Halogenuros - in HD

[aj]

Hm, the Blip.tv footage looks OK to me. Amazing images for a technology reconnaisance. So much detail in these 4x5mm frames. Steady too.

[reflex]

....take it easy....

No, I think I'll speak my mind instead. I am rather passionate about this.

There is something wrong with the file - inexplicably duplicated frames. I initially watched it on blip, then downloaded the HD version and single stepped through the frames to confirm that there wasn't a technical issue on my machine. Others in this thread have reported similar results, so it's a replicable problem across multiple OS platforms.

Halogenuros has been touted as a film that demonstrated the capabilities of Super 8. But it is spoiled by short "artistic" shots, extensive digital enhancement (image stabilization, dust removal, color grading), CG , multi-layer compositing and digital slow motion. The image is inexplicably cropped from 4:3, as well.

[Daniel]

Hello Reflex,
(1)
The encoded file is made directly from a FCP 1920x1080 "uncompressed" 4:2:2 timelime, that's how I managed to get the short-feature out of the Quantel eQ station using the live LUT that is only active on the HD-SDI (or SDI output). Please take in consideration that from a bit rate that is higher that 1 Gb/s the compression (Compressor sub-software) did create a 1280x720 file at a bit rate of around 5 Mb/s................. This compression is not AVC Intraframe only scheme! , It's an interframe and intraframe compression.
Having said that, please consider that many shots were timestretched on the original timeline.

(2)
Halogenuros, got initial finantial ressources from FONDART (Chile) to develop a technical and artistical investigation called "Técnica de Intermediario Digital (2k) aplicada al formato de cine super-8", in english, this name would be aprox : Digital Intermediate technique (2k res.) applied to super-8 format.
The short-film explores the format capabilities through the possibilities that gives today digital intermediate. It was also stated on the original plan that the primary release would be 35mm prints with 1.85:1 image aspect ratio (normal aspect ratio for cinemas in Chile..... final aspect ratio was 1.86:1)...... and this was achieved in October 2006. So from my point of view it is normal to have cropped the image , and to have tryed to stabilize everything that could be stablized. The digital compositions are a maneer to demonstrate that the format (... i can only speak from this experience and workflow) have a lot of colour information, and quite good resolution, so that keying operations are possible. Of course 50D ("01") perfomed much better than 200T ("17").
Of course anyone is free to demonstrate other aspects or other charecteristics of the format.


André, thank you for your comments, the technical report should be available this week at http://www.filmhalogenuros.cl (but in spanish).


Alan, I am glad you found this a good work.
Concerning Spirit Datacine, official tech. information from Thomson explains that only 45% of the CCD photosites are being used when operating in 2k scannning mode. Because, Thomson did not made a specific super-8 gate (with proper optics), so an adapter is placed on top of the 16mm lens gate.....
So to be precise, this short film have 2k resolution on the vertical axis (1556 lines, cropped then to 1.86:1) but on the horizontal axis, only 45% of the 1920 photosites of the luminance are being used this means around 864 pixels interoplated to 1920 and then 2048, as Spirit Datacine1 is neither truly 2k on the horizontal axis. For R,G,B CCD arrays Phil. Loranchet (book. Le Cinéma Numérique, 2000, Editions Dujarric) explains that Spirit Datacine have larger colour photosites, 960 for each primary colour instead of 1920 smaller photosites (such as in the luminance CCD array).

So, of course that better resolution can be achieved from super-8, but at the time of this short-film was scanned (2004-2006), it appeared that Spirit Datacine to be a good option for the project. Beyond resolution, I feel that the fact that Spirit Datacine1 could output RGB 10 bit log files, was a good thing for colour grading. (of course considering that LUT was needed, as the footage was mantained as logarithmic through all the post from Spirit Datacine to Arri Laser....... for the 35mm theatrical version).

DHI

[Daniel]

when operating in 2k scannning mode

to precise : the Spirit Datacine, only uses 45% of the CCD arrays (4 arrays), on any mode : 2k scanning, hdtv or sdtv transfer......... due to the magnification factor of the 16mm LGA.

[mattias]

Other people over the internet have found it is ok for watching.

oh for crying out loud this is not a pissing contest, we're trying to help. if you don't remove it or fix it it's your loss. you will go down in history as the guy who managed to burn thousands of dollars on something completely worthless and came close to taking the entire super 8 format with you. congrats.

/matt

[mattias]

André, thank you for your comments

you hypocrite. why didn't you say that you only wanted unconditional praise and was gonna get extremely defensive and annoying if anyone didn't comply?

/matt

[Daniel]

Have a nice day mattias

[Uppsala BildTeknik]

Oh, well. So much for this high quality clip, to show super8 at its best.
Suffering from incorrect framerate, stuttering and all.

If you cannot see the micro-pauses with your own eyes, I guess it is difficult to correct the problem. You might reproduce it without noticing. :?

[Uppsala BildTeknik]

You know that dancing girl in the clip? Import the SD version to your NLE (HD is of no importance for the pulldown), jump to the dancing girl and step frame-by-frame. Sometimes you get replicat frames and sometimes several are new in a row. And sometimes her movements take a BIG jump, I guess at least one frame is skipped when this occurs.

