4:4:4 10bit single CMOS HD project

Obin Olson · July 28th, 2005, 10:33 PM

next week we start production of 2 30sec commercial projects using the reelstream output and a production version of the micro35 adaptor...should get some nice results i hope...i will post images/video when we get it....pcb boards are being done up for our 1080p hd project....;)'

btw as for the ccd vs. cmos....if they are both raw data capture at or above 10bit they are both very good quality imho...just look at the dvx footage so far...and the rolling shutter is a large issue with cmos...we are doing our current design around a 1inch format ccd....its 16mm diag. in size

Jason Rodriguez · July 28th, 2005, 11:00 PM

Quote:

No 35 DOF, rubbish no -HD even build up from an
industrial cam with Altasens has WITHOUT any adapter 35 mm DOF, FACT.

You better watch yourself there Ron.

The fact is that depending on how open your aperture is, you're going to get a "similar" DOF to 35mm at a smaller aperture.

For example, if you had a lens that was a f1.3 on a 2/3" sensor(there are lenses for 16mm our there that can do that), then you're going to get similar DOF to 35mm optics at around f2.8. And if you've ever shot with 35mm optics, you'd know that f2.8 is pretty shallow DOF.

You don't need a 35mm sensor to get that "35mm" DOF, nor do you need an adapter. What you need is at least a 2/3" sensor (which there are many out there), and shoot at apertures larger than f1.6.

If you don't believe me, then run down to your local Barnes and Noble, and go get a copy of the American Cinematographer's Handbook. Look up their DOF charts for 16mm, 2/3" CCD/CMOS, and 35mm, and you'll see where the two cross-over, or where the actual "in-focus" areas match for a given set of focal lengths.

Keith Wakeham · July 29th, 2005, 10:23 AM

Thats some good info Jason. I think a lot of us forget that sensor size isn't everything and the optics used determines a lot about what kind of picture your going to get.

As for the DOF argument a some high school physics will tell you that any sensor size can have the exact same focus effect with a lens with a different focal length. Reality causes things to get more difficult and usually expensive because of the transmissive properties of glass. To get more light to your sensor your gonna have to shell out some more green.

I wonder if their is any sort of DOF wiki out their on the web.

Omar Saad · July 29th, 2005, 12:31 PM

Is the rolling shutter artifacting still an issue with the faster readout on the new Altsens? Has anyone seen video from this chip? Obin arent you using the si-1920HD?

Keith Wakeham · July 30th, 2005, 07:18 AM

Their are a lot of issues to contend with on the altasens sensor if you dig deep enough. At 60p or even 48p the rolling shutter artifact shouldn't be a big issue but their is the issue of what results because the sensor has to be running at such a high speed. Big data. Sensor has to run around 150mhz @12bit = 214 MB/s, less wanted data but that is the raw data rate including blanking.

So no matter what you need something capable of writing that speed and running at that clock rate. If you only capture ever other frame you need to be able to write at that rate to drives or to a buffer and read the buffer to the drives. Means twice the number of drives or much more complicated hardware. The buffer would have to be at least a frame so you need 3mb for a frame.

That is only one of the many issues. All sensors have their own set of issues it seems, and the altasens has a trade off of less electronics being required initially but will need more processing and more things to make it work in a hardware design.

Omar Saad · August 1st, 2005, 12:53 AM

I have been reading up as much as I can on the reelstream dvx mod, I am a little unsure of a couple of things however; First of all are the ccds producing a real hd image or is it just a rezzed up version of sd?

I have looked over the clips posted and they seem to look very good, so maybe I am answering my own question. What has really struck me however is the notion that this may perhaps be a better solution than the Altsens single cmos camera. Correct me if I am wrong....a camera like the si-1920hd with the altsens chip will produce a 1440x1080p image off a single chip after the bayer filter, correct? Acording to the reelstream site the dvx mod will put out a 1540x990 image when at 4.4.4 RGB, 3ccds. True the sensors are smaller but is'nt the image going to at least be the same quality as the altsens camera? Like I said I am not sure that I am on the right track with this, just wondering what everyone else thinks.

thanks,
Omar

Keith Wakeham · August 1st, 2005, 03:36 PM

The the output is not "true hd" in terms of pixel per pixel, it is upscaled based on pixel shift which is similar to debayer to my understanding. So once you crop to 16:9 is acts more like a bayer 720p camera to my understanding.

Omar Saad · August 1st, 2005, 04:13 PM

Keith, thanks for the response....that seems to make more sense. It still looks great though. How is your camera project going? Did you ever get a new website up?

Juan M. M. Fiebelkorn · August 2nd, 2005, 12:28 AM

Imagine it like being something like the new Sony HDV camera but recording frame by frame uncompressed and with 12 bit bitdepth.
Forgot to mention the "progressive part"...

Keith Wakeham · August 2nd, 2005, 03:43 PM

I don't want to disclose anything at this point about what I'm currently doing so their won't be a new website in a while. I'm just going to say I'm being quiet and working on something.

