4:4:4 10bit single CMOS HD project

Wayne Morellini · January 5th, 2005, 05:30 AM

<<<-- Originally posted by Obin Olson : We are now moving the image preview/display off the CPU and into the GPU of our motherboard...this will allow much more CPU time for other things like Histogram etc in realtime..not to mention the use of a lower-cost slower CPU if we want..this work is being done as I speak and we are still awaiting the 64bit card for capture of the native raw 10bit 1080p footage..any day now...64bit ...arrgg I hate waiting !!
-->>

Good to hear the progress, I have been talking to Rob privately about it several months ago, after the response I got to suggesting GPU programming here before. I passed on information on two or three GPU API systems to them. Do we have any idea how small, and the breakdown, of the processor load is for single chip 720p is now?

I understand that upcoming (2005) new mobile versions of 0.9 micron AMD processors will also run at respectively low powered levels (can't rememeber but in the 20watt range, maybe sub 20). VIA will be releasing new 2+GHZ and dual core processors in a few months, hopefully in the same package. This how ever does not help old weak GPU and ethernet systems, but eventually they should be replaced. The power requirement is 3.5watt at 1 Ghz now (so a dual core will be 7w per 2Ghz). Still irrelavent for our purposes, but hopefully they get there act together.

<<<-- Originally posted by Obin Olson : I am now thinking that I may just downsize the 1080P image after capture to 720p because workflow with 720 would be much easer then 1080p..thoughts anyone? -->>>

I have often wondered about image problems at 720p or SD preview. There are certain electrical interferences that can mar an image that may go unoticed in the 720p and SD downsized images. These might rarely happen in reality, but it might go unoticed for a while.

Obin Olson · January 6th, 2005, 02:55 PM

I have often wondered about image problems at 720p or SD preview. There are certain electrical interferences that can mar an image that may go unoticed in the 720p and SD downsized images. These might rarely happen in reality, but it might go unoticed for a while.

What?! I don't understand that one Wayne...

Wayne Morellini · January 6th, 2005, 04:06 PM

Strong electromagnetic feild, electrical/power faults might put up a fine pattern of niose pattern wobbling through the image (or any other fault) that goes unoticed when averaged out in the downscaling of the preview. Also undesired fine visual detail, paterns that might cause certain strobing, staircasing due to the bayer filter pattern. Just thin possibility, otherwise there is nothing wrong with it.

Jason Rodriguez · January 6th, 2005, 04:11 PM

a good camera product should never have those problems, not if it's properly engineered/constructed.

Wayne Morellini · January 7th, 2005, 02:57 AM

Circumstances would be normally exceptional with a very well made and conditioned eng camera. Buit with simple cheap cameras attached to not so reliable PC's, who knows. But don't worry about it, Obin asked for comment, and those were the only two things that I knew of.

Jason Rodriguez · January 7th, 2005, 07:33 AM

I understand Wayne, but that would only be a problem with poorly designed circuitry and power-sources/clocks with CCD's. Since many of these new chips are actually iSoc's, (imager system-on-a-chip), the A/D converter, etc. are already enclosed inside the CMOS sensor, and all you should be getting off the chip are digital signals, so everything's already "digitized". Unless the CMOS chip was manufactured badly, or somebody poorly designed the power ciruitry to the CMOS chip, you shouldn't get RF interference or banding.

Joshua Starnes · January 7th, 2005, 11:33 AM

On the FX-1 front, Kaku Ito's (sp?) been running some tests and it looks like you can capture uncompressed 8-bit 1080i HD from the chip out through the component out and directly onto a hard drive, bypassing the MPEG-2 compression altogether.

Obin Olson · January 7th, 2005, 09:03 PM

We are having good luck with the GPU use for display of pixel data..I have no update to play with yet but my programmer is happy...this is a good sign! ;) more later

Wayne Morellini · January 8th, 2005, 07:30 AM

<<<-- Originally posted by Joshua Starnes : On the FX-1 front, Kaku Ito's (sp?) been running some tests and it looks like you can capture uncompressed 8-bit 1080i HD from the chip out through the component out and directly onto a hard drive, bypassing the MPEG-2 compression altogether. -->>>

I posted a simular thing before, but I wasn't sure of the interpretation of the data, has it been conclusively proved with on screen tests, and is there a link?

