Perry Paolantonio

This is not accurate. It's a longstanding bit of incorrect "common knowledge" that 16mm is 2k, 35mm is 4k, etc, etc, but it's simply untrue. There is no way one can say X film gauge = Y pixel resolution, because film isn't made of pixels. That's an apples to oranges comparison. Film grain is an organic shape. It's not a square pixel. If you want to get an accurate representation of that grain, you need more pixels to properly render all the subtle details, curves and variations. Let's pretend you're a scientist working with nanoscale-level material. Are you going to use an optical microscope like you'd use in a biology lab, or are you going to use a scanning electron microscope, if you want to get an accurate picture of what you're looking at? Sure - in practice, if you look at a 2k scan projected with a 2k projector, and a 4k scan projected with a 4k projector, assuming both are correctly calibrated and you're viewing at the correct distances you *shouldn't* see any difference between the two. But if you get close up, you will. It's the same as 1080p and 720p. Once you're past a certain distance from the screen, your eye simply can't resolve enough detail to discern a difference. But if you increase the screen size or get closer to it (imagine a 60" television 6 feet from a couch, not an uncommon situation), then you clearly see the differences. So what does that extra resolution get you? Better grain resolution, as mentioned. Since the grain *is* the image, why wouldn't you want this? Even if the end product is destined for a lower resolution, you're better off scanning at a higher res and downscaling it anyway. More photosites on the sensor mean you have more points picking up the fine details and that can mean not only increased sharpness, but slightly better dynamic range (again because you're able to render a single grain of film with more than one or two pixels like you'd get with a lower res scan). But much of this arguing about resolution is kind of silly (the number of pixels are only one of many factors here - optics, dynamic range, transport stability of both camera and scanner, and registration quality are all arguably more important), I'll say that to me the most compelling reason to scan to 4k is to avoid having to scale up later. 4k televisions are under $1000 now, and they're going to replace HDTVs (unlike all that 3D nonsense being pushed on consumers a couple years ago). When you have a gigantic 4k screen and you're close to it, you're going to see softness from upscaling the image from a 2k scan. 8k TV isn't that far off. The less you have to blow up the image in the future, the better, and there's no real downside to downconversions for use on current systems. It makes no sense to me to *not* scan at 4k right now based on this argument alone. -perry

I personally scanned both of these examples. These are the ones I mentioned above, posted by Friedemann Wachsmuth. Both were shot on the Logmar, both scanned on our ScanStation - the first one you link to was scanned on the newer 5k version of the scanner, the second one, which is about a year older, was done on our old 2k scanner (same machine, different sensors). In both cases, Friedemann post-stabilized the image. In both cases, the scans were done before Lasergraphics released a software update that fixes horizontal registration, which is why he did the post-scan stabilization. Neither scan exhibits any rocking until they're stabilized in software later. What you are seeing on those Vimeo clips is not what the scan looked like. That said, that stabilization step is no longer necessary because it's now done during the scan using the horizontal lines of the perf to do vertical registration and the left (perf-side) edge for horizontal registration. There is *no* rocking problem we're aware of in the scanner, which means this issue is either something to do with the camera (maybe the edge guide wasn't engaged?) or something that Pro8mm or Moises did after the scan was complete. But I can assure you that there is no rocking of the image in the scan itself. This issue is a red herring, in my opinion. -perry

I wouldn't hold your breath on any changes to the stock. Kodak claims the stock is within the SMPTE spec for Super 8, which allows for this level of inconsistent perf placement -- something that's been baked into the format since the beginning (see my post above with a link to film from 1969 that has a slightly different irregular perf pattern). Do you really think they're going to bother spending resources on this at this stage of Super 8's life cycle? Remember, we're talking about Kodak here - not exactly known for their forward thinking... There are two problems seen in the video at the beginning of this thread: 1) The perforation wiggles back and forth. Two or three months ago, the whole frame would have wiggled and the perf would have been stable. A software update from Lasergraphics now uses the edge of the film to stabilize the image while scanning. There is NO post-scan manipulation necessary to do this. The perf movement you see is there because the film edges are now stable. What you're getting here is a more accurate scan because it reflects the state of the actual film and the whacky perf positions from frame to frame -- the scan shows how the strip of film is constructed. Again, nothing needs to happen here other than selecting the type of optical registration you want, before initiating the scan. It's a real-time operation done while the film is being scanned. Since the perf will be cropped out anyway, all of this is kind of moot - it's behind the scenes stuff done by the scanner that just happens to be exposed if you overscan the way the film above was overscanned. 2) The rotational rocking motion: the source of this is unknown. But we've scanned film from 4-5 different Logmar cameras so far and have yet to see anything resembling that rotation. Carl is correct that the problem seems to be either: A) something specific to that camera -or- B) something that was done AFTER the scan and before upload to Vimeo. With the horizontal weave fixed IN THE SCANNER, in REALTIME, at this point, I don't think it matters that much that the perforations are imperfect. Other scanner manufacturers will probably implement a similar fix soon, and eventually this problem will just go away and we'll all have stable scans. The rocking motion in this scan, though, is curious and I'd like to know more about where that came from. We've never seen it on our ScanStation and we've scanned hundreds of thousands of feet of S8 on that machine. -perry

