CRT film-outs

[Paul Bruening]

Made up separately and contact printed on. But if you know a clever trick with Dolby or DTS, I'm all ears.

[Dirk DeJonghe]

With the slow method you will also have the latent image fading to contend with, count about 1.5 printer points per 24 hours of aging, not necessarily the same in all layers. The end of your roll will be noticeably lighter than the beginning of your roll.

[Paul Bruening]

With the slow method you will also have the latent image fading to contend with, count about 1.5 printer points per 24 hours of aging, not necessarily the same in all layers. The end of your roll will be noticeably lighter than the beginning of your roll.

 

Would a long warm-up time, the capping shutter and a little delay between images solve that?

[Paul Bruening]

Doesn't the slow exposure approach solve most of the CRT problems as well as the LCD problems? I mean, except for the slightly wonky incoming voltage issues that I get from the county power coop.

 

We've been here isolating the very fine and precise technical details of this system. On the flip side, how much can go wrong that no one ever notices?

[Paul Bruening]

I just now checked these 21" TF CRTs. They have 4 way convergence adjustment in their on-board menus.

[Phil Rhodes]

Bear in mind that you will probably get somewhat better EHT regulation performance if you let them settle for half a second before the exposure; sudden demands can make things "bounce", but they often recover thereafter.

 

P

[Paul Bruening]

Has anyone seen or seen a price on CMO's 56" 3840 X 2160 plasma screen?

[Paul Bruening]

Fellas,

 

Interesting news. I just got the DG5 up and running. It's only cranking at 12 Hz. I've got some kind of wrinkle in the PNY NVidia FX3500 card to iron out. I should be running at 48 Hz. Here's the interesting part: I did a spot reading with my Spectra L-558 Cine meter against 18% gray. It read 6 fps at f1.4 on ASA 3. This thing is way brighter than my best CRTs. If I could time the shutter to hit the same refreshes every time at 6 fps continuous and use my Nikon 85MM lens at f1.4, I could grind out 3 1st gen, 4K prints per day off this thing. That's 90 per month.

 

I had to go back and do the math twice just to be sure. That's 3, 1st generation, shot on ASA 3 print stock, 2 hr. feature, 4K sourced prints per day! With load times the number probably comes out to 2 prints per day. 1 per day if I plan on sleeping. But, still.

 

How can I make that happen?

[Phil Rhodes]

Well, we've still got to pass the possibly not-inconsiderable "does it look any (*^@ing good" barrier, but that's mainly a matter of LUTs (NV drivers, photoshop, AVIsynth, Decklink, etc, although the trick is really in creating these LUTs rather than imposing them) and probably some trial and error. Shoot sequences of numbered frames, examine on a lightbox, etc.

 

The problem I see is still one of timing, although it may be less of a concern on a TFT. Software may be the easiest approach, if the black level on the TFT is good enough.

 

P

[Paul Bruening]

Well, we've still got to pass the possibly not-inconsiderable "does it look any (*^@ing good" barrier, but that's mainly a matter of LUTs (NV drivers, photoshop, AVIsynth, Decklink, etc, although the trick is really in creating these LUTs rather than imposing them) and probably some trial and error. Shoot sequences of numbered frames, examine on a lightbox, etc.

 

The problem I see is still one of timing, although it may be less of a concern on a TFT. Software may be the easiest approach, if the black level on the TFT is good enough.

 

P

 

I got some Fujichrome ASA50 and 400 Friday afternoon. I had to drive all the way up to Memphis to get it. Only one store in the whole city had any slide film in stock. At least there's still a place up there that runs E6. I'll shoot those tomorrow and process them on Monday. My 8 ft. wide screen ought to be big enough to analyze them. I also have a microscope so I can determine how much of the DG5's grid pattern makes it into the film's grain/dye matrix. I have to use a X10 loupe to see the grid. It's most visible on solid colors of the test pattern. It's otherwise invisible to my naked eye. I'm crossing my fingers on the results.

 

I'm also wondering if multiple generations of film printing are part of the way the source grid gets hidden through blending. It might be too much final resolution for a 1st gen print.

[Phil Rhodes]

I'd be tempted to hide the grid by defocusing slightly. It's not as if imaging the TFT matrix is actually getting you extra resolution. The "correct" solution would be a carefully-tuned optical low-pass filter.

