Home Made HD Cinema Cameras - Technical Discussion

Wayne Morellini · February 15th, 2005, 12:29 PM

Single chip 4:4:4 versus 3 chip compressed (is it 4:2:0 or 4:2:2), hmm. Yes RAW should still be better.

MiniDV has had the capability to record ten times as much for a while, it is just not implemented. So maybe they can improve it (but please for a full sized DVCPRO type tape). Is this the dvx100 or the 400 replacement, or cheap 900 alternative?

If they can match FX1 price I think I will think about buying, I have priority methords around the small chip problems. One thing that is a must is a conventional detachable lense mount (we would be going after an adaptor).

Has anybody got any news links to this?

Wayne.

Wayne Morellini · February 15th, 2005, 12:38 PM

<<<-- Originally posted by Jacques Mersereau :
DVCPRO HD is also fairly heavily compressed (though it looks good) at 6:1 and I am unsure as to the color -->>>

Just occured to me, that this 6:1 is using simular technology to the DV codec?? If so that will not be much better than 25mb/s Mpg2 codec (except perhaps in motion artifacts). If this is 6:1 Mpeg2, then it is reasonably good. I wonder is it variable comrpession for motion artifacts, as the PS2 can take a variable rate. Are we still talking about 8 bit?

Thanks

Wayne.

Joshua Starnes · February 15th, 2005, 12:42 PM

<<<-- Originally posted by Jacques Mersereau : DVCPRO HD is also fairly heavily compressed (though it looks good) at 6:1 and I am unsure as to the color sampling.
HDCAM is something like 3:1:0 (?) -->>>

I believe that DVCProHD samples at 4:2:2, the same as original HDCAM.

And while it doesn't record the same no. of lines that it displays (just like the Varicam and the CineAlta) the image you do get is still pretty darn good.

<<<-- Originally posted by Wayne Morellini : If they can match FX1 price I think I will think about buying, I have priority methords around the small chip problems. One thing that is a must is a conventional detachable lense mount -->>>

It's not the FX-1 price they're trying to compete with, it's the Z1, so think around $7K-$9K, which is still pretty good. Really, with the camera they're offering, $4K just isn't realisitic.

<<<-- Originally posted by Wayne Morellini : If this is 6:1 Mpeg2, then it is reasonably good. I wonder is it variable comrpession for motion artifacts, as the PS2 can take a variable rate. Are we still talking about 8 bit? -->>>

I'm not sure, but I don't think DVCProHD uses Mpeg2 to encode. DVCPro is its own standard with its own codec. Someone else would know better.

We are still talking about 8-bit, however. The DVCProHD standard (at the moment anyway, they could change it the way Sony changed HDCAM) is 8-bit 4:2:2.

Still, anyway you slice it, it's a hell of a lot better than DV or HDV, and after everything P+K and the Do-It-Yourselfer's have accomplished with lens adaptors, putting good glass on it shouldn't be a problem.

Wayne Morellini · February 15th, 2005, 12:59 PM

Thanks.

Actually, if they did do sub $5K it would be realistic once they sold the P2 packs for $5K a time ;(

So what sort of image format do they record? How about handling sports action, explosions, and dry forests with billions of long leaves whitish one side, green on he other flashing around in the breeze (like we have here) with things like varaible comrpession (for motion)?

Aaron Shaw · February 15th, 2005, 01:11 PM

You won't have the motion problems that HDV has. The format compresses each frame individually. It's the same idea as DV compression - just less compressed.

Wayne Morellini · February 15th, 2005, 01:50 PM

Yes, DV is around 6:1 compression, how is it less comrpessed than DV? I actually don't like the comrpession of DV either, so something twice as good would be good (even if just on the DV motion problems).

Thanks

Wayne.

Rob Lohman · February 16th, 2005, 06:46 AM

I think you mixed things a bit. DV is 5:1, the camera above seems
to be 6:1 (higher levels). The end quality is what counts though...

Wayne Morellini · February 16th, 2005, 07:24 AM

So seriously, we are saying that it is not an enhancement over the DV condec, just more resolution? Lower compression for 720p would be good.

Aaron Shaw · February 16th, 2005, 11:38 AM

It is lower compression. It's a 4:2:2 codec. It would be impossible, I would think, to have higher compression on a codec with more information. Who knows though.

In any case, there's twice as much color information as in DV which is a very, very good thing.

Wayne Morellini · February 18th, 2005, 10:54 AM

I understand that this is the 960*1080 one, that puts a bit of an unfortunate sting in the tail.

Wayne.

