Is an HD Super 8 transfer overkill?

Forum covering all aspects of small gauge cinematography! This is the main discussion forum.

Moderator: Andreas Wideroe

Post Reply
carllooper
Senior member
Posts: 1206
Joined: Wed Nov 03, 2010 1:00 am
Real name: Carl Looper

Re: Is an HD Super 8 transfer overkill?

Post by carllooper »

Uppsala BildTeknik wrote:If all you are comparing is scaling of images back and forth then fine. It has nothing to do with neither a SD transfer nor a HD transfer.
That is not all I'm doing. What I'm doing, as distinct from whatever you think I'm doing, has everything to do with SD and HD transfers. How do you think transfer systems are built in the first place? Do you think someone just assembles such machines by trial and error? They are built on theory.
carllooper wrote:...I'm only scanning a small window on the film. The scanning resolution of 24K you mention is what I said the scanning resolution would be if I were scanning the whole frame.
Uppsala BildTeknik wrote:So once again all you have are theoretical scaling data. No real life transfers, no real life results and no real life conclusions.
Real life transfer systems are built on theory. My real life system is built on the same theory as any other real life system. I have done real life transfers with real life results.
Uppsala BildTeknik wrote:Transfer one hour to 24K, transfer the same hour to 4K. Compare results, figure out how you will use the material, if you have use for the extra details and if it is worth the hassle of dealing with a 24K file. My guess is that you won´t ever even end up doing a real 24K transfer. So what is the point of the whole "24K test" then?
The whole universe doesn't revolve around your particular use case. The purpose of the tests, as I've already said, is to understand the nature of the FILM (which is not well defined) rather than the scanning system (which is well defined). And from such design an optimum transfer system which balances budget, quality and redundancy. The optimum transfer system won't be a 24K system. You don't have to guess that. As I already said, the information gathered by the 24K tests will be information used in designing the optimum system. I don't yet know what that optimum system will be until I complete the tests on the FILM. It is the FILM which is not well defined. It is the FILM which requires testing, using the best techniques at my disposal.

Do you think scientists looking at FILM using an electron microscope is overkill? And if not why not? I use electron mircroscope images of FILM in my research. The structure of the silver halide crystals and the dye clouds informs how I understand the FILM. It is the FILM being analysed, not the transfer system.

My transfer system will probably end up somewhere between 4K and 8K. However the ultra high def tests are in relation to how to process those scans. They inform the processing algorithms.

Perhaps the problem you are having in understanding what I'm doing is because you are under the false impression that upscaling and downscaling are no different from each other. Upscaling adds artificial information, through interpolation. But downscaling is very different. Downscaling removes information. It doesn't add any new artificial information.

All I'm doing is removing from an HD scan that information which makes it HD. The remaining information is SD. I then interoplate the SD information back up to HD. The difference between the original HD and the SD interpolated back to HD becomes the important information. It tells me something about the FILM, rather than anything about the transfer system. I already understand the transfer system. The transfer system is a well defined system. The FILM is what is being analysed, not the transfer system.

Hope that helps to understand the theory. If there is any problem with this, it is not because it is theoretical, but because there is something wrong with the theory. I don't know what is wrong with the theory and neither do you. As far as I can tell there is nothing wrong with the theory. It's really quite a simple theory. There isn't much to understand. There isn't that much that could be wrong with it. You should look at the theories I have to deal with day in and day out as a software developer. They are a million times more complex than this.

Carl
Carl Looper
http://artistfilmworkshop.org/
Will2
Senior member
Posts: 1983
Joined: Mon Sep 26, 2005 12:18 am
Real name: Will Montgomery
Location: Dallas, TX
Contact:

Re: Is an HD Super 8 transfer overkill?

Post by Will2 »

Uppsala BildTeknik wrote: Transfer one hour to 24K, transfer the same hour to 4K. Compare results, figure out how you will use the material, if you have use for the extra details and if it is worth the hassle of dealing with a 24K file.
I've always been amazed by my friend at ILM telling me that until a couple years ago or for very special effects shots, they did everything in HD resolution; not even 2K. This would have included some good looking movies as well as the Star Wars films. :D

The reason is simple; money. Manipulating anything larger just takes exponentially more time and therefore more money. And if there's one thing Lucas cares about it seems to be money.

Not that this has anything to do with what Carl is saying, so please continue.
User avatar
Nicholas Kovats
Posts: 772
Joined: Sat Mar 25, 2006 7:21 pm
Real name: Nicholas Kovats
Location: Toronto, Canada

Re: Is an HD Super 8 transfer overkill?

Post by Nicholas Kovats »

Carl,

Tremendous work. I happen to know it's real from the lab results you have shared. Thank you for exploring, testing, prodding, theorizing and extracting significant information from the fundamental 1 micron silver halide crystal and dye cloud structures also known as film!