We agree on this. You can clearly see the rocking motion in the clip Moises posted. This is not something I've seen in any other logmar clips that we've scanned, so I agree one of these two options is probably the culprit. I can't think of anything else that might cause that effect. -perry

I didn't have a ton of time to do a search through old footage, but I grabbed some reels of my own family's home movies, which happen to be sitting on a shelf at the office. I picked a random reel and scanned through it. All of the film I looked at was Kodachrome II Type A, which was manufactured from 1965-1974. This one was manufactured in 1969, according to the date code burned into the film. Here's a link to a few seconds of the film as a 2k ProRes file, zooming in on just the perf and film edge (which is why it's a bit soft, since there's some digital blow-up happening at this zoom level): https://www.dropbox.com/s/3wivwy3s5hz8h4l/1969kodachrome.mov?dl=0 As you can see, the perforation is moving relative to the film edge. it is *not* as extreme as what we see on better cameras, but there's definitely inconsistent perf placement. The pattern is actually a little different than what we see in most batches, which result in a sawtooth wave repeated every 5 perfs. What we see here is that perfs 1 and 6 are the same, and are off from 2, 3, 4, and 5, which seem to be the same. That is, in a 5-punch perforator, the first one is off, while the next 4 are the same. What we see in modern film looks as if the punch is at a slight angle relative to the edge of the film, thus the sawtooth pattern. With older footage, perf inconsistencies are less common, but they are there. They manifest in slightly different patterns than what we see in modern stock, and in many cases with older film, they're basically perfect. This says to me that the manufacturing of the film has gotten sloppier than it once was, but that it was never 100% perfect and that this issue has been there pretty much since the beginning. -perry

I think there are a couple things being argued here that are getting all tangled up together. First, let's talk about the horizontal weaving. Perforation edge to film edge distance varies in Kodak Super 8, frame by frame, on a 5-frame cycle. That is, frames 1 and 6 are the same distance from the edge of the film, but frames 1 and 2 are not. When the camera is edge guided (pin registration doesn't really come into play with this if the pin is small enough not to move the film), a scanner that doesn't also edge-guide will result in an image that weaves back and forth. This is what the ScanStation produced up until the past month or so, when software edge guiding was added. We agree on this, no? Second: Rocking - I'm not convinced the rocking exists in the scan. I can say this because we've scanned a bunch of footage from the Logmar and have never seen it in our scans. The only time we've seen this was in the test films Friedemann Wachsmuth scanned with us and posted to Vimeo a few months ago, but that footage was post-processed and stabilized by him. The gentle rocking you see there IS NOT in the original scan, which I am looking at right now. Nor is it in the subsequent scan we did of the same film, after getting the software update from Lasergraphics that fixes the horizontal weave issue. No rotation. It's not there. I also looked at some other Logmar footage we've scanned for other clients, and there is no rocking on their scans. The two examples I've seen of this are the one at the top of this thread and the one Friedemann posted in May or June. We know Friedemann's version included a post-scan stabilization pass. We know Moises did not stabilize this film based on the initial post in this thread. Do we know if any post-scan stabilization happened at Pro8mm? Can Moises confirm that? Third: Edge guiding. The Scanstation uses the left (perf-side) edge for horizontal registration. The software places it along a perfect vertical line. If the film were to pass through the camera at a 60 degree angle, the software would align it to a 90 degree angle (not that that could happen, just using an extreme example to illustrate). Using the right edge of the frame is entirely unnecessary, because the left edge is being used to correct for both position and rotation at the same time. This was confirmed with Lasergraphics this morning. -perry

