davidferguson

Online stream for those without a LW AM receiver: https://www.youtube.com/watch?v=Ugd8G5w-Sfo

See my top level comment for more info on this, but the Aladdin scan used in the article was from a 35mm trailer that's been scanned on an unknown scanner, and had unknown processing applied to it. It's not really possible to compare anything other than resolution and artefacts in the two images.

And it was made by a lab that made choices on processing and developing times, that can quite easily affect the resulting image. You hope that labs are reasonably standard across the board and calibrate frequently, but even processing two copies of the same material in a lab, one after the other will result in images that look different if projected side by side. This is why it's probably impossible to made new prints of 3-strip-cinerama films now, the knowledge and number of labs that can do this are near zero.

TL;DR: Linking to YouTube trailer scans as comparisons for colour is misleading and not accurate.

---

see the 35 mm trailer for reference

The article makes heavy use of referring to scans of trailers to show what colours, grain, sharpness, etc. looked like. This is quite problematic, because you are replying on a scan done by someone on the Internet to accurately depict what something looked like in a commercial cinema. Now, I am not a colour scientist (far from it!), but I am a motion picture film hobbyist and so can speak a bit about some of the potential issues.

When projected in a movie theatre, light is generated by a short-arc xenon lamp. This has a very particular output light spectrum, and the entire movie process is calibrated and designed to work with this. The reflectors (mirrors) in the lamphouse are tuned to it, the films are colour graded for it, and then the film recorders (cameras) are calibrated knowing that this will be how it is shown.

When a film is scanned, it is not lit by a xenon short-arc lamp, instead various other illumination methods are used depending on the scanner. CRTs and LEDs are common. Commercial scanners are, on the whole, designed to scan negative film. It's where the money is - and so they are setup to work with that, which is very different to positive movie release film stock. Scanners therefore have different profiles to try and capture the different film stocks, but in general, today's workflow involves scanning something in, and then colour correcting post-scan, to meet an artist's expectations/desires.

Scanning and accurately capturing what is on a piece of film is something that is really quite challenging, and not something that any commercial scanner today does, or claims to do.

The YouTube channels referenced are FT Depot, and 35mm Movie Trailers Scans. FT Depot uses a Lasergraphics 6.5K HDR scanner, which is a quite high end one today. It does have profiles for individual film stocks, so you can set that and then get a good scan, but even the sales brochure of it says:

Many common negative film types are carefully characterized at Lasergraphics to allow our scanning software to compensate for variation. The result is more accurate color reproduction and less time spent color grading.

Note that it says that less time is spent colour grading - it is still not expected that it will accurately capture exactly what was on the film. It also specifies negative, I don't know whether it has positive stock profiles as I am not lucky enough to have worked with one - for this, I will assume it does.

The "scanner" used by 35mm Movie Trailers Scans is a DIY, homemade film scanner that (I think, at least the last time I spoke to them) uses an IMX-183 sensor. They have both a colour sensor and a monochrome sensor, I am not sure what was used to capture the scans linked in the video. Regardless of what was used, in such a scanner that doesn't have the benefit of film stock profiles, etc. there is no way to create a scan that accurately captures what was on the film, without some serious calibration and processing which isn't being done here. At best, you can make a scan, and then manually adjust it by eye afterwards to what you think looks good, or what you think the film looks like, but without doing this on a colour calibrated display with the original projected side-by-side for reference, this is not going to be that close to what it actually looked like.

Now, I don't want to come off as bashing a DIY scanner - I have made one too, and they are great! I love seeing the scans from them, especially old adverts, logos, snipes, etc. that aren't available anywhere else. But, it is not controversial at all to say that this is not colour calibrated in any way, and in no way reflects what one actually saw in a cinema when that trailer was projected.

All this is to say that statements like the following in the article are pretty misleading - as the differences may not be attributable to the direct-digital-release process at all, and could just be that a camera white balance was set wrong, or some post processing to what "looked good" came out different to the original:

At times, especially in the colors, they’re almost unrecognizable

Compared to the theatrical release, the look had changed. It was sharp and grainless, and the colors were kind of different

I don't disagree with the premise of the article - recording an image to film, and then scanning it in for a release _will_ result in a different look to doing a direct-digital workflow. That's why major Hollywood films spend money recording and scanning film to get the "film look" (although that's another can of worms!). It's just not an accurate comparison to put two images side by side, when one is of a trailer scan of unknown accuracy.

