why does digital share the limitation as film in this regard...

[Bryan Colvin]

I've been thinking recently, since these cameras are basically emulating a shutter, there is no physical shutter correct? (i don't own one so if i'm wrong please correct me)

we should be able to achieve physically impossible (because its digital not physical) shutter angles/speeds. I guess with the current firmware this may not be possible, but i wonder why we don't make this possible.

let me give an example to make it more clear.

if you wanted crazy motion blur on a shot, and you had your shutter set to (using shutter speed, not angle for this example) 1/4th of a second. shooting at 24fps. I know this is impossible on film, BUT this is digital... why can't we have a way to make this possible. In my mind I can see it working very similar to a guitar delay pedal, where there is a feedback loop. I'm sure it wouldn't be easy, but it seems completely feasible if someone thought about how to do it.

if you want to shoot extreme slow motion at 300fps, you need a SHIT LOAD of light now to make it happen! I know this is just how photography/cinematography works, but we're talking about a DIGITAL camera, run by software, not by physicality. In The Matrix, (shot film obviously), bill pope talks about how much light was required to pull off the slow motion shots. its crazy! what if this sensor chip feedback loop type of solution was figured out letting the light to affect multiple frames over time. We could have LOW LIGHT slow motion.

I just feel like ppl are so much wanting to emulate film that they're not thinking out of the box on this, and realizing it could totally be possible for this to happen.

What do you guys think?

 

[Eric Cameron]

Theoretically I suppose you could reset the sensor at a rate slower than your frame length, but there would be a lot of issues. The sensor would be resetting at different points in different frames, unless you found a matching cadence for the reset and the framerate. Even with a matching cadence, you would still have the sensor refreshing in the MIDDLE of capturing a frame, which would no doubt lead to super crazy artifacts... honestly this just doesn't seem to be in the realm of reason. You can do something similar to what you are suggesting in post by adding luminance between neighboring frames in a weighted manner, which would in all likelihood lead to fewer (but still there) artifacts. Either way it wouldn't be so much good for low light as it would just be a novel effect.

 

[Bryan Colvin]

i guess what im bringing up would require additional hardware in the camera. im not asking for a hack to make it work, no body wants artifacts! I don't think its just a novel effect, i think it would be an advantage of digital film making.

I guess what I'm saying isn't "low light slow motion" necessarily. Just if you are shooting super slow motion you wouldn't need ridiculous amounts of light.

 

[Onar Stangeland]

What you are describing her was possible with the Canon XL-1 back in the late nineties...

 

[Larry Kelly]

yes slow shutter as it's called in Sony land. On my ex1 I can shoot with slower than 180deg shutter, lots of smear but for static shots you can shoot at night without the same level of noise.

 

[David Rasberry]

I don't know how you would cram more than 1/24th seconds of exposure (360 degree shutter ) into a frame that is recorded in 1/24th of a second. Physics only bends so far.

 

[Christopher Mills]

How about a step printing style post process.. shoot at 6fps, "print" each frame 4 times.

 

[David Mullen ASC]

If you are taking 24 pictures per second, it's impossible for each of them to be exposed for more than 1/24th of a second unless you build a time machine. Some cameras allow super slow shutter speeds but basically do this by reducing the capture rate to allow longer exposure times.

 

[Nick Pasquariello]

It's not impossible. You just have to reset how you think about the pacing of individual exposures.

First, it helps to think about light as what it is: discrete amounts of photons. Like little balls that are bombarding the sensor at a constant rate (assuming nothing in frame is moving). If the sensor is "always on", always taking in light, then over the course of a second, for a 24fps shot at 180 degree shutter, the sensor is taking in 48 times the amount of photons that it needs to get in order to record a single frame.

If you wanted to draaaaaag the motion, couldn't you compile the frames? So, first set it to 360 degree shutter. Then record 24 frames. Only, the first frame is actually a summation (or average) of frames 1-6. The second frame is a summation of 2-7. The third is 3-8.

So the camera would sort of have to do this in a two stage process. For each 1/24th of a second, it would have to copy the frame into RAM, and then use that frame in summation or averaging involved in the next 5 1/24ths of a second as well. This is all assuming you want a 1/4 second drag. If you want a 1 second drag, then you're keeping 24 frames in RAM and summing or averaging 24 frames together.

Currently not possible in the firmware, but not technologically impossible. You can do it in post, at any rate.

 

[David Mullen ASC]

It's not impossible. You just have to reset how you think about the pacing of individual exposures.

First, it helps to think about light as what it is: discrete amounts of photons. Like little balls that are bombarding the sensor at a constant rate (assuming nothing in frame is moving). If the sensor is "always on", always taking in light, then over the course of a second, for a 24fps shot at 180 degree shutter, the sensor is taking in 48 times the amount of photons that it needs to get in order to record a single frame.

If you wanted to draaaaaag the motion, couldn't you compile the frames? So, first set it to 360 degree shutter. Then record 24 frames. Only, the first frame is actually a summation (or average) of frames 1-6. The second frame is a summation of 2-7. The third is 3-8.

