Add physical shutter to 5D2 to eliminate wobble, Dan Vance style? at DVinfo.net

Paul Martin · December 14th, 2008, 05:41 PM

How about putting a physical shutter of some kind in front of the lens mount on the 5D2 to block light from the CMOS sensor during the read/reset phase of sensor readout?

Dan Vance knows all about this stuff: VanceCam VC25P Camcorder

According to the FAQs on the website of a company that makes Liquid crystal shutters, LCD Shutters from Liquid Crystal Technologies. Our liquid crystal shutters can solve many different design challenges., LCD shutters are less than ideal for cameras.

Seems like a half-disc spinning at 1800 RPMs in sync with the camera's pulses is the way to go.

Using medium format lenses with long focal flanges would allow for the added bulk between the lens and the 5D2; the Mamiya RB67 flange focal distance is 112mm.

180 degree shutter and few/no rolling shutter artifacts would be nice.

Sounds plausible, no?

James Miller · December 15th, 2008, 01:54 AM

Hi Paul,
Even if possible, and I would say not with this rig, by their reference you would loose 45% light.

Paul Martin · December 15th, 2008, 02:22 PM

Quote:

Originally Posted by James Miller

Hi Paul,
Even if possible, and I would say not with this rig, by their reference you would loose 45% light.

Uh, yeah, that's why I said a liquid crystal shutter would likely be less than ideal...

However, the 5D2 is damn sensitive. I would sure as hell trade a stop of sensitivity for wobble free footage.

Still, I think a large diameter spinning half (or variably-sized) disc is the way to go, Mitchell NC style.

Paul Martin · December 15th, 2008, 02:23 PM

Here's the email I just wrote to Zacuto, Redrock Micro, Brevis, Cinemek, and Shoot35:

I have followed the 35mm adapter revolution since Agus Casse first posted some grabs on DVinfo.net. I instantly knew that DoF control was what I was needed for my digital filmmaking.

I tried making some adapters myself but was never satisfied. About a year ago I realized that the new generation of DSLRs with Live View meant that pretty soon full-frame and APS-C cameras would soon be shooting video.

Sure enough, the D90 and 5D2 are now here, but they both suffer from wobble due to the rolling shutter readout of their CMOS sensors. I have a 5D2 and it's incredible, but it pretty much can't be handheld at 50mm and up. Here and here I posted the idea of putting a mechanical shutter between the 5D2 and a medium format lens (to allow for the space required for the added shutter), to block light from the sensor during sensor readout. Non-reflex film cameras like the Eyemo work this way.

A spinning half-disc synced to the camera's pulses might do the trick. There are also liquid crystal shutters, but they aren't fully transmissive nor fully opaque.

If you guys built such a "shutter adapter" that actually prevented wobble, you'd have plenty of customers: 5D2 customers, Red One customers, and certainly D90 customers, if it was inexpensive enough.

A spinning shutter would also still cause rolling shutter artifacts, but they would be like the RS artifacts we see in movies shot on Arris, for example. So long as the diameter of the spinning disc was big enough, the transit time of the leading edge of the disc would be faster than the readout transit time of the sensor, thus reducing RS artifacts, especially wobble.

Jim Jannard was just saying that the RS artifacts we're used to seeing in movies shot on film cameras make the image more dynamic. I don't agree, but hey, if a spinning shutter adapter could make a video camera look even more like an Arri, by replacing nasty wobble artifacts with more subtle RS artifacts that were used to seeing in 35mm films, I'm sure it would be welcomed.

I know that if it can be done, you're the guys to do it.

The hybrid DSLR/videocamera isn't going away. Unfortunately, rolling shutter CMOS readout probably isn't going to be replaced by global readout any time soon.

That's where you guys come in!

Thanks for taking the time to read this.

I look forward to any correspondence.

Evan Donn · December 15th, 2008, 06:35 PM

I don't think this would work, precisely because of the rolling shutter. The VanceCam used CCDs, not CMOS...

For the sync to work the mechanical shutter would have to be open in the middle of the electronic shutter's exposure... but the electronic shutter isn't 'open' all at once, it's scanning from top to bottom. So at what point do you open the mechanical shutter? Is there any point at which the entire sensor is reading out simultaneously? If so, I assume it's going to limit the effective frame rate to a fraction of the total scan time which means you're probably looking at relatively high shutter speeds, at least relative to the normal shutter speeds used for film.

