4:4:4 10bit single CMOS HD project

Wayne Morellini · December 12th, 2004, 12:31 AM

IBIS already does 450Hz sensor (under $1000) but I am interested in cheap sensor with a bit extra speed for 120fps speed (for special effect ts buffered to main board memory).

The pricing seems to be great, nearly the same as the original pricing (I think it was $2795), down from those $4K+ pricing.

<<<-- Originally posted by Ben Syverson :Wayne, what exactly is your idea for better than 2:1? Remember that this is 2:1 in terms of the Bayer, not the final RGB. If you can get better than 2:1 on a Bayer image, I'd love to hear how.... -->

I have a link to it in the technical thread above, I have been discussing it with David there. The principle is simple, transmission and tape comrpession format have error correction overheads, and compress over one to x frames, all that limits comrpession ratio. So I suggested (as we had upto 8GB on board (now 32MB on the head) that we could do file like compression accross the lot. I posted a link to the best file compression product with image comrpession mode before. We assume that video lossless codecs are the best, but how many frames do they compress across, one, maybe 6? For us we can go beyond that (if we have the processing power), as I said go to the link above and have look.

Thanks

Wayne.

Wayne Morellini · December 12th, 2004, 12:44 AM

I suppose, if you could set the Alatsens to 120fps 720p mode you could get it, but you have to use a special decoder for the distorted bayer pattern. For this sort of effect, that ussually looks weird, that is acceptable anyway. But if you are going to show at 1080 60fps (which some people want) you will want 120fps to seamlessly slowmo that on screen "kiss", or head turn etc. I suppose we don't need hat quality anyway, we can digitally take the 60fps and use it at 24fps on 60Hz screen.

Of note: In recent months I remember reading somewhere of some rpocess to get one photon to charge two electrons. That is equivalent to a QE of 200%.

If this new IBIS can cu it as good as a Micron, hen hopefully they use pixel shift ;).

Now does this 3 chip camera use microlense, that would give us trouble when using large aperature optics, or does it just use the new well method?

Régine Weinberg · December 13th, 2004, 02:30 AM

Microlinses are much more a pain
with wide angle, proof
on a Canon 16 M still cam, wide angle is shit
on a Kodak 14 M there are no lenses,
they are other problems, moire, anti alaising
but hell sharp even with wide angle
only you have to tweak each pic
on a PC with Kodakt software.

Wayne Morellini · December 14th, 2004, 12:34 AM

By the way, I have found out that IBIS5a sensor in around 100 euro whole sale. So cameras on it could be quiet cheap.

I also was over at the HD for indies site and saw a news listing for a 8 drive Sata RAID array that would do 500Mb/s, so I imagine 4 drives should be able to do 250MB's. So a three chip Alatsens, is 150+ MB's, a three chip Ibis would be 75+ MB's, both with pixel shift. So does that mean you can times the res by 4, which means it beats Bayer 8 and 4 mpixel in data requirements, virtually no light loss? Sorry, just waking up from dream land, ahh 3 chip, pixel shift, mmm, sleep again....

Juan M. M. Fiebelkorn · December 14th, 2004, 02:50 AM

I don't believe microlenses could become a problem even using a wide angle lens which was prepared for a resolution of around 160 line pairs per millimeter.
We are just using 2 Mpix sensors not 14 Mpix ....

Wayne Morellini · December 15th, 2004, 10:14 AM

I posted an update to the technical thread by mistake, to do wih fast laptop drives benchmarks, 80GB 1.8inch, and Windows XP PDA.

I would like to ask a question though about pixel shift.

It has occured to me in the distant past, to shift two of the primary colours, rather than just the green value that the DVX100 does. This basically gives 3 times the resoltuion in each direction, but it occurs to me that I haven't heard of it being used like that, and there maybe a good reason for that. So would it work, or is there some problem with doing that?

Thanks.

Wayne.