Seen my first FX1 today, if it could only do progressive 1080, 12 bit, HDV2/uncompressed, for $3000, I would go for it.
-----

Other news, from what camcorder info is saying, it looks like Panasonic is using pixel shift on many of it's 3 chip cameras. Get this the new model has around 420K gross pixels (that is 420K/3 CCD's) or around 140K pixels per CCD. It must use the 3 chip pixel shift I have been talkign about, becauses it uses that to get 1.3MP stills (140K*9). It will be interesting to see what latitude and low light performance this camera will get with such relatively large pixels. Now, if the Juan mod was appliable, it would be ussuable in 1.3Mp doco's. I would still like to see a Altasens with three chip pixel shift ;) (Imax ;)
-----

Jason, it is not as simple as that, but it doesn't matter, it is just a consideration.

Joshua Starnes · January 9th, 2005, 12:57 AM

I posted a simular thing before, but I wasn't sure of the interpretation of the data, has it been conclusively proved with on screen tests, and is there a link?

I don't think it has been conclusively proved, but the first round of tests seem to be good. There are screen tests, where you can download the (very large) uncompressed footage from his website. He has a link to it in his thread on the subject in the Alternative Imaging methods.

Dan Diaconu · January 9th, 2005, 10:08 AM

I hope this might be of interest to some:

http://www.panavisionsvi.com/imagers_Quad.htm

Wayne Morellini · January 10th, 2005, 06:24 AM

Sorry, I didn't realise you meant there was a thread on it here.

Valeriu Campan · January 10th, 2005, 06:33 AM

You can find here the link for the component out recorded out of FX-1:
http://www.dvinfo.net/conf/showthrea...threadid=37427

Wayne Morellini · January 10th, 2005, 06:49 AM

The comments makes me wonder if the camera is outputing some pre corrected pixel shifted veresion of 1080p.

Wayne Morellini · January 10th, 2005, 10:58 PM

I got around to posting a adaptor screen technology (face plates), and image down scaling technology for direct sensor chip connection (tapers). Good stuff, hi res, high transmission options. The post is here:

http://www.dvinfo.net/conf/showthread.php?s=&postid=261489#post261489

It has been a saga and a half, I went to look up another technology, I have over 160pages in my bookmark (which were open all at once), of nearly three hundred pages from just one company, that could have fitted into 10 pages (if the web pages had been designed in British style rather than European), this is getting too much for my health.

January 10th, 2005, 10:58 PM	#2370
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	35mm adaptor I got around to posting a adaptor screen technology (face plates), and image down scaling technology for direct sensor chip connection (tapers). Good stuff, hi res, high transmission options. The post is here: http://www.dvinfo.net/conf/showthread.php?s=&postid=261489#post261489 It has been a saga and a half, I went to look up another technology, I have over 160pages in my bookmark (which were open all at once), of nearly three hundred pages from just one company, that could have fitted into 10 pages (if the web pages had been designed in British style rather than European), this is getting too much for my health.

January 5th, 2005, 05:30 AM	#2356
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	<<<-- Originally posted by Obin Olson : We are now moving the image preview/display off the CPU and into the GPU of our motherboard...this will allow much more CPU time for other things like Histogram etc in realtime..not to mention the use of a lower-cost slower CPU if we want..this work is being done as I speak and we are still awaiting the 64bit card for capture of the native raw 10bit 1080p footage..any day now...64bit ...arrgg I hate waiting !! -->> Good to hear the progress, I have been talking to Rob privately about it several months ago, after the response I got to suggesting GPU programming here before. I passed on information on two or three GPU API systems to them. Do we have any idea how small, and the breakdown, of the processor load is for single chip 720p is now? I understand that upcoming (2005) new mobile versions of 0.9 micron AMD processors will also run at respectively low powered levels (can't rememeber but in the 20watt range, maybe sub 20). VIA will be releasing new 2+GHZ and dual core processors in a few months, hopefully in the same package. This how ever does not help old weak GPU and ethernet systems, but eventually they should be replaced. The power requirement is 3.5watt at 1 Ghz now (so a dual core will be 7w per 2Ghz). Still irrelavent for our purposes, but hopefully they get there act together. <<<-- Originally posted by Obin Olson : I am now thinking that I may just downsize the 1080P image after capture to 720p because workflow with 720 would be much easer then 1080p..thoughts anyone? -->>> I have often wondered about image problems at 720p or SD preview. There are certain electrical interferences that can mar an image that may go unoticed in the 720p and SD downsized images. These might rarely happen in reality, but it might go unoticed for a while.

January 6th, 2005, 02:55 PM	#2357
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	I have often wondered about image problems at 720p or SD preview. There are certain electrical interferences that can mar an image that may go unoticed in the 720p and SD downsized images. These might rarely happen in reality, but it might go unoticed for a while. What?! I don't understand that one Wayne...

