Dan Baxter

Film doesn't have a gamut as wide as Rec.2020 and Rec.2020 isn't full gamut either. Anyway we know how to force it to do full gamut... but what we don't know yet is whether it matters at all to the final quality. That will require further testing with the right type of film. There's a WIP manual for it now that's available to users... previously you're correct there was very little in the way of documentation. I'm really not following you here Carl. You can't just scan a film and expect the scan to look exactly like a brand new projection print without any colour work. For many scanners grading is difficult, try grading a faded film off a GodenEye sometime - you won't be impressed.

Hey Pete, good to see you back! Yeah we've discovered some new stuff with the ScanStation that I know you'll be interested in (it pertains to a problem you literally told me about in person haha). 🙂 I'll shoot you an email and fill you in on some other interesting things I know you'll be interested in. Pertaining to the OP's original question - Kodak's NYC lab has recently replaced their Cintel with a ScanStation for dailies.

Don't make conclusions from this discussion. You can make it go a little bit beyond the Rec.709 colour space and get wide gamut using very specific scan settings, but we're still not sure if it really matters or not. It isn't clipping anything, it seems to be that for most formats what it does is use a LUT to convert the full gamut into Rec.709. That isn't necessarily a bad thing because if it didn't do that you'd get clipping if you did no post colour correction. Please have a look sometime at how the Blackmagic Cintel works - Blackmagic supplies the user with LUTs they have to use to convert the scan into a usable colour space in Resolve, so that the LG does this under-the-hood for you isn't necessarily a bad thing, it doesn't mean you're losing anything, it doesn't stop you doing HDR grades, it doesn't appear to be irreversible, and it cuts down the post-scan work required to make a deliverable file. That is to say if the scan has 10-12 true stops of dynamic range in it, it doesn't appear to be losing any of the dynamic range in the Rec.709 conversion. Also trust me on this: lots of Lasergraphics users think they know more about them than they actually do, we've discovered stuff that nobody else has known. I'm not talking about the obvious stuff that any professional company knows about. Yes there's an absence of under-the-hood settings, but again that's not really a negative. Even with a scanner as seemingly basic as a ScanStation in terms of software settings, the average user still has enough trouble as it is making them do consistent quality work. If you could change every under-the-hood setting you would quickly break it. As an example - let's say you need to scan without the diffusion concealing the base scratches because you have a job that requires you see all the film damage: a ScanStation can't do that. If you put something in front of the light to to make the effect go away then it gets confused and it can no longer find the perfs to register the frame. So that's an under-the-hood setting and changing it would require rewriting the capture software and providing an entirely new light-box with a software setting allowing you use the diffusion to hide damage or to make it visible. If enough users wanted that feature I'm sure they could add it, but most users would just use another scanner for jobs like that. So the point is that just because there's an under-the-hood setting that you can't change, it doesn't necessarily mean that it's a bad thing - if you want complete control over everything you can always commission a DCS Xena!

I assume you mean something like this? Just look at how he's handling the customer's film (running the film through his bare fingers while it's being transferred)!! What he's doing with the laptop is using it in place of a VCR or DVD recorder - that is he's taking s-video input and pressing "record" basically. He is right, most of them won't be in working condition especially for $400. Although that's about the only thing he's correct about, I definitely wouldn't be taking advice from someone doing commercial transfer work in 2022 using Elmos and Tobins! I seem to recall that this company (GotMemories) deletes comments that are critical about their services from their Youtube comments. The CCD imagers will develop faults overtime, read this thread. Resolution isn't even the main issue - although they only have a 250-line camcorder not 450 lines (unless it was replaced with an aftermarket camcorder). Some models have s-video output like in that video, but others are composite-only video. The ones sold in North America are NTSC, and the ones sold in Europe and Australia are PAL. If the camcorder works well (which it probably doesn't now after all this time) it'll still be crushing the blacks and getting poor colour reproduction as even if they have aftermarket ones they're ancient and some 10 years older than the tech inside the Tobin TVT-16 Telecine. It also doesn't appear to be designed to handle negatives at all so if you were doing negatives you'd likely have to work out the colour inversion yourself.

Oh joy, once again Perry is mistaken. Yes they do re-sell them, and one of my mates came very close to buying a used ScanStation Personal recommended to him by Steve Klenk. He held out as he saw that it was a dead product and purchased a full one instead.