Here's what I don't understand about your argument. I've been shooting Super 8 for 25 years. I have always seen horizontal weave in the film. That's with film that was projected, scanned, and telecine'd. It's demonstrably worse with modern scanners that don't do what the ScanStation is doing now (edge alignment for horizontal positioning), because they're more precise and that exposes the inconsistently placed perfs. But horizontal weave in Super 8 has always been there, with just about any camera/projector setup I've ever used. When I was in college, I shot Super 8 on probably a dozen different cameras (ranging from my own crappy Chinon to just about every Nizo ever made, Canons, Nikons, etc), and projected them on everything from small consumer projectors through really nice onces like the Elmo GS1200. They all had horizontal weave, even on shots where we had the camera locked down on a beefy Sachtler fluid head tripod. I just don't understand how you can say the scanner is the problem here, when the problem is clearly with the manufacturing of the film. And don't get me wrong, I'm not bashing Super 8 - I love the format and still shoot in it. But it was conceived of as a cheap and convenient consumer format. It was never meant to have the level of precision you'd expect from 16 or 35, and that's just fine. It is what it is. But placing all the blame on the manufacturers of scanners and saying that the film itself is fine simply isn't correct, because this is an issue that's always been there - even when projected. -perry

Depending on the scanner being used and the pricing, why not do both? We can scan to DPX and ProRes simultaneously on our ScanStation, and it's only a little more per foot for the second file. No reason not to do 4k DPX for archiving and 4k ProRes for working with. I mean, unless you don't need archival copies, then I'd just stick with ProRes 4444 or something similar. -perry

Please explain, then, why 35mm, 16mm film and regular 8 (basically 16mm film) don't have this problem. At all. It's a flaw in the Super 8 film. It's within spec because Kodak's spec was sloppy. It was sloppy because it was meant to be cheap and fast, not precise. It's good scanners and good cameras that expose this flaw. -perry

This argument best applies to newly shot film. One of the reasons we chose the ScanStation was because it does all of this optically and can easily scan the most screwed up, shrunken film you can imagine. We ran some film through the scanner last week that the client had been told was unscannable. I wasn't even sure we could do it, because the film was so badly shrunken and cupped. But you know what? the only reels we couldn't deal with were the ones that had fused into a solid block of acetate. All of the rest of the film went through the scanner, and despite the extreme shrinkage, we were able to get really nice 4k scans of what was left on it. Mechanical edge guides would have been useless in this case, because the film was so warped, a spring loaded guide would have pushed the high edge up higher. If one existed, I wouldn't use a pressure plate with that film, because it likely would have caused further damage to the film. In terms of the pin registration, you can't put shrunken film into a mechanical pin-registration system without risk of damage (or without using specially modified pins, which are by design narrower and less precise). So in terms of pin reg, optical is always more accurate when you're talking about scanners of this resolution. -perry

That's *precisely* what the recent software update for the ScanStation does. In fact, you could argue that, much like optical pin registration, the way it's being done in this scanner is actually more accurate than if it were done mechanically. -perry

I really don't think this is a fair characterization of what's going on here, and I don't think blaming others (besides Kodak) is constructive. The scanner - in this case a ScanStation, the same machine we use at Gamma Ray Digital, is using both the perforation and the edges of the film to stabilize the image during the scan. It uses the perf for vertical stabilization and it uses the edges of the film for horizontal. The latter is a new feature added a month or so ago, and it works quite nicely. Previously, the perforation would be in a fixed vertical/horizontal position and the edges of the frame would weave back and forth in a pattern corresponding to the perforation's position relative to the edge of the film. Kodak's perfs are to the SMPTE spec (which I presume they wrote), but precision on the order of 16mm or 35mm pin registered film probably never even crossed the original engineer's minds back in the 1960s. Super 8 was conceived of as a consumer format, not a pro format. In 2015, we're pushing it in directions Kodak never imagined - Logmar included! The latest versions of the ScanStation software added the film-edge stabilization as an option for film in which this problem occurs. This is effectively the same as the edge-guides in the Logmar camera. Now, the film held laterally stable while only the vertical edges of the perfs are used to stabilize on that axis. In our experience (having one of the first installed ScanStations), only very stable high end cameras exhibit this problem, either because they use edge guiding or because they have really solid film paths. Canons, Nikons, Leicinas, Logmars, and a few others show this problem. We've seen it with film going back to the 1960s, so it's always been there, it just hasn't been exposed until the advent of pin registered Super 8 scanning. As for all the complaints about the weave - remember that overscan like this is not typically how the film will be displayed. The weave is *really* obvious when you can see the perfs, but perfs are not normally seen. Once you crop the film edges and perf out, it's damn-near rock solid. And that's pretty impressive for a format that was never meant to do what we're doing with it. One other thing - there is a scratch in this film. I don't believe that was caused by the scanner, simply because the only part of the scanner that ever touches the center of the film is the capstan roller, and because it's a capstan, driving the film forward, it doesn't slip over the frame and couldn't cause a scratch like this. The film path through the gate itself is a V-groove, so the only part of the film that's touching the gate itself are the edges - the upper points of the V. My guess is that the scratch is from the cartridge, camera, or in processing (or possibly in breakdown/setup of the film before/after processing). -perry

