Is an HD Super 8 transfer overkill?

[Uppsala BildTeknik]

...spacial integration preserves the temporal integration of film grain, resulting in the same kind of perceived increase in sharpness...

Hah, you guys know how to go into details... I just know a HD transfer looks better on a big HDTV if compared to a SD transfer. So by that I come to the conclusion that HD is not overkill for super8.

But it is interesting to read your discussion! :mrgreen:

[mattias]

yeah, some people also try to avoid the question of how the world started by blaming it on god, without reflecting for a second over the fact that they just asked new questions instead of providing an answer. ignorance is bliss but it doesn't give us airplanes or penecillin. :-)

[carllooper]

yes i don't mean sensor noise but integration of grain. but no i don't mean that spacial integration improves sharpness, quite the contrary as you say, i mean that spacial integration preserves the temporal integration of film grain, resulting in the same kind of perceived increase in sharpness. I'm not quite sure if the level is the same or even close, but the more we discuss the closer to the truth we're likely to get. ;-)

Cool. I'll talk about this futher in a paper rather than here, and post a link. And we can pick it up from there.

cheers
Carl

[mattias]

btw kent, as others have pointed out, there are too many things except transfer quality going on in your comparison. what you need to do in order to judge the resolution only is to take one of your hd clips, scale it to pal, then back up, using high quality scaling, maybe throw in an sd transfer too but make sure you scale it the same way and watch it through the same hd signal path. judging sd quality on a tv that deinterlaces all sd content, and god knows how it scales then, doesn't really make sense. you gotta isolate the parameter you want to study.

[carllooper]

Hah, you guys know how to go into details... I just know a HD transfer looks better on a big HDTV if compared to a SD transfer. So by that I come to the conclusion that HD is not overkill for super8.

If it didn't cost anyone, in terms of time, money, or resources, to scan film at 1000K then we'd all be doing that. Why not? Better safe than sorry.

But it does cost - so it makes sense to work out, in a cheap manner, (such as using some information and maths), where the cutoff point might be, so that we can save some money and spend it on something else we in which we might be interested, such as eating dinner.

Every single particle in the original film matters.

Since the smallest particle is 0.2 um, then the apparent answer is that we'd require at least the following definition scan to reproduce every single particle.

28450 x 21100 pixels

But when we feed this into the system equation for both a lens @ 400 px/mm (200 lp/mm) and film @ 5000 px/mm (0.2 um particles) we get:

R = sqrt( 1/( 1/400^2 + 1/5000^2 ))
= 399 px/mm
= 2270 x 1684 px

In other words, no matter how high the definition of the film, we can't get a signal that is any better than the resolution of the lens (which we rated at 400 px/mm)

So we can define a scanning limit in terms of the limit of the lenses we're using. Any higher scanning won't increase the spatial resolution. Each pixel will catch all the particles within it's view and sum them together. So there is no loss of tonal definition. That loss would only occur if you scanned at low definitions where you would start to see contours along the quantisation boundarys (as in when you do a large gaussian blur on an 8 bit image). At low definitions you need more bits per pixel. Or use noise to mask the boundaries.

Here are the lens resolutions from a 2007 research paper, and associated S8 pixel resolutions. Since lens resolutions are rated at 30% contrast, they therefore extend beyond such rated resolutions, but where zero contrast/visiblity occurs I don't know. So the pixel resolutions given here are for cut offs at the 30% point. If you don't want to lose that extra bit of detail you would need to increase the pixel resolutions given here. By how much is your guess is as good as mine. But 1.3X is probably sufficient.