He doesn't think my opinions are valid, probably best to leave it at that. Two of my friends have them, and they're essential if you want to do archive work. For an individual like you or hobbyist the main advantage would be that you don't have to spin them up like hard drives, so a cheap used LTO-6 or LTO-5 drive might be an option, but at that point it's no longer cheaper than hard drives per terabyte it's just a more stable and safer storage medium. If you want to experiment just look for a cheap used drive under $300 or so, set it up and if you like it you can invest more in a higher spec drive. If you can buy an LTO-8 drive for $1,000 then that's an absolute bargain and you'll save money over hard drives in no time.

Very interesting Robert. Nice form-factor, but it's still on feet whereas Arriscan comes on casters as you know and you can wheel it around. I know that sounds like a small thing, but I do know that one my mates has to take his Arri to the client on occasions where they have sensitive film they don't want to let out, so transportability can be important. Nice design though and it seems to be full-featured which would include their DFT soundtrack reader which Arri doesn't have.

Yes they did and that comparison is over a decade old and useless now. Don't ask me why LG still uses it because it makes their own product look worse than it currently is: you're not going to see noise like that without HDR scanning any more on either of them, plus they've both moved from CCD to CMOS so we're talking about a fundamentally different type of sensor noise. Also I'm pretty sure that the ScanStation doesn't do true HDR scanning, what the Director does is two or three exposures at different intensities to bring out the details in the dense regions: a little bit more in the highlights for negs and a bit more in the shadows for print. The ScanStation does two flashes very quickly at the same intensity, it's just designed to reduce noise really it can't bring out extra detail the way that the Director's 2-flash or 3-flash scanning can. Perry's in error. The early Arris were not designed for print that's correct, the later ones are. They have a setting for print and do a great job. They had 2-flash HDR from the very start I believe, or if not from the start early on. That's what you're seeing in those pictures: film that's too dense for the scanners that they were scanned on largely due to the imagers of the time. Those old CCD imagers needed to do a high pass and a low pass to compensate. You won't see noisy scans like that off normal prints now. Again the website is useless if you're wanting to know about Arri vs Director. LG is highlighting completely the wrong thing there. Director has options for all formats: 35mm,16mm, 8mm, and also 28/17.5/9.5 I'm pretty sure. It also has soundtrack readers that the Arri doesn't have (though it can't do 35mm mag). The Arri has a wetgate which the director doesn't, although that's only relevant in a wetlab setting. At the time of that comparison (2010?) the Director was 2K and 35/16 only, was a sprocket-driven scanner, completely different imaging tech, completely different light, etc. Director I think is easier to operate than Arriscan: Arri runs on linux, LG on Windows 10. Arri has interchangable transports: it has a sprocket transport and a sprocketless one, but the Director only has one transport now which is sprocketless (the old ones had sprocket transports). Pretty sure even now the Director has way more overscan wasting resolution compared with Arri. So unless you have a time machine and wish to travel back in time to 2010, all the meaningful differences are very different now to what they were then.

Just to clarify for you, his scanner is one of my friends. We haven't reached the same conclusion he has. But yes LG can be a pain with communication. We can test ScanStation vs ScanStation (same model run the same way) at some point, no need to test against completely different scanners. If Robino sends over a colour chart we could re-scan it on another ScanStation quite quickly and then it'd be obvious if one wasn't getting perfect colour. We can already test the same gen Sony Pregius in other scanners as it is, there's really no need to bother Robert unless it would be useful for him to do a comparison perhaps against his SS Personal - but as he already has a Xena I think the value would be low. There's no clipping in the colour scopes. Yep, my mate's not having an "issue", he just wants to make sure it's working as best as it can.