I stumbled across the article about the ThunderScan in about 2012 when looking for info about ImageWriter II upgrades, and have been slightly obsessed ever since. It's such a brilliant idea - a higher resolution scanner, that was far lower in cost than its competitors, achieved by reusing the paper transport that most customers already had.

I'm lucky enough to own two working ThunderScans now (and one third one that I needed the software driver from). They work exactly as advertised, and it's a joy to see them zip across the page, digitising line by line.

The software by Hertzfeld is another joy to use. The scrolling, which Hertzfeld calls "inertial scrolling" in that article, is now familiar to us all who have used touchscreen devices. It's funny to think that the feature that wowed so many at the 2007 iPhone launch actually existed all the way back in 1984, designed by one of the key creators of the Macintosh.

I wish there were more creative hacks like this - I just know that if a company tried to do something similar today, the printer manufacturer would instantly roll out an update to break this functionality.

It's fantastic work you've done. As someone who works at a older software company (founded early 80s), I'm sad that there isn't a push internally for us to make our old software source available, or even just the binaries available!

What sort of tactics did you use to convince them? Maybe I can apply them to where I work too...

Sometimes it's forced upon cinema owners. There's only one company making non-laser cinema projectors these days, and so laser is really the only tech in town when it comes to replacing an ageing or broken projector.

Thanks for the advice, it's appreciated! Any ideas how to get in touch? If not I'll try my luck at public twitter (which has had a 50/50 success rate for me in the past...)

Oh you worked on How to Train Your Dragon? That's my favourite movie in the world!

Really random follow up question - there's a How to Train Your Dragon advert that I assume played before the first film in cinemas, advertising the cinema's popcorn and other concessions, but the only two copies on YouTube are in really terrible quality. (one in English https://www.youtube.com/watch?v=f37fI14Ne0I, the other in Finnish I think https://www.youtube.com/watch?v=W0Hr4mD3G1Q). I really really want a high quality version of this, and I'm wondering if you know of anyone at DreamWorks who I can contact about this? I've tried tweeting at a couple of people but unsurprisingly heard nothing back, so I'm maybe looking for someone in marketing?

I have a Newton MessagePad 110, got it off eBay some years ago for a cheap price.

It still works, and I’m going to take it to EMF Camp this weekend and send a fax using it!

If you try hard enough you can fit more! I've seen a projector with all of 35mm optical, 35mm mag, DTS, Dolby Digital and SDDS. Was a sight to behold!

There was no demand, because you can't buy 35mm prints, unlike 16mm and 8mm. And additionally, in order to finance the purchase of digital projectors, many cinema owners went with something called "Virtual Print Fees" (VPFs), where the cost was paid off in time. However, most of these mandated that the film projectors were either removed and destroyed, or disabled so they couldn't run film. Why? Well this ensured that everyone switched to digital and couldn't go back, so the VPFs were still paid. Very sad, hearing about all these machines being damaged.

Is that a Bang & Olufsen? They commonly had a similar system of main and "extra" speakers.

Indeed it is! Films come on 20 minute reels, and from the late 80s onwards were spliced together onto one big reel (or platter) for playback. However the splice often wasn't perfect, and the reel ends are most likely to contain dust/dirt/scratches, which could knock out the digital audio and revert playback to the analog track. This obviously isn't ideal, so Dolby had the idea of encoding the area at the end of each reel throughout the entire reel too, which is stored in a cache in the processor. Then, when the end of the reel is reached (you can tell this because each block has a sequence number), the processor uses the cached data to ensure a nice changeover.

There is actually a reason to do this - and people do. For long term archival film is king - it's store and forget, unlike hard drives that require migration every few years. So people write digital data to film, either in the form of text documents as images, or some digtal encoding of the data, and then it's put into long term storage.