As for the sony hdv camera, the ccd's in the fx1 have about 3 times as many pixels in comparison to the dvx100 but I don't think they are pixel shifted. The sony hdv camera would likely deliever a superior resolution if it was modded like the reel-stream dvx but at a much much higher datarate.

Frank Vrionis · August 2nd, 2005, 09:36 PM

but the Sony is only interlaced.

what exactly is pixel shift? from my understanding there's a pseudo version that doesn't yield 'true' resolution - many still digital cameras use it.

and then there's the one that the XL1 used and did produce extra detail/resolution.

Keith Wakeham · August 3rd, 2005, 04:12 AM

Pixel shift is only something native to 3 ccd. Most ccd's are aligned perfectly in 3ccd so that all the pixels are lined up perfectly in relation to the other ccd's. Pixel shift is when one of the ccds is physicaly moved so its no longer in line. Normally the green is "shifted" so that it lies in the middle between the red and blue, so a green pixel is overlapping 2 red and blue pixels.

Lots of info in the forums, just have to search for it and it will answer questions like this.

Régine Weinberg · August 10th, 2005, 07:11 AM

http://www.iwill.net/product_2.asp?p_id=76&sp=Y

it, way cheaper as a full blown laptop

Noah Yuan-Vogel · August 10th, 2005, 01:29 PM

Obin, you have worked with the si3300rgb, right? Do you have any samples of the video it produces that show how good/bad the rolling shutter is at a 48MHz pixelrate? Could you post some, or otherwise make available to me, such video (I have some webspace available if you need). I am considering buying a camera with the same sensor but I am very concerned about the rolling shutter, but I havent even seen what it looks like to figure out what is acceptable and what is not in terms of motion artifact. I just really want to get a good idea of what we are talking about when it comes to rolling shutter cameras being bad for digital cinema. I'm just surprised it could be so unusable, micron.com lists the sensor as being appropriate for still photography (in which rolling shutter artifacts are also not acceptable) so i would expect it to be alright for video.

I am interested in the sensor in the si1300...

kyle granger? any rolling shutter artifact samples?

Anyone else who can help out with this, please let me know. I'm sure there are other people who are also interested in seeing sample video as well.

Jason Rodriguez · August 14th, 2005, 08:00 PM

Unless all you do are whip-pans back-and-forth, if you get the sensor to go at 48fps and then drop every other frame, the rolling shutter artifacts are un-noticeable. You won't notice the difference on any "normal" professional pan. But like I said, if you whip the camera around, you'll see it Iif you know what you're looking for), although typically there's so much motion blur on those whip-pans, that again, unless you know what to look for, you won't notice the rolling shutter.

All-in-all, it's not a problem.

BTW, you're not going to be able to clock the chip high-enough to run the chip at full-resolution and get this "drop-every-other-frame" mode. You're going to have to run at 1280x720 (windowing).

Give Steve a call over at Silicon Imaging, he can tell you a lot more about the chip, and their camera is pretty reasonably priced.

August 10th, 2005, 07:11 AM	#2953
Régine Weinberg Major Player Join Date: Jan 2004 Location: Bordeaux, going to Bangkok, 2011 Posts: 232	Very small barebone computer http://www.iwill.net/product_2.asp?p_id=76&sp=Y it, way cheaper as a full blown laptop

July 28th, 2005, 10:33 PM	#2941
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	next week we start production of 2 30sec commercial projects using the reelstream output and a production version of the micro35 adaptor...should get some nice results i hope...i will post images/video when we get it....pcb boards are being done up for our 1080p hd project....;)' btw as for the ccd vs. cmos....if they are both raw data capture at or above 10bit they are both very good quality imho...just look at the dvx footage so far...and the rolling shutter is a large issue with cmos...we are doing our current design around a 1inch format ccd....its 16mm diag. in size

July 29th, 2005, 10:23 AM	#2943
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	Thats some good info Jason. I think a lot of us forget that sensor size isn't everything and the optics used determines a lot about what kind of picture your going to get. As for the DOF argument a some high school physics will tell you that any sensor size can have the exact same focus effect with a lens with a different focal length. Reality causes things to get more difficult and usually expensive because of the transmissive properties of glass. To get more light to your sensor your gonna have to shell out some more green. I wonder if their is any sort of DOF wiki out their on the web.

July 29th, 2005, 12:31 PM	#2944
Omar Saad Regular Crew Join Date: Dec 2003 Location: San Luis Obispo, CA Posts: 38	Is the rolling shutter artifacting still an issue with the faster readout on the new Altsens? Has anyone seen video from this chip? Obin arent you using the si-1920HD?

July 30th, 2005, 07:18 AM	#2945
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	Their are a lot of issues to contend with on the altasens sensor if you dig deep enough. At 60p or even 48p the rolling shutter artifact shouldn't be a big issue but their is the issue of what results because the sensor has to be running at such a high speed. Big data. Sensor has to run around 150mhz @12bit = 214 MB/s, less wanted data but that is the raw data rate including blanking. So no matter what you need something capable of writing that speed and running at that clock rate. If you only capture ever other frame you need to be able to write at that rate to drives or to a buffer and read the buffer to the drives. Means twice the number of drives or much more complicated hardware. The buffer would have to be at least a frame so you need 3mb for a frame. That is only one of the many issues. All sensors have their own set of issues it seems, and the altasens has a trade off of less electronics being required initially but will need more processing and more things to make it work in a hardware design.