Juan M. M. Fiebelkorn · December 15th, 2004, 10:34 PM

I guess shifting just the green is the optimum.That way green pixels are getting the intensity values that fall on the in-between spaces of the R/B sensor pixels, Giving you something like a 100% fillfactor (with a three sensors covering the same area)
I can't see what would be the advantage of shifting two of them if you don't have empty areas to cover .....

Wayne Morellini · December 15th, 2004, 11:23 PM

But are there any diadvantages to shifting two, like do we loose effectiveness in defining colour, pixels or brightness. I am interested in this because it allows a progressive SD 3 chip to do 1080p.

Another question what is the most sensitive sensors/camera for somehing like this, and what are the best value cheap ones.

I think Pana may have used it in the DVX100 not for future HD, as some suppose but to get rid or artifacting from inter pixel spaces etc.

Juan M. M. Fiebelkorn · December 16th, 2004, 12:11 AM

I think you won't be getting anybetter resolution, just that.
I say this because, with the green sensor shifting you are covering all the empty spaces available.....going beyond that you will be just superimposing pixels again, Am I clear now?

I don't really know, what I can tell you is the DVX100 is rated at around 640 ASA and I can asure that is true.My experience with it during 2 months confirmed it (you know going shooting through dark streets and being able to see more through the camera than what your naked eyes can see is an amazing thing ,mostly when we are not talking about a night vision device or a night-shot handycam :) )

There isn't a thing like "artifacting from inter pixel spaces " (in fact something related in some way with that is what we call "aliasing").
If it existed monochrome cameras would be full of them....which is not the actual case...

Wayne Morellini · December 16th, 2004, 01:16 PM

What I meant was, when there is incomplete fill between a pixel, and it shows up in a image (because not enough pixels are used to missaliagn with the format) in things like edges that move from pixel to pixel and blink out temporaily as they do.

Yes, on those camcorder info tests there was a vast differnce between the Sony's and the Pana in low light, but in other tests, when they spent time in manual control, they got the near the same lighting levels. I see the raw tests shows a lot of range, compared to normal, I wonder if the camera actually can go lower in low light.

The Pixel shift gets it's resolution from being able to mathematically sample the difference from the overlapping pixels forming artificial sub pixel quads, so you can mathematically sumise what the diffrence means on each of these quads, to form 4 new pixels, instead of interpixel spacing, giving the extra resolution. That is what I meant, that the reason 3 missaligned pixels are not used, might be some fundamental break down in quality of the samples etc.

Thanks

Bye.

Obin Olson · December 16th, 2004, 10:32 PM

Well well I crawl out from under a rock just in time to get blinded by the sun:

"We have made significant progress with Altasense camera. It will have powerful FPGA with power PC processor inside. It will also have 32 Mega bytes of image buffer memory. The interface will be Giga-Ethernet. We are aiming for 60 fps with loss-less compression. The good news is that components cost will be low. We hope to provide low introductory price, even lower than what we thought earlier. The negative news is that March is the new date for release."

but that's OK after 1 week of having UPS "delay" my package it's finally in the hands of ouraprogrammer for final work..we will be using the 3300rgb camera for now and upgrade to the ALtasense VERY soon..I wish I had more "info" but all is quiet before the storm right?

keep up the merry mood.

Steve..EPix 64bit word?

Régine Weinberg · December 18th, 2004, 08:19 AM

Dear Obin

sonds amazing the Alta news really !!!

Dear Wayne: In a three CCD cam you have a R, G, B ccd stupid but, take it this way Pana is shifting green horizontal and vertical
Canon only horizontal, Canon CCD's are smaller Canon want's to boost resolution. Why only pana does it, nobody knows, rumors telling me Pana wanted to do other things maybe. Shifting all three sounds crazy but why not - crazy does not mean can't be done.

