Is an HD Super 8 transfer overkill?

[granfer]

Well I don't know about HD transfers being "overkill"... but this thread certainly is!

We still don't know what quality of film he wants to transfer (or have I missed that bit?).

[christoph]

At the risk of beating a dead horse, what really defines the "resolution of the format"?

well, people might have different definitions, but mine is:

the maximum resolution a certain format can hold (usually in linepairs, even though this alone doesn't tell the full story). so in this case, how many linepairs can we count in a digital image of PAL size (720x576 for those lucky enough not to live in NTSC land ;)

as soon as you bring cameras or monitors into the discussion i'd talk about a resolution of a *system*, not the format, because it's not my problem if other people use crappy TV sets.. otherwise even the viewing conditions (daytime, nighttime) would affect the resolution of a format, which doesnt make sense to me.

if you'd shoot a test target with a camera with 720x576 sensor, you would only get about 70% of the resolution then if you shoot the same target with a 1400x1000 pixel camera and downsample it to 720x576. so a PAL image can store more information then a 720x576 imager can provide (this has to do with sampling theory). this is one of the reasons that all high-end camera systems actually have a sensor with more then PAL resolution. however, even digibeta cameras with 800line sensors will usually filter out high-frequeny detail to avoid flickering on CRT monitors, so they don't take advantage of the full PAL resolution either. so if you use a HD camera and downrez using high-quality algorithm (lancos, or sync etc) you'll get more detail.

and before somebody shouts again "so you need a HD transfer after all", this is why i said with inexpensive equipment it's of a definite benefit. the old PAL (ie non-HD) flashscan only gives you about 400lines resolution even with a test target. the flashscanHD downsampled to PAL will give over 500lines, and the native flashscanHD 720p frame will give about 600lines. but a spirit downsampled to 720p will give close to 700lines, and if output to 1080p about 1000lines.

that's why i prefer to keep the resolution of a format and of a system separate. because a 720p transfer from different systems will all have different resolution.

since super8 only has about 600lines even if well shot, a oversampled PAL image does a very good job in holding most of the information. an el cheapo PAL transfer will only give you about 300-400 lines though (because of poor lens, sensor and signal processing), which as you know looks like crap.

++ c.

[Will2]

Super 8 transfers done in HD on the same system by the same colorist are certainly more sharp vs. an SD transfer on the same system.

So overkill = NO.

Worth it?
Depends on what you're paying for it, how well the footage is shot.

I just sent 1500 feet of Regular 8 and Super 8 over to Debenham Media Group for transfer. $25 per 50' reel for HD using their mega-modified Rank machines for HD output. I'll report back on quality when I get it.

[christoph]

I just sent 1500 feet of Regular 8 and Super 8 over to Debenham Media Group for transfer. $25 per 50' reel for HD using their mega-modified Rank machines for HD output. I'll report back on quality when I get it.

well, looking at the sample clips on their website, their system has a true resolution of just slightly higher then PAL.

they do have very nice colors and contrast though, and a pretty good degraining system, so overall, the image looks very nice and the price is fair if they take as much care with your film as with their samples.
++ c.

[MovieStuff]

as soon as you bring cameras or monitors into the discussion i'd talk about a resolution of a *system*, not the format, because it's not my problem if other people use crappy TV sets......

But it doesn't matter how theoretically good the original PAL signal is since the only effective way to capture and work with it is via a flawed SD system which inherently causes the greatest degradation of the image, both in terms of resolution as well as color and contrast. You are not going to get 600 lines out of a standard def DVD, regardless of how the original image was captured and no matter how good the "system" is that you are watching it on. In fact, you will be lucky to maintain even 450 lines during a live up-rezz to a good HD monitor and the colors and contrast will suffer during that conversion, compared to watching the same SD material on an SD CRT, which has a wider color and contrast range. So when you talk about the format, you sort of have to include the intended system of reproduction. I mean, if no one is going to see it, then what's the point?

