Episode 73 - ROB HARDY - Cinematographer (9 replies and 5 comments)
Team Deakins sits down with English cinematographer, Rob Hardy for a great, in-depth conversation. We cover many subjects, including his work on Boy A with John Crowley, the experience of the movie, Red Riding, with Michael Winterbottom, and his frequent collaboration with Alex Garland. He also talks about working with a Steadicam, operating the camera, mixing light sources and his choice of lenses per project. We even get into his experience on Mission Impossible. And we don’t forget to talk about the series DEVs. It’s a special inside look at how he works and thinks. You don’t want to miss it!
RECOMMENDED EPISODE VIEWING: Boy A, Devs, Ex Machina
Please post further discussion and comments below.
Enjoyed the podcast! Well, I've enjoyed all of them so far. 🙂
One note... early in this podcast Rob Hardy shares some advice he was given regarding the two ways to become a cinematographer.
I think it'd be cool if (Roger/James) could add a "best advice given" question or "most memorable advice given" etc. to your regular insightful repertoire of guest questioning.
I'd guess that these little nuggets of advice that you only get from working with someone who has more experience than you, are the kinds of gems that can make a huge difference in a career. And, they are often not the sort of knowledge/wisdom you could ever pick up from a class or a book.
To me one of the most interesting things Hardy said was that Ex Machina wasn't extensively previsualized and they'd do blocking on the day. It's a four-hander and one of the characters required CGI for every shot! I wonder how that worked out with VFX and budgeting.
Yes, but that character was always more of a 'real' character than a CG one. The CG was an adjustment to what was real and it would have required vastly more complex work to shoot the character as a blue screen element and comp her in afterwards. I'm not sure what the alternative might have been?
I have a question – in the interview, Rob talked about the 'texture' of tungsten light, as opposed to the texture of LED light sources. Also about how tungsten behaves on glass.
I'm intrigued, but don't understand what he meant by texture in this context?
I have an understanding of how relative size and direction of light source affect the look of the texture of what they're illuminating, but had never occurred to me an equivalent size LED placed in the same position would give a different texture to a tungsten bulb? Although I'm not sure that's what he was getting at...
My understanding is that 'texture' in this context refers to the spectral range of a tungsten source against that of an LED as seen in the waveform. I may be wrong, but that is my take on the difference between the two sources.
Thanks for replying.
The spectral range is to do with the CRI, right?
I'm surprised the difference is clearly visible, although I'm a stills photographer and have hardly used LED at all. Don't know many stills people who use it much. The norm is some tungsten, some daylight and a fair bit of flash (which of course has no relevance for motion).
I've heard a few people say LED still feels 'synthetic' compared to tungsten.
An LED seems to have less range of color, more like spikes of the primaries with less 'roll off', if that makes sense. It probably doesn't but that is what it feels like to me!
Have you ever tested gels or diffusion with LEDs, or does that defeat the point of them?
It depends on the colour you're trying to achieve. As Roger said above an LED has a spiky discontinuous spectrum. A camera has a spectral response curve (parts of the spectrum it can resolve).
Here is an incredibly crude example. In that case say you program your RGB, RGBWW etc LED to achieve a yellow colour. It analyses it and adds a certain amount of red and a certain amount of green. Now say our hypothetical LED emits a RED that is out of the 'range' of the spectral response curve of the camera however the Green it emits frequency is within the range of the camera. The camera will then perceive that Yellow light as Green. Now if you considered the LED source a white light emitter and gelled it to achieve said colour. It would with less fault. However that then requires gelling all the fixtures and doesn't allow for easy colour change.
However of course this isn't always the case. It depends on the certain colour, LED fixture and camera.
To also demonstrate this and show why an LED fixture could appear synthetic. Here is the spectral power distribution of a store bought RGB LED downlight designed for household use - https://imgur.com/KWaUT6d»
Here is a spectral response curve of a respected cinema camera - https://imgur.com/qcalRn1»
Compare the two, how much the camera can resolve in capture to the frequency bands of light emitted from the LED fixture. (if I remember correctly) A wider band in the red often lends itself nicely to skin tones. Under this fixture they may look fairly synthetic.
Thanks for that , good explanation to something i have been struggling to understand.
Will correct a few things and add a few things.
This is only for particular colours - that's why camera tests are so important when working with LED fixtures.
If you look at a spectral power distribution chart of a random household LED you'll see the pronounced spikes Roger talked about above. (From my understanding) The issue with this is that only a limited gamut of colours can be resolved under such a limited frequency bands of light. This could potentially give a synthetic feel. To counteract this LED manufacturers try and even out the spectrum (I believe for skin tones a lot of thought is given to the red bandwidth). However if you compare an LED spectral chart to that of tungsten. You'll see a massive difference.
Measuring LED's is another fun topic as most people talk about 'CRI'. That is measuring the spectral power distribution comparison to the spectral response curve of our photopic vision. The spectral response curve of a camera changes from manufacturer, sensor design and even down to different bodies (however differentiation between camera bodies is often minuet). As noted above the spectral response curve differentiates between sensor designs which in some cases differentiates heavily from our eyes.
There are better measurement tools such as TLCI, even one developed by the academy comparing the spectrum of an LED to that of tungsten light. It's just a question of how useful they are if you're trying to find a constant measuring tool for something with two altering variables. I do wonder is it best to choose the LED with the closest spectral power distribution to that of the response of the camera? I wouldn't know but it sounds logical to me. Some new LED fixtures are coming with a setting to choose which camera you are shooting on and they attempt to match it.
Great to have that so clearly explained. Thanks so much for taking the time and effort.