LG aren't clear about what's going on under the hood, but they may have a reason. There's no clipping in the colour scopes so it doesn't look like anything is clipped it may just be transformed from sRGB space using an LUT which is not necessarily a bad thing. The way that colour comes off most scanners involves a lot more work, one operator who runs an older model Scanity (he doesn't know exactly how old it is as he's just an employee) told me basically "The ScanStation is something else if that's really the way colour comes in the scan. We spend a lot of time on colour correction with the Scanity." In short, there's no evidence that anything the raw hardware is capable of is actually missing in the scan, and comparing against other scanners would be the wrong way to go about it even if you buy the same camera and lens and put it in a XENA as you would need to take out the LG light and program it to work exactly the same way otherwise it isn't like-for-like. The Blackmagic Cintel doesn't go straight to Rec709, but they do still provide their own import LUTs and something about the import for prints has been proven to be problematic (whereas it isn't for negatives). To get to Rec709 involves a secondary colour LUT that the user has to create themselves, so more work. That's the same for most Bayer scanners, they capture raw or if they don't they may mess up the quality (for example the Moviestuff software crushes the capture and you lose detail). What LG is doing isn't necessarily a bad thing if nothing is lost or if there's minimal colour detail lost. The scans are only as good as the person operating the scanner anyway. The ScanStation isn't exclusively for restoration work, it is a trusty reliable all-round workhorse. It can be used in different ways and it does what it's told. You can tell it to give the film its best effort, or you can tell it to make some quick proxies. If you're using it to inspect film it has a very nice built-in editing table and when a splice opens it reliably stops the scan and tells you to fix it. Some of the older scanners if you walked away from them to make a quick cup of coffee while they're doing their work at 7+ seconds per frame would un-spool the film onto the floor if you weren't there to babysit it when a splice opens. Many other scanners break down a lot, it doesn't. It'll scan on reels or cores it doesn't mind. It'll handle almost any film including film that's too badly warped for other scanners. Most of the time it goes straight to a deliverable format which is more than what most other scanners will do ("DPX-only"). There's a zillion different uses for them, you don't ever have to use it for restoration and you could still derive value out of it. As way of example, MemoryLab purchased their one pretty much just for home movie scanning. They had a choice to buy the Archivist, but when I asked last year why they chose the ScanStation they said it represented better value for them and they felt the increase in the cost was justified and they wanted maximum resolution for 8mm. We'll get to the bottom of the colour gamut eventually. In professional settings, and this comes from multiple users not just one, most clients want Prores which is compressing beyond DNG or DPX anyway. For many the standard is to offer Prores XQ or HQ as the standard choices, it's made so that it can make the compression for you with minimal detail loss.

Sigh, had Perry not blocked me he would have seen I corrected my error. Other thing I would note is that the time between generations is increasing. When it was only 2 years then 3 generations made more sense, but you can still buy LTO-6 drives brand new and the time between generations is increasing now (4 years LTO-8 to -9 and it wasn't double capacity) so it's not surprising given the media only lasts 20-30 years when stored per spec. And they most certainly won't be manufacturing the drives that can read them in 30 years. You can still buy LTO-6 brand new, I think that's it though don't take that as gospel as I didn't check every manufacturer. Also LTO-6 and below drives can be had for peanuts now, that said if LTO-9 makes the most financial sense for your use over LTO-8 then of course go for it. LTO-7 Type-M requires an LTO-8 drive.

Apologies I always seem to misquote specs. LTO 1-7 can read/write the last two generations and read 3 generations down (e.g. LTO-7 can read LTO-5 but not write). LTO-7 can not read Type-M. LTO-8 and LTO-9 can read/write the last two generations only and not read 3rd generation down. LTO-8 can read/write LTO-7 Type-M. I think adding read capacity 3 generations down now adds too much to the cost of the drive.

Film fades, but stored well it lasts a lot longer than these tapes do. Once digital media is unreadable it's useless. Note that LTO-9 doesn't support reading LTO-7, so unless you need LTO-9 probably LTO-8 represents better value as it will cover LTO8-6 including Type-M.

It's not used for production, it's used for long-term archival storage. Cheaper than hard drives (the drives themselves are expensive, but the media is cheap per terabyte), and you don't need to read them every year like hard drives. No one uses M-Disc for professional archiving it's not a proven thing, plus the capacity is only at consumer-level quality the reason to use LTO tapes is to back-up scans or restorations at 4K or even higher like 6K or 8K 16-bit DPX which is the master copy that makes everything else (the DCP, bluray, back to 35mm and whatever else). As Dennis says the best archival format is film not digital, but if you're investing in a serious restoration you'll want an archival backup of the final restoration and probably also of the raw scans.