So the camera would sort of have to do this in a two stage process. For each 1/24th of a second, it would have to copy the frame into RAM, and then use that frame in summation or averaging involved in the next 5 1/24ths of a second as well. This is all assuming you want a 1/4 second drag. If you want a 1 second drag, then you're keeping 24 frames in RAM and summing or averaging 24 frames together.

Currently not possible in the firmware, but not technologically impossible. You can do it in post, at any rate.

But you aren't actually taking a series of single pictures with a longer exposure time than the frame rate allows, you are combining frames and their individual exposure times, essentially building an effect (which could be done in post to film footage, it's not limited to digital.)

 

[Nick Pasquariello]

But you aren't actually taking a series of single pictures with a longer exposure time than the frame rate allows, you are combining frames and their individual exposure times, essentially building an effect (which could be done in post to film footage, it's not limited to digital.)

Yes, but many of the things we do on set can be duplicated or expanded upon in post. Diffusion filters are a prime example. Or even the frame-summing and frame-averaging that are now available with the Timelapse functionality in Build 4 for Epic/Scarlet: those functions could be done in post, but instead are in-camera because it's simpler (and more helpful to see closer to what you are going to capture when you're on location rather than at your computer a week later).

Also, the net effect of what I lay out is 'taking a series of single pictures with a longer exposure time than the frame rate allows', which is something that wouldn't be possible with film, but is possible with digital due to the ability to "hold on" to a reading of photons in your RAM for a duration of time that is longer than a single frame.

I'm not saying it *should* be done, nor is it a thing that a Red camera can currently do. I'm just saying, it's actually a thing that is possible from an engineering standpoint.

 

[Onar Stangeland]

How about a step printing style post process.. shoot at 6fps, "print" each frame 4 times.

If I'm not wrong, this is exatly what happens in the Canon XL and Sony EX cams!

 

[Art Chong]

If we're talking about getting "longer" than 1/24th of a second shutter into 24fps, I have an idea:


What if you use a mirror rig, but with no stereo offset or convergence angle? Meaning two cameras looking straight forward, on the exact same line.

Set each camera to 12 fps, 360 degree shutter, and offset each camera's recording start time by 1/24th of a second. Combine the two "eyes" of 12 fps in post in alternating L1,R1,L2,R2 sequence.

You'd get something like:

Frame 1 (Frame 1 from Cam 1): recorded 0.00000 - 0.08333 seconds

Frame 2 (Frame 1 from Cam 2): recorded 0.04167 - 0.12500 seconds

Frame 3 (Frame 2 from Cam 1): recorded 0.08333 - 0.16667 seconds

Frame 4 (Frame 2 from Cam 2): recorded 0.12500 - 0.20833 seconds.

And so on.


The effect would be an ultra-smeary, probably a little ghost-like image, where the motion blur from one frame overlaps the next. Long exposure in motion without speed-up or slow-down. It's like the opposite of 180 degree shutter. Instead of half the motion blur of a 360 degree shutter, it's double the motion blur of a 360 degree shutter, in it's 720 degree shutter!

Maybe this could be a big budget dream sequence

 

[Terry VerHaar]

If I am not mistaken, the techniques dscribed wouldn't work because, at some point, you'd simply run out of past frames to sum. Because you are attempting to represent more than, say 1/24th of a second, in that same amount of time, wou'd eventually arrive at "the present" in your summing and not have the data you need to record per your 1/24th + parameters. Essentially, you'd eventually "run out of the past" to work with.

:huh:

 

[Art Chong]

If I am not mistaken, the techniques dscribed wouldn't work because, at some point, you'd simply run out of past frames to sum. Because you are attempting to represent more than, say 1/24th of a second, in that same amount of time, wou'd eventually arrive at "the present" in your summing and not have the data you need to record per your 1/24th + parameters. Essentially, you'd eventually "run out of the past" to work with.

:huh:

If you're referring to my method:

You have 2x time recorded because you're using 2 cameras in a stereo rig. There is no "past". Each camera records 12fps at 1/12th second or 360 degree shutter. You record things in real time, but with two tracks in parallel, offset by recording start time. The offset is what gives you overlapping exposure times between frames.

If you look at the math, you can see that each PLAYBACK FRAME overlaps. I think that's where you're thinking that I'd run out of the "Past", but they're from two cameras. If you look at the RECORD times per camera, you'll see that each record sequentially, with no overlap. I'll attempt to illustrate this by rearranging the frames and adding additional labels:

Frame 1 (Frame 1 from Cam 1): recorded 0.00000 - 0.08333 seconds - playback frame 1, recorded frame 1 from camera 1
Frame 3 (Frame 2 from Cam 1): recorded 0.08333 - 0.16667 seconds - playback frame 3, recorded frame 2 from camera 1

total recording time: 0.16667 seconds

Frame 2 (Frame 1 from Cam 2): recorded 0.04167 - 0.12500 seconds - playback frame 2, recorded frame 1 from camera 2
Frame 4 (Frame 2 from Cam 2): recorded 0.12500 - 0.20833 seconds - playback frame 4, recorded frame 2 from camera 2

total recording time: 0.16667 seconds.