 

Reversal is likely to be fairly hi-con, which may hide any tendency toward milkiness the display has due to its comparatively low dynamic range. Do Kodak still make the "deeper blacks" print stock?

[Paul Bruening]

I also got 4 rolls of negative stock. I'll shoot those for the milkiness thing.

[Paul Bruening]

I've got another bomb for your gourd, Phil. Given the precise registration of my Micky what are the chances I could shoot the stereo analog, Dolby Digital and DTS right beside each image. The analog track would require an extremely fine adjustment top and bottom in the gate to fine tune the gaps and double exposures out. I think that's doable as a matter of physics. The software might take some doing. But I imagine After Effects could combine all of the images, once created, through macros. Since I can only use 3200 pixels out of the available 3840 for scope (squeezed) images, there's a little space to spare on one side available for sound tracks. I could live with 1st gen prints only being 3K if I could do the sound at the same time. I'm wondering how much the cut-out of the gate might effect the film plane, though. Usually, sound is recorded to film while it is u-wound around a drum, ensuring it's distance to the sound heads.

[Paul Bruening]

I've done the measurements on picture area compromise when including soundtracks on the DG5:

 

When DD, DTS and AS, soundtrack takes 982 of 3840; picture takes 2858 of 3840.

 

When DTS and AS, soundtrack takes 703 of 3840; picture takes 3137 of 3840.

 

So, that's more or less a 3K image with every kind of sound that a Cinevator can do.

[Paul Bruening]

Well, Phil,

 

The numbers aren't going in the right direction. Measurements on the vertical frame come out to this:

 

A gate area in the sound track slightly large enough to allow overlap exposure (precise, slightly overlapping double exposure is fine since tracks are all opaque color or full transparency) will take 2400 of the DG5's vertical 2400. That leaves 1947 of 2400 for a standard scope picture frame. That makes the horizontal measurement only 2596 wide (using a 4:3 frame ratio. I'll have to redo the numbers for modern anamorphic projection). Still, a 2.5K wide 1st gen print with Dolby Digital, DTS and analog stereo sound track ain't bad for a DIY rig. I can't see Spielberg cutting me any slack on the DTS. Same goes for Dolby. All the same, analog stereo is free and I think the software to create the overlapping images could be scrounged.

 

 

EDIT: 2668 wide for modern anamorphic projection.

 

That means sound track makes the picture resolution 81.12% compared to a picture only (scope) resolution with sound tracks added through contact printing (recalculated modern anamorphic frame is 3289 X 2400 on DG5). I'm okay with the idea of an 81% picture resolution when gaining soundtracks.

[John Sprung]

I very much doubt that you could live with the inevitable small registration errors where the sound track crosses the frame lines. You could simulate such errors using the free Audacity sound editing program.

 

 

 

 

-- J.S.

[Paul Bruening]

Precision linear stages could make the micro adjustments. The gate can be precision ground for the overlapped parts of the track. I guess the issue is whether a Mitchell's registration is good enough. I know it is absolutely flawless on Frankenmitchell. I haven't tested my scanner's NC yet. It's a pellicle reflex. So, I wouldn't have to worry about any rackover slop. But, it does cut the possible frame rate down to 3 fps. Toothed belt feed from the motor would isolate vibration. I can put the whole rig on a piece of whopping big, scrap I-beam. What else other than temperature variations could pose a problem?

[Phil Rhodes]

what are the chances I could shoot the stereo analog, Dolby Digital and DTS right beside each image

 

I know about as much about film sound as I do about the culinary techniques of prehistoric India, but my immediate reaction is "pretty slim". I don't think the registration required for picture stability is anything even remotely like that required for glitch-free sound, and I think you'd get a pronounced 24Hz rattle in it as the sound head kept going over the joins.

 

Not only that, but the digital tracks are often encoded with non-public algorithms; I don't think you'll fin that Dolby or DTS are willing to give you the paperwork on how it's supposed to work. I know that SDDS, which is less of an issue now, uses or used ATRAC compression, as seen on Minidisc, which is certainly non-public. The on-print DTS timecode I suppose might be reverse-engineerable, but then how're you going to make the disc - it's a CD-ROM, but it's another proprietary lossy codec. Dolby Digital is (I believe) AC-3, and there are (not very good) open implementations of that, but I still don't think you'd end up with a readable track. The data is duplicated on both sides of the film at widely-spaced intervals but it's offset by a whole number of frames so your joins would be in the same place every time (even if you wanted - shudder - to rely on error correction to save you).

