Perry Paolantonio

The hard part is going to be finding a place with a 16mm film recorder. I'm fairly certain the one at Metropolis in NY can do 16mm - Jack Rizzo told me a couple years ago that theirs uses an Oxberry movement, so it's a bit more modular than most recorders, which are usually 35mm only. For a filmout, I'd definitely recommend scanning at 4k, though - you want all the resolution you can get on the digital side of the equation.

^ What David said.

If you want to show it on HD, then yes you'd pillarbox. But why limit yourself to that format in the transfer? If you're going to the trouble of transferring and conforming A/B rolls, you might as well do it at a higher resolution now, even if the immediate need is just for HD. That way you have a higher res version in your back pocket already, should you need it in the future... As for the pillarboxing/stretching: If you scale the 4:3 scan down to HD and pillarbox, you're making a 16:9 image, but the picture area is a 4:3 window in the center. The black just pads it out to 1920x1080, and becomes part of the HD picture. If you show that on an HDTV it wouldn't stretch because as far as the TV is concerned, that's full frame 16:9, even if about 25% of the frame is actually the black pillarboxing.

"Supervised Transfer" and an hourly rate suggests it's a telecine or telecine-based system, like the Spirit: real-time with color correction done at the time of transfer. This is different from data scanning, where you're just capturing the film flat to DPX or some other file format, for color correction later. With data scanning, the client isn't usually present for this part of the process because it's slow and uninteresting and doesn't involve any creative decision making. The Avid is their edit system, not their film transfer system. Presumably, the supervised transfer (On the Spirit) would include conforming the A/B rolls into the final, contiguous film, but it may depend on their setup and whether your cut neg has any optical effects (fades, dissolves, etc). I can't say for sure but I'm guessing they're charging for the transfer and color grade in the first part, and then the ingest/conform/layback in the second. The fact that it's 4:3, at least in my mind, would completely eliminate 1080p transfers from consideration because you'd have to pillarbox the image and ultimately, you're not getting a very high res transfer that way. If you ever wanted to make, say, a 2k DCP, you'd have to scale up significantly, which is something to avoid if possible. In the past few years, it's become much more affordable to scan at high resolution on scanners that have much better resolution and registration than the Spirit, so it's worth looking into. It's not something you need to do in a supervised way so you could send that part out, and then work with a local colorist to do the grading. That's something we do for a lot of clients who aren't local, and I know many other scanning services do the same. With the recent proliferation of Resolve systems out there, most of our clients either do their own grading and conforming, or they bring it to someone local to them. In fact, a lot of people are doing 4k scans with us, grading at 4k in Resolve, and outputting to HD or 2k (but hanging onto the 4k graded scan for when they need it - an inevitability, what with consumers starting to get their hands on 4k hardware in significant numbers). -perry

Is the original footage 4:3 or are you shooting Super 16? In the case of S16, you're not gaining a ton by going to 2k. But with 4:3 there's a compelling argument for scanning at 2k vs HD. With a 4:3 HD scan, you're getting an image area of 1440x1080, but with a 2k scan, 2048x1556. That's more than double the resolution. If the source is Super 16, you're getting more with 2k, but not double. Everything Dirk says is correct as well, though these days a 2k scan shouldn't really cost too much more than HD. We also do a special price for A/B rolls, because the easiest way to handle them is to scan each roll straight through, and then edit out the slug in a nonlinear edit or grading system later (rather than assembling during the scan). So we charge full price for the A roll and a discounted rate for the B roll. We do this a lot at 2k and 4k for 16mm, actually. -perry

hah - posts crossed. This, of course, assumes two things: 1) that you're using a sprocket-based transfer system 2) that the system in use is capable of zooming out enough to include the area between the perforations Neither of these things is an issue on a modern scanner, but your mileage may vary if you're talking about older telecine-based transfers. -perry

According to Wikipedia it's 11.66mm x 6.15mm -- http://en.wikipedia.org/wiki/16_mm_film#Ultra_16_mm Though I'm not sure this is an absolute, given that Ultra16 isn't an official standard.

Er.. your Aatons and Arris are professional cameras, built to a different standard than consumer gear. They cost tens of thousands of dollars new, compared to a few hundred for a S8 camera - cameras that were mass produced and used a lot of plastic. Even your Scoopic and R16 could be classed as low-end professional cameras, so you'd expect better build quality than most Super 8 cameras.