Look at this: http://www.natemc.com/tech001b.php
so we kan convert the 1.8" Toshiba dis to a 0,1,5 or 10 arry

and to all sorry but to have a huge LCD screen as Drake does it beside the cam I'm not callin a viewfinder

Ronald
I've been some day's of to
NORWAY SAILING, NO EMAIL
BUT LOT OF FUN

Obin Olson · December 18th, 2004, 09:09 AM

I will have a 8inch touch screen with playback and live preview..later on I will ad a viewfinder if I can ever find a good microdisplay that works from a VGA input

Régine Weinberg · December 18th, 2004, 09:47 AM

Hm
which resolution can you have with the 8 inch touch screen ?

Régine Weinberg · December 18th, 2004, 10:02 AM

a raid with the Toshiba, converted to Sata look there please
http://www.xpcgear.com/hrr1820a.html

can do the bandwith for the altasens, guess thats the way Kinetta will have his 10 disk's. To do all this with FGPA would be a nightmare.

found once a small portable fuelcell doing 320 W, looking for a small mobo with PCIx to have the mobo with the disks in something shaped like a 35mm film magazine could be

anyway the cam has not to have the look as a Video cam if it looks like a russion 16mm or 35mm a prototype is quite easy to build and

going wild, crazy

all you know, but it's only 5 pm here. no red wine only tea and cake, but crazy Bolex had a rotating prism 20/80 % that was the best and Arri did the same to have a viewfinder.

To have a mechanical shutter is not bad with cmos, we get rid of all rolling shutter problems, Vance did it on the Vance cam, it's on his page.

we would have a 90-10 prism and something as a night vision device, to get reely cheep, that would be a B%W viewfinder exact as the lens . We only need a small screen to have some cam parameters and it would be done. a decent high res electronic viewfinder is not a cheapo a prism doing 90/10 can be done , it's not an all digital approch but why should it be ? Id should be a HD indie cam, and I'm used to an Arri like viewfinder all life, thousend's too - to change that will not be wise and B&W would be okay sometimes even much better

As a TFT or LCD like Drake is not camera man like sorry.

December 18th, 2004, 10:02 AM	#2325
Régine Weinberg Major Player Join Date: Jan 2004 Location: Bordeaux, going to Bangkok, 2011 Posts: 232	1.8 inch harddisk a raid with the Toshiba, converted to Sata look there please http://www.xpcgear.com/hrr1820a.html can do the bandwith for the altasens, guess thats the way Kinetta will have his 10 disk's. To do all this with FGPA would be a nightmare. found once a small portable fuelcell doing 320 W, looking for a small mobo with PCIx to have the mobo with the disks in something shaped like a 35mm film magazine could be anyway the cam has not to have the look as a Video cam if it looks like a russion 16mm or 35mm a prototype is quite easy to build and going wild, crazy all you know, but it's only 5 pm here. no red wine only tea and cake, but crazy Bolex had a rotating prism 20/80 % that was the best and Arri did the same to have a viewfinder. To have a mechanical shutter is not bad with cmos, we get rid of all rolling shutter problems, Vance did it on the Vance cam, it's on his page. we would have a 90-10 prism and something as a night vision device, to get reely cheep, that would be a B%W viewfinder exact as the lens . We only need a small screen to have some cam parameters and it would be done. a decent high res electronic viewfinder is not a cheapo a prism doing 90/10 can be done , it's not an all digital approch but why should it be ? Id should be a HD indie cam, and I'm used to an Arri like viewfinder all life, thousend's too - to change that will not be wise and B&W would be okay sometimes even much better As a TFT or LCD like Drake is not camera man like sorry.

December 12th, 2004, 12:31 AM	#2311
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	IBIS already does 450Hz sensor (under $1000) but I am interested in cheap sensor with a bit extra speed for 120fps speed (for special effect ts buffered to main board memory). The pricing seems to be great, nearly the same as the original pricing (I think it was $2795), down from those $4K+ pricing. <<<-- Originally posted by Ben Syverson :Wayne, what exactly is your idea for better than 2:1? Remember that this is 2:1 in terms of the Bayer, not the final RGB. If you can get better than 2:1 on a Bayer image, I'd love to hear how.... --> I have a link to it in the technical thread above, I have been discussing it with David there. The principle is simple, transmission and tape comrpession format have error correction overheads, and compress over one to x frames, all that limits comrpession ratio. So I suggested (as we had upto 8GB on board (now 32MB on the head) that we could do file like compression accross the lot. I posted a link to the best file compression product with image comrpession mode before. We assume that video lossless codecs are the best, but how many frames do they compress across, one, maybe 6? For us we can go beyond that (if we have the processing power), as I said go to the link above and have look. Thanks Wayne.

