Is Super 16 solvable in HD ?
by Herve LefelBefore a packed room, the meeting was opened by CST President Pierre-William Glenn, who reminded us that not all shooting was digital yet, and of his commitment to photochemical, although he stated that digital is established in post-production, and will be in projection in a nearby future.
He argued that in "the complex system that can be digital ’capture’, faster is not also cheaper" (capture : a fraudulent maneuver, this delightful neologism having nothing to do with shooting, can we conclude that digital has "captured" the legacy of photochemical ?)
The origin of the joint CST-FICAM working group is due to the rejection by some broadcasters of Deliverables from films shot in Super 16, as Laurent Hebert, CST general manager and moderator, reminded us.
First, some technical reminders.
To make an image in 16 by 9 high definition, you need :
- A resolution of 1920 x 1080 = 2 073 600 pixels (ie 5 times the SD signal of 720 x 576 = 400 000 pixels)
- An uncompressed format called "native" which is digitized in 4:4:4 RGB 10bit log, which yields 1024 shaded values in each color, which can be recorded on HDCAM SR tape or as data on a disk drive.
- In a 16 by 9 format, the photochemical surface of Super 16 is equivalent to 77.733 square mm (11.76 mm x 6.61 mm) and using a 4:4:4 RGB 10 bit scan at 2K resolution, it allows for a quantity of pixels equal to 2 157 960 (1960 width x 1101 height).
This is why, shooting in the superior quality of a native 4:4:4 RGB 10 bit uncompressed HD signal is explicitly recommended to safeguard the quality of the image resolution in recording, and for all subsequent manipulation of post until the final phase of the "master".
The tests were presented on different monitors (all LCD) :
- Two BT-LH 2550, control monitors for TV production
- A Samsung 32 inch "HD ready" (not full HD), for TV viewers
- A BVM-L230w, reference monitor for color timing
- A Mac 23 inch control monitor for production or editing
These monitors were balanced but not specifically calibrated. The first video screening was from an SR reference master. Then Marie-Pierre Moreuil (Kodak), Françoise Noyon-Kirch (CST) and Patrick Leplat (Panavision Alga Techno) summed up the process of shooting and postproduction for us :
- Shooting Equipment : Aaton Xtera camera, Zeiss T1.3, Master Prime and Ultra Prime lenses.
- Film : Kodak 7201 50 Daylight, Fuji 8663 250 Daylight, Kodak 7219 500 Tungsten.
The sequences were composed of static shots, wide, medium, tight, and pans, in day exterior, day interior and night interior and exterior, with lighting diagrams explained by Francoise Noyon-Kirch (in the absence of Guy Famechon, the cinematographer of the tests, Francoise pointed out that the lighting equipment used, although at a minimum, was that typical of that used on TV movie shoots).
After this screening, a first round table was organized. Jean-Yves Carabot pointed out that he had also refused Deliverables, but without specifying whether the native medium was S16 or HDCAM and (surprise !) that he had no statistics on this subject. Same story with the France Television channel, no statistics, but also a nagging obsession on the fact that the S16 might have more noise than HDCAM, although this also depends on the mode of transmission : ADSL (what speed ?), TNT (where do you live ?), cable, etc.
Patrick Leplat noted in this regard that, with the MPEG4 codec, the viewer can receive rates between 4 Mbps and 14 Mbps with an average of 8. In fact, the desire of broadcasters would be to keep a homogeneity to the image regardless of the initial medium. Could one call this a formatting of the image ?
Christian Bourguignon acknowledged that, paradoxically, despite the noise, since the cessation of SD broadcast, S16 had as many qualities as native HD.
Luc Beraud, meanwhile, stated his dismay with digital : dailies on computer, compression and decompression during editing, evoking the cost of HD, the superior maneuverability and ergonomics of S16 equipment... familiar and recurring arguments. Bertrand Mouly took up these elements, adding that S16 had increased in quality (granulation, image texture) and that the choice depended on the film you were making, mentioning in passing the superior latitude of photochemical.
