Friday, July 04, 2008

The HD Masters, part 3

D-Cinema

Image acquisition as well as the production/post process for all-digital cinema was discussed at some length. It seems to be coming increasingly clear that single chip (i.e. Bayer-filtered) cameras are not regarded as suitable for high-end imaging. In fact Sky has said that for this reason the Red camera is not acceptable for programme makers. As an aside I recently saw some Red pictures at a facility that is cutting a drama for the Beeb and they were having great trouble getting rid of the noise (and hence v.objectionable compression artifacts) in the blue channel - you'd not suffer that with a three-chip camera.
With this in mind it seems that the Sony F23 and Thompson Viper will carry the bulk of D-Cinema work for the high-end.

Schubin's Metropolitan Opera

Although the man himself didn't talk about it several speakers made reference to how the New York Metropolitan Opera has been transmitting it's live productions to European cinemas for the last couple of years. There was much talk of liberating theatre owners to do much more innovative programming with HD football coverage, music shows and the like.

3D imagery

Although I've always found 3D imagery a bit intense and (for me) it detracts from a good film (rather than adding to the enjoyment) there has been a lot of development work going on recently - the Dolby systems seems to be the one that will be widely adopted. The Hanna Montana concert movie was shot for $9m and took $65m in first week in only 800 theatres! There are more than twenty 3D films in production/post at the moment but they mostly seem confined to the pre-teen market.

NHK's SpaceCam!

An "Earth-rise," or the rising Earth over the Moon, was first captured by the Apollo project. The Earth rising image taken by the KAGUYA on November 7, 2007, was not a full Earth-rise (i.e. not all of the globe was seen in shining blue.). This time, a "full Earth-rise" was taken by the onboard HDTV cameras in faraway space, some 380,000 km away from the Earth. It was also very precious because it was one of only two chances in a year for the KAGUYA to capture a Full Earth-Rise when the orbits of the Moon, the Earth, the Sun and the KAGUYA are all lined up.
The shooting was performed by the KAGUYA's onboard HDTV for space use, which was developed by NHK. The movie data was received at JAXA, then processed by NHK.

One of NHK's project engineers walked us through the modifications they had to do to a Sony 950 to make it work in space - this for me was probably the most interesting session of the conference. They then showed us the final movie and it was fantastic.

There were several other interesting sessions including how the BBC Bristol's Natural History Unit has moved over entirely to electronic acquisition. They had some stunning footage from the next BBC animal show 'Life' - the sequence of dolphins herding fish off Madagascar was amazing. Get the BBC HD channel (or download from your favorite torrent site!). Over all I think 2008 is the year HD breaks through - with Sky and Virgin offering services and FreeviewHD and FreeSat just around the corner things look good. However - getting back to what Mark Schubin said about what is technically appropriate I heard an interesting comment;
A well authored SD DVD played on a decent up-scaling DVD player gets 'dangerously close' to the quality of 12Mbit DVB H.264

Labels:

Wednesday, July 02, 2008

The HD Masters, part 2

"I want my 1080P"

I'm glad that the Beeb and Sky got behind 1080i as a start with 1080p as the eventual intention. Here are a few of the observations the speakers made on that subject;

At high compression rates (especially sub 8-Mbits MPEG4) progressive pictures look better than interlaced - A lot of the broadcasters who got behind 720P bang on about this but it seems a poor reason to settle on a technical spec that is 'better' because it avoids the current limitations of the technology better than a visually better option.

For production purposes 200Mbits should be considered the minimum. JPEG2000 is very good at these data rates and subsequent delivery at H.264/AVC-MPEG4 (even at 8Mbits) shows no artifacts.

Sky are assuming that as soon as 880MBit HDCamSR machines (Sony HDW-5800) become commonplace they will specify 1080P as their delivery spec.

Part of the 1080P spec included 24P! EBU tech document 3321 covers this.

BluRay has of course been 1080P from the outset - the chap from the 'packaged media' (the new name of DVDs?!) industry body told us that they assume they have a five-year head start on the broadcaster and they intend to make BluRay a premium format using this as their main marketing point.
Currently BluRay (both disks & machines) are outselling DVD (for a comparable point in the format's life, say 1997) by six to one.

Coding, Mux & delivery;

In the past decade the venerable MPEG2 transport stream has done good work - but a combination of newer codecs - some variant of MPEG4 - probably the AVC or H.264 codecs (which subjectively halves the required bitrate at both SD & HD) and better multiplexing (moving the QAM64 and DVB-T2 transmission) means that there could be a lot more space for channels. However - with the analogue switch-off nearly on us and the whole country buying MPEG2 decoders & TVs another change (without the associated benefit of moving to HD) seems unlikely.