August 1st, 2005, 12:53 AM	#2946
Omar Saad Regular Crew Join Date: Dec 2003 Location: San Luis Obispo, CA Posts: 38	I have been reading up as much as I can on the reelstream dvx mod, I am a little unsure of a couple of things however; First of all are the ccds producing a real hd image or is it just a rezzed up version of sd? I have looked over the clips posted and they seem to look very good, so maybe I am answering my own question. What has really struck me however is the notion that this may perhaps be a better solution than the Altsens single cmos camera. Correct me if I am wrong....a camera like the si-1920hd with the altsens chip will produce a 1440x1080p image off a single chip after the bayer filter, correct? Acording to the reelstream site the dvx mod will put out a 1540x990 image when at 4.4.4 RGB, 3ccds. True the sensors are smaller but is'nt the image going to at least be the same quality as the altsens camera? Like I said I am not sure that I am on the right track with this, just wondering what everyone else thinks. thanks, Omar

August 1st, 2005, 03:36 PM	#2947
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	The the output is not "true hd" in terms of pixel per pixel, it is upscaled based on pixel shift which is similar to debayer to my understanding. So once you crop to 16:9 is acts more like a bayer 720p camera to my understanding.

August 1st, 2005, 04:13 PM	#2948
Omar Saad Regular Crew Join Date: Dec 2003 Location: San Luis Obispo, CA Posts: 38	Keith, thanks for the response....that seems to make more sense. It still looks great though. How is your camera project going? Did you ever get a new website up?

August 2nd, 2005, 12:28 AM	#2949
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	Imagine it like being something like the new Sony HDV camera but recording frame by frame uncompressed and with 12 bit bitdepth. Forgot to mention the "progressive part"...

August 2nd, 2005, 03:43 PM	#2950
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	I don't want to disclose anything at this point about what I'm currently doing so their won't be a new website in a while. I'm just going to say I'm being quiet and working on something. As for the sony hdv camera, the ccd's in the fx1 have about 3 times as many pixels in comparison to the dvx100 but I don't think they are pixel shifted. The sony hdv camera would likely deliever a superior resolution if it was modded like the reel-stream dvx but at a much much higher datarate.

August 2nd, 2005, 09:36 PM	#2951
Frank Vrionis Regular Crew Join Date: Feb 2005 Posts: 93	but the Sony is only interlaced. what exactly is pixel shift? from my understanding there's a pseudo version that doesn't yield 'true' resolution - many still digital cameras use it. and then there's the one that the XL1 used and did produce extra detail/resolution.

August 3rd, 2005, 04:12 AM	#2952
Keith Wakeham Major Player Join Date: Jan 2005 Location: St. John's, NL, Canada Posts: 416	Pixel shift is only something native to 3 ccd. Most ccd's are aligned perfectly in 3ccd so that all the pixels are lined up perfectly in relation to the other ccd's. Pixel shift is when one of the ccds is physicaly moved so its no longer in line. Normally the green is "shifted" so that it lies in the middle between the red and blue, so a green pixel is overlapping 2 red and blue pixels. Lots of info in the forums, just have to search for it and it will answer questions like this.

August 10th, 2005, 01:29 PM	#2954
Noah Yuan-Vogel Major Player Join Date: Dec 2004 Location: New York City Posts: 613	Obin, you have worked with the si3300rgb, right? Do you have any samples of the video it produces that show how good/bad the rolling shutter is at a 48MHz pixelrate? Could you post some, or otherwise make available to me, such video (I have some webspace available if you need). I am considering buying a camera with the same sensor but I am very concerned about the rolling shutter, but I havent even seen what it looks like to figure out what is acceptable and what is not in terms of motion artifact. I just really want to get a good idea of what we are talking about when it comes to rolling shutter cameras being bad for digital cinema. I'm just surprised it could be so unusable, micron.com lists the sensor as being appropriate for still photography (in which rolling shutter artifacts are also not acceptable) so i would expect it to be alright for video. I am interested in the sensor in the si1300... kyle granger? any rolling shutter artifact samples? Anyone else who can help out with this, please let me know. I'm sure there are other people who are also interested in seeing sample video as well.

August 14th, 2005, 08:00 PM	#2955
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	Unless all you do are whip-pans back-and-forth, if you get the sensor to go at 48fps and then drop every other frame, the rolling shutter artifacts are un-noticeable. You won't notice the difference on any "normal" professional pan. But like I said, if you whip the camera around, you'll see it Iif you know what you're looking for), although typically there's so much motion blur on those whip-pans, that again, unless you know what to look for, you won't notice the rolling shutter. All-in-all, it's not a problem. BTW, you're not going to be able to clock the chip high-enough to run the chip at full-resolution and get this "drop-every-other-frame" mode. You're going to have to run at 1280x720 (windowing). Give Steve a call over at Silicon Imaging, he can tell you a lot more about the chip, and their camera is pretty reasonably priced.