Well I haven't seen the results off one, but 90 CRI is not particularly high and it wouldn't surprise me if the new cameras have visible sensor noise. The low-cost scanners in the $3K-12K range were originally designed for low-end work where the quality wasn't important. Moviestuff designed theirs for archives where the quality wasn't essential but they just needed to catalogue their holdings, and the Tobins were designed for the home movie to DVD market in the 00's (there's a little bit of information from the guy that designed them here). The Tobins cost $3600 new. The thing to take note about those 00's low-cost scanners is they had a particular use case in mind because the companies that were making the professional scanners weren't paying those markets any attention, which is clear if you ask how many professional scanners in the 00's could do 8mm at all? They were all focused on restoration (the scanning manufacturers that is) but some of the scanning companies were also looking at new markets like archives and government work. In 2010's you saw an entirely new class of scanners that were professional or at least semi-professional and suitable for those markets (and also for dailies), and which didn't cost upwards of $500,000. So the purpose that Moviestuff and Tobin were designed for, while those markets certainly still exist they now have scanning systems designed for them that are a much better fit for that purpose today. You can see in the 2007 thread that Clive Tobin said "It is a specialist 1CCD camera which is adequate for old home movies." Adequate for the market in 2007 is how the original designer describes it - it's now 2022. For non-commercial work such as hands-on hobbyists they can get value out of some of the cheaper scanners today like the Pictor Pro, or even a Moviestuff if they want to put in serious efforts to improve them. But it does take serious work, even just building a new light and fitting it in is not an exercise for a novice user. For commercial work you really want something more capable now because it affects your workflow. If your budget allows for a Pictor Pro and that's the best you can afford and it's for commercial work, then sure that may be the best entry choice and then you could invest in another scanner down the line. The downside is you wouldn't be able to do 16mm until you can buy another scanner, but if 16mm is not essential for the time being then you would definitely find it better than starting with a Retroscan for 8mm work.

Yeah it's like this: (That's an Archivist).

You can just look at the lights (when the scanner is OFF, obviously as the lights are as bright as the sun!) The sphere is completely enclosed in a Cintel or Arriscan but you can see it because the top of the light is just clear glass. Look at an angle into the enclosure and you'll see the sphere. I assume they would be enclosed in other scanners as well to keep dust off it. The HDS+ has this blob that they call a "lens", and I think it was based on the Cine2Digits diffusion cone (Frank's the one that made the light originally that Filmfabriek use).

Sure, but if we're talking about versatility if the scanner goes to a delivery format you can also do in-person scanning sessions for clients and they can take their scans with them right away. The only way to get it out is to rewash the film, but you want a lab that really knows what they're doing (generally they create their own systems out of film processors because there are no suitable machines designed for rewashing that are actually safe for archival handling of film). You can see the difference in a scan quite easily for embedded vs on-the-surface dirt as the embedded dirt won't be as solid black. What I mean is with diffusion alone. The Cintels currently have an integrating sphere although they will be moving to a new light that is more similar to the ScanStation cube starting with the next model by the sound of it on their website. For scratch concealment the integrating spheres are the best, but they scatter light 180 degrees whereas the integrating cube scatters it less and directs more light from the light source to the gate. In practical terms that means a shorter exposure for the same amount of light and therefore less motion-blur at faster speeds. You have to run the HDS+ at a slower speed to use it don't you? It's got a capstan right after the gate is that correct? That's the design flaw for wet-gate scanning, the capstan should be further away to allow time for your fluid to dry. If you look at the Pictor/Pictor Pro you can see the capstan is located before the gate and before the wetgate sponges. You'd also normally use air knives to assist drying (if you look at the table above Perc has a low evaporation rate whereas Isopropanol has a high evaporation rate). Even if you had a wet lab wetgate scanning with Perc (or the organic equivalent) requires serious engineering, that's another reason why the simpler solutions have their place.