Roger

[christoph]

hi roger,

i think there's not much sense in continuing in this topic.
i've done my fair share of tests as i'm sure have you. obviously we come to different conclusions, maybe because you have to deal with NTSC which is quite a bit worse then PAL, maybe because we have different preferences.

happens all the time, just yesterday there was a yellow-red card for brazil that i'm dead sure wasn't justified, but the TV commentary thought different even after the 5th replay. we were clearly watching the same event, came to totally different conclusions.

wow, i managed to get this even more off-topic, but this phenomenon is very fascinating to me.. i mean, how some people can think that nuklear power plants are a clever idea still totally beats me.

christoph, ducks and hides

[Andreas Wideroe]

At the risk of beating a dead horse, what really defines the "resolution of the format"?

In my world this is how much information the film itself can hold no matter what camera or lenses were used.

If you're counting linepairs on a S8 neg film you would get about 3K if I don't remember wrong.

Reg8 would hold about 2K.

/Andreas

[MovieStuff]

At the risk of beating a dead horse, what really defines the "resolution of the format"?

In my world this is how much information the film itself can hold no matter what camera or lenses were used.

If you're counting linepairs on a S8 neg film you would get about 3K if I don't remember wrong.

Reg8 would hold about 2K.

/Andreas

Understood but I was actually referring to video which, unlike film, can not be viewed directly. As such, the final display resolution of SD video can be widely varied. Since you can't really view the video without the final display, then the delivery/display medium must be taken into account when talking about the effective, useful resolution of a given video format, particularly SD when compared to HD. Otherwise, all video formats are pretty much just as good as the other, if viewed off the original tape on a professional monitor. And we know that simply isn't true.

Roger

[christoph]

If you're counting linepairs on a S8 neg film you would get about 3K if I don't remember wrong.

well, while i just said that people can have different opinions, this is most definitely incorrect ;)

btw, how do you get an image on the film that you can measure without any optical system? a printing lens can be optimized more then a camera lens, but it's still limited somehow.

anyway, even assuming you have a near perfect lens, and shoot on modern camera negative film, you'll be very hard pressed to get 1K out of a super8 frame (btw, if film is usually measured in K=1024 horizontal while video in lines vertical).

since you wont believe me, here a few numbers ;)
the MTF chart of Vision2 50D:
http://motion.kodak.com/motion/uploaded ... 01_mtf.gif

they stop the chart at 80 cicles (line pairs) since meaningfull sharpness is around 50% MTF, that would result in 50lp/mm, ie 0.5 to 0.6K on a super8 frame.
even if we extrapolate for a MTF of 5% (which means barely visible structure) and assume about 150lp/mm that's only about 1.6K.

so if you actually *count* line pairs, you won't see much more then 1K on 50D and quite a bit less on 200T. and when you use a real super8 camera and shoot a test charts, it's more like 0.7K.

now if you say that we need to have 3K to hold all the grain structure of a super8 image that''s hard to argue, because any analog system needs indefinite bandwith/storage to reproduce identically in digital form. it's easiest to see in audio. some peope will claim that 44.1khz/16bit is not enough, other say 48/24 is enough, others say you need 196/24 to accuratelly capture the detail in an LP record.
++ c.

[Will2]

well, looking at the sample clips on their website, their system has a true resolution of just slightly higher then PAL.

Can you tell directly from their test footage? I wonder if I should send a SMPTE test reel in to get a more accurate analysis. I'll call Kodak and see if they still sell calibration/test reels anymore.

I can't debate your logic because it makes my head hurt, but perhaps there's a difference in perceived sharpness/resolution rather than an actual technical difference. Their test clips look like higher resolution to without actually measuring.

I'm most interested in their HD R8 transfers as most other transfer places seem to have a significant up charge for R8.

[MovieStuff]

.... perhaps there's a difference in perceived sharpness/resolution rather than an actual technical difference.