The other problem I see with this is the camera's electronic shutter speed - ideally you'd get the sensor running at 1/30 and then use the mechanical shutter to set a higher speed, but there's no consistent way to set and lock a shutter speed on the 5D.

Paul Martin · December 16th, 2008, 11:49 AM

Quote:

Originally Posted by Evan Donn

For the sync to work the mechanical shutter would have to be open in the middle of the electronic shutter's exposure... but the electronic shutter isn't 'open' all at once, it's scanning from top to bottom. So at what point do you open the mechanical shutter? Is there any point at which the entire sensor is reading out simultaneously? If so, I assume it's going to limit the effective frame rate to a fraction of the total scan time which means you're probably looking at relatively high shutter speeds, at least relative to the normal shutter speeds used for film.

1. Lock camera at 1/30th (put hand over lens, press *). Now every pixel is integrating for 1/30th of a second, before being read and reset in a rolling manner. In this mode, pixels are read, reset, and immediately begin integrating again, as there is no time left over to wait for the start of the next frame.

2. Spin a shutter in front of the sensor such that shutter is spinning at 30 revolutions per second.

Using a 180 degree shutter, the shutter will be blocking the sensor 50% of the time, even though the pixels will be integrating for slightly less than 100% of the time--slightly less because of the very short time a given pixel is not integrating because it is instead being read and reset.

The system needs to be synchronized such that the mechanical shutter moves into place, blocking the sensor, prior to the commencement of the rolling reading and resetting of each pixel.

If it takes X milliseconds from the time the shutter begins to block the sensor to the time the sensor is completely blocked, then the temporal disparity between the top and bottom of the resultant image is X ms. Going to the trouble of making a mechanical shutter is only worthwhile if X is less than the time it takes for the sensor's own electronic rolling shutter to traverse the sensor. With a shutter large enough in diameter, therefore having a very fast leading edge when spinning at 30 revolutions per second, I do believe this is possible.

Perhaps there is also the possibility of using a stepper motor and pulse width modulation to operate the shutter, such that the 180 degree shutter moves very rapidly into place, stands idle for as great a part of 1/60th of second as is possible, then very rapidly moves out of place. In this manner, the shutter would still spin, but in a pulsed fashion, rapidly accelerating and deaccelerating so as to minimize transit time of the shutter edge over the sensor.

But I'm no electrical engineer.

Dan Vance is though, and he's hooking me up!

Awesome guy!

He's sending me a liquid crystal shutter and a syncing circuit.

Thanks Dan!

If it works, I'll start on a spinner.

Paul Martin · December 18th, 2008, 05:37 PM

I shot a clip while a strobe went off.

http://www.vimeo.com/2568049

Most of the frame was exposed by the flash.

I started experimenting with shooting clips with the 5D2 at various shutter speeds, recording the strobe of my 20D going off.

Only while shooting with a 1/30 shutter were there any frames of the 5D2 video that were uniformly exposed by the strobe. About a third of the strobe bursts were essentially "synced" in this way; the rest of the strobe bursts were spread across 2 frames.

The fact that some of the frames in the video are evenly illuminated by the strobe when the 5D2 is shooting at 1/30th means that there does exist a moment where every pixel on the sensor is integrating simultaneously. Therefore, the 5D2 is capable of being globally shuttered.

The interesting thing about a strobe going off while shooting, like in the clip above, is this: you could be shaking the camera around like crazy, making the image look like a big swirling pool of mud, but then the strobe goes off, and you have one perfectly crisp frame.

December 14th, 2008, 05:41 PM	#1
Paul Martin Regular Crew Join Date: Dec 2002 Location: San Francisco, CA Posts: 47	Add physical shutter to 5D2 to eliminate wobble, Dan Vance style? How about putting a physical shutter of some kind in front of the lens mount on the 5D2 to block light from the CMOS sensor during the read/reset phase of sensor readout? Dan Vance knows all about this stuff: VanceCam VC25P Camcorder According to the FAQs on the website of a company that makes Liquid crystal shutters, LCD Shutters from Liquid Crystal Technologies. Our liquid crystal shutters can solve many different design challenges., LCD shutters are less than ideal for cameras. Seems like a half-disc spinning at 1800 RPMs in sync with the camera's pulses is the way to go. Using medium format lenses with long focal flanges would allow for the added bulk between the lens and the 5D2; the Mamiya RB67 flange focal distance is 112mm. 180 degree shutter and few/no rolling shutter artifacts would be nice. Sounds plausible, no?