To the extent that telecine is a key element in the chain of postproduction, Christopher Massie informed us that in fact, 60% of refusals of Deliverables were due to sound problems, linked to Dolby E. He emphasized the heterogeneity of shooting modes (HD, HDV, etc.., S16), the disparate means of production, and added that the number of codecs do not facilitate solutions.
Intermission
The second part of the morning consisted in a more detailed presentation of the tools with different flavors. Marie-Pierre Moreuil came back to the evolution of S16 : emulsion thickness, fine-grain and flat grains (they were previously cubic), a latitude of 13 stops while a Sony F35 only has 11, more flexibility in timing, a range from ISO 50 to 500 (the digital cameras range from 200 to 320) and, finally, a resolution of 2.2 million pixels.
The evolution of S16 cameras, the Aaton Xtera instead of the XTR, the arrival of the Arriflex 416 in replacement of the SR, the optical quality of fixed lenses (Zeiss and Cooke), have contributed to the revival of S16, stressed Patrick Leplat.
Francoise Noyon-Kirch came back, for her part, to how these tests were shot and exposed (- 1, 0, +1). Christian Lurin (Eclair Lab) gave the detail of the post workflow, namely a Spirit DataCine HD telecine at 4:2:2 8-bit (not 4:4:4 10 bit, which is pricier), with color timed dailies, which is generally the case with TV movies, and then editing, auto assembly, transfer to HDCAM SR, final timing on da Vinci 2K, correction of defects, that is to say DVNR (Digital Video Noise Reduction) --- is this systematic ? — And finally Dolby E and the Deliverable.
The second screening was a variation of the first edit with exposure variations of normal, normal + 1 stop, normal - 1 stop, then going on to two compression algorithms performed as follows :
Compression mode
- Software : ATEME KFE 2
- Codec : MPEG-4/AVC (H264)
- Resolution : 1440 x 1080, on screen 1920 x 1080
- CBR (Constant Bit Rate)
- Compression rates :
- 4 Mbps
- 14 MbpsMichel Danger came back to the noise and grain in S16 and admitted that he always went through a DVNR for the Deliverables, which makes us wonder about the margin of maneuver left to cinematographers and directors, and he attacked the poor quality of the Mpeg4 codec.
Bernard Cassan came back to the quality of the broadcast network (picture freezes don’t only happen in ADSL, but also on TNT... deep in the backwoods) and secondly on the quality of a good telecine with a good initial negative.
Karim Canam put things pragmatically, using specific examples, he pointed out that the choice between HD and S16 depended on the relationship with the broadcaster, and especially on the type of film made. It was easier for him to shoot a light fiction in some neighborhoods in S16 than in HD, thus agreeing with the position of directors and cinematographers. He reminded us that in production, we are always on the cutting edge, always asking the question of how far can we go.
Patrick Leplat emphasized the diversity of different processes, the importance of adapting the telecine digitizing to the downstream tools and diversity of postproduction workflows. He noted finally that there was a lack artifacts displayed in the tests screened.
Marie-Pierre Moreuil reminded us that you had to use the right film at the right time and that there was a wide range in sensitivity for all proposed shooting conditions : it is certain that using (for ease, or gain in time) a 500T in full sunlight has visible consequences (noise, image grain) on a 26 inch HD monitor with a compression of 4 Mbps
In conclusion, if some issues were not addressed (processing inconsistencies in the lab, for example), it is noteworthy that broadcasting conditions are an important parameter and that the recurrence of arguments on the post workflow is the same as with features.
But the attitude of the broadcasters is at best ambiguous : they refuse things in French Deliverables that they tolerate in American television series and MOWs (filmed in 35 or HD). Are they also willing to pay the price of quality (4:4:4 telecine, for example) ?
Lawrence Hebert reminded us that a film’s finances is composed of costs but also, for others, of income, and that expertise is related to experience, and therefore training.
Translated by Benjamin B