 

I think you may be reaching a bit far there.

 

P

[Paul Bruening]

I know about as much about film sound as I do about the culinary techniques of prehistoric India, but my immediate reaction is "pretty slim". I don't think the registration required for picture stability is anything even remotely like that required for glitch-free sound, and I think you'd get a pronounced 24Hz rattle in it as the sound head kept going over the joins.

 

Not only that, but the digital tracks are often encoded with non-public algorithms; I don't think you'll fin that Dolby or DTS are willing to give you the paperwork on how it's supposed to work. I know that SDDS, which is less of an issue now, uses or used ATRAC compression, as seen on Minidisc, which is certainly non-public. The on-print DTS timecode I suppose might be reverse-engineerable, but then how're you going to make the disc - it's a CD-ROM, but it's another proprietary lossy codec. Dolby Digital is (I believe) AC-3, and there are (not very good) open implementations of that, but I still don't think you'd end up with a readable track. The data is duplicated on both sides of the film at widely-spaced intervals but it's offset by a whole number of frames so your joins would be in the same place every time (even if you wanted - shudder - to rely on error correction to save you).

 

I think you may be reaching a bit far there.

 

P

 

First, I agree about the DTS and Dolby. Even If I could afford pirated software, I'd get busted and sued in short order. Concerning the analog stereo, my thinking is that instead of butt-joins I'd overlap the exposures. That's just a matter of a clockmaker cutting out the gate to allow the overlapping exposure. I think the track images can be precisely copied and pasted in software so that the ends overlap each other with the overlaps matched and then double exposed. It would have to be done with macros anyway since the doing the job by hand would be prohibitively lengthy. Since the track area is either opaque or transparent it can't really be overexposed anyway. The placement of the tracks on the DG5 will be pixel precise down to 1/2400. The issue is film registration. Can the Mitchell handle it? What percent error do you think the track could sustain given that the error will be buried in the overlaps.

 

I'd rather make the critical decisions now, than pursue something that will never work out anyway.

 

Magnatech International quoted me a price of $65,000 for their sound recorder unit. It might as well have been 65 million. I was hoping for something more like $3,000. I could turn 150 crack tricks to buy that.

 

...and I was hoping to get some nice, dinosaur curry recipes from you.

[Paul Bruening]

Dolby's the stuff between perfs. Of all the formats it would be the easiest to expose to film through this approach. I don't know what would be involved in creating the images, though. SDDS is the one that runs continuously down both edges. It is pretty much impossible to pull off this way. But, it's on the outs anyway.

[Paul Bruening]

The on-print DTS timecode I suppose might be reverse-engineerable, but then how're you going to make the disc - it's a CD-ROM, but it's another proprietary lossy codec. Dolby Digital is (I believe) AC-3, and there are (not very good) open implementations of that, but I still don't think you'd end up with a readable track.

 

Have you heard of any grabbable versions of the DTS codec that are close enough? The time code could be bought as one film copy, scanned and saved for standardized, future use. Then the CD or DVD parts conformed to that universal timecode sample. Legality is the thing to deal with.

 

How far off are the grabbable versions of AC-3? Can they be turned into the data stamps found on optical tracks?

[Phil Rhodes]

With a not-insignificant reverse-engineering effort, possibly, but I still view all of this sound stuff as impractical.

 

If a data block is the area between two sprocket holes, there are three per frame, but more accurately there are two and two-halves per frame, and the half-areas would suffer from overlap problems just as badly as the analogue track. The data is encoded as (obviously) a 2D barcode, almost certainly obfuscated against exactly this sort of reverse-engineering and involving nontrivial data compression techniques. My answer to the question "can the Mitchell handle it" is "I don't really know but almost certainly not".

 

You might get away with putting the DTS timecode on the print, but you'd then have to figure out how to make the disc to go with it.

 

Basically, $65k is probably quite cheap to solve this.

 

P

[Paul Bruening]

With a not-insignificant reverse-engineering effort, possibly, but I still view all of this sound stuff as impractical.