December 15th, 2008, 06:35 PM	#5
Evan Donn Major Player Join Date: Feb 2006 Location: Philadelphia Posts: 795	I don't think this would work, precisely because of the rolling shutter. The VanceCam used CCDs, not CMOS... For the sync to work the mechanical shutter would have to be open in the middle of the electronic shutter's exposure... but the electronic shutter isn't 'open' all at once, it's scanning from top to bottom. So at what point do you open the mechanical shutter? Is there any point at which the entire sensor is reading out simultaneously? If so, I assume it's going to limit the effective frame rate to a fraction of the total scan time which means you're probably looking at relatively high shutter speeds, at least relative to the normal shutter speeds used for film. The other problem I see with this is the camera's electronic shutter speed - ideally you'd get the sensor running at 1/30 and then use the mechanical shutter to set a higher speed, but there's no consistent way to set and lock a shutter speed on the 5D. __________________ My latest short documentary: "Four Pauls: Bring the Hat Back!"

December 18th, 2008, 05:37 PM	#7
Paul Martin Regular Crew Join Date: Dec 2002 Location: San Francisco, CA Posts: 47	5D2, strobe light, interesting result I shot a clip while a strobe went off. http://www.vimeo.com/2568049 Most of the frame was exposed by the flash. I started experimenting with shooting clips with the 5D2 at various shutter speeds, recording the strobe of my 20D going off. Only while shooting with a 1/30 shutter were there any frames of the 5D2 video that were uniformly exposed by the strobe. About a third of the strobe bursts were essentially "synced" in this way; the rest of the strobe bursts were spread across 2 frames. The fact that some of the frames in the video are evenly illuminated by the strobe when the 5D2 is shooting at 1/30th means that there does exist a moment where every pixel on the sensor is integrating simultaneously. Therefore, the 5D2 is capable of being globally shuttered. The interesting thing about a strobe going off while shooting, like in the clip above, is this: you could be shaking the camera around like crazy, making the image look like a big swirling pool of mud, but then the strobe goes off, and you have one perfectly crisp frame.

December 15th, 2008, 01:54 AM	#2
James Miller Major Player Join Date: Mar 2007 Location: Sussex, UK Posts: 317	Hi Paul, Even if possible, and I would say not with this rig, by their reference you would loose 45% light.

December 15th, 2008, 02:23 PM	#4
Paul Martin Regular Crew Join Date: Dec 2002 Location: San Francisco, CA Posts: 47	Here's the email I just wrote to Zacuto, Redrock Micro, Brevis, Cinemek, and Shoot35: I have followed the 35mm adapter revolution since Agus Casse first posted some grabs on DVinfo.net. I instantly knew that DoF control was what I was needed for my digital filmmaking. I tried making some adapters myself but was never satisfied. About a year ago I realized that the new generation of DSLRs with Live View meant that pretty soon full-frame and APS-C cameras would soon be shooting video. Sure enough, the D90 and 5D2 are now here, but they both suffer from wobble due to the rolling shutter readout of their CMOS sensors. I have a 5D2 and it's incredible, but it pretty much can't be handheld at 50mm and up. Here and here I posted the idea of putting a mechanical shutter between the 5D2 and a medium format lens (to allow for the space required for the added shutter), to block light from the sensor during sensor readout. Non-reflex film cameras like the Eyemo work this way. A spinning half-disc synced to the camera's pulses might do the trick. There are also liquid crystal shutters, but they aren't fully transmissive nor fully opaque. If you guys built such a "shutter adapter" that actually prevented wobble, you'd have plenty of customers: 5D2 customers, Red One customers, and certainly D90 customers, if it was inexpensive enough. A spinning shutter would also still cause rolling shutter artifacts, but they would be like the RS artifacts we see in movies shot on Arris, for example. So long as the diameter of the spinning disc was big enough, the transit time of the leading edge of the disc would be faster than the readout transit time of the sensor, thus reducing RS artifacts, especially wobble. Jim Jannard was just saying that the RS artifacts we're used to seeing in movies shot on film cameras make the image more dynamic. I don't agree, but hey, if a spinning shutter adapter could make a video camera look even more like an Arri, by replacing nasty wobble artifacts with more subtle RS artifacts that were used to seeing in 35mm films, I'm sure it would be welcomed. I know that if it can be done, you're the guys to do it. The hybrid DSLR/videocamera isn't going away. Unfortunately, rolling shutter CMOS readout probably isn't going to be replaced by global readout any time soon. That's where you guys come in! Thanks for taking the time to read this. I look forward to any correspondence.