GitHub did this with its Archive Program https://archiveprogram.github.com/, and Piql https://www.piql.com/ is a company that specialise in doing this. They're the company that created the Cinevator, a high speed data to film transfer machine that was originally used for creating film prints, but since film is kinda dead these days, they've pivoted to archival storage.

Absolutely true and I agree with you. Something else to back up that it's not quite AC3 is that I've captured the raw signal that goes into the decoder box's "AC3-decoding card", and it's not quite what I'd expect for AC3. You can learn more about that and download the raw signal capture at https://fanrestore.com/thread-2633-post-73179.html#pid73179

I've captured Dolby Digital data to rip the audio from the film before. I don't (yet) have a SDDS reader, but when I get one it's something I want to try.

My knowledge comes from someone who helped develop Dolby Digital on film so I'm pretty sure it's correct.

All correct (as far as I'm aware), except for the unfortunately common inaccuracy that Dolby Digital on film is AC3. Sadly it's not - it's an earlier iteration of that technology that's incompatible. It must be pretty similar, though.

Yes, a few 35mm films had 4 track mag audio. The only one I can think of off the top of my head was Fantasia.

Although interestingly, Dolby Digital on film isn't exactly AC3 - it's an earlier version that's incompatible.

Dynamic loading was indeed a software update of the processor using the data carried on film. If a film contained an update, it would be contained at the start of a film. The processor would detect this, read the software update, reboot (and revert into mono analogue audio for this) and then continue off using the updated software.

There are 3 blocks every 128 on the film that contain the software update (and/or other data channels, or are unused). They're also used for something called the "Splice cache" which I can discuss if you want. The rest are the audio blocks.

As far as I'm aware, Dolby only used this once - to update from what they called version EC9 to EC11. I've also heard that it didn't work well with the DA10 - their first processor - and this may be part of the reason.

TLDR: If you know about audio encoding and/or want to help reverse engineer this format, I'd be very interested in hearing from you!

----

Oh, a topic I know a bit about!

To add a few things onto the stuff mentioned in the article: - A lot of places mention Dolby Digital on film being AC3 encoded. This is sadly not true - ish. Dolby were developing AC3 for film and other applications, but the version that ended up going on film isn't quite AC3 - although I suspect it's very close. To quote someone at Dolby:

The audio coding used is actually a pre-cursor of what became AC3, and in fact, two different versions of the codec were used during the lifetime of Dolby Digital on film. We referred internally to these as EC9 and EC11. I don’t remember the exact characteristics of these codecs but unfortunately they are not compatible with AC3

- The article mentions Reed-Solomon encoding being used - this is indeed the case, but more specifically a concatenated code meaning that two Reed-Soloman instances are applied one after the other. However because nothing is nice and simple here, some of the bits are only protected by one of the R-S layers and not both, but I'm not sure which at the moment.

- The article questions how the bytes are stored - this is actually mentioned in the patent. The bytes are arranged in 2x4 groups, ie 2 horizontal, 4 vertical, and therefore each block is 38 bytes wide, and 19 bytes high. Interestingly the picture in the article is rotated - you can tell by the Dolby logo being the wrong way round.

- Something not mentioned in the article or the patent - before the Reed-Solomon decoding is applied, all the bits are XORed against a constant in order to break up large black or white areas which would be hard to detect. Because the same value is used during encoding as decoding, the original data is recovered. Sadly, I don't know what this constant value is.

I would love there to be a open source decoder for the Dolby Digital on film, but at the moment I'm at the limit of my knowledge on AC3 and audio signal processing. If anyone knows more, or wants to help, I'd be delighted! You can reach me on email `fergusondavid6@gmail.com` or discord `davidferguson#2018`

And actually, spam is quite interesting to some people. It certainly gives a flavour of what early-2000s internet was like, and what happens when spam filters aren't good.

That's nice, but it doesn't scale. Google only let you solve a few (5 or so) audio captchas in quick succession before you're banned for a while, so it's no good for us.

It's going alphabetically through the group's people have nominated, and once that's done, it'll go by group member count.

You can only join a very small number of groups this way before it just stops working - we did try this out early in the archive process

Don't suppose you can find a way to unban our accounts...? ;)