November 30th Announcement Feedback

[joshua csehak]

For capturing HDR motion, at low frame rates, why couldn't you just not reset after the read and say read three times in a row?

Hmm, could be tricky with any kind of movement, cause each frame would have elements offset by a little bit (depending on how much movement). But it might be that the software could mush them together to end up with a pleasing motion blur. You'd have to, actually, cause otherwise you'd be shooting with a really tight shutter angle, I think.

Thinking out loud here, if you had a guy running across the screen and a stationary camera, shooting at 48fps leaving the gate open the whole time (if that's possible), and each frame was getting alternately overexposed and underexposed by a stop... You'd probably want to wind up with two streams that you could combine in post into one 24fps file, and one stream would be offset 1/48th of a second later than the other. So the background would look like a regular HDR, but the guy might look weird.

The other way to do it, I suppose, would be some kind of debayering-type thing where the sensor was like a checkerboard, and the even-numbered pixels would be set to underexpose by a stop, and the odd-numbered ones to overexpose, and then you'd wind up with two streams that were exactly the same timewise, but at half the resolution of a normal recording. That'd be pretty cool, actually. If it's even possible.

[paulherrin]

I thought it would be cool to capture HDR motion, but there are some levels that technology hasn't reached yet. One is that I don't believe a true HDR monitor exists yet.

Scotty, I was specifically referring to the tone-mapping HDR processes, not necessarily 32-bit final output... So from a computer tech perspective they wouldn't be true HDR but we don't want to get into all the linguistics... Simply put, it would be nice to get a little bonus sometimes for scenes that do have a high dynamic range of light. We're not talking about those janky lookin "HDR's" that make one cry - we're talking about better replication of the eye's natural response to light, used in a delicate, effective, and intentional manner... i.e. something Graeme might just enjoy. (:

For example...

Shot near the Spanish Steps in Rome, Italy.

For capturing HDR motion, at low frame rates, why couldn't you just not reset after the read and say read three times in a row?

I've thought about this as well... however, as joshua said there will be motion issues in many cases. maybe not unresolvable, but issues just the same. and honestly i don't know enough about how the camera works but I'm not sure that you could adjust shutter speeds (or ND's or whatever would be used for alternate frames) in that amount of time. i just don't know.

AND even with the exact same shot is going to give you motion issues at different shutter speeds, obviously, when dealing with quick motion especially.

But something with ND's might work... Like you start out overexposed a bit, and it ND's down from there. I have nothing whatsoever to back up how it could actually, practically be accomplished... I'm just saying. haha

You could also use a mirroring rig (for 3d) with zero interaxial and do the same thing and ND one of the cameras.

However, I would reallllllllly like to try using simply the RAW data at different exposures.
Anyone like to supply some test footage?

[Chris Kenny]

Not necessarily. It'll be more baked but there is nothing to prevent RED from outputting an unprocessed RGB image. Any LUTs applied to the footage would be on top of a flat RGB anyway so you might as well encode it as metadata.

This would be ideal -- footage that's debayered but otherwise unprocessed. Basically raw RGB, like a Viper in FilmStream mode. You end up with footage that's a lot easier to work with than footage that requires a full debayer, but that still gives you most of the benefits of shooting raw.

[Nik Harper]

Yeah I imagine you plug one end of an ethernet cable into the brain and one end into your computer and the media mounts as an external drive. Pretty cool.

Heckyeah! That's so awesome!

[Chris Kenny]

What would be really slick is a "background" footage downloading feature, where the camera constantly and transparently backed up footage over ethernet whenever it could, automatically pausing whenever you had to roll a take or whenever you pulled the ethernet cable, and picking up where it left off when it could.

[Todd M.]

What would be really slick is a "background" footage downloading feature, where the camera constantly and transparently backed up footage over ethernet whenever it could, automatically pausing whenever you had to roll a take or whenever you pulled the ethernet cable, and picking up where it left off when it could.

[Steven Caesare]

Yeah I imagine you plug one end of an ethernet cable into the brain and one end into your computer and the media mounts as an external drive. Pretty cool.

I'm not so sure.

Ethernet is typically not a mass storage interconnect in this type of scenario, like USB mass-storage is. The exception is iSCSI, which is non-trivial and won't be plug-n-play for 99.9% of systems out there.

I suspect a more likely scenario is that the camera gets plugged in to your network (rather than directly into your computer), and you go from there.

The camera woyuld then get an IP address, and provide some method of interacting with it. This could be a small web interface, and/or hopefully a command interface (scriptable!). These would allow you to browse and transfer data from the camera media. (And also provide for remote control).

It's possible that the camera could also implment an NFS or CIFS stack and share the data out that way, so it could be mounted as a volume from your computer... that's the closest you'd get to having it show up as an external HDD, but I think that a slightly less likely scenario.... but would be cool if included.

All pure speculation, of course (other than the camera-control via commands over the ethernet link, which has been confirmed).

-sc

[Curran Giddens]

Scotty, I was specifically referring to the tone-mapping HDR processes.

The new Arri cameras do some interesting tricks to get High(er) Dynamic Range video:


Lots of secret sauce in the sensor, most of it we cannot talk about.

One thing we can talk about is the Dual Gain Architecture. We are taking the signal from each pixel, and splitting it into two different pathways. One with high, one with low gain. So out of 32 signal paths from the sensor to the A/D converters now we have made 64. Each goes into a 14 bit A/D converter, and then the high gain and the low gain 14 bit images are being re-combined in the camera electronics into a 16 bit image with meaningful brightness data in all 165 bits. HDR.

[Felix van Oost]

Just a question, are you guys at RED expecting to be able to cope with all the orders for the Scarlet 2/3" Fixed when it comes out, or will there end up being a 1 year waiting list like for the RED One? I know it's not easy to answer, but what are your plans for this?

[PaulClements]

What would be really slick is a "background" footage downloading feature, where the camera constantly and transparently backed up footage over ethernet whenever it could, automatically pausing whenever you had to roll a take or whenever you pulled the ethernet cable, and picking up where it left off when it could.

Would you really want another cable hanging off the camera? It'd be better if you could mount two harddrives and have it make two copies as it recorded, take one off and when you put it back on have the camera duplicate any missing shots back and checksum them both whilst the camera isn't rolling. Though frankly if the drives were reliable enough I don't see why you couldn't just walk away with two copies each day without doing any backups whatsoever.

Paul