Talking to John Galt is like talking to a living Wikipedia. Ask him a question and as you work through the answer a term comes up that requires more explanation, so we follow the link into that explanation, only to find another term appears, linking to yet another explanation… It can take a while to get back to the original question and usually you make it, but occasionally there is mutual recognition that we are lost – at one point he says, “I’ve done it again, haven’t I?”
He wears his glasses on a string round his neck, and repeats pronouns and conjunctions – “… and, and, and…” – which gives him the air of a professor. It probably comes from a need to regroup as he mentally simplifies the high jargon of sensor design into a language that ordinary mortals can understand.
Galt, who was in Auckland during Queen’s Birthday weekend with cinematographer Dean Semler (see the July issue for his profile), gets agitated if you try to call him the designer of the Panavision Genesis camera, which debuted in 2004 as a practical digital alternative to 35mm film. It has been used most recently in New Zealand on Underworld 3, and will be used on Wizard’s First Rule.
But his proper title of Senior Vice President of Something-Terribly-Important at Panavision in Los Angeles is way too cumbersome and, whether he likes it or not, he is widely regarded as the father of the camera system. Galt gets even more agitated as he talks about the hijacking of megapixels by marketing departments using what he calls bogus maths…

Megapixel madness
Perhaps ordinary mortals latch on to easily remembered numbers like megapixels. He says he opens presentations by asking how many grains of silver in a frame of film. Then he asks whether anyone has ever thought about this, and follows up by questioning why we care how many pixels there are in a camera sensor.
Nevertheless, just in case we do care, Galt is scathing about cameras that claim to be 4K. He says a 4K camera sensor would have 4096 photo-sites across the frame for each of the colours red, green and blue (apparently in a camera sensor pixels are called photo-sites). That is about 12,300 samples across the frame. If you multiply 12,300 by the number of vertical photo-sites in a super 35mm [1:178] frame, it works out at about 28 million photo-sites in all. Therefore, a camera that has less than that number of photo-sites is not actually shooting 4K.
He says this is because they mostly use a Beyer pattern sensor, which has two luminance photo-sites for every red photo-site and blue photo-site. The image is then processed – and pixels are interpolated (constructed) to constitute a 4K signal. These types of cameras are similar to metrology cameras, he explains. They are not designed for motion picture applications but for measurement. High-speed HD cameras have a similar spectral response.
“As I often say it’s because they use these things for motion analysis and they don’t really care that much about colour because crash-test dummies don’t bleed,” he jokes. “So the reality is that when people talk about a 4K camera it’s completely bogus. But people latch on to that. Genesis, which we consider to be a 2K camera, actually has 12.4 million photo-sites.
“If you go back to 1976, the researcher who invented the Beyer pattern was a Kodak researcher by the name of Bryce Beyer, and in his patent he basically describes a sensor that would have twice the number of luminance elements as colour elements. That’s right there in his patent. Because the human visual system is biased towards green, about 70% of what we see as luminance is in the green channel. So if you measure the response of these kinds of cameras, the green channel is the same as the human visual response. That’s how they work, and how they get very good resolution but really lousy colour.”

Galt says that as technology develops, the temporariness of the artefacts of our civilisation increases: “We have stone carved tablets that are 6000+ years old but 90% of every piece of motion picture film that has been shot since the beginning of motion picture has gone, 50% of colour film has gone.”

According to Galt, Genesis enables you to intercut with any colour film emulsion that exists but, when you try and expand the colour space of Beyer pattern cameras, the cinematographer is not going to get the kind of colour they are expecting to either intercut with or emulate film emulsions.
He lightly calls it a big dark dirty secret that the best 35mm cine lenses in the world are not 4K. This applies to all the manufacturers. He says they are optimised for contrast and sharpness, which is desirable for motion picture use, not for maximum resolution used for aerial photography or to find terrorists in the desert. Motion picture lens designs aim for very high contrast at relatively low spatial frequencies because the human eye has limited resolution – particularly in the dark gloom where we watch a movie.
He sees the lenses as a resolution bottleneck right now. “We are between 10% and 15% of the diffraction limit on high-end lenses,” he says. “The physics of optics has been around for several hundred years and we are reaching the limit of classical optics, so if you want to go beyond what we are doing just now, you’d better make the sensor a hell of a lot bigger, so it would be Vistavision or 65mm. That’s the only way.”