December 12th, 2004, 12:44 AM	#2312
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I suppose, if you could set the Alatsens to 120fps 720p mode you could get it, but you have to use a special decoder for the distorted bayer pattern. For this sort of effect, that ussually looks weird, that is acceptable anyway. But if you are going to show at 1080 60fps (which some people want) you will want 120fps to seamlessly slowmo that on screen "kiss", or head turn etc. I suppose we don't need hat quality anyway, we can digitally take the 60fps and use it at 24fps on 60Hz screen. Of note: In recent months I remember reading somewhere of some rpocess to get one photon to charge two electrons. That is equivalent to a QE of 200%. If this new IBIS can cu it as good as a Micron, hen hopefully they use pixel shift ;). Now does this 3 chip camera use microlense, that would give us trouble when using large aperature optics, or does it just use the new well method?

December 13th, 2004, 02:30 AM	#2313
Régine Weinberg Major Player Join Date: Jan 2004 Location: Bordeaux, going to Bangkok, 2011 Posts: 232	Microlinses are much more a pain with wide angle, proof on a Canon 16 M still cam, wide angle is shit on a Kodak 14 M there are no lenses, they are other problems, moire, anti alaising but hell sharp even with wide angle only you have to tweak each pic on a PC with Kodakt software.

December 14th, 2004, 12:34 AM	#2314
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	By the way, I have found out that IBIS5a sensor in around 100 euro whole sale. So cameras on it could be quiet cheap. I also was over at the HD for indies site and saw a news listing for a 8 drive Sata RAID array that would do 500Mb/s, so I imagine 4 drives should be able to do 250MB's. So a three chip Alatsens, is 150+ MB's, a three chip Ibis would be 75+ MB's, both with pixel shift. So does that mean you can times the res by 4, which means it beats Bayer 8 and 4 mpixel in data requirements, virtually no light loss? Sorry, just waking up from dream land, ahh 3 chip, pixel shift, mmm, sleep again....

December 14th, 2004, 02:50 AM	#2315
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	I don't believe microlenses could become a problem even using a wide angle lens which was prepared for a resolution of around 160 line pairs per millimeter. We are just using 2 Mpix sensors not 14 Mpix ....

December 15th, 2004, 10:14 AM	#2316
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I posted an update to the technical thread by mistake, to do wih fast laptop drives benchmarks, 80GB 1.8inch, and Windows XP PDA. I would like to ask a question though about pixel shift. It has occured to me in the distant past, to shift two of the primary colours, rather than just the green value that the DVX100 does. This basically gives 3 times the resoltuion in each direction, but it occurs to me that I haven't heard of it being used like that, and there maybe a good reason for that. So would it work, or is there some problem with doing that? Thanks. Wayne.

December 15th, 2004, 10:34 PM	#2317
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	I guess shifting just the green is the optimum.That way green pixels are getting the intensity values that fall on the in-between spaces of the R/B sensor pixels, Giving you something like a 100% fillfactor (with a three sensors covering the same area) I can't see what would be the advantage of shifting two of them if you don't have empty areas to cover .....

December 15th, 2004, 11:23 PM	#2318
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	But are there any diadvantages to shifting two, like do we loose effectiveness in defining colour, pixels or brightness. I am interested in this because it allows a progressive SD 3 chip to do 1080p. Another question what is the most sensitive sensors/camera for somehing like this, and what are the best value cheap ones. I think Pana may have used it in the DVX100 not for future HD, as some suppose but to get rid or artifacting from inter pixel spaces etc.