But what are you going to do if I am able to originate even more information over a wider area per square millimetre for your valuable work? :)
Nicholas Kovats
Shoot film! facebook.com/UltraPan8WidescreenFilm
MitchPerkins
Posts: 41
Joined: Tue Aug 30, 2011 8:36 pm
Real name: Mitch Perkins
Location: Toronto, Ontario, Canada

Re: Is an HD Super 8 transfer overkill?

Post by MitchPerkins »

People, please! You're annoying Nigel. Stop annoying Nigel.

M
carllooper
Senior member
Posts: 1206
Joined: Wed Nov 03, 2010 1:00 am
Real name: Carl Looper

Re: Is an HD Super 8 transfer overkill?

Post by carllooper »

freedom4kids wrote:But what are you going to do if I am able to originate even more information over a wider area per square millimetre for your valuable work? :)
An important consideration (and this underpins Roger's thought experiment) is that no matter how high the scanning definition it's always going to be better (in terms of definition) than a lower definition scan. The best a lower definition scan can do is interpolate the information it does not possess, and not only that, such interpolation can cost.

This is precisely because a higher definition scan is higher definition, by definition!

So if you see a signal in the difference between HD and upscaled SD, in and of itself that doesn't tell you anything new. You should expect to see a signal no matter how high the scanning resolution, all the way down to the quantum mechanics. Because this will be a function of the intrinsic difference between a higher def scan and anything lower. It's a self fulfilling prophecy. The only reason you might not see a signal will be due to noise but only noise. That, however, is interesting in itself. Noise is mathematically undefined. It can, however, be described in terms of statistics, which is the next best thing. But we'll leave that aside for the moment (it is an incredibly important attribute of film but one that needs separate attention).

So what you are looking for is not just a signal as such, but evidence of a sharp signal. But not just a sharp signal. Soft signals in the FILM will always come out as sharp signals when you do the HD:SD comparison (whichever method you use). They can give you a false positive, in terms of what you looking for: the resolving power of the FILM. You have to be careful to distinguish between the measurement system and what is being measured.

For testing the resolution of the FILM what you need is an external reference, which is the purpose of a resolution chart, or "Modulation Transfer Function" chart.

By photographing such a chart on FILM, and then analysing a high defintion scan of the FILM, you can distinguish between signals that are just sharp by definition (due to the intrinsic difference between HD and SD), and those that belong (or fail to belong) to the sharpness of the chart (which is independant of everything else).

It should be noted by those who have something against theory that a resolution chart is a completely artificial signal. It is purely theoretical. But that's precisely what makes it so useful. It is well defined. One doesn't need to test the chart because one has precisely defined it in the first place. Just like the measurement system. It is well defined.

So the FILM (one hopes) gets "trapped" between a well defined chart and a well defined measurement system. It's properties, within these constraints, can be analysed and compared with what a lower definition measurement might otherwise interpolate. So long as the chart and scanning resolution are sufficiently high one should be able to close in on the ideal scanning resolution. But I don't yet know what is a "sufficiently high scanning resolution". I haven't yet found it. All I know is that Super8 @ 2.5K (or 16mm @ 5K, or 35mm @ 10K) is not yet sufficiently high to confirm whether those resolutions are in fact ideal scanning resolutions. I can, however, confirm that they are better (and not just intrinsically so) than Super8 scanned at 2K, 1K or 0.7K scan.

What makes the resolution of FILM difficult to re-establish in the digital age is actually that which digital processing allows. For example, convolution based sharpening filters can tease out sharper signals from the FILM that were otherwise soft to the naked eye. These operate better as one increases the scanning resolution. So there are additional complications introduced by our new found abilities to not only digitally analyse FILM at higher definitions, but to digitally "understand" that analysis as well.

Now quite apart from the resolution of the FILM is the relationship between the FILM's signal and noise (or grain), irregardless of whether the signal component is soft or sharp (low frequency or high). This can be explored in ultra high definition scans without requiring a resolution chart because the interest is no longer resolution. There are other attributes of FILM (whether implicit or explicit) such as grain, dynamic range and colour that can be explored. More involved analysis is required, or becomes possible here.

The purpose of all of this can vary. It can answer the thread question, but it can also (as in my particular case) answer design questions for an ideal digital transfer and processing system, ie. one that satisfies budget constraints, avoiding overkill, improved quality, and perhaps above everything else: curiosity.

Carl
Carl Looper
http://artistfilmworkshop.org/
carllooper
Senior member
Posts: 1206
Joined: Wed Nov 03, 2010 1:00 am
Real name: Carl Looper

Re: Is an HD Super 8 transfer overkill?