 

If a data block is the area between two sprocket holes, there are three per frame, but more accurately there are two and two-halves per frame, and the half-areas would suffer from overlap problems just as badly as the analogue track. The data is encoded as (obviously) a 2D barcode, almost certainly obfuscated against exactly this sort of reverse-engineering and involving nontrivial data compression techniques. My answer to the question "can the Mitchell handle it" is "I don't really know but almost certainly not".

 

You might get away with putting the DTS timecode on the print, but you'd then have to figure out how to make the disc to go with it.

 

Basically, $65k is probably quite cheap to solve this.

 

P

 

It would take a committed facility to make any money back on a $65,000 piece of equipment. The quote I got two years ago on a sound roll shot and processed was just under $500 per 2,000 ft. for analog stereo only. It's way more economical for me to go that way than to buy a sound recorder even if I had the money. I was just trying to beat the $500 ($2,000 for my current 4 roller feature) costs using the DG5.

 

I guess the only solution is to shoot a registration roll on the NC and measure the variation under microscope.

[Keith Walters]

I think you’d do well to steer clear of digital stereo for the time being. For ordinary program material it’s really hard for most people to pick the difference between analog and digital film sound, (although many people like to maintain the conceit that it makes an enormous difference to their self-styled “golden ears”.

 

The reality is that apart from DTS (which hardly anybody uses) digital sound tracks have only a very short operational life in the average multiplex projector, and most operators switch permanently to the analog track by the end of the second week or so. As the digital track starts to wear, the decoder more and more frequently switches back to the analog tracks as it encounters digital dropouts. This can produce jarring jumps in the sound field, and generally the more modest performance of the analog track is considered a lesser evil than the irritation of it switching between the analog and digital signals all the time.

 

My suggestion is to record the analog tracks in much the same manner as they are read during projection. That is, taking a separate loop of film “downstream” from the imaging section, where its movement can be stabilized via sprocketed rollers.

 

My scheme involves using a red laser module and an octagonal scanning mirror/motor assembly from a junked laser printer to generate a bright red line on a groundglass screen. This line would then be focussed onto the sound area of the film via a short focal length lens, possibly one salvaged from a junked compact digital camera.

 

As the film moves past the lens, suitably masked, the red line would produce a continuous cyan-coloured strip in the sound track area.

Control circuitry for the laser would produce two small gaps in the red line which would correspond to the left and right audio tracks. The left and right audio signal would modulate the widths of the two gaps in the red line, producing the actual stereo audio signal on the film. Most of this could be most practically achieved with simple analog circuitry

 

The practicality of this depends on the speed at which you could run the film. The refresh rate of the red line can be regarded as a carrier frequency, and so for realtime DAT-quality sound sampled at 48kHz, the octagonal mirror would have to rotate at 6,000 times per second, so as to give 48,000 sweeps per second.

However if the film ran through at 1 fps (ie 1/24th speed) the mirror would only have to run at 250 revolutions per second or 15,000RPM, somewhat more manageable. The audio would need to be slowed to 1/24th as well, but this is a trivial task these days. Faster frame rates should certainly be possible.

 

 

I don’t know whether it is possible to get the “formula” for Dolby-A encoding with surround enhancements, but in my experience, Dolby analog decoders produce quite convincing results even with "vintage" stereo music tracks from an MP3 Player, which is how I used to test them, when I couldn’t (or couldn’t be arsed to) lace up a projector!

 

Dolby analog surround sound encoding is a pretty simple affair, it’s the decoders that do all the work. The upshot of all this is that most people are going to be happy as long as there is something that sort of sounds like surround sound, if they care at all. If they want 13-channel SDDS, well, that’s probably not your market…

[Keith Walters]

The data is duplicated on both sides of the film at widely-spaced intervals but it's offset by a whole number of frames so your joins would be in the same place every time (even if you wanted - shudder - to rely on error correction to save you).

The Dolby decoders actually use a sort of modified digital camera to "photograph" each block of data and then auto-align it using the "double-D" symbol and other non-changing framing blocks as a reference, hence the system is pretty tolerant of registration errors.

However actually getting the blocks of data onto a printing negative is a lengthy procedure requiring specialized precision equipment.

I doubt there is any way a "garage" operation could ever achieve this.