Wayne Morellini · February 19th, 2005, 05:16 AM

I have been reading of an 300GB drive that spanks the Raptor drive in performance.

http://www.storagereview.com/php/ben...bench_sort.php

http://www.storagereview.com/article...8D300L0_2.html

I have also figured a new scheme that should help the capture schemes maximise drive performance and get more.

As we know the outer tracks record faster than the inner tracks, as that is the bottle neck the max throughput is number of drives * inner track performance. But if you start recording across all the drives at once, but recording from the inner track on half the drives, and the outer track on the other half, you get a higher average performance in between the inner and outer track rating (which can be quiet high). So you can easily increase the data rate of a 200MB/s system, and do wonders on a two drive systems.

Now the other things is that the benchmark used in some of these max datarate tests looks like a read benchmark, rather than testing write performance. What goes here?

Wayne Morellini · February 19th, 2005, 05:30 AM

http://www.gizmag.com.au/go/2539/

Fuel cells are the way to go for us, but have been limited and expensive, this looks like a real break through. I don't know their release date, so maybe it will miss the first round of fuel cells (really soon).

I also remember, at the begining of the year (tomshardware news), one company (I think mentioned in the origional Phyorg article I posted here) has shown off a prototype of their 1 terabyte optical drive.

Wayne Morellini · February 19th, 2005, 09:40 AM

New 320GB drive from Western Digital, fast.

http://www.tomshardware.com/storage/...3200jb-05.html

Rai Orz · February 22nd, 2005, 03:38 AM

Wayne:
In a other thread you wrote this:<<<--Looking at the Lupa 4000 as well, interesting.-->>>

We use now the fillfactory 2/3" IBIS5a sensor for our DRAKE camera. We capture 1280x720 at 24fps with it.
One goal for us is (at the end of this year) a real full size 35mm sensor camera with min. 1920x1080. So lets play with the LUPA 4000 data sheet:
max. 15fps!! To slow. No, its a 4M/Pixel ship. At 1920X1080 it runs at round 27fps, global shutter. At 24fps we can capture some more lines (usefull to show outside areas at viewfinders). Is more light sensity, 4 times higher than IBIS5a. But the best thing is the size: It comes with a realy full 35mm size. A sample sensor (not a camera) cost under $ 1000,-. Bad news is: At this time, they produced only mono versions. Okay, with RGB prism its ready for a 3Chip camera. (3x 2Mpixel = 6Mpixel) . This means at 24fps 8Bit = 150MB/Sec, 10Bit = 187MB/Sec, 12Bit = 234MB/Sec.
Well, i see a diffent way: Lets call it a electronic bayer filter. We have 3 Sensors, but we do a software pixel mix, same way optical bayer filter work. At the end we have only 2M/pixel, same as a single color sensor. With those camera we can every time switch to 6M/pixel (or something between it). A feature for the future.

Wayne Morellini · February 23rd, 2005, 05:40 AM

Rai

I wasn't really getting at anything, with that statement, just stating an interesting chip you mentioned. I have done most of the same calculations, and yes, at half res you are cooking for 1080p. I would be interested in a real picture/sensitivity performance comparison between it and Altasens (beyond the aperture and rolling shutter issues).

If you really are going to combine 3 chip to do artificial bayer, then remember my suggestion of getting better accuracy out of bayer (and also as a compression method) by recording the difference in the bayer assumption to the real value as well as bayer. (the assumption being the non tested primary colours at each location, a sort of upscaling). Still, you could also bin to lower resolution, use half, or third, shift to get hi-res, and still get lower than Bayer rate.

I have been working on techniques that could be used to apply a bayer pattern on a monochrome sensor. But a simpler one is, on a DOF adaptor you have a projection screen, on one side you could put a bayer pattern filter for single chip. Problem is thermal/shock shifting alignment, and that you are already using a 35mm chip.

My interest in 3 chip is:

A) With true pixel shift, getting better picture and image/light pickup than Bayer for same datarate.
b) Getting higher resolution for post 1080p cinema picture and Imax type theatres.

This leads to reduced cost and increased performance over 6-18mpixel video sensor. For top end camera, good single chip cameras being mid end product, and something like Micron etc etc being lowend. So I advocate these three levels for different consumer levels out there. I am currently looking into wide aperature problem solutions.