It's worth getting something like a Tobin TSC Speed Checker to verify if the motor is truly good. Good to have no matter what, but it'll tell you without requiring any film if your motor is holding sync speed. -perry

We're currently working on the digital restoration of a fairly well known 1970s horror film that was shot on 16mm but primarily distributed on 35. We've got 16mm and 35mm intermediate elements for scanning, but it seems the 16mm intermediates (CRIs) were actually reductions from the 35mm, because the 35mm frames actually slightly more picture, indicating the 16mm film was cropped. You'd think 16mm intermediates would be contact prints made from the 16mm A/B rolls (which are nowhere to be found), but apparently not. David is right - the path an older film like this takes to what you see on YouTube can be a long and very weird one, full of strange twists. It's like detective work sometimes, trying to figure out what's what!

I personally prefer Rivas-style splicers, and that's what we use here for any splicing of 16mm. The tape is pre-perforated so it takes a little practice to get the technique down, but the splicers are solid and easily available on ebay. They look like this: http://www.ebay.com/itm/Lot-of-2-QTS-Ediquip-16mm-Film-Splicer-Model-No-1060-1-Used-/231582138554?pt=LH_DefaultDomain_0&hash=item35eb5f00ba Note there are two there - one splices straight and the other diagonally. You don't want the diagonal splicer, that's for 16mm fullcoat audio tape. You want the straight type. Also, these are in rough shape, but these splicers are super easy to take apart with just a screwdriver or two, and all the parts can be easily cleaned and lubricated, so they're simple to maintain. Lots of people like the Ciro guillotine splicers, and they're fine for new film, but they're much more expensive. The advantage is that they don't use pre-perfed tape, and they're less tricky to use. I prefer tape, because you can undo the splice if you need to and it's less destructive than cement splices. -perry

Here's the thing. If you're paying $25k to scan your dailies, you're either doing the work in the wrong place, or you're talking about insane amounts of film. A little over 20 years ago, I made a 16mm documentary from raw stock to finished print. It was only a 12 minute film. I worked some deals on the stock, had free access to cameras through my school, and happened to work at the lab, so I got heavily discounted processing, timing and printing. I had to pay full price for my optical track, because the lab didn't do that in-house. that 12 minute film cost me a grand total of about $3000 over the course of a year, and that's with discounts and upgrades to my Mac LCIII to handle 16bit sound so I could do my own audio mix ($500 for a 500MB hard drive!). Had I paid full price for processing, rented sync sound cameras and tape decks, and had I paid for a facility to do my mix, I would have been into it for well over $5000, most likely. Film --the whole process of making one-- has always been an expensive proposition. The only way to do it on a shoestring is to work deals, and frankly, most businesses in the industry are willing to do that, if you're reasonable about expectations and willing to be flexible. The good news, at least in terms of DI and scanning, is that the insanely specialized hardware to do this is becoming much more inexpensive, and that's allowing services like ours to offer very reasonably priced scans. I haven't heard of anyone working in 16mm doing a $25K scan job in a while, unless you're talking major productions like some television shows or bigger-budget features. -perry