Post by carllooper »

What I might also add, is that the presence of a difference signal in a hi-def scan of FILM (compared to a lo-def scan) is still significant in it's own right, irregardless of whether the signal represents a sharp signal or a soft signal.

While such a signal is expected anyway, due to the intrinsic difference between a hi-def and lower-def scan, it is also a signal that is REAL. It is not a signal that belongs to the measurement system. It is not an interpolated signal. It is there in the FILM, whether the source signal was sharp or soft.

If you were to apply the same methodology to a digitally acquired image there will be a point beyond which the difference signal becomes zero. There is a point beyond which there is no difference between your re-sampling of the digital image and an interpolation.

For example, if you re-sample a native 4K digital image at 8K and compare that to a the 4K interpolated up to 8K - THERE IS NO DIFFERENCE. It is exactly the SAME THING. And the difference signal will be ZERO.

Since there are NON-ZERO difference signals (and not just noise) at the equivalent of 16mm @ 24K (even if they represent soft signals) it means there is more information in the 16mm film than there is in any native digital image less than 24K. There may be (probably is) more but that is as far as I've gone so far.

The real issue is what to do with that information. It is REAL information. A REAL signal. But if you don't do anything with that REAL signal, one may as well substitute an interpolated signal (or a lower def signal). They'll look the same. No sharper.

So that's the real challenge. How to digitally process that additional information, obtained though a high-def scan, so that the result is better (and therefore better looking) than a lower def or interpolated substitute.

Carl
Carl Looper
http://artistfilmworkshop.org/
User avatar
Uppsala BildTeknik
Senior member
Posts: 2261
Joined: Thu Jun 24, 2004 7:20 am
Location: Sweden, Alunda
Contact:

Re: Is an HD Super 8 transfer overkill?

Post by Uppsala BildTeknik »

Wow. If you put the same time and effort into making a real SD scan, like you do into writing marathon posts that are unreadable on a internetforum, you would actually have a real test in your hands. :)
carllooper
Senior member
Posts: 1206
Joined: Wed Nov 03, 2010 1:00 am
Real name: Carl Looper

Re: Is an HD Super 8 transfer overkill?

Post by carllooper »

Uppsala BildTeknik wrote:Wow. If you put the same time and effort into making a real SD scan, like you do into writing marathon posts that are unreadable on a internetforum, you would actually have a real test in your hands. :)
Just because you can't read it Kent doesn't mean someone else would have the same problem.

Be that as it may I do appreciate your comments anyway, in a perverse sort of way, because if nothing else they have prodded me into being a bit more clear about what I'm doing, not that you seem to be getting much out of it, unfortunately.

And I do in fact put just as much work (in fact more) into developing what I'm otherwise discussing. But it's a good to take a break from that, and everything else I do, from time to time, to talk about it.

Your theory about my methods is interesting. Your theory is that because my method is not the same as your method it must therefore be wrong. So here's a lesson for you Kent. See if you can work out what might be wrong with that theory. Once you've worked that out you might find yourself a little wiser and perhaps even a touch happier.

Like I was saying about Nigel's call to action, why should action and talk be opposed? Just because that's the fevered philosophy of Hollywood action heroes doesn't mean it's correct.

Carl

ps. why would I want to do a "real" SD scan when I can get a much better SD scan from the HD? And if I can already get a better SD scan from the HD, then surely if my original HD scan proves already better than the excellent SD scan derived from it, it's going to be better than a bad (ie. "real") SD scan that wasn't.
Last edited by carllooper on Fri Jun 29, 2012 8:36 am, edited 1 time in total.
Carl Looper
http://artistfilmworkshop.org/
User avatar
Uppsala BildTeknik
Senior member
Posts: 2261
Joined: Thu Jun 24, 2004 7:20 am
Location: Sweden, Alunda
Contact:

Re: Is an HD Super 8 transfer overkill?

Post by Uppsala BildTeknik »

carllooper wrote:Your theory is that because my method is not the same as your method it must therefore be wrong.
No. Your method is flawed because your intention was to compare a SD transfer to a HD transfer, but you took the shortcut and skipped getting a true SD transfer to have as a base for your comparison.

So you are, again, not comparing a SD transfer to a HD transfer. You are comparing a HD transfer up/downscaled to different resolutions. And that will not give you the same results you would get from comparing a true SD transfer to a true HD transfer.

That is all I am saying, and it has nothing to do with whatever methods I have used in order to compare transfers.
carllooper
Senior member
Posts: 1206
Joined: Wed Nov 03, 2010 1:00 am
Real name: Carl Looper

Re: Is an HD Super 8 transfer overkill?