January 6th, 2005, 04:06 PM	#2358
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Strong electromagnetic feild, electrical/power faults might put up a fine pattern of niose pattern wobbling through the image (or any other fault) that goes unoticed when averaged out in the downscaling of the preview. Also undesired fine visual detail, paterns that might cause certain strobing, staircasing due to the bayer filter pattern. Just thin possibility, otherwise there is nothing wrong with it.

January 6th, 2005, 04:11 PM	#2359
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	a good camera product should never have those problems, not if it's properly engineered/constructed.

January 7th, 2005, 02:57 AM	#2360
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Circumstances would be normally exceptional with a very well made and conditioned eng camera. Buit with simple cheap cameras attached to not so reliable PC's, who knows. But don't worry about it, Obin asked for comment, and those were the only two things that I knew of.

January 7th, 2005, 07:33 AM	#2361
Jason Rodriguez Trustee Join Date: Mar 2003 Location: Virginia Beach, VA Posts: 1,095	I understand Wayne, but that would only be a problem with poorly designed circuitry and power-sources/clocks with CCD's. Since many of these new chips are actually iSoc's, (imager system-on-a-chip), the A/D converter, etc. are already enclosed inside the CMOS sensor, and all you should be getting off the chip are digital signals, so everything's already "digitized". Unless the CMOS chip was manufactured badly, or somebody poorly designed the power ciruitry to the CMOS chip, you shouldn't get RF interference or banding.

January 7th, 2005, 11:33 AM	#2362
Joshua Starnes Major Player Join Date: Jun 2004 Location: Houston, TX Posts: 356	On the FX-1 front, Kaku Ito's (sp?) been running some tests and it looks like you can capture uncompressed 8-bit 1080i HD from the chip out through the component out and directly onto a hard drive, bypassing the MPEG-2 compression altogether.

January 7th, 2005, 09:03 PM	#2363
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	We are having good luck with the GPU use for display of pixel data..I have no update to play with yet but my programmer is happy...this is a good sign! ;) more later

January 8th, 2005, 07:30 AM	#2364
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	<<<-- Originally posted by Joshua Starnes : On the FX-1 front, Kaku Ito's (sp?) been running some tests and it looks like you can capture uncompressed 8-bit 1080i HD from the chip out through the component out and directly onto a hard drive, bypassing the MPEG-2 compression altogether. -->>> I posted a simular thing before, but I wasn't sure of the interpretation of the data, has it been conclusively proved with on screen tests, and is there a link? Seen my first FX1 today, if it could only do progressive 1080, 12 bit, HDV2/uncompressed, for $3000, I would go for it. ----- Other news, from what camcorder info is saying, it looks like Panasonic is using pixel shift on many of it's 3 chip cameras. Get this the new model has around 420K gross pixels (that is 420K/3 CCD's) or around 140K pixels per CCD. It must use the 3 chip pixel shift I have been talkign about, becauses it uses that to get 1.3MP stills (140K*9). It will be interesting to see what latitude and low light performance this camera will get with such relatively large pixels. Now, if the Juan mod was appliable, it would be ussuable in 1.3Mp doco's. I would still like to see a Altasens with three chip pixel shift ;) (Imax ;) ----- Jason, it is not as simple as that, but it doesn't matter, it is just a consideration.

January 9th, 2005, 12:57 AM	#2365
Joshua Starnes Major Player Join Date: Jun 2004 Location: Houston, TX Posts: 356	I posted a simular thing before, but I wasn't sure of the interpretation of the data, has it been conclusively proved with on screen tests, and is there a link? I don't think it has been conclusively proved, but the first round of tests seem to be good. There are screen tests, where you can download the (very large) uncompressed footage from his website. He has a link to it in his thread on the subject in the Alternative Imaging methods.

January 9th, 2005, 10:08 AM	#2366
Dan Diaconu Major Player Join Date: Dec 2004 Location: Vancouver, Canada Posts: 804	I hope this might be of interest to some: http://www.panavisionsvi.com/imagers_Quad.htm

January 10th, 2005, 06:24 AM	#2367
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Sorry, I didn't realise you meant there was a thread on it here.

January 10th, 2005, 06:33 AM	#2368
Valeriu Campan Major Player Join Date: Mar 2004 Location: Melbourne, Australia Posts: 223	You can find here the link for the component out recorded out of FX-1: http://www.dvinfo.net/conf/showthrea...threadid=37427

January 10th, 2005, 06:49 AM	#2369
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	The comments makes me wonder if the camera is outputing some pre corrected pixel shifted veresion of 1080p.