Making sense of sensors
In the arena of camera design, Galt says it has been very much a race to get CMOS (complementary metal-oxide semiconductor) sensors that are as good as CCDs (charge-coupled devices). He says industry is not quite there yet. CCDs have ruled because they have a very high fill factor (this is the percentage of the area of the sensor that is devoted to collecting light). However, CCDs must bucket the charge and put it through an external analogue-to-digital converter. The way Galt puts it, the beauty of CMOS is on-chip processing, which means the sensor can get the charge out quicker and ultimately power consumption is much less – about 20% of what a CCD would be.
Galt explains the limits to dynamic range in digital cameras by noting the signal-to-noise ratio of the latest CCD and CMOS imagers is about 12db better than the film emulsion, “but remember, that film emulsion has this wonderful characteristic of non-linearity. What happens is you have got a toe that aesthetically people like but really it is inertia, it takes a certain number of photons to make the silver developable. In film you run out of developable silver and that is where the shoulder comes from; as you expose more and more light there are fewer and fewer silver grains that are capable of being developed.”
CMOS and CCD devices are linear and, even though they have a greater signal-to-noise ratio than a film emulsion, the scene dynamic range is limited because the photo-site eventually saturates. “So if you look at a camera like Genesis, you have five stops above 18% grey, which is half-way between black and white, and you have five stops below it – a range of 10 stops. Film emulsion has much less than that on the straight line but probably another couple of stops on the top – that you can’t actually recover, but it will give you a graceful exit.”

Coming soon?
Galt says Panavision has just developed a CMOS sensor called Dynamax-35. It has the ability to output a film-like S curve over at least 16 stops – way beyond any film emulsion except those made back in the ’50s to shoot nuclear explosions.
When questioned about how long it will take to turn the new sensor into a production camera Galt becomes vague. He will admit that the Genesis took four years to develop and put in the field, but says they have learned a lot about that kind of process, hinting that he expects the process on a new camera to be quicker.
Panavision recently released a Solid State Recorder SSR-1 as an alternative to the Sony SRW1 video tape decks widely used on the Genesis camera.
“Everything I have done with Sony on electronic cinematography and then within Panavision, we have looked at film camera as the model,” explains Galt. “So the idea was to create something with a size and weight that would be equivalent to a 400ft magazine.
“The reason we chose flash memory was size, weight, and power consumption. A lot of people are using flash with massive amounts of compression, [but] I don’t think that is a good idea, not for motion picture applications,” he says. “We could have done mild compression and no one would have ever have known about it, and it would have saved on memory. The reason we didn’t do it was power consumption, because the compression engine would have cost us another 10 or 12 watts, we think. One of the bizarre things about programmed electronics is that you don’t actually know the power consumption until you finish the firmware.
“So we were aiming for 20 watts, and it ended up being 30. We knew a compression engine would have added 50% to our estimate, maybe more – who knows – so we made it uncompressed to keep the power consumption down.
“I personally have no problem with compression if it is the right compression algorithm to maybe even 5 to 1, but others in the production and post community feel otherwise.”

Storage strategies
Galt urges care about the LTO (linear tape-open) tape widely used for backups and archiving in the post community and large institutions. He says each version of the tape format from LTO-1 through to LTO-4 is faster with a higher capacity, but is ‘backward compatible’ to only two generations. This means everything recorded on LTO-1 is only good for playback on LTO-3 machines and will not playback on LTO-4 machines.
“This is the only strategy we have right now for backing up data. What it really means is that every five years you have to re-record everything you’ve got.”
He says that whether you have originated in film or digital, you should archive on silver-based photographic films, because that is still the single thing that we know will last the longest.
“Very few people do it because it is expensive, but we don’t have a truly archival technology that has replaced silver halide emulsion yet. The beauty of a piece of film is that in 50 or 100 years somebody will hold it up to the light and see an image. Now a piece of digital tape – I can’t imagine the technology that will be used but I am sure there will be technology that will allow them to reconstruct that, and they will think we were incredibly silly and naïve.
“All we can work on is what we know right now, and what we know right now is that if someone gave us a mystery piece of digital or analogue tape, the ability to recover the information out of that is next to impossible if you don’t know what it is.”
He says that as technology develops, the temporariness of the artefacts of our civilisation increases.
“We have stone carved tablets that are 6000+ years old but 90% of every piece of motion picture film that has been shot since the beginning of motion picture has gone, 50% of colour film has gone.
“One of my projects at Sony was restoration software development for fixing old movies, and one of my proudest achievements in the group was that we did the first digital restoration of a live action motion picture that was for all intent and purpose lost. So on one hand our digital technology tends to create incredible impermanence, but on the other hand it’s allowed us to restore things that would otherwise never be seen again – at least not close to their original pristine condition.”
Right now he says the Dynamax 35 sensor development is the most exciting thing he has been involved with in a long, long time but, from a technology standpoint, “we are just scratching the surface”.

© Copyright Onfilm magazine,
July 2008 www.onfilm.co.nz