Old lens (1840-1930)
20 lp/mm
228 x 169 px

Average lens
40 lp/mm
375 x 338 px

Very good large format lens, or many 35mm lenses at f/8
60 lp/mm
683 x 506 px

Excellent large format lens, Schnieder 150 APO Symmar at f5.6, f/8
80 lp/mm
910 x 675 px

Superior 35mm format lens, many second tier lenses at f/8
100 lp/mm
1138 x 844 px

Outstanding 35mm lens, Nikor & Canon 50mm & 85mm lenses at f/8 on a tripod, superior processing, film only, no prints
120 lp/mm
1366 x 1012 px

Exceptional 35mm lens: Leica or Zeiss 35mm or medium format lens
140 lp/mm
1593 x 1182 px

Best possible 35mm lens (you wont find one apparently)
200 lp/mm
2276 x 1688 px

For comparison here is a Super 8 frame (TriX Canon 1014 lens), scanned at 3K, but downsampled for comparison at:

a. 2276 x 1688 (as in a scan for 200 lp/mm)
Downsampled from 3K
http://members.iinet.net.au/~carllooper ... ible35.jpg

b. 1138 x 844 (as in scan for 100 lp/mm)
Downsampled from 3K then back up to above 2.3K for comparison.
http://members.iinet.net.au/~carllooper ... rior35.jpg

Carl

[mattias]

i think small format lenses usually have greater resolving power per mm though, there are super cheap 1/4" 1080p video cameras that take exceptionally sharp pictures, which suggests at least 100 l/mm. they usually have problems with contrast, flare, color balance, chromatic aberration and so on instead, but they have to prioritize something, and at that small sensor size it has to be sharpness. the cheap cctv lenses i'm currently looking at for my beaulieu are rated at 3 megapixel for 1/2" (slightly larger than super 8), which means 150 l/mm (300px/mm).

the scanner also has a lens, in the case of the flashscan a regular camera lens i think? a lens designed for a certain subject distance with a fixed aperture could probably be made much sharper at a pretty low cost, i mean a cheap neg scanner with a plastic lens can be extremely sharp. it would take some volume though. kent, what lens does the flashscan hd use?

[MovieStuff]

.... judging sd quality on a tv that deinterlaces all sd content, and god knows how it scales then, doesn't really make sense. ...

But that's how people will be watching their SD content and the fact that you recognize there is a widely varying factor involved in scaling on a TV only shows how important this is. You have to take into account the release and display methods. I mean, if you want to look at numbers only, then theoretically any 950 line SD broadcast camera has what it takes to capture nearly all the resolution in a frame of 8mm film. Only there is no common capture and release SD format that will maintain the original SD resolution of that camera. By the time you capture, edit in DV (which is typical) and then release on a regular DVD, that drops to 350-400 lines or often less. And film yields a higher degree of detail at running speed than it does on a single frame, which is possible to maintain on an HD workflow but is limited to by the coarse, fixed pixel pattern of the SD workflow. Worse, as you have just noted, SD upscaled on an HD monitor is often terrible even though the same HD monitor may reproduce HD content just fine.

So you really do have to take into account how any image will be displayed. When I paint on commission, I need to know if the painting will be viewed in tungsten, such as in a study, or daylight, such as in a hall next to a window. That determines the color balance of my mixes. I also need to know what distance the painting will be viewed; above a fireplace where people are further away or on a living room wall where they can get up close and personal. I can't just paint one way and expect the painting to look the same in every situation. That would foolish. In the same sense, knowing how your transfers will be viewed can make a big difference.

I agree with Kent. HD isn't overkill if it dependably "looks" better when viewed on an HD monitor, regardless of the reason.

Roger

[mattias]

But that's how people will be watching their SD content

yeah but this thread is about transfer, not viewing. if you just want to overcome the limitations of modern tv sets you can simply scan in sd and upscale it to hd before viewing. in fact as you're suggesting that's probably a good idea if you're going to work in compressed formats, downscaled compressed hd is still higher def than dv, very close to uncompressed sd. but again that has very little to do with what resolution you scan at.

[Uppsala BildTeknik]

kent, what lens does the flashscan hd use?

I have no idea.

btw kent, as others have pointed out, there are too many things except transfer quality going on in your comparison.

I´m just comparing a SD transfer to a HD transfer, simple as that. And as long as the SD looks better, then it looks better.

I will leave the nit-picking about all the details around this to you guys. You can upscale/downscale and argue about lines per millimeter and whatnot.