It says they do on the website. The main benefit is in the workflow not the overall quality as they're based on the same technology (their own proprietary LED lights designed for Bayer scanning with Sony Preigus sensors). You're transcoding overnight, the LGs scan straight to Prores. You do have flexibility to do things that the LGs won't let you do like re-position the camera. Sure, the QC is important, and the LG helps you with that because it can do a second output (or a third, a fourth, a fifth) at the time of scanning. So you can make a 2K Prores HQ for QC and check that instead of having to check 5K or 6.5K files. Real wetgate is a wet-lab service and uses Perc, the only true wetgate chemical. You can get a similar effect with other chemicals, but Perc most closely matches the refractive index of film which is why other chemicals have more unwanted side-effects. Anyway there's nothing stopping you from making a simple wetgate solution for a LG using a safe chemical like Film-Guard or Isopropyl or Lumina scanning fluid and lots of people have done it. My friend who is a Filmfabriek rep would probably say that method can get you 95% of the full effect, but also so can just cleaning the film first and using a scanner with diffusion. Using chemicals that don't match the refractive index of film as closely though degrades the quality - so you lose fidelity such as contrast or dynamic range. The integrating spheres have the best diffusion for scratch concealment, but the light-cubes in the Archivist or ScanStation or the blob that Filmfabriek has get you perhaps 90% of the effect that the integrating sphere will give you. I've seen this first-hand, there are scratches that completely disappear on an Arriscan or even on a Blackmagic Cintel are not fully concealed on a ScanStation. Everyone I know that has ScanStations has their own wet-gate solutions. Yes it's nice that FF has a simple solution that's designed from the get-go so you can use it right away, but it's really not as complicated as you're making out. If someone can't make their own I would question their ability to use it for consistent professional work. My apologies Robert if it sounded insulting, what I meant was I'm aware of many companies that have 5K ScanStations (the full ones not SSPs) with no HDR that are more than likely running on the original version of the software they came with in 2015-2017, and are used for commercial work. They typically charge exactly the same rates that the company across the street with a current model charges because there's no difference in the amount of work involved in using them. The cost is based on labour not equipment. How they're used makes a huge difference - in the default settings the artificial sharpening is so aggressive that you can literally see the Bayer mask in scans! LG is not innocent in this, until 2020 the "full" ScanStation still came with the JAI camera and you had to pay a lot extra for the 6.5K Sony-Emergent. So there will be users that purchased them in 2019-2020 that didn't know any better in terms of the quality of the camera and may have thought that it's just an increase in resolution rather than significantly better camera for scanning. People have a hard time understanding dynamic range, they find it easier to understand resolution, that's why I usually talk about colour fidelity because it paints a more visual picture that people can understand. Thankfully they all come with proper cameras now and the JAI is no longer an option. They need better documentation, and Steve Klenk has even admitted this to one of my friends - he says it's a work in progress. The Arriscan XT has a big thick manual, so does the Blackmagic Cintel if you were to print off the PDF. Very simple things that an inexperienced user can get wrong are outlined there, and despite what others here say I have seen first-hand (relayed through friends that have shown me) or been told second-hand about numerous simple errors that simply would not be made if there was documentation you could show the scanning company. One of the most common is you're told there's no artificial sharpening when "filtering" is on 0.4 (the default setting). Most end-clients can't tell if something is off they are totally reliant on the company to do it right, but it would be a lot easier for the knowledgeable client if they could show their scanning company documentation that backs up what they say - and everyone benefits then because the company ups their game then all their other clients now get better scans. I'm reminded here of a time around 2014 that someone sent off his home movies to a company with an original 2K ScanStation and he said "wow these scans look worse than a restroscan" so he told them to do it again, and they did and then he was satisfied with the results. There may still have been further room for improvement, he would not have been able to tell but the moral of the story there is that company clearly knew they could do better work as their re-scan was a night-and-day difference (mind you they were very clearly priced for low-end work they were not charging what you'd expect in 2014 for professional quality, but that doesn't mean the client understands that because the client thinks he's shopped around and found a great deal). There's a lot of companies like that, just because the scanner is expensive doesn't mean they care about QC. By the way did they give (sell) you HDR for the Archivist?

Not really, that's a bit of a myth that I think is spread from the low-end home movie market where they justify using Tobins or Retroscans by saying "this is professional quality and without having to spend $250K on a Lasergraphics ScanStation". I'm friends with one of the reps for Filmfabriek so it's hardly in my own interests to be talking-up their largest competitor, but this myth that the price makes it unattainable for low-end work is just not true. I don't think LG designed them originally for professional work at all, if they did they did a poor job because all the default settings are for low-end work and not what you want for restoration. They have a bunch of cool features that lower the output quality, and the ones that increase the output quality to get you the best work are not intuitive or documented or explained in the training. For example, manually slowing down the scan if there's motion-blur: that's not intuitive to an average user, not covered in training to my knowledge, not documented, and it requires an operator who knows how film should look intuitively and will know if something looks off. There are numerous companies that take shortcuts with them because they think it makes no difference. Professional companies use them to do consistent quality work, but that's only one use-case and they would be in the minority of users. There are many different use cases where you'll make back the investment in about a year, it's not like the ye-olde slow DPX-only scanners that cost $1M and the investment would take a multi-year timeframe to payoff charging scanning rates that only the high-end clients can afford (which limits the market you can use it for). You don't necessarily need high-end clients. One of my friends has a client that does low-value films regularly on his ScanStation, and the client is very happy because they get colour correction that wasn't possible with the way they were having them scanned previously. That's not a high-end client. That's just a normal small company. Sure the upfront cost is a lot, but it's not as much as you're making it out to be in terms of a business asset for delivering a service. Compare the equipment costs in other industries - CNC routers for example. Also $170K isn't the entry cost, the entry cost is the cost of the Archivist which is around $45K or so and includes the host computer. You could have one of those for a year and then trade-up to a full ScanStation for 35mm. Or you just configure the ScanStation with less features - if all your work is 35mm then just buy the 35mm gate and don't buy the 16mm or 8mm gates until later. In short I do not think it's an exclusively high-end scanner at all, you just have to look at how the majority of them are used in the real world to see that. Yeah that's another option is to outsource that work.

