might). Rather, registration is being done entirely by digital stabilisation, because while the camera uses a +2 offset perf for pin-registration (as per smpte standard), the scanner doesn't. The scanner uses the adjacent sprocket (non-standard). I'm currently building an optical printer, for blowing up Super8 to 16mm, so will be making sure I exploit the camera's pin-registration, ie. use the +2 offset perf for registration.

It remains to be seen whether vagaries in the Kodak perf are actually a problem. Perfs are allowed to weave left/right as perfs are not used for horizontal registration (in camera or projector). Rather the edge of the film is used for that - the other edge being on a spring guide to allow for vagaries in the width of the film. For vertical registration It's only the top/bottom of the perf that is used.

 

One approach a post pipeline might pursue (and I might pursue this) is to have the film overscanned so that not only is the edge of the film in the capture but parts of the adjacent frames as well. Software can then use stitching methods to stitch a number of captures together (so to speak because really it's just computation of some offset vectors) in order to relate a given frame to the +2 perf with which it was otherwise mechanically registered.

 

C

Edited by Carl Looper
Link to comment
Share on other sites

A downside with the proposed stitching method is that it would necessitate a corresponding decrease in pixels allocated to the image area.

 

So it will need to be tested - to see if the benefit of reaching the correct registration sprocket outweighs the cost of such (the decrease in pixels allocated to the image area).

 

Using the frameline for registration isn't a bad solution. Constraining the stabilisation to translational motion will get rid of the rocking. But it remains to be seen how arbitrary shots fare. Friedemann's use of a tripod helps the tracker. And that inadvertant grit in the gate would also help considerably (the irony).

 

But what will be the situation with a clean gate and an erratic shot?

 

Or a night time shot in which one can't see the frame line. Indeed that last example would also fustrate the stitching proposal.

 

I guess that's where the artist steps in. As much as we might like to geek out out on automation, the artist remains (as they always were) a necessary part of the process. So in addition to other things one might do on a shot by shot basis (colour, exposure etc) registration would be part of that. The art of registration.

 

C

Link to comment
Share on other sites

 

Great looking footage, I would still guess Super 8 if I had to but it is really close to regular 16mm.

 

 

It's quite revealing the comparisons that have been made to 16mm. Its not as if the work was inviting or requiring such comparison.

 

That it does so needs some deconstruction.

 

Another way of saying the result is "really close to regular 16mm" is to say "16mm is really close to the result". They mean the same thing.

 

In other words the real issue isn't why the Super8 looks as good as it does, but why 16mm transfers are not looking as good as they could, ie. twice as good.

 

In other words, where the critique should really lie is with 16mm. Why is it being used as the metric for judging Super8 if it can't even live up to it's own assumed distance from Super8?

 

Super8 is an alternative to 16mm. Not a substitute. This has become my mantra of late.

 

C

Edited by Carl Looper
Link to comment
Share on other sites

The fault is not in 16mm. Projecting Super8 and 16mm will reveal the quite distinct differences between the two.

 

But it's in terms of how they transfer to digital. What actually happens in this process?. In other words the real issue is not the difference in terms of film, but the difference in terms of digital: the digital metrics otherwise used to characterise analog film. There is something flawed in such.

 

C

Link to comment
Share on other sites

I would have to agree this is the best S8 footage I have ever seen. They have succeeded in building the ultimate S8 camera. I can think of a lot of purposes this camera would serve. But I think there are far more people like me who use film for non dialog projects, art projects, installations, music videos, home movies, stylish cut-aways, personal fun ect... where the two main advantages this camera offers is rock solid registration and wider frame for better HD transfers. I think they could open up a wider market with an additional pared down model that could be made for less? without the on board audio, without the wifi and high tech digital stuff. Just focus on the same transport system with the same gate, and some variable frame rates from single to 54fps would be all that a lot of us need.

 

That said, my issue is since i already have a nice Bolex Rex4 with a full set of primes, i'm much better off cropping that image to 16x9 if sound sync is not an issue. If you're doing weddings or have a small production company, then this is a great camera at a good price. I just think it misses the vast majority of the niche at the current but justified price tag,

Link to comment
Share on other sites

I agree. I love everything about this camera but would almost never use any of its features other than the transport and when available the 200ft magazine.

 

Very rarely do I ever have any desire to sync audio with my super 8 and when I do, external sources are more than sufficient.

 

After the magazine and transport the only other thing I could see as an advantage for me is the video tap... But a version with just a regular optical viewfinder would be nice.

 

Again, not to take down an amazing camera with amazing features. It's a work of art and amazing engineering. It's just too much for me and most other super 8 users I know.

Link to comment
Share on other sites

Hi Anthony,

 

....without the on board audio, without the wifi and high tech digital stuff. Just focus on the same transport system with the same gate, and some variable frame rates from single to 54fps would be all that a lot of us need.

 

The electronics in our S8 camera alone is about 600€ nothing more! And that's counting the WiFi, Audio, LCD etc.

so even if we stripped all that out and made the camera with a hand-crank and crosshair viewfinder like the Ikonoskop 16mm the cost saving towards the end user would be so little that people would just complain about the high price and the lack of features.