You should invite Chirstoph to this discussion, he is very good at details regarding image quality and resolution too. He transfers his super8 to HD, and at the same time he was arguing in another thread in this forum that HD for super8 was overkill. I can´t bent my head around that...

If I´d ask you the same, next time you will get your next super8 project telecined, will you choose SD? If you choose HD, you can´t argue that SD is good enough...

[mattias]

I´m just comparing a SD transfer to a HD transfer, simple as that.

yes, but you're *also* comparing how your tv displays sd content with how it displays hd content. it's not "nitpicking" to try and isolate the variable you're trying to measure, it's standard procedure in any test, even simple-as-that observations. if you're also interested in how tv sets display content you can easily devise a separate test for that where transfer quality isn't a confounder. as for testing transfer quality alone it's super easy so why are you so reluctant? just take that sd clip, upres it with compressor with scaling set to best, then do the comparison. what's so scary about knowledge?

[MovieStuff]

But that's how people will be watching their SD content

yeah but this thread is about transfer, not viewing.

LOL. Isn't that kind of like someone that can't read joining a writers group? Reading and writing are as related as capture and display.

if you just want to overcome the limitations of modern tv sets you can simply scan in sd and upscale it to hd before viewing.

We did that very thing using an HD transfer house so we'd have a bench mark for comparison when designing our own HD units. What you're saying is true but only if the display monitor is so small that the HD format is incidental to the process. On a large HD monitor or HD projector, SD upscaled to HD does not look the same, no matter what method you use. I mean, we can argue about why it should but it simply doesn't; at least not dependably enough that you, I and others will always see the same quality result.

Now, you can argue that the display method should be left out of this discussion but there is a reason that standard def broadcast cameras had about double the resolution than the final display would allow. They knew that, by the time the image reached the viewer, there would be degradation of the image for a variety of unpredictable reasons. Likewise, if the results of upscaling will not dependably overcome the worst of common HD display methods that still allow original HD content to look good, then we're really talking about a "what if" scenario that always works in theory but, in practice, unreliably.

To me this whole discussion is a lot like arguing that the sky isn't really blue; it just looks that way because of light scatter and refraction through miles and miles of air, blah, blah, blah. But if you ask the average person on a sunny, cloudless day what color the sky is, they will say it's blue, even if you know why it really isn't. Perception is everything. Footage captured and displayed in HD will consistently look better than footage captured in SD and then displayed in HD. That's reason enough to capture in HD. Any other "what if" scenarios are too unpredictable to be useful, even if they make for an interesting discussion.

Roger

[Uppsala BildTeknik]

So Mattias, you will get SD transfers for your super8 stuff in the future. Correct?

[mattias]

Isn't that kind of like someone that can't read joining a writers group?

no, it's like a dyslexic person who says a book sucks because he can't read it.

i'm sure you've done exactly the test i'm suggesting, i was commenting on the fact that kent apparently hasn't, and more so that he doesn't even understand why it's necessary. since you apprently do i'm not sure what you're trying to tell me.

[MovieStuff]

i'm sure you've done exactly the test i'm suggesting, i was commenting on the fact that kent apparently hasn't, and more so that he doesn't even understand why it's necessary. since you apprently do i'm not sure what you're trying to tell me.

I'm saying that upscaling SD to HD can work if the display isn't too large. But you can't always presume the HD monitor will be small. And, to be fair to your position, you can't always assume the HD monitor will be so large that SD scaled to HD won't work, either. But if you scan in HD to begin with, then you will get consistent results whether the monitor is big or small. I think this is a pretty easy concept to understand.

Roger

[mattias]

So Mattias, you will get SD transfers for your super8 stuff in the future. Correct?

there are other issues, especially with the flashscans. roger has been touching the subject here, and if i have material that has (in "pixel equivalents") around 720x576 pixel resolution and i get to chose between a scanner that produces around 400x300 and records at 720x576 and one that does maybe 700x500 and records at 1280x720 i will chose the latter. i'm not arguing that your flashscan hd isn't better than your flashscan sd, i'm just questioning whether it's really capturing much more than pal res from the film.