December 16th, 2004, 12:11 AM	#2319
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	I think you won't be getting anybetter resolution, just that. I say this because, with the green sensor shifting you are covering all the empty spaces available.....going beyond that you will be just superimposing pixels again, Am I clear now? I don't really know, what I can tell you is the DVX100 is rated at around 640 ASA and I can asure that is true.My experience with it during 2 months confirmed it (you know going shooting through dark streets and being able to see more through the camera than what your naked eyes can see is an amazing thing ,mostly when we are not talking about a night vision device or a night-shot handycam :) ) There isn't a thing like "artifacting from inter pixel spaces " (in fact something related in some way with that is what we call "aliasing"). If it existed monochrome cameras would be full of them....which is not the actual case...

December 16th, 2004, 01:16 PM	#2320
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	What I meant was, when there is incomplete fill between a pixel, and it shows up in a image (because not enough pixels are used to missaliagn with the format) in things like edges that move from pixel to pixel and blink out temporaily as they do. Yes, on those camcorder info tests there was a vast differnce between the Sony's and the Pana in low light, but in other tests, when they spent time in manual control, they got the near the same lighting levels. I see the raw tests shows a lot of range, compared to normal, I wonder if the camera actually can go lower in low light. The Pixel shift gets it's resolution from being able to mathematically sample the difference from the overlapping pixels forming artificial sub pixel quads, so you can mathematically sumise what the diffrence means on each of these quads, to form 4 new pixels, instead of interpixel spacing, giving the extra resolution. That is what I meant, that the reason 3 missaligned pixels are not used, might be some fundamental break down in quality of the samples etc. Thanks Bye.

December 16th, 2004, 10:32 PM	#2321
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	Well well I crawl out from under a rock just in time to get blinded by the sun: "We have made significant progress with Altasense camera. It will have powerful FPGA with power PC processor inside. It will also have 32 Mega bytes of image buffer memory. The interface will be Giga-Ethernet. We are aiming for 60 fps with loss-less compression. The good news is that components cost will be low. We hope to provide low introductory price, even lower than what we thought earlier. The negative news is that March is the new date for release." but that's OK after 1 week of having UPS "delay" my package it's finally in the hands of ouraprogrammer for final work..we will be using the 3300rgb camera for now and upgrade to the ALtasense VERY soon..I wish I had more "info" but all is quiet before the storm right? keep up the merry mood. Steve..EPix 64bit word?

December 18th, 2004, 08:19 AM	#2322
Régine Weinberg Major Player Join Date: Jan 2004 Location: Bordeaux, going to Bangkok, 2011 Posts: 232	Dear Obin sonds amazing the Alta news really !!! Dear Wayne: In a three CCD cam you have a R, G, B ccd stupid but, take it this way Pana is shifting green horizontal and vertical Canon only horizontal, Canon CCD's are smaller Canon want's to boost resolution. Why only pana does it, nobody knows, rumors telling me Pana wanted to do other things maybe. Shifting all three sounds crazy but why not - crazy does not mean can't be done. Look at this: http://www.natemc.com/tech001b.php so we kan convert the 1.8" Toshiba dis to a 0,1,5 or 10 arry and to all sorry but to have a huge LCD screen as Drake does it beside the cam I'm not callin a viewfinder Ronald I've been some day's of to NORWAY SAILING, NO EMAIL BUT LOT OF FUN

December 18th, 2004, 09:09 AM	#2323
Obin Olson Trustee Join Date: Jan 2003 Location: Wilmington NC Posts: 1,414	I will have a 8inch touch screen with playback and live preview..later on I will ad a viewfinder if I can ever find a good microdisplay that works from a VGA input

December 18th, 2004, 09:47 AM	#2324
Régine Weinberg Major Player Join Date: Jan 2004 Location: Bordeaux, going to Bangkok, 2011 Posts: 232	Hm which resolution can you have with the 8 inch touch screen ?