If you are curious, you could check. If you don´t care, why bother?

[Daniel]

Import the SD version to your NLE

Hello Uppsala BildTeknik

But as I mentionned, the compression method for the internet delivery was not intraframe-only. That's why placing that H.264 footage on your timeline is likely to differs from the original timeline. This is a lossy compression scheme. This is not I-frames "group of pictures". There is B (bi-directionnal) and P (predicted) frames as well.


From Apple source :
http://www.apple.com/quicktime/technologies/h264/

The H.264 encoder features for example :

- Advanced frame reordering (B-frame) support to more efficiently represent movie data.
- Improved intraframe prediction for more efficient compression of details and gradients in high-motion video.
- Increased precision in motion estimation for crisp reconstruction of objects in motion.
- Flexible block sizes in motion estimation for more efficient encoding of complicated motion in areas of fine detail.





For me, the H.264 provided the possibility to reproduce well colour, mid-gray and contrast range.
Compression ratio is high, bit rate is low, and there is a lot of movement on the images, there is also visible grain structure on every part of each image , which is in definitive a complex task for the encoder, and so artefacts are possible, mainly because the compression method (to meet the internet delivery specifications compatible to iPod/Apple TV) is a lossy-one.

With a lossy compression, when decompressing, the data retrieved might be different from the original , "but close enough to be useful in some ways" is for example how a source like Wikipedia refers as of today, to lossy compression terminology.

Thank you anyway for your attention.

[Uppsala BildTeknik]

But as I mentionned, the compression method for the internet delivery was not intraframe-only. That's why placing that H.264 footage on your timeline is likely to differs from the original timeline. This is a lossy compression scheme. This is not I-frames "group of pictures". There is B (bi-directionnal) and P (predicted) frames as well.

Oh, but I know quite a lot about compression, H.264, GOP structures, I frames and all.
But the pulldown should not, and will not, be affected by long GOP compression. Somewhere, somehow, the playback speed was altered by someone or something.

No lossy compression scheme repeats complete images, not even H.264.
No lossy compression scheme skips frames, as this clip seems to do.

If you look at the clips I have on my webpage at http://www.uppsalabildteknik.com/english/?page=131 or at http://www.uppsalabildteknik.com/english/?page=134 (check Lisa Lindal), you will see that the playback is much smoother. These clips are all encoded with lossy GOP structures.

If you click on the arrow "to the right" in the music video webpage you will find a music video from "Moto Boy". This was shot with 18fps, I believe. Also encoded with lossy GOP structures.

[Daniel]

I am not a H.264 encoder specialist, thank you for this informative message .
From what I understand there are some significant differences between H.264 and H.264-intra (AVC-intra).......... This was encoded H.264. The skip/repeating of frames might have occured somewhere before.

Then if it is not the encoder (and I agree that there is something strange with those frozen frames) it might be before, in any of the steps done from the archive-in to the HD-SDI to HD-SDI routing between eQ and the ingest to FCP through Kona3 board.

After looking on Quicktime frame by frame, I recognize there is something strange, on this two internet versions, with some frames repeated, I didn't understood firstly what you were referring, as there is many repeated frames as time-stretch was used on part of the footage.

Anyway, the files will be maintained like these, for this internet release, the overall quality is ok. This short-film is ONLY one among many professional super-8 projects in many place in the world today and yesterday. This do not pretend to be the best quality possible for super-8,

Thank you, sorry for the misunderstanding and all the best,
DHI

[VideoFred]

For example on the first keying sequence, the actress footage was shot 18fps and slowed down to 24fps with replicate method.

OK Daniel, but this sequence should give you the following pattern:
ABCC DEFF GHII JKLL MNOO PQRR

Every 3th frame should be repeated once. Once only .
And correct me if I am wrong, but this is not slowing down play speed.
It is changing frame rate while preserving play speed.

You could have created slow motion by slowing down play speed to 12fps and then repeating every frame (once, not twice!) to have a final 24fps output.

Anyhow... Congratulations with the project!
But this frame rate thing is realy not good for the project.
I would change it if I where you.

Fred.

[Daniel]

Hi Fred,

thanks and yes you are correct that's exactly what was done on the keyiing footage of the actress , shot 18 fps, but preserved that frame rate, repeating one frame each three frames. (the first part of that footage before the keying takes place, when only the actress is there with the Rosco Digicomp blue fabric behind her, is different as it was shot 18 fps, but I made a "ramp", i mean not lineal time stretch, that's why some part of that footage may have more or less repeated frames (on the original eQ station).

concerning the frame rate, frankly, from my point of view, it is quite ok.... it is not "unwatchable", neither "jerkier than a 9fps cell phone", at least that's my opinion..... I can understand that if this file is seen on a PAL monitor at 25fps, adding one frame each second, or two field each second, will only make this issue more evident, however this internet delivery is intended to be seen at 24fps, and if seen at 24fps for example on a computer screen, the footage is quite smooth, some time to time the repeated frame issue is more noticeable, but nothing so bad. Thank you for your recommendation, but at this stage, I will not and can't make another version. This project has been running without fresh funds for many time, each on-line operation is slow and expensive, as the archives are on HD RGB and the timeline is 1920x1032 uncompressed RGB 10 bit log.