Post by carllooper »

carllooper wrote:Your theory is that because my method is not the same as your method it must therefore be wrong.
Uppsala BildTeknik wrote:No. Your method is flawed because your intention was to compare a SD transfer to a HD transfer, but you took the shortcut and skipped getting a true SD transfer to have as a base for your comparison.
Like I said, your theory is that my method is wrong because it isn't the same as your method. But more than that, as I can now see, you think my intention was to do what you did. If that was my intention I would have done it. I didn't. Your problem is that I "skipped getting a true SD transfer", ie. that I didn't do what you did. So what? I used a different method. I wanted to use a different method. And I did.
Uppsala BildTeknik wrote:So you are, again, not comparing a SD transfer to a HD transfer. You are comparing a HD transfer up/downscaled to different resolutions. And that will not give you the same results you would get from comparing a true SD transfer to a true HD transfer.
Yes, that's right, I'm not doing what you did. And I also get different results. That's to be expected.
Uppsala BildTeknik wrote:That is all I am saying, and it has nothing to do with whatever methods I have used in order to compare transfers.
Yes it does. Everything you've just said just confirms what I've already said, that the problem you have with my method is that it's not the same as yours.

Why would I want to do a "real" SD scan when I can get a much better SD scan from the HD? And if I can already get a better SD scan from the HD, then surely if my original HD scan proves better than the otherwise excellent SD scan derived from it, it's already going to be better than a bad (ie. "real") SD scan that wasn't.

I guess you could try and argue that a "real" SD scan might be better than an real HD>SD scan. But all things being equal (same film, same lens) I have no idea how you could argue that.

Carl
Last edited by carllooper on Fri Jun 29, 2012 9:14 am, edited 2 times in total.
Carl Looper
http://artistfilmworkshop.org/
User avatar
Uppsala BildTeknik
Senior member
Posts: 2261
Joined: Thu Jun 24, 2004 7:20 am
Location: Sweden, Alunda
Contact:

Re: Is an HD Super 8 transfer overkill?

Post by Uppsala BildTeknik »

carllooper wrote:Why would I want to do a "real" SD scan when I can get a much better SD scan from the HD?
Quite simply because you can never get a real life SD scan from a HD transfer, what you get is a downscaled HD transfer. Wich, in essence, is not the same as a SD transfer, because you already explained it in your post that the results will not be the same.

I can see this is going nowhere, again... But I guess we agree that your test will not give the same results as a test comparing a SD transfer to a HD transfer would give. So, we actually agree here! :mrgreen:
carllooper
Senior member
Posts: 1206
Joined: Wed Nov 03, 2010 1:00 am
Real name: Carl Looper

Re: Is an HD Super 8 transfer overkill?

Post by carllooper »

Uppsala BildTeknik wrote:
carllooper wrote:Why would I want to do a "real" SD scan when I can get a much better SD scan from the HD?
Quite simply because you can never get a real life SD scan from a HD transfer, what you get is a downscaled HD transfer. Wich, in essence, is not the same as a SD transfer, because you already explained it in your post that the results will not be the same.

I can see this is going nowhere, again... But I guess we agree that your test will not give the same results as a test comparing a SD transfer to a HD transfer would give. So, we actually agree here! :mrgreen:
Thank God. :mrgreen:

Carl
Carl Looper
http://artistfilmworkshop.org/
User avatar
Uppsala BildTeknik
Senior member
Posts: 2261
Joined: Thu Jun 24, 2004 7:20 am
Location: Sweden, Alunda
Contact:

Re: Is an HD Super 8 transfer overkill?

Post by Uppsala BildTeknik »

And what method is the correct one then... Just kidding! 8) :lol:

Your method is the correct method for you, of course. [handhake] :mrgreen:
User avatar
Nigel
Senior member
Posts: 2775
Joined: Sat Oct 05, 2002 10:14 am
Real name: Adam
Location: Lost

Re: Is an HD Super 8 transfer overkill?

Post by Nigel »

I was going to weigh in on this but after I typed a paragraph I laughed and said "Fuck It".

Good Luck
carllooper
Senior member
Posts: 1206
Joined: Wed Nov 03, 2010 1:00 am
Real name: Carl Looper

Re: Is an HD Super 8 transfer overkill?

Post by carllooper »

Uppsala BildTeknik wrote:And what method is the correct one then... Just kidding! 8) :lol:

Your method is the correct method for you, of course. [handhake] :mrgreen:
I'm going to take some advice from Nigel and shoot some film, well actually shoot some video (the film in my fridge is already earmarked for something else), to see if I can't otherwise explain/express the methods and theories I'm using in a more visual way.

[handshake]

cheers
Carl
Carl Looper
http://artistfilmworkshop.org/
Post Reply