You're talking about 30-40 year old cameras in many cases, and in a consumer format, no less. These were probably never regularly serviced when used by their original owners (except in rare circumstances, i'd think). Electronics that old could be subject to a bunch of issues: solder joints that come undone, blown capacitors, frayed wires and short circuits. Not to mention corrosion from batteries left in the camera for years on end... Mechanically, there can be issues as well. There are moving parts in these cameras and in anything with moving parts, things are bound to break sooner or later. Especially when the parts are mostly plastic... "Belts" (usually just tiny rubber bands) don't last forever either, and those are often problematic in older cameras. Super 8 cameras, with a few exceptions (I'd say Beaulieu, Canon, Nikon, Nizo, Bauer and a few others) weren't usually built for the long haul. I'm not sure I'd classify Elmo as "excellent" - they were good cameras (I still have a working one that was in my attic for 20 years and fired right up last time I used it), but you know - these things are old!

Another quick update: The PCB arrived this morning for the RGB+IR lamphouse. This one is a bit big, but fits the Imagica. I am toying with tweaking the design (seems every time I finish a design I learn some more tricks that could minimize the size. Now I'm obsessed with shrinking it!). In any case, this will use SMD components, which I'm about to learn how to work with for the first time. Should be fun. I'm hoping to have it fully assembled in the next couple of days. https://www.flickr.com/photos/friolator/17037539117/in/set-72157644369553789

Very possible that it's a match. Phil Vigeant from Pro8mm told me that their modified camera uses a battery that outputs 3.6V and 7.2V -perry

I think all Nizos had Schneider-Kreuznach lenses. Maybe it was a modification?

one resistor per channel would probably work, but after talking to some EE friends, the recommendation was to keep them separate. I think this is a pretty good explanation of why: http://electronics.stackexchange.com/questions/22291/why-exactly-cant-a-single-resistor-be-used-for-many-parallel-leds I mean, it'd probably be ok, but it's not that big a deal to have multiple resistors if I'm using an array for them. 5 extra minutes of soldering, pretty much. Yeah - the Arduino PWM pins are being used for this so you can vary the intensity of each color independently to either do pure R, G and B with a mono camera, or mix a pretty good white for a color camera or black and white film. -perry

this isn't going to win any beauty contests, but here's the first prototype of the LED array for my scanner: https://www.flickr.com/photos/friolator/16936609298/in/set-72157644369553789 This board holds 25 RGB LEDs, which allows color to be mixed from the Arduino. At the moment, each color is on full blast, though it's not truly white since each color has a slightly different voltage requirement. A bit more tweaking is needed, but not much. What you don't see is that the other half of the breadboard has another 11 LEDs wired up, so it's actually a 36-LED array. That should provide more than enough light in the final version. Also, these aren't diffuse LEDs, so they're not really useful in the scanner at the moment. this is just a proof of concept. And it only uses 7W of electricity (Compare that to a Spirit!). It'll be much prettier when it's done - a custom printed circuit board, surface mount LEDs, integrated resistor arrays and a single connector to the Arduino (probably coax with 9-pin connectors)

Machine vision cameras like this run several thousand dollars, usually. The test camera I'm using in the 35mm scanner (again, just for testing purposes) runs about $1000 including the frame grabber. We'll swap this out for a CameraLink frame grabber when the scanner is closer to being finished, and at that time, can use just about any Camera that has a cameralink interface. If you're willing to work with a used camera, they do come up on ebay from time to time. There are lots of low resolution ones there now, the good high end cameras are harder to come by and are more expensive.

gah. cursed edit time limit on this forum! I tried to edit my last post, but I'm too slow. that link is to the photo in my photostream, not in the imagica-specific photo set. these are all the scanner photos so far: https://www.flickr.com/photos/friolator/sets/72157644369553789/ Near the end you can see how the relays were all rewired using cat5 to breakout connectors on DIN rails. It's a really nice, neat way to do this. My original layouts re-used some of the old wiring, but I was having all kinds of problems. It just made more sense to rip it out and start all that from scratch. One of the things I decided to do was to make it possible to shut the scanner off from software using the Arduino and relays. That way if a slow scan went late after hours, I could program it to shut the whole machine down, or log in remotely and do it from home. -perry