February 16th, 2005, 06:46 AM	#322
Rob Lohman RED Code Chef Join Date: Oct 2001 Location: Holland Posts: 12,514	I think you mixed things a bit. DV is 5:1, the camera above seems to be 6:1 (higher levels). The end quality is what counts though... __________________ Rob Lohman, visuar@iname.com DV Info Wrangler & RED Code Chef Join the DV Challenge \| Lady X Search DVinfo.net for quick answers \| Buy from the best: DVinfo.net sponsors

February 19th, 2005, 05:16 AM	#326
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	New fast big hard drive & capture performance improvement technique I have been reading of an 300GB drive that spanks the Raptor drive in performance. http://www.storagereview.com/php/ben...bench_sort.php http://www.storagereview.com/article...8D300L0_2.html I have also figured a new scheme that should help the capture schemes maximise drive performance and get more. As we know the outer tracks record faster than the inner tracks, as that is the bottle neck the max throughput is number of drives * inner track performance. But if you start recording across all the drives at once, but recording from the inner track on half the drives, and the outer track on the other half, you get a higher average performance in between the inner and outer track rating (which can be quiet high). So you can easily increase the data rate of a 200MB/s system, and do wonders on a two drive systems. Now the other things is that the benchmark used in some of these max datarate tests looks like a read benchmark, rather than testing write performance. What goes here?

February 19th, 2005, 05:30 AM	#327
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	New improved fuel cell, and Terabyte Optical disk. http://www.gizmag.com.au/go/2539/ Fuel cells are the way to go for us, but have been limited and expensive, this looks like a real break through. I don't know their release date, so maybe it will miss the first round of fuel cells (really soon). I also remember, at the begining of the year (tomshardware news), one company (I think mentioned in the origional Phyorg article I posted here) has shown off a prototype of their 1 terabyte optical drive.

February 15th, 2005, 12:29 PM	#316
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Single chip 4:4:4 versus 3 chip compressed (is it 4:2:0 or 4:2:2), hmm. Yes RAW should still be better. MiniDV has had the capability to record ten times as much for a while, it is just not implemented. So maybe they can improve it (but please for a full sized DVCPRO type tape). Is this the dvx100 or the 400 replacement, or cheap 900 alternative? If they can match FX1 price I think I will think about buying, I have priority methords around the small chip problems. One thing that is a must is a conventional detachable lense mount (we would be going after an adaptor). Has anybody got any news links to this? Wayne.

February 15th, 2005, 12:38 PM	#317
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	<<<-- Originally posted by Jacques Mersereau : DVCPRO HD is also fairly heavily compressed (though it looks good) at 6:1 and I am unsure as to the color -->>> Just occured to me, that this 6:1 is using simular technology to the DV codec?? If so that will not be much better than 25mb/s Mpg2 codec (except perhaps in motion artifacts). If this is 6:1 Mpeg2, then it is reasonably good. I wonder is it variable comrpession for motion artifacts, as the PS2 can take a variable rate. Are we still talking about 8 bit? Thanks Wayne.

February 15th, 2005, 12:42 PM	#318
Joshua Starnes Major Player Join Date: Jun 2004 Location: Houston, TX Posts: 356	<<<-- Originally posted by Jacques Mersereau : DVCPRO HD is also fairly heavily compressed (though it looks good) at 6:1 and I am unsure as to the color sampling. HDCAM is something like 3:1:0 (?) -->>> I believe that DVCProHD samples at 4:2:2, the same as original HDCAM. And while it doesn't record the same no. of lines that it displays (just like the Varicam and the CineAlta) the image you do get is still pretty darn good. <<<-- Originally posted by Wayne Morellini : If they can match FX1 price I think I will think about buying, I have priority methords around the small chip problems. One thing that is a must is a conventional detachable lense mount -->>> It's not the FX-1 price they're trying to compete with, it's the Z1, so think around $7K-$9K, which is still pretty good. Really, with the camera they're offering, $4K just isn't realisitic. <<<-- Originally posted by Wayne Morellini : If this is 6:1 Mpeg2, then it is reasonably good. I wonder is it variable comrpession for motion artifacts, as the PS2 can take a variable rate. Are we still talking about 8 bit? -->>> I'm not sure, but I don't think DVCProHD uses Mpeg2 to encode. DVCPro is its own standard with its own codec. Someone else would know better. We are still talking about 8-bit, however. The DVCProHD standard (at the moment anyway, they could change it the way Sony changed HDCAM) is 8-bit 4:2:2. Still, anyway you slice it, it's a hell of a lot better than DV or HDV, and after everything P+K and the Do-It-Yourselfer's have accomplished with lens adaptors, putting good glass on it shouldn't be a problem.