after analysing some files, i think that the issue could come from the latest archive, that was an archive done after the film-out DPX exportation. this project runned on extremely difficult conditions on the on-line platform in Chile, because the available storage on that eQ station was and is, not very big. May be 80% of the time, I have been working with video-free space of about 1% to 5% or so, this means that in those conditions, the platform can make a mistake on renders, exportation or may be archive process. Big part of the time on that eQ suite, was in fact archive in-archive out operation, but in theory the archive process is not a generation-loss, as the same DPX image is placed on external storage onto a "archive wrap".

In my understanding the 35mm version is not affected at all, as it comes from earlier stage , and nobody up to now have mentionned any problem like this on the 35mm prints that were the main result and the spirit of this project. Having said that it is a fact that there was a drop (line drop) on the opening fade of the last sequence, and a few corrupted frames on one vegetation shot, but very localized and not reallydisturbing, and definitely not comparable to the phenomen that Christoph firstly mentionned on those internet versions.

For budget reasons, the only new version will be done for the blu-ray disc, if blu ray versions are made, and will be done from earlier archive , not only for this repeated frames issues, but also because I would like to try to color grade for lineal display , from archives that comes before the final grading for 35mm.
This got more complex, only because, I was not able to mantain always the history on the archives. Actually I have 1TB of storage for this project for the original DPX scans and the archives. The best would have be to plan two workflows from the befinning 1/ film-out (maintaining log files and using LUT for grading, and then exporting to film-out without LUT enabled to avoid double contrast and double saturatio etc), 2/ all other lineal-display delivery as video-master, internet versions etc..... but the lack of budget in some steps, and other factors, led me to always give priority to the main workflow, the 35mm film-out to 2242 stock through Arri-Laser.

Let me know if you are watching this at 24fps, or converting it some way to PAL 25fps.

Good bye
Daniel

[Daniel]

Hi Fred,

thanks and yes you are correct that's exactly what was done on the keyiing footage of the actress , shot 18 fps, but preserved that frame rate, repeating one frame each three frames. (the first part of that footage before the keying takes place, when only the actress is there with the Rosco Digicomp blue fabric behind her, is different as it was shot 18 fps, but I made a "ramp", i mean not lineal time stretch, that's why some part of that footage may have more or less repeated frames (on the original eQ station).

concerning the frame rate, frankly, from my point of view, it is quite ok.... it is not "unwatchable", neither like "jerkier than a 9fps cell phone", at least that's my opinion..... I can understand that if this file is seen on a PAL monitor at 25fps, adding one frame each second, or two field each second, will only make this issue more evident, however this internet delivery is intended to be seen at 24fps, and if seen at 24fps for example on a computer screen, the footage is quite smooth, some time to time the repeated frame issue is more noticeable, but nothing so bad. Thank you for your recommendation, but at this stage, I will not and can't make another version. This project has been running without fresh funds for many time, each on-line operation is slow and expensive, as the archives are on HD RGB and the timeline is 1920x1032 uncompressed RGB 10 bit log.

after analysing some files, i think that the issue could come from the latest archive, that was an archive done after the film-out DPX exportation. this project runned on extremely difficult conditions on the on-line platform in Chile, because the available storage on that eQ station was and is, not very big. May be 80% of the time, I have been working with video-free space of about 1% to 5% or so, this means that in those conditions, the platform can make a mistake on renders, exportation or may be archive process. Big part of the time on that eQ suite, was in fact archive in-archive out operation, but in theory the archive process is not a generation-loss, as the same DPX image is placed on external storage onto a "archive wrap".

In my understanding the 35mm version is not affected at all, as it comes from earlier stage , and nobody up to now have mentionned any problem like this on the 35mm prints that were the main result and the spirit of this project. Having said that it is a fact that there was a drop (line drop) on the opening fade of the last sequence, and a few corrupted frames on one vegetation shot, but very localized and not reallydisturbing, and definitely not comparable to the phenomen that Christoph firstly mentionned on those internet versions.

For budget reasons, the only new version will be done for the blu-ray disc, if blu ray versions are made, and will be done from earlier archive , not only for this repeated frames issues, but also because I would like to try to color grade for lineal display , from archives that comes before the final grading for 35mm.
This got more complex, only because, I was not able to mantain always the history on the archives. Actually I have 1TB of storage for this project for the original DPX scans and the archives. The best would have be to plan two workflows from the befinning 1/ film-out (maintaining log files and using LUT for grading, and then exporting to film-out without LUT enabled to avoid double contrast and double saturatio etc), 2/ all other lineal-display delivery as video-master, internet versions etc..... but the lack of budget in some steps, and other factors, led me to always give priority to the main workflow, the 35mm film-out to 2242 stock through Arri-Laser.

Let me know if you are watching this at 24fps, or converting it some way to PAL 25fps.

Good bye
Daniel