The representative of OFCOM did a good presentation of how Mux-B will be liberated when the analogue switch-off is complete. This will be a dedicated HD Mux and since it will start with DVB-T2 power, QAM64 channel-packing and AVC we'll be able to pack four 9Mbit HD channels in. One will go to the Beeb and the others will probably be awarded to ITV-HD, Channel 4 and five-HD.
As the other Muxes move over the DVB-T2/QAM64 (but remaining at MPEG-2 TS for compatibility) - remember OnDgital was QAM64 and it wasn't until the Beeb re-launched as FreeView that they switched to QAM16 for more reliable coverage. The T2 spec allows for more transmission power and any current Yagi aerial will deal with that.

Currently none of the manufacturers offer a StatMux for MPEG-4 - BBC R&D have one they've built themselves and Sky reckon they'll be testing some prototypes before the end of the year. Since Mux-B becomes available Q4 of 2009 it seems there is a bit of time.

HD Production observations;

There were several broadcasters in attendance - here are a few of the things they said;
The EBU maintain that for the current crop of manufacturer production codecs at 1080i nobody should go below 185Mbits - that's the highest data rate that Avid DNxHD offers and you can drive Avid ProRes a lot lower! Where does that leave DVCProHD?!
France-2 had a very interesting set of footage from the Roland Gaross tennis last year. They had the same court shot at 16x9 HD and 16x9 HD shot-to-protect 4x3. The latter was so much busier as the cameraman tried to keep the action the centre of frame. It was all a lot tighter focused and I realised how much more like the world 16x9 is. Of course when you come to transmit stuff shot-to-protect 4x3 the codecs fall apart!

Labels:

Tuesday, July 01, 2008

The HD Masters, part 1

I went to this conference last week and it was superb - lots of interesting speakers and demonstrations. I made copious notes and will blog them over the next few days - click the HD Masters 2008 tag for all of them together

Opening notes;

In 1988 I saw my first demonstration of HD at Television Centre - organised by Metro Video (remember them) they had recorded nature and football footage using the Sony 1250 line 1" analogue VTR (30 Mhz RGB recording). At the time I'd never seen such clear electronic pictures - I can remember telling one of my fellow trainee engineers that it "....was like looking out of a window".
After that my next brush with HD was in 1994 when the company I was working at did the OB for a job at the Royal Opera House for NHK - it was a five camera shoot with five ISO Sony HDVRS-1000 machines - again, a 1" format with decks that looked like BVH-3000 machines. There were several engineers from Sony to cosset the machine and again, the pictures were splendid! The Cameras (Thompson, IIRC) were tube models (which at the time we were getting used to CCD cameras in our SD studios - mostly BVP-7s) and they had an 'old master' quality to them - very rich blacks.
Anyhow - after that, aside from a few graphic/film-type jobs I was involved with I didn't do much HD work until HD was firmly established around six or seven years ago. SInce then it has, clearly, been on the rise with new formats (SR, HDV, Red Camera etc etc.) and what with Sky/Virgin et al offering paid for services for the last year or so and the Beeb's successful DVB-T trials from summer 2006 (which I captured and watched avidly) it seems like 2008 is the year HD goes mainstream.

Mark Shubin;

If you've not listened to The Schubin Report and you're in television engineering then shame on you! Mark is an old-school engineer and gentleman who has a real handle on not only the technology but the importance of where it all fits into the workflow and how people are in and around TV. He gave the opening presentation and talked a lot about how we have to let go of old assumptions about human vision - the Lechner distance (what most eye tests have been based on for a century) doesn't quite tell the whole story and for most living room SD would be adequate on sub-50" displays if our vision was only as good as 1 arc-minute. Tests have revealed this not to be the case with the perception of reality being greater at 1080 than 720 than 576 lines. Mark is a great proponent of what is technically possible against what is appropriate and he use the best expression I heard all conference; "...if you can tell the difference then you're lying!" in relation to colourimetry.
He also had some great examples of how people are 'trained not to tell the difference' - the best example was the opera singer who was used as the voice in the thirties to demonstrate how good wax cylinder recordings were confessed that she'd trained herself to sound like the recordings of her during the 'blind' tests!

Labels:


 
Phil's technical blog