February 15th, 2005, 12:59 PM	#319
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Thanks. Actually, if they did do sub $5K it would be realistic once they sold the P2 packs for $5K a time ;( So what sort of image format do they record? How about handling sports action, explosions, and dry forests with billions of long leaves whitish one side, green on he other flashing around in the breeze (like we have here) with things like varaible comrpession (for motion)?

February 15th, 2005, 01:11 PM	#320
Aaron Shaw Major Player Join Date: Jul 2004 Posts: 587	You won't have the motion problems that HDV has. The format compresses each frame individually. It's the same idea as DV compression - just less compressed.

February 15th, 2005, 01:50 PM	#321
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Yes, DV is around 6:1 compression, how is it less comrpessed than DV? I actually don't like the comrpession of DV either, so something twice as good would be good (even if just on the DV motion problems). Thanks Wayne.

February 16th, 2005, 07:24 AM	#323
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	So seriously, we are saying that it is not an enhancement over the DV condec, just more resolution? Lower compression for 720p would be good.

February 16th, 2005, 11:38 AM	#324
Aaron Shaw Major Player Join Date: Jul 2004 Posts: 587	It is lower compression. It's a 4:2:2 codec. It would be impossible, I would think, to have higher compression on a codec with more information. Who knows though. In any case, there's twice as much color information as in DV which is a very, very good thing.

February 18th, 2005, 10:54 AM	#325
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	I understand that this is the 960*1080 one, that puts a bit of an unfortunate sting in the tail. Wayne.

February 19th, 2005, 09:40 AM	#328
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	New 320GB drive from Western Digital, fast. http://www.tomshardware.com/storage/...3200jb-05.html

February 22nd, 2005, 03:38 AM	#329
Rai Orz Regular Crew Join Date: Jun 2004 Location: Germany Posts: 136	Wayne: In a other thread you wrote this:<<<--Looking at the Lupa 4000 as well, interesting.-->>> We use now the fillfactory 2/3" IBIS5a sensor for our DRAKE camera. We capture 1280x720 at 24fps with it. One goal for us is (at the end of this year) a real full size 35mm sensor camera with min. 1920x1080. So lets play with the LUPA 4000 data sheet: max. 15fps!! To slow. No, its a 4M/Pixel ship. At 1920X1080 it runs at round 27fps, global shutter. At 24fps we can capture some more lines (usefull to show outside areas at viewfinders). Is more light sensity, 4 times higher than IBIS5a. But the best thing is the size: It comes with a realy full 35mm size. A sample sensor (not a camera) cost under $ 1000,-. Bad news is: At this time, they produced only mono versions. Okay, with RGB prism its ready for a 3Chip camera. (3x 2Mpixel = 6Mpixel) . This means at 24fps 8Bit = 150MB/Sec, 10Bit = 187MB/Sec, 12Bit = 234MB/Sec. Well, i see a diffent way: Lets call it a electronic bayer filter. We have 3 Sensors, but we do a software pixel mix, same way optical bayer filter work. At the end we have only 2M/pixel, same as a single color sensor. With those camera we can every time switch to 6M/pixel (or something between it). A feature for the future.

February 23rd, 2005, 05:40 AM	#330
Wayne Morellini Inner Circle Join Date: May 2003 Location: Australia Posts: 2,762	Rai I wasn't really getting at anything, with that statement, just stating an interesting chip you mentioned. I have done most of the same calculations, and yes, at half res you are cooking for 1080p. I would be interested in a real picture/sensitivity performance comparison between it and Altasens (beyond the aperture and rolling shutter issues). If you really are going to combine 3 chip to do artificial bayer, then remember my suggestion of getting better accuracy out of bayer (and also as a compression method) by recording the difference in the bayer assumption to the real value as well as bayer. (the assumption being the non tested primary colours at each location, a sort of upscaling). Still, you could also bin to lower resolution, use half, or third, shift to get hi-res, and still get lower than Bayer rate. I have been working on techniques that could be used to apply a bayer pattern on a monochrome sensor. But a simpler one is, on a DOF adaptor you have a projection screen, on one side you could put a bayer pattern filter for single chip. Problem is thermal/shock shifting alignment, and that you are already using a 35mm chip. My interest in 3 chip is: A) With true pixel shift, getting better picture and image/light pickup than Bayer for same datarate. b) Getting higher resolution for post 1080p cinema picture and Imax type theatres. This leads to reduced cost and increased performance over 6-18mpixel video sensor. For top end camera, good single chip cameras being mid end product, and something like Micron etc etc being lowend. So I advocate these three levels for different consumer levels out there. I am currently looking into wide aperature problem solutions.