This, the final part of Zacuto’s HDSLR versus 35mm film web series has a number of eye-openers.
If you haven’t seen it already don’t read any further!
I’m going to go into quite a bit of detail on what I think is the conclusion, for me anyway, of this shootout.
So, spoilers ahead…
[HR][/HR]Let’s go! First off – congratulations to Zacuto and everyone involved on a great series. These guys are pioneers in the HDSLR revolution, and we all have them to thank for spreading the word and getting knowledge out about what these cameras are capable of, and for getting so many talented shooters involved, from within the ASC or from main-street. I love their production standards, their kit and their personalities in general. Lovely people! Another stand out for me was the animation at the end – unfortunately being English and a bit stupid I cannot pronounce his name, or find it on Vimeo! So you’ll just have to watch the episode!!
(UPDATE: thank you to ‘xStep’ for sending me the details about Carlos Lascano’s work. The animation is on Vimeo, and I have added it to the top of this article. It’s some of the best stop motion I’ve ever seen, let alone from the EOS 7D. Also this animation starred in a past episode of Critics, I remember it well – [url]http://www.vimeo.com/877053 – Carlos is a star in the making)
Shane Hurbut’s short ‘The Last 3 Minutes‘ and the Eurovision music video are also worth a look, the former for creativity (one of my favourite HDSLR videos) and the latter for green screen work.
So onto the technical side.
Right now HDSLRs record in a modern codec, which is nether-the-less a bit rubbish. H.264 is a bit like what JPEG is to RAW in digital stills terms, but with an even bigger gap in dynamic range (latitude) and compression. This is because H.264 is a video codec with very sophisticated motion-related algorithms. Like predictive frames that aren’t quite real. Like small chunks of a frame estimated completely, to save space (macro blocking).
All kind of stuff is going on with these modern codecs and it results in a softer and more digital looking artefact-ridden image, with a lot of the raw information thrown away. This is fine for the consumer, or for the web where the finishing format is almost certainly going to be H.264 anyway – but for the filmmaker, the source recording format is essential – we like to keep as much image quality on our drives as possible and not sacrifice anything. This light our cameras captures is sacred. It bounces around this the entire Universe, this mathematical construct we live in and is caught by our lenses like a butterfly in a net, graciously allowing us to paint our emotions up on the silver-screen for others to feel. It’s amazing, and a privilege. So we don’t want an electronic box of tricks in between us and the lens if it is going to add it’s own digital signature.
We want to capture ‘reality’.
Ironically this is particularly important for green-screen work, like the kind Philip Bloom is helping to do on the set of Red Tails. Both Bloom and Rick McCallum were involved in the Zacuto shootout, so they know – and we do now – that the 7D is currently the best for green-screen work! It has a very high bit rate codec and a softer look, so it produces cleaner footage – at least at lower ISOs.
Now many of us already know that H.264 is a tricksy modern codec that has a certain digital ‘look’, but what the Zacuto test confirms is what the scale of improvement is going to be in image quality when we get raw video, and also from when the image processing side of things improves.
Currently things have a long way to go. The EOS cameras had moire and aliasing, in the form of coloured rainbows on fine detail in the blow up. The GH1 uses a slightly more sophisticated down-scaling method and a lower resolution sensor, so suffers less in this regard, but it has it’s own issues with the AVCHD implementation, causing motion artefacts.
The good news is I believe we are now close to HDSLRs which will output a clean uncompressed HD-SDI or live HDMI feed, so image processing (i.e, downscaling the full sensor to 2K or 4K HD) will be essential. I know the Panasonic AF100, based on the GH1’s sensor will do it HD-SDI so the HDSLR sensors are here already and in December so will the image processors. That’s when the cosh will really be put on Red’s more expensive offerings.
The only doubts I have concern two things – firstly the manufacturer’s will to put live HD-SDI out in consumer cameras, and then there are the power-demands, heat issues for such a small, compact body.
In such a small package (which for most shooters runs off small batteries out in the field) – power is a major thing to be considered.
Now HDSLRs, despite being used by many from the ASC / CSC, HDSLRS remain primarily consumer cameras – most consumers don’t shoot on film sets, so things like live uncompressed HD-SDI output out may not even be on the spec list for consumer HDSLRs even when the technology is right there. Strategically, the manufacturers may want to leave it only for the more expensive kit (like the AF100)
I hope this isn’t the case, but we’ll just have to wait and see on that one.
This episode contained a resolution test. Kodak and Fujifilm 35mm film processed to 4K and 2K was compared to the HDSLRs. Kodak has had an edge throughout over Fujifilm. All the HDSLRs in the resolution test didn’t really compete with film when blown up because it exaggerates all the codec nastiness.
Although in reality the difference in perception is much less for the average cinema-goer sat way back from the screen and not freeze-framing. This is because spatial resolution increases with motion, so resolution in film is quite unlike digital photography for example.
I don’t mind 2K for the time being – the main advances are with the codec, raw output and image processing hardware.
Because as we saw in this episode, when the scene switches to the H.264 footage, or in the case of the Nikon D3S – the older MJPEG codec – things go to slightly to hell in comparison! The magic of the modern codec starts to unravel a bit, and you see the workings of the technology – the macro blocking, and things like that. Blown up, the HDSLR footage was nowhere near as clean as film, although it had less grain.
The Nikon D3S does have one advantage due to it’s poor codec. It’s grain is quite organic. and the grain is kept by the codec. I’ve always found MJPEG mode on the GH1 for example, to have much more noise than AVCHD. The H.264 compression really does remove a lot of data (a lot of it grain) and looks baked in. This can be a good thing if you have a noisy sensor, but on the Nikon D3S this is such a shame, because the sensor delivers a really film like grain and low noise even at high ISOs.
It’s also interesting to see how much raw improves the latitude of HDSLRs. The sensor is capable of being a lot closer to film’s dynamic range than the codec allows it to be.
The only confusing part of the episode, for me, was the discussion on Apple’s editing format Pro Res. Is it good or not? Or should you edit H.264 natively? I am undecided if their conclusion was that it’s best to use the best quality Pro Res (like 4-4-2 HQ) setting possible, or just avoid processing the H.264 footage entirely / as little as possible? Unless I heard incorrectly there was a comment on footage shown ‘this has saved you going through Pro Res!’ and then a later one from Philip Bloom which seemed to contradict that. Interesting.
Encoding and transcoding footage from my GH1, involves two steps – not just the conversion from H.264 based AVCHD but also de-interlacing and pull-down due to the 60i wrapper. The native 24p firmware update does seem to remove one layer of necessary processing and improve image quality slightly – but I am yet to properly shoot with it in order to confirm this.
The 550D / T2i was in the test for the first time, and it looks great. As good as 7D, roughly equal. But the 7D and T2i were both noticeably the softest looking of all the HDSLRs, with a bit too much sharpening going on with the T2i especially. But I like both cameras as much as any other. They’re still fantastic!
The GH1 held up very well again, although some of the framing seemed a bit off with the shots. Its very difficult to make a 2x crop camera look similar to 35mm though.
None of this pixel peeping however takes away from the fact that HDSLRs are bringing a huge, sweeping, barrier demolishing change to filmmaking and allowing many more talents to get themselves noticed than before. Coupled with the video revolution on the internet (as broadband gets faster and faster) and we have the perfect recipe for a new golden-age in film.
Except this time it won’t be film that’s used!
Viva la revolution.