Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

General Question About CMOS Rolling Shutter

There's some interesting research on distributed processing of sensor data. Once you get past the linearity of data streams. We're kinda locked into an 2D manufacturing paradigm right now, which gives us an XY orientation of rows and columns for reading the photosites. If we could break the z axis on the sensor's backside the possibility exists to segment the sensor into smaller zones of discreet data stems. This is a whole can of worms as you can imagine, not the least of which is precision nanoscale manufacturing processes and thermal mitigation. But in theory- breaking the third dimension in planar sensors into a true mosaic of concurrent processing could yield sensors that make Red's 617 look like a flip video camera. Imagine a 617 that yields Phantom scale frame rates. Pretty sexy...
 
The biggest issue is going to be interconnect. The more you add, the more you take away the window of light hitting the active area. With how most are made now, you only need one bus the width of the sensor. If you had 2 sets of AD converters, then you'd need 2 busses and 2 sets of control lines. There are always trade-offs.

The problem with localized reading is that if you don't do it all at once, you end up introducing other artifacts. Imagine if the sensor was split into 4 areas and you read them all sequentially as individual sensors. Sure, you could read faster, but at the same time, for really fast shutter speeds and fast camera movements, you'd have what would look like tearing in the images. At least with a full rolling shutter, it looks a bit more natural, and in some cases on fast pans on stationary visuals, you can fix some of it up in post easily.

I still think the global shutter is the best approach myself.
 
The biggest issue is going to be interconnect. The more you add, the more you take away the window of light hitting the active area. With how most are made now, you only need one bus the width of the sensor. If you had 2 sets of AD converters, then you'd need 2 busses and 2 sets of control lines. There are always trade-offs.

The problem with localized reading is that if you don't do it all at once, you end up introducing other artifacts. Imagine if the sensor was split into 4 areas and you read them all sequentially as individual sensors. Sure, you could read faster, but at the same time, for really fast shutter speeds and fast camera movements, you'd have what would look like tearing in the images. At least with a full rolling shutter, it looks a bit more natural, and in some cases on fast pans on stationary visuals, you can fix some of it up in post easily.

I still think the global shutter is the best approach myself.

Rick - thanks for weighing in. I see what your saying and agree. It could get weird! I have done a little more research here and have also uncovered some comments from Jim suggesting that in order to best emulate film's own "rolling shutter" effect, you might not want a global shutter, just a fast read-reset on an electronic rolling shutter. In any case, it does seem as if RED will succeed in getting the read-reset times so very low that there really aren't any issues to worry about.
 
well dude in that case u would get the 360 degree shutter motion,not the 180 degree shutter and it wont be very cinematic,film reacts to light in that way at an instant like a global shutter,all photosites at once but in that case there is an external mechanical 180 degree shutter.
 
The phantom flex and Weisscam HS-2 both have global shutters on CMOS sensors and do incredibly high frame rates. Not 5K, and neither is cheap... but it's certainly possible.
 
well dude in that case u would get the 360 degree shutter motion,not the 180 degree shutter and it wont be very cinematic,film reacts to light in that way at an instant like a global shutter,all photosites at once but in that case there is an external mechanical 180 degree shutter.

That really isn't the way a film camera does it. The mechanical shutter "wipes" across the film in a similar way to how a CMOS sensor is read. The trick to CMOS is getting the read-reset time similar to the mechanical shutter. EPIC does that.

Just so you know... there is skew in a film camera with a mechanical shutter, one of the many reasons film has "character".

Jim
 
[...]The trick to CMOS is getting the read-reset time similar to the mechanical shutter. EPIC does that.[...]


Jim

OK. Epic does that in motion, and that alone is revolutionary.

But stills mechanical shutters can typically handle 1/2000 shutter speeds, which is necessary to freeze fast action. However, this would require a read 10x faster than 5 msec. Stills mode reset could still be 5 msec (for a ~200 fps burst !!) or even longer. If RED has achieved this shutter speed for use in stills mode with even a 20 fps burst it would definitely be the ultimate sports, action, and wildlife stills camera. Seems it would also be patentable, since AFAIK this has never been done with a CMOS sensor - at least not close to 12 MP. I thought I read on another recent thread that individual lines read almost instantaneously and there is a short but significant pause between line reads. Could that pause be shortened significantly - at least for bursts in stills mode? 10x shorter read and 10x longer reset for 1/2000, ~20 fps would be a nice tradeoff for stills.