No magical hand-waving. Both cameras record equal amounts of real time, with slightly different start times.


If you're referring to Nick's method:

I think Nick's method is something like record everything at 96fps and blend the frames in post, putting 4 frames of recorded time into 1 frame of playback time with some overlap in the blending. Playback frame 1 blends recorded frames 1-4, playback frame 2 blends recorded frames 3-6, or something like that. You'd have to come up with a formula that gets you realtime playback in the end. Of course, you'd have to interpolate some motion blur between frames, so they don't just look like 4 pictures stacked on top of each other, which is David's point, that a recorded frame isn't a sequence of ultra-short frames blended together. The sensor picks up X amount of light and constructs a frame with that.


I still think this could potentially produce a really cool, wispy, dream look.

 

[Todd T]

This feature is in the latest (beta) firmware build; Just enable Frame summing or Averaging

 

[Terry VerHaar]

Art - sorry... no, I was not referring to your method but rather the method proposed at the front of the thread suggesting combining of frames (in one camera).

 

[Nick Pasquariello]

I think Nick's method is something like record everything at 96fps and blend the frames in post, putting 4 frames of recorded time into 1 frame of playback time with some overlap in the blending. Playback frame 1 blends recorded frames 1-4, playback frame 2 blends recorded frames 3-6, or something like that. You'd have to come up with a formula that gets you realtime playback in the end. Of course, you'd have to interpolate some motion blur between frames, so they don't just look like 4 pictures stacked on top of each other, which is David's point, that a recorded frame isn't a sequence of ultra-short frames blended together. The sensor picks up X amount of light and constructs a frame with that.

Sorry, no, the method I described would be shot at 24fps both in terms of timebase and editbase, and the smearing would occur in-camera. But rather than "Record a frame ---> Write that frame ---> Record next frame ---> Write next frame" it would be "Record a frame ---> Record a frame ---> Record a frame ---> Record a frame ---> Record a frame ---> Record a frame ---> Sum or average those 6 frames ---> Write a frame ---> Record a frame ---> Sum or average the most recent 6 frames ---> Write a frame ---> etc."

The whole point of doing this is that you get a trail of motion. A longer motion blur than just 180 or 360 degrees allows. It would be 2160 degree shutter. As in a quarter-second shutter without making it so that the shortest possible frame-length was a quarter second. Again, all this math and numbers are based on the OP's original scenario of having a 1/4 second shutter for each frame in a 24fps clip.

 

[Bryan Colvin]

that's cool that some of the cheaper older cameras were able to do this!

Although to the rest of you, I think what almost ALL of you missed was what I said in my original post.

I KNOW ITS PHYSICALLY IMPOSSIBLE with NORMAL, what you're USED TO, cinematography. Go back to my simple example of the GUITAR DELAY PEDAL in my original post.

What is this effect? It is echoing time with a feedback loop giving you that IMPOSSIBLE capture of audio, HOW'D THEY DO THAT?!?!?!. This technology has existed for a long time now. All I'm saying is figuring out how to add a feedback loop next to the sensor that is capturing light the same way these guitar effects pedals work.

It seems like the same thing as a guitar pedal. You have the A->D converter, and then where you process the data there is a space for a feedback loop to contribute to the data BEFORE output (writing to capture media). Its just data at this point. We're not talking about what you're used to with a film camera. Think outside the box here guys, this is DIGITAL!

I also don't know if this SHOULD be done... but to me, it seems like all you'd be doing is giving you more creative freedom by having this available. If you want to just say "IT CANT BE DONE AND SHOULDN'T BE DONE BECAUSE IM NOT USED TO IT" then you're missing what i'm saying!

Maybe someone could build a modular hardware plugin device that would allow for cinema FX pedals if you will, inserted right before writing to the capture media. I KNOW you could do things in post and a lot of people wouldn't want to bake in something like this... but it just seems like another creative brush and i wanted to share my idea on it and see if anyone else would be excited by something like this. Thanks for reading and sharing your thoughts thus far! But please read this post and re-think what i'm saying before you post agian saying "IT CANT BE DONE!" i say.. "YES IT CAN!"

 

[DL Fitch]

Dude, nobody is telling you it can't be done because they aren't used to it, but have moved far beyond your 'guitar pedal' analogy to a discussion of real implementations - an existing form of which can be found in the 4.x firmwares. In rare occasions this may be something you'd want to preview in the field, but the main reason you wouldn't want to bake-in the effect is because it can be achieved easily (and with much better control) in post.

After Effects allows a 720-degree shutter, and in Maya it's infinite, so you can at least get a idea of how this would look by using a simulated shutter.