Reel-Stream Sold

John Papadopoulos · December 6th, 2007, 07:49 AM

Andromeda was 4:2:0 of course since there were just enough pixels to cover that. But all single sensor devices (dslrs, digital cinema etc) are 4:2:0 and the single Sony CMOS line is a lot worse than 4:2:0. The HVX is also a true 4:2:0 in 1080 mode, 640x540 true chroma, less than what a 4:2:0 1440x1080 format has actually.

The difference in uncompressed solutions marked as 4:4:4 is that the chroma is debayered or pixel shifted, and no color sampling is applied after that. So it is very smooth and there are no chroma artifacts like the ones you would get in a 4:2:0 video format. That doesn't increase actual resolution though, it just produces better results, free from artifacts. It's not 4:4:4 though. But I can relate to manufacturers who say their equipment is 4:4:4. When people hear 4:2:0 they think of chroma sampling artifacts they see on their HDV or DV cameras, not true chroma resolution. DSLR and digital cinema cameras does not have that problem, so it wouldn't be fair to call them 4:2:0.

The HVX will also look better than higher resolution 1080 HDV chroma because of compression but that has nothing do do with the chroma pixel count.

IMHO, the weaks links are sensor technloogy (CCD vs CMOS), then compression (less vs more), then resolution (more vs less).

The HVX will not look good on paper compared to Sony CMOS solutions, but the actual result will usually be better in everything except resolution. The CCDs provides motion quality and the compression scheme preserves that and also provides better chroma quality compared to HDV.

2K from Hydra could not possibly be better than about 550-620 lines per picture height in 16:9. But it would be uncompressed which is more important than resolution IMHO.

14stops of latitude is impossible in the sensors we have today, especially in the older sensors used in these ccd cameras. A little more than 10 is closer to the truth and the highest quality CCD and CMOS sensors will do 11.5 if you have the light (not every shooting scenario). CMOS sensors should be grateful when they achieve more than 7-8 stops in real life use because they need much more light in order to get their dynamic range spec compared to CCD. How often do you have 2000lux available? And if you can contain the highlight of a bright window in your interior shot, how much detail and how much noise is in your interior shadows?

The Andromeda used pixel shifting so it's not comparable to bayer. In bayer you have ideal positioning of your pixels, in a pixel shifted solution you have to pray for manufacturing parameters than are completely unpredictable (sub pixel precision sensor/ccd block assembly). In bayer you can get 800lines on a 1000pixel sensor. In a 500pixel pixel shifted solution you should be very happy if you get 550-600. If any company wants to claim that they can mount ccds on a prism with 0.1 microns precision I would be happy to prove the equipment required for that does not exist.

From personal experience, the horizontal resolution of a Sony FX1, a camera than only shifts horizontally is very unpredictable. I have seen from 550 to 800 on a chart. That is about 970 to 1400 pixels, which is (surprise!) close to 960 (sensor resolution or 0% luma advantge) and 1440 (format resolution or 50% luma advantage). But an FX1 that is better than 600-650 lines horizontally is very rare. More than 20% of advantage over sensor resolution should not be expected.

December 6th, 2007, 07:49 AM	#16
John Papadopoulos Major Player Join Date: Nov 2007 Location: Athens Greece Posts: 336	Andromeda was 4:2:0 of course since there were just enough pixels to cover that. But all single sensor devices (dslrs, digital cinema etc) are 4:2:0 and the single Sony CMOS line is a lot worse than 4:2:0. The HVX is also a true 4:2:0 in 1080 mode, 640x540 true chroma, less than what a 4:2:0 1440x1080 format has actually. The difference in uncompressed solutions marked as 4:4:4 is that the chroma is debayered or pixel shifted, and no color sampling is applied after that. So it is very smooth and there are no chroma artifacts like the ones you would get in a 4:2:0 video format. That doesn't increase actual resolution though, it just produces better results, free from artifacts. It's not 4:4:4 though. But I can relate to manufacturers who say their equipment is 4:4:4. When people hear 4:2:0 they think of chroma sampling artifacts they see on their HDV or DV cameras, not true chroma resolution. DSLR and digital cinema cameras does not have that problem, so it wouldn't be fair to call them 4:2:0. The HVX will also look better than higher resolution 1080 HDV chroma because of compression but that has nothing do do with the chroma pixel count. IMHO, the weaks links are sensor technloogy (CCD vs CMOS), then compression (less vs more), then resolution (more vs less). The HVX will not look good on paper compared to Sony CMOS solutions, but the actual result will usually be better in everything except resolution. The CCDs provides motion quality and the compression scheme preserves that and also provides better chroma quality compared to HDV. 2K from Hydra could not possibly be better than about 550-620 lines per picture height in 16:9. But it would be uncompressed which is more important than resolution IMHO. 14stops of latitude is impossible in the sensors we have today, especially in the older sensors used in these ccd cameras. A little more than 10 is closer to the truth and the highest quality CCD and CMOS sensors will do 11.5 if you have the light (not every shooting scenario). CMOS sensors should be grateful when they achieve more than 7-8 stops in real life use because they need much more light in order to get their dynamic range spec compared to CCD. How often do you have 2000lux available? And if you can contain the highlight of a bright window in your interior shot, how much detail and how much noise is in your interior shadows? The Andromeda used pixel shifting so it's not comparable to bayer. In bayer you have ideal positioning of your pixels, in a pixel shifted solution you have to pray for manufacturing parameters than are completely unpredictable (sub pixel precision sensor/ccd block assembly). In bayer you can get 800lines on a 1000pixel sensor. In a 500pixel pixel shifted solution you should be very happy if you get 550-600. If any company wants to claim that they can mount ccds on a prism with 0.1 microns precision I would be happy to prove the equipment required for that does not exist. From personal experience, the horizontal resolution of a Sony FX1, a camera than only shifts horizontally is very unpredictable. I have seen from 550 to 800 on a chart. That is about 970 to 1400 pixels, which is (surprise!) close to 960 (sensor resolution or 0% luma advantge) and 1440 (format resolution or 50% luma advantage). But an FX1 that is better than 600-650 lines horizontally is very rare. More than 20% of advantage over sensor resolution should not be expected.