Bummer. Mine are right inside the door: https://www.flickr.com/photos/friolator/14053271300/in/set-72157644369553789 (more photos of the scanner in various states in that photoset, too). The lens in it is the same as in your XE - a Nikkor 95mm printing lens, so about as sharp as they get. That lens is actually worth more than I paid for the scanner! -perry

Does yours have the shinto prayer cards inside too? I'm not superstitious, but I figured I should leave them there. Visions of the Brady Bunch in Hawaii... Damn straight! $12 on amazon. Definitely not suitable for actual work, but good enough for preliminary testing while the real thing is being built. I'm still debating whether I should do RGB + IR. Kind of leaning towards going for it - I mean, while I'm going to the trouble of building my own LED array, if I can squeeze a second matrix of IR emitters in there, then I can make dustmaps for restoration, too. -perry

The advantage of working with an existing transport is that all the engineering was done for me. I just have to hack together the pieces. So there are a ton of sensors on the scanner for different things: perf counting, film tension, gate position, etc. The arduino makes it pretty easy to read all these sensors and react accordingly. In the case of frame counting, there are two perf sensors that, when they're in the correct position, tell you it's safe to engage the registration pins. But I'm using one of them as a counter (essentially it's the same idea as your LED/photosensor on the projector shutter blades, only it's a proximity sensor that goes to the off state when the perf is in front of the sensor). So tracking frames is fairly straightforward, just by keeping a count of where I am and then dividing by the number of perfs per frame. Ebay. I was watching it for almost a year and decided to just go for it. The only downside of this model is that I can't use 2000' reels on it. But I'm going to remove the top cover to the scanner and mount some new reel hubs a little farther out, then connect them to the torque motors that keep film tension via a belt or chain drive. So basically, I'm just moving the center hub a few inches and letting the platters overhang the chassis a bit. I've got a couple of old 35mm analysis projectors someone gave me, with lots of rollers and similar parts I can scavenge for that. Cool. I looked at OpenCV, but decided to go with the software that works with the frame grabber, and I'm glad I did. It can do a fair bit of stuff with the image in the frame buffer while it's in memory, like flipping it, scaling, etc. That's pretty fast. I can also pass this off to ImageMagick in memory, where I would merge the three channels into an RGB image and then write it to disk. So theoretically, it should be reasonably quick. The current camera is limited to about 5fps, and I think right now the bottleneck is my transport speeds, which aren't very fast. I'm guessing that once everything is fine-tuned. it'll be able to do somewhere between 1 and 2 fps at full resolution, with pin registration With ImageMagick, I believe I can also do some stabilization, so I may have the gates enlarged to include more of the perfs than you currently see, to give me something to do optical pin registration on. That should speed things up and make it safer for archival film. -perry

Heh. You should see the shelf full of bits and pieces I took out of it. Really all that's left is the transport. The original design had a robotic platform under the scanner deck with the line sensor. Behind the film was a fibre optic snake that went down to the halogen lamp (just 24v projector bulbs!). The sensor and the light snake would sweep the film in sync with one another. There's a bit of diffusion between the lamp and the gate, but it's just photo diffusion material. for now I'm leaving it be to see how even my final lamphouse design is, but i've been looking at diffusers like the ones in the Edmund catalog if it's not good enough. I'm in the process of prototyping the LED array, and once the kinks are worked out, I'll get a PCB made for it and get it installed. Right now, for testing, I'm using a 24v LED MR16 bulb I bought from a RV supply house. Other than being sickly green, it's actually surprisingly even. So I'm kind of modeling my array after it, but mine will let me control the individual channels. -perry

