I Bloom
03-15-2007, 08:15 PM
I've been giving some thought to the issue of not being able to overcrank the RED camera at 4K without an expensive raw output and storage. When this first became clear to me it was a big downer. I guess I was missing the bigger picture and the bright side which was that the camera is doing 4K wavelet compression in camera and that 27mb/s 4K footage is going to become a very viable and managable origination format just like DVCPro HD is now which is really exciting.
At the same time I'm the type of shooter who likes to overcrank things just a little bit all the time, I often film my MOS shots in the gray area between 24 and 36 fps wear the audience isn't quiet sure whether its slow motion or just a very graceful world. And so I've done some thinking about the red and 4K and here are some thoughts:
1. Renting extra equipment to overcrank isn't a solution that I relish. Neither is being tethered just for those shots. The solution to overcrank some shots is often a very subtle one, it happens sometimes in the spur of the moment and often (definitely) is not neccessarily something that is decided in preproduction. So it would be better instead to have an onboard way of doing this. So that the operator can just make it happen.
2. But switching to 2K windowed is an unattractive tradeoff, at least in theory, maybe we'll all see 2K windowed cut with 4K and all think its beautiful and just fine.
3. But some limitations on overcranked shots would be ok. Often slow motion shots don't last that long, its still useful to have the overcranking option even if these takes can only be say 30 seconds long.
So one solution, and this is something I'm sure the RED team has already thought about but I'd like to just open up some discussion, is some kind of RAM drive (the fast kind of RAM that looses all its data when you cut the power) that attaches to the camera instead of flash memory or an optical port and this drive is essentially a buffer, that stores as much RAW data as possible and then feeds it back to a hard drive at a slower rate (this is of course how the Phantom cameras work). This would be better still if it fead the RAW data back through the compression algorithm and then wrote it to the hard drive as Redcode compressed. Having this buffer attached to the camera could essentially make it like a RED-Highspeed body akin to a highspeed film body.
But there are a lot of technical issues with this, first off is capacity. Lets say it's reasonable that a highspeed module could have 8GB of RAM storage. Then by my math you could store about 25 seconds of RED raw uncompressed (323mb/s) before the buffer would need to be purged. That amount of capacity might be useful enough to create a large enough market to make developing and producing such a buffer worthwhile for RED, but maybe it might be too limiting for the relative cost of all that RAM and the added infrastructure to make it work.
Now we get into some speculation: What I know about software gives me some good guesses but without the hard data, (and we're not going to ever get the hard data, this is all proprietary stuff) but what if the camera could instead start the work of compressing the footage (within the limitations of the processor) say to 150mb/s and then complete the work once the buffer was full, then would a full minute of capacity make the buffer a viable thing. Perhaps its not possible to break the compression task into stages or its not possible for the camera to regurgitate footage in the way I've described without a complete redesign, who knows! (Graeme probably knows, but he's not saying)
In any case in the spirit of helping make this thing better I'm hoping to open this up for discussion just to see what people think and how important this is to them. Most likely the Red Two camera will run at 4 times the speed and do 4K compressed at 120fps and we'll never think of it again but until then this is really about being worried about limitations on creativity and little tradeoffs for quality.
Ian Bloom
http://www.ianbloom.com
At the same time I'm the type of shooter who likes to overcrank things just a little bit all the time, I often film my MOS shots in the gray area between 24 and 36 fps wear the audience isn't quiet sure whether its slow motion or just a very graceful world. And so I've done some thinking about the red and 4K and here are some thoughts:
1. Renting extra equipment to overcrank isn't a solution that I relish. Neither is being tethered just for those shots. The solution to overcrank some shots is often a very subtle one, it happens sometimes in the spur of the moment and often (definitely) is not neccessarily something that is decided in preproduction. So it would be better instead to have an onboard way of doing this. So that the operator can just make it happen.
2. But switching to 2K windowed is an unattractive tradeoff, at least in theory, maybe we'll all see 2K windowed cut with 4K and all think its beautiful and just fine.
3. But some limitations on overcranked shots would be ok. Often slow motion shots don't last that long, its still useful to have the overcranking option even if these takes can only be say 30 seconds long.
So one solution, and this is something I'm sure the RED team has already thought about but I'd like to just open up some discussion, is some kind of RAM drive (the fast kind of RAM that looses all its data when you cut the power) that attaches to the camera instead of flash memory or an optical port and this drive is essentially a buffer, that stores as much RAW data as possible and then feeds it back to a hard drive at a slower rate (this is of course how the Phantom cameras work). This would be better still if it fead the RAW data back through the compression algorithm and then wrote it to the hard drive as Redcode compressed. Having this buffer attached to the camera could essentially make it like a RED-Highspeed body akin to a highspeed film body.
But there are a lot of technical issues with this, first off is capacity. Lets say it's reasonable that a highspeed module could have 8GB of RAM storage. Then by my math you could store about 25 seconds of RED raw uncompressed (323mb/s) before the buffer would need to be purged. That amount of capacity might be useful enough to create a large enough market to make developing and producing such a buffer worthwhile for RED, but maybe it might be too limiting for the relative cost of all that RAM and the added infrastructure to make it work.
Now we get into some speculation: What I know about software gives me some good guesses but without the hard data, (and we're not going to ever get the hard data, this is all proprietary stuff) but what if the camera could instead start the work of compressing the footage (within the limitations of the processor) say to 150mb/s and then complete the work once the buffer was full, then would a full minute of capacity make the buffer a viable thing. Perhaps its not possible to break the compression task into stages or its not possible for the camera to regurgitate footage in the way I've described without a complete redesign, who knows! (Graeme probably knows, but he's not saying)
In any case in the spirit of helping make this thing better I'm hoping to open this up for discussion just to see what people think and how important this is to them. Most likely the Red Two camera will run at 4 times the speed and do 4K compressed at 120fps and we'll never think of it again but until then this is really about being worried about limitations on creativity and little tradeoffs for quality.
Ian Bloom
http://www.ianbloom.com