Bear in mind that with cameras like the R10, Logmar, Beaulieu, Leicina, some Canons (and even the lowly Quarz) that have very steady transports, you'll either have to use a scanner like the ScanStation or do post-scan stabilization, assuming you use a scanner that does optical pin-registration. To my knowledge, the ScanStation is now the only one that properly handles Kodak's messed up perf placement on Super 8, which causes jerky back and forth motion on film shot on high end cameras, when the scanner uses the perfs to digitally pin-register the scan. You can see an illustration of the perf issue here: ...basically, they're not consistently placed, relative to the edge of the film, but the problem only really shows up when the film is shot in a pin-registered camera like the Logmar, or a camera with a very stable transport, like the others mentioned above. That, in combination with a pin-registered scanner results in an annoyingly jittery image that requires post-scan stabilization, and that can have negative effects on the image quality if not done correctly (blur, pixel interpolation, etc). The footage above was actually scanned on our ScanStation a while back. It's from one of the very early prototype Logmar cameras. Lasergraphics recently released a recent fix to address this problem in their scanner - the perf is still used for vertical registration but now the edges of the film are used for horizontal, and that all but eliminates the jitter problem. Now, you see the opposite effect, which is more representative of what's actually happening on the film: the film is steady (left to right), but the perf jitters back and forth in the pattern seen above). I don't think the Kinetta, Flashscan or others that use optical registration are doing this, so your mileage may vary with those machines. -perry

In a word ...Don't. Feed the site the highest quality you can, even if that means you have to wait overnight to upload. H.264 is highly compressed even at high bit rates, and is strips out a lot of color information as well. Doesn't matter if *that* file looks good on your computer because Vimeo, Youtube and whoever will recompress it further, to their specs, after you've uploaded it. Think of it is as a photocopy of a photocopy. Generation loss and all that. Try uploading something like a ProRes 422 file (even regular ProRes 422 should give you an improvement, with a little file size savings over HQ). Also, higher resolution matters here. If you can upload 2k, you get better HD when they recompress it. Same with YouTube. -perry

...but doesn't actually exist in the wild except for a few slick photos of the models who apparently run these machines. It's been close to a year and a half since they were first announced at NAB 2014, with no sign of them. As of today, vaporware, pretty much. Lasergraphics has a stripped down version of The ScanStation now that starts at about $50k, I believe. limited in several ways but I'm sure more feature packed than the Cintel or Moviestuff units. -perry

Take a look at the date on this thread - 10 years ago almost to the day! Things have changed a bit in the past decade, so it seems unfair to criticize someone for something they said back then... That said, while scanning motion picture film on a flatbed is doable, it's a needlessly tedious process. There are much better ways to handle this kind of thing in 2015 using simple, cheap motion control systems, LED lighting and industrial machine vision cameras, scanning frame by frame, and avoiding the need for all all that post-scan processing work. Also, telecine for 8mm and 16mm films is basically done at this point. There are several scanners on the market now that can do 2k or higher resolution scans for small gauge film, and all do an excellent job. Anyone using a telecine for this is pretty much limited to HD at the most, unless it's heavily modified, and even then, it's subject to all the limits of that kind of a system.

A lot of the films we work on for North American DVD and Blu-ray release are film-originated. Most were converted to 25fps for broadcast in Europe or other PAL regions. The films themselves were typically made at 24, but the best available masters are frequently at 25. It's something we've become pretty adept at dealing with over the years, because we never know what we're going to get. A month ago we got 4 films that were all at 25 (but were originally shot at 24). Last week, three more for the same client that were transferred in Europe from the film masters to 24 (so we need to make 23.98 masters for DVD/BD here). -perry

Most modern time stretch algorithms do pitch correction. We do this all the time, converting from 25fps masters to 23.98 (often for clients who can only get the good masters of US films from European distributors, so they're at 25). Even tools like Soundtrack Pro do a nice job of pitch correcting. You'd never know the speed was changed unless you were watching a clock. -perry

That's a good point, and I should have been clearer in my original post - I'd avoid older *telecine*-style transfer systems, for the reasons David enumerated. Even the Scanity can have problems with splices, simply because it's a line scanner with a continuous motion transport. Our Northlight is like 12 years old and slow as hell, but produces beautiful scans. Unlike the Spirit, though, it's an intermittent transport, so it's not subject to the frame warping that happens on machines like the Spirit/Shadow/Scanity. -perry

Hi Mayer, We don't do processing, but if you'd like a quote on scanning, send me a message (or use the contact form on our web site. Personally, I'd avoid transferring on older hardware like the Spirit. You will get better results from a modern large-sensor scanner like the Xena or our 5k ScanStation, even if you're only outputting to 2k. -perry

Same here - my ACL 1.5 always kept perfect sync. I've recorded audio with a Nagra 4, Tascam DAT and even a cheap handheld digital voice recorder, and all were dead on when synced up. As long as both devices (camera and audio recorder) have crystal sync, you're good to go. -perry