Cool. When I'm done with it (hopefully in a few weeks), I'll post some footage of my 35mm scanner. It's an older Imagica scanner, with the guts removed. Using an Arduino Mega to control it, currently building a custom LED (RGB) lamphouse. It'll be slow, but will do multi-flash per color RGB scanning with a mono sensor. I'm using a cheap color machine vision camera right now (4.6k), but it's just for prototyping. Once it's done, I'll swap it out for a higher end camera. The camera's frame grabber board has a separate C++ library that gives you full control over most of its functionality, so I bought that from the manufacturer. As long as I use one of their CameraLink boards in the final version. it's just a swap-out replacement and will work with nearly any CameraLink camera. The control software is being built in RealStudio (currently Xojo, but I use an older version). I built a whole serial command and response system in the Arduino, and a library of transport functions that you can call from the application. Basically, I send it a serial message and it does what I tell it, then reports back. The scanner I'm basing this on had 5-phase stepper motors and motor drivers (big external boxes, but conceptually similar to the one you're using), so I'm controlling steppers for forward and reverse motion, lens focus, camera platform focus, and pin-registration/pressure plate. Kind of mind blowing what you can do with a $20 controller like the Mega. The plan is to get PCBs made for this once I'm satisfied it's working properly, then mount everything in a 1 rack unit box inside the chassis. It's taken far too long to get to this point, but it's nearly done and now it's becoming a lot more fun. What are you using to invert the negative and remove the orange cast? is that a filter in OpenCV? I'll be doing all the image processing in memory using ImageMagick, so that's not particularly hard to deal with, it'll just take some calibration to get it just right. Here's some early footage of mine, taken a few weeks ago, with the transport finally responding as expected to my commands: https://www.youtube.com/embed/YWuAcmAf2ww -perry

I agree with Mark. There must be something up with your scan, and you're not going to get very good results with one of these setups. What was wrong with your initial scan? Can you post some frame grabs? What kind of scanner was used? -perry

Since you're on Windows, making a ProRes file is a bit more complicated (reading ProRes isn't an issue, but making them is less straightforward than on a mac). Resolve, for instance, can't export a ProRes file because Blackmagic doesn't offer that as an option. If we're not scanning directly to ProRes (our scanner can do that, even though it's Windows based), we use ffmpeg to do conversions - not user friendly but it works. There are some limitations, however. Premiere might export it out, though - I'd be surprised if Adobe doesn't include this (you'd need to have Quicktime installed first, of course). Resolve (at least version 11) can import your files on Windows. We have two systems: the full version and a light version, and both will take the 5k file (though the Lite version will limit your output to UHD - you need the full version to round trip 5k->5k or to output true 4k. In resolve, just point it to the folder of DPX files, and it appears as if it was a single file. Just remember - these files *won't* play back with speeds near real time at that resolution, as DPX, off of your external drives. On the RAID in our Resolve system, 16bit 5k files play at about 5-6fps. This is just playback with no color correction or other effects - so the low speed is almost entirely due to disk speed. That's on a RAID that can easily move data at 1.5GB/second. If you're playing back off of an external drive I'd expect slower speeds. 5k ProRes 4444 might play back ok if you have a fast PC - the disk speed won't be as much of an issue, but ProRes is CPU-bound and 5k 4444 files requires a fairly hefty CPU to decode in real time. -perry

Memory isn't as much the problem as disk space is. Your 50' reel is just under 450GB as 5k, 16bit DPX. ~115MB/frame. So the main issue will be disk space, not RAM. As Phil said, it's not loading the whole movie into RAM, just a few frames at a time usually. 5k DPX won't play back smoothly off external portable drives, you'd need a RAID capable of over 3GB/second in order to play it in real time. The usual workflow for files this large is to make smaller proxy files (you could do this from After Effects, at something like 2k, where you'd need speed more on the order of 500MB/second for DPX, or just a fraction of that for something like ProRes) and edit/grade those proxies. Then when you're finished, relink the media in your timeline up to the 5k version, set up your final crop, and render out the final version at 4k. -perry

Hi Scott, I haven't used Premiere in years, but in After Effects, when you bring in an image sequence it shows up in your bin as if it's a single file. If you right click on the file there's an option labeled "Interpret Footage." there you can set the frame rate. You might see if there's something similar in Premiere - Adobe is usually pretty good about re-using features from application to application. If you can do this, then that's where you'd set your frame rate for the incoming DPX sequence, which will appear in the user interface as if it was any other movie format. If you bring that into a 24fps timeline, you'd have to either repeat frames (like a step print process would have done back in the day), or interpolate up to 24. I'd avoid interpolation, but that's just my preference. On the crop - your scans were done with overscan. They can be scanned directly at 4k with a slight crop just inside the frame lines, if that's what you want. Saves you a step and gets you smaller files to work with. Phil - you'd be surprised how much is actually on the film. 5k is not a format we scan to very often, except for archives that want to capture the full frame plus perfs and film edges, but there's subtle detail there that's lost in a 2k or HD scan. -perry