Well, there is an inherent difference in the mathmatical resolution per frame and the perceived accumulative resolution when film is running at full speed. Video uses a fixed pixel pattern and therefore has the same resolution per frame as it does at running speed. But film has a random grain pattern that never repeats itself. At running speed, the more film frames/grain that overlap on the retina of your eye, the higher the perceived level of detail compared to a single film frame. That's why old 8mm film looks less grainy at running speed than it does on a freeze frame when you stop the projector. You don't need to measure it in line pairs, etc. Just project some film and look at it. Regardless of the film format, film benefits from an increase in detail at running speed that video doesn't. That's also why I find comparing the mathmatical resolution of single video frames and single film frames rather pointless. That's not how they're viewed in the real world.

Roger

[christoph]

well, looking at the sample clips on their website, their system has a true resolution of just slightly higher then PAL.

Can you tell directly from their test footage?

well, with a bit of fiddling i can..

but make a guess for yourself, which one of these crops has the highest resolution, and how much?








and the same at 300% (open in different browser tabs for best comparison)
http://dl.dropbox.com/u/4067280/filmsho ... e_A-3x.png
http://dl.dropbox.com/u/4067280/filmsho ... e_B-3x.png
http://dl.dropbox.com/u/4067280/filmsho ... e_C-3x.png
http://dl.dropbox.com/u/4067280/filmsho ... e_D-3x.png

++ c.
[edit: changed the magnification to 300% to make it easier to see, just make sure your browser doesn't zoom then down]
i should also add a full frame of the original test clip:
http://dl.dropbox.com/u/4067280/filmsho ... e-full.png

[christoph]

Just project some film and look at it. Regardless of the film format, film benefits from an increase in detail at running speed that video doesn't.

well, that's true in theory, but in practice the poor registration of the projector will actually reduce the resolution. this is a problem even with a good telecine unless you stabilize the footage digitally afterwards.

the grain reduction effect is definitely a big benefit on running film compared to still frames though.

btw, this is not just something i made up myself, there are several studies by some really good people which looked at that because scanning at high resolution is expensive.
they found that 2K on 35mm is enough because on the projected screen, the image will only have about 1K resolution anyway (a big part being registration issues).

this is slowly starting to change now, because visual fx benefit from footage with more detail, and because of digital projection coming soon, so the original negative is scanned at 4K (which equals about 1K in super8). most of the time the final master is still 2K though, but it still has the benefit of oversampling.

++ c.

[MovieStuff]

well, that's true in theory, but in practice the poor registration of the projector will actually reduce the resolution.

If you have a projector with poor registration. But, as long as the projector is working correctly, you can see the increase in detail on most any given home movie projector at running speed. Look at locked off footage of a road sign that is unreadable when a regular 8mm projector is on a still frame but is totally legible when the film is running at full speed. It's actually pretty obvious and more apparent on smaller formats than larger formats like 35mm that already have good detail on a single film frame. That is also why viewing stills from an SD transfer and an HD transfer are good for comparing color and contrast but tell you little about the level of detail you will ultimately see at running speed.

My point is that even if one accepted SD video and 8mm film as having the same resolution per frame, the level of viewable detail for film increases as the film runs faster. But, on video, it doesn't matter how fast you run the video frames, the level of perceptible detail is no better than on a still frame. Thus, if you scan your 8mm footage to SD video, you can never get past the limited, coarse fixed pixel pattern inherent in the SD video system, which only gets worse as you take it through editing and disc authoring and then up-rezzing on an HD system. Scanning directly to HD avoids those issues and allows the best chance of maintaining more of the accumulative resolution that film has to offer.

Roger

[christoph]

ok, i propose an open challenge ;)

you shoot a test chart
http://www.bealecorner.com/trv900/respat/EIA1956.pdf
with the best super8 camera and film stock you can find, under optimal conditions, and transfer in your HD machine.

i shoot a still image of the same chart and downscale it to PAL and 1K resolution.

and then we count lines :)
(btw, i've already done this test, so it's kinda pointless)

i should add that i love super8 precisely *because* of the softness, the grain, and the slight unsteadiness (well, and the highlights)... but any attempt to compare it in terms of *resolution* with modern digital media is bound to fail.

++ christoph