That's the key difference between the Blackmagic Cintel, the Filmfabriek scanners, and a Lasergraphics. With a modern LG whether that's a ScanStation a Director or their cheaper models (ScanStation Personal/Archivist) you can scan straight to a deliverable format. From what I understand most scanning clients want Prores XQ or Prores HQ unless they're home movie clients in which case they often want MP4/Bluray (which you do need to transcode for full quality). So for MP4 it's a bit more work, but it's not as much as if you were starting from DNG/camera raw. The other thing that affects the time is the prep work (cleaning the film). A wet lab with perc-converted ultrasonic cleaners is the most efficient, but even with those you may need multiple passes on really dirty film to get it clean (other options include HFE-converted ultrasonics, other large non-ultrasonic cleaners like Kodak P-200 with Isopropanol or Naptha, or a Kelmar cleaner with Film-Guard or other choice of solvent). Many of the post-production houses only use safe chemicals which are generally less efficient and drives up the cost if they have to spend longer cleaning the film. Kodak had a list on their website, but they've removed the page. Here's a screenshot: What's helpful is the Kodak list clearly shows you the efficacy of the solvent against the cost. There's a lot left out of the list, like comparing the efficiency of an ultrasonic Lipsner-Smith against a non-ultrasonic cleaner with the same chemical. Also the Lipsner-Smith ultrasonics were never designed for use outside of labs so they're complicated to use, or to put it another way they're not idiot proof. You can buy refurbished Lipsner-Smith ultrasonics from MMT where they're converted to use a chemical other than Trike, and if they're in the hands of someone that knows what they're doing they're safe for film, but if someone doesn't know what they're doing they can easily get the tension wrong and damage film. There's also the complication that they're not portable. They weren't designed for scanning, they were designed for wet labs well before digital scanning was a thing. So maybe you're doing work for a client that won't let out their film and you have to take your scanning equipment to them, even if you have Perc-converted ultrasonics you'll need another option in that scenario and that will also affect the price you can quote if it drives up the time it takes to do the work.

That is completely not how it works. BMW shipping cars with software-locked features is shocking (as it should be) to everyone. But we are talking about vastly different quantities, and if you bother to look at the Arri pricing they don't charge extra for software features. Arri you're talking about a 2004 scanner with a user-base at least two orders of magnitude smaller than BMW. Are you seriously saying you'd expect to have the latest and greatest for free? As Perry says the scanning speed changes and for the ScanStation's at the hardware level it's fundamentally limited by the camera. To get the best scan though, you may have to slow down the scanner anyway and that's the so-called secret. All continuous-motion scanners can have motion blur and while 15fs may be fine for most film, it isn't the case for all film. I won't share exact details of this on open forum, but I do have examples of where companies with 5K model ScanStations (obsolete now) are charging rates that would make even Pro8mm blush. So please do not think the rates you see published by some companies is necessarily fair or normal. With anything to do with film, put on your critical-thinking cap and be as sceptical as anything.

Well green light gets you the magenta emulsion layer with some limited (ideally nil) cross-talk from the others, I've never heard of magenta/green being difficult to capture it's usually yellow/blue that's difficult. That looks like a photo from an optical printer?

They're using LED lights specifically designed for scanning film, not off-the-shelf products. You're describing the old lights used 20 years ago. Like this : Full-spectrum light with dichroic filters. If doing it this way was the best way to do it today that's how it would be done in the best modern scanners - but they don't because this is an old solution engineered prior to the LED technology available today. The solution was already engineered, the earliest Arriscan (launched in 2004) probably had a very similar light. It's also not all that complicated really, I cannot see this system significantly adding to the cost of production. I can think of one reason not to do it, and that's flicker. How do you eliminate the flicker inherent in Xenon lights? Another problem is brightness - Xenon lights lose their brightness and become dimmer over their lifespan, for projection that's not so important but for scanning you want a consistent brightness.

Robert, interesting though your vid is, it has noting to do with the question. Bayer scanner question not RGB.