We haven't heard a lot about EPIC stills other than pulling from motion. Is there another revolutionary advance yet to be announced? If so, I hope Scarlet has it too. 5 MP RAW stills with good DR could look very nice.
 
@Steve - FWIW, RED ONE does do shutter speeds of 1/2000th. I can only guess Epic does that or better. IMHO, you might be confusing shutter speeds and read-reset times. They are not the same thing. As near as I can tell, a fast shutter speed is accomplished by a given line of pixels turning "on" (reset) then being read and turning "off" within that 1/2000th of a second (as an example). That process happens line by line "rolling" down the shutter. (To be sure, line 2 can "fire" before line 1 turns "off," etc.). The read-reset time is how long it takes for this whole "roll" to take place from top to bottom of shutter and then start over. If the shutter speed is very fast, you simply have more "off" time during each roll. READ-RESET time has more of an effect on frame rate than it does on shutter speed.

I am not an engineer and I certainly could be mistaken on some of this but these are the things I have been investigating lately and this is what I (think!) I have learned. :cheers2:
 
OK. Epic does that in motion, and that alone is revolutionary.

But stills mechanical shutters can typically handle 1/2000 shutter speeds, which is necessary to freeze fast action. However, this would require a read 10x faster than 5 msec. Stills mode reset could still be 5 msec (for a ~200 fps burst !!) or even longer. If RED has achieved this shutter speed for use in stills mode with even a 20 fps burst it would definitely be the ultimate sports, action, and wildlife stills camera. Seems it would also be patentable, since AFAIK this has never been done with a CMOS sensor - at least not close to 12 MP. I thought I read on another recent thread that individual lines read almost instantaneously and there is a short but significant pause between line reads. Could that pause be shortened significantly - at least for bursts in stills mode? 10x shorter read and 10x longer reset for 1/2000, ~20 fps would be a nice tradeoff for stills.

RED One can do speeds up to 1/10000th.
Epic is able to do greater than that and at 5k 120fps.
 
Well Terry and Deanan, that's great! My main hope now is that Scarlet will have a 1/2000 or faster shutter speed with a ~20 fps or better stills burst. Thanks!
 
Well Terry and Deanan, that's great! My main hope now is that Scarlet will have a 1/2000 or faster shutter speed with a ~20 fps or better stills burst. Thanks!

150fps for Scarlet @ 3k and at least 1/2000th.
 
150fps for Scarlet @ 3k and at least 1/2000th.

Deanan - thanks so very much for taking what I know to be very valuable time for weighing in on this. I know you guys live and breathe this stuff but some of us are still noodling it out!

One thing that still is unclear to me - (you knew this was coming!) - seems like the frame rate is limited by the read-reset time in that you could only do the reciprocal of the read-reset time; i.e. if it takes 5 ms to read-reset, you could only do that 200 time per second and your frame rate would thus be limited to 200. Is that correct?

Thanks. :bigear:
 
Deanan - thanks so very much for taking what I know to be very valuable time for weighing in on this. I know you guys live and breathe this stuff but some of us are still noodling it out!

One thing that still is unclear to me - (you knew this was coming!) - seems like the frame rate is limited by the read-reset time in that you could only do the reciprocal of the read-reset time; i.e. if it takes 5 ms to read-reset, you could only do that 200 time per second and your frame rate would thus be limited to 200. Is that correct?

Thanks. :bigear:

Frame rate is limited by a couple of factors rather than just read/reset.
Same as with Red One.
 
Frame rate is limited by a couple of factors rather than just read/reset.
Same as with Red One.

Hmmmm... cryptic answer. :001_smile: I don't blame you.

I guess I was pondering the theoretical upper limit of the frame rate as it relates to read-reset. :coolgleamA:

And thanks again!
 
RED One can do speeds up to 1/10000th.
Epic is able to do greater than that and at 5k 120fps.

Not accusing anyone, but why is it RED ONE MX isn't able to do that too? Same sensor... :001_smile:
 
Back
Top