 

Our real problem is that our minimum ordering quantity of metal is 50 sets. We must manufacture batches of fifty cameras each time and we need to pay everything upfront with our vendors hence we need to rely on a prepayment model in order to build cameras (just like we did with the beta round) and that model has a time element in it as well as we need to have sold almost all the cameras in advance before we can start to build them so if the cameras are slow to sell it can take a long time before we can start.

 

If we had the luxury of building 1.000 cameras the price could be dropped significantly but we would never be able to get under the 2.000€ level even with those numbers as the cost of manufacturing in Europe is simply too high and with the precision involved and required to build these cameras there's only a limited amount of companies that can do it and they know they can charge for it - our camera for example is build in the same factory as PhaseOne and Hasselblad.

 

 

Best regards

Lasse

Link to comment
Share on other sites

From a Kodak paper on sharpness:

 

"The “sharpness” of a film is the subjective perception of good edge distinction between details in a photograph."
page 56, http://motion.kodak.com/motion/uploadedFiles/US_plugins_acrobat_en_motion_newsletters_filmEss_06_Characteristics_of_Film.pdf

 

The paper goes on to say that an objective measure of sharpness is facilitated by measurement of the modulation transfer function. However there are many ways of measuring this function, each giving different results. So in general an MTF measured for a given stock is not so much describing the stock itself per se, but providing a common reference for the comparison of stocks. So long as a set of stocks are measured using the same method of measuring MTF, comparison of the MTFs will provide for a comparison of the film stocks.

 

But in terms of the number of pixels required to digitally reconstruct a particular stock in terms of it's sharpness, the MTF doesn't really provide anything. There is no criteria for establishing where one should nominate a cut off. For anywhere one can nominate such a cuttoff, one can equally nominate a higher frequency, beyond the cuttoff, that might still be usefully encoded.

 

The MTF curve itself is, in principle, a plot of the film's response to all frequencies, ie. up to infinitely high frequencies, with corresponding infinitely small responses (we can suppose). So it is up to other criteria to nominate a response level below which we would not allocate any more pixels. The MTF data itself doesn't provide us with any way of nominating which response level should be nominated as a lower bound.

 

Futhermore, the metric used in an MTF is typically in terms of cycles per unit of space (eg. line pairs per mm) which can lead to the incorrect assumption that if we had some way of nominating a lower bound, we would do so in terms of the same metric, a byproduct of which can be the incorrect assumption that a smaller gauge would require less pixels be allocated compared to film of a larger gauge. But there is nothing whatsover in MTF data for motivating such an assumption.

 

To put it simply, any candidate method for nominating a response threshold should treat the gauge as an independant variable, corresponding to the variation in magnification between gauges - where the resulting threshold for one gauge would not necessarily be the same threshold created for another gauge.

 

C

Edited by Carl Looper
Link to comment
Share on other sites

Hey Friedemann, I just watched your footage again at full resolution on my 4K TV at a distance of about 7' or 8', and I really was blown away all over again! It really is as sharp and clear and detailed as The Walking Dead (super16) or really anything on TV these days. (At least in the center of the frame, you can definitely feel it get soft around the edges, but that's obviously because the lenses weren't designed to shoot that wide).

 

But also, the grain! I'd love to see something in ProRes, but it looks great from across the room! Really pleasant textures in the sky and grass especially. I'm really even more impressed than the first time I saw the footage. Too cool.

Link to comment
Share on other sites

 

It looks like film to me,

 

I agree.

 

Not only does it look like film (and distinctly so) it is film - or at least that aspect of film which transfers to video. And while this might not be of interest to the casual observer who finds it easy to confuse it with video, it is important to those who otherwise create work. Anyone experienced in digital post will tell you that film sourced data provides far more flexibility than direct-to-video data. You have much better control over colour, gamma, contrast, etc. The data responds in a way that is very artist friendly. And Super8 sourced imagery is no exception.

 

C

Link to comment
Share on other sites

  • Premium Member

Make sure to watch this in highest quality and fullscreen:

https://youtu.be/3Nh9BTMWj9M

 

Or better download this version and watch it offline:

https://drive.google.com/file/d/0B9PJ3Kp3qOe4UklQZ0NrM1dNaVk/view?usp=sharing

 

If anybody interested, I can also share an ungraded ProRes 4444 snippet.

 

All the tripod shots where shot with the Optivaron (recognisable by the slight vignette in top right, since I did not take the lens hood off). All handheld shots (mostly in the middle) were taken with the Fujinon lens.

 

The incredible folks from http://www.gammaraydigital.com/did the scanning, once in 2K and once in 4K. Unfortunately, my Mac Pro (late 2008) is a bit too old to handle 4K video, so I had to grade the 2K version. It still looks amazing though, make sure to download the 2K version for full quality!

 

The clip was stabilized (tracking marker put on the frameline) since the scanner software used currently does not register based on the +2 sprocket hole, but on the sprocket hole next to the visible frame. You can see how bad the sprockets in this cartridge were if you watch the weaving sprocket hole on the left. But watch how image and frameline are like welded together: Absolutely no jitter, weaving and a sharpnes I have never seen before with any other Super 8 camera. This is due to the registration pin -- so thanks to Denmark for finally making a S8 camera that has this professional feature!

 

 

If you don't have a Logmar yet, order one now. I tell you, the achievable quality is abso**(obscenity removed)**inlutely addictive. :)

 

Hi Friedemann. I know this is 50D which is fine-grained to begin with, but did you have any grain-removal done?

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...