35mm RGB stripped sensor

[Krstic Zoran]

RGB stripped sensor in F35 and SRW-9000PL has twice the vertical resolution and three times the horizontal resolution of a normal 1920x1080 HD raster.Three times the horizontal resoluotion of HD is logical.With this approach Sony made advantage in comparison with Bayer pattern used by RED,ARRI,... because RGB stripped sensor avoid interpolation to make full color picture.

 

But, what is purpose of twice the vertical resolution (2x1080 px = 2160 px)? Here I found that:"...It was one normal row of photosites filtered for each color and one row ND'd to capture bright information per color".

http://www.cinematography.com/index.php?showtopic=28026

 

Very logical answer,but is it truth!?

Today at presentation of Sony SRW-9000PL I asked about that and presenter told me oppositely explanation...

 

He said that no one ND was used over half of CCD's array.His theory is based on Gain function..."...It was one normal row of photosites filtered for each color and one row is amplified to capture information at low light (per color)". Amplification is made by Gain of +3dB...half array is read by 0 dB gain...second half is read by gain value of +3dB.

 

New theory sounds like a ALEXA readout mechanism...

 

There are two opposite explanations...what is the real story behind Sony's stripped CCD!?

Edited February 18, 2011 by Krstic Zoran

[David Mullen ASC]

RGB stripped sensor in F35 and SRW-9000PL has twice the vertical resolution and three times the horizontal resolution of a normal 1920x1080 HD raster.Three times the horizontal resoluotion of HD is logical.With this approach Sony made advantage in comparison with Bayer pattern used by RED,ARRI,... because RGB stripped sensor avoid interpolation to make full color picture.

 

But, what is purpose of twice the vertical resolution (2x1080 px = 2160 px)? Here I found that:"...It was one normal row of photosites filtered for each color and one row ND'd to capture bright information per color".

http://www.cinematography.com/index.php?showtopic=28026

 

Very logical answer,but is it truth!?

Today at presentation of Sony SRW-9000PL I asked about that and presenter told me oppositely explanation...

 

He said that no one ND was used over half of CCD's array.His theory is based on Gain function..."...It was one normal row of photosites filtered for each color and one row is amplified to capture information at low light (per color)". Amplification is made by Gain of +3dB...half array is read by 0 dB gain...second half is read by gain value of +3dB.

 

New theory sounds like a ALEXA readout mechanism...

 

There are two opposite explanations...what is the real story behind Sony's stripped CCD!?

 

It could be gain instead of ND on the second row of photosites, I was sort of making a layman's guess... my point was just that there are twice as many photosites as needed to get 2MP per color (the sensor is 12MP instead of 6MP) because of this feature of the extra row. But you wouldn't amplify gain to improve highlight detail, you'd do it to improve shadow detail. I thought the second row was for overexposure information not underexposure information but I guess either way you are increasing DR a little bit (3db is a half-stop.)

 

The thing is that you can't really get the resolution benefit of a 12MP sensor beyond 1080P/2K finishing because the next row of the same color is too far apart unlike in a Bayer pattern. In other words, the final resolution is HD. And the RGB striped pattern does create some color moire issues. The main benefit is mostly the fast, high-quality conversion to RGB that the sensor allows in real time. Also, there is an equal amount of resolution per color.

[John Sprung]

The one advantage that the stripe design has over Bayer is that the Nyquist limit is the same for all three primaries. That turns out to not be such a big deal in actual practice. Sony's abandoning the stripes and going with a new idea, Bayer rotated 45 degrees, in their new Q67 chips. I don't believe they're in any available cameras, Soon Only Not Yet.... ;-)

 

I asked John Gault of Panavision about the two rows per output line on the Genesis, he says they're identical, they don't do any ND or gain tricks, they're just intended to double the vertical Nyquist limit.

 

 

 

-- J.S.

[David Mullen ASC]

The one advantage that the stripe design has over Bayer is that the Nyquist limit is the same for all three primaries. That turns out to not be such a big deal in actual practice. Sony's abandoning the stripes and going with a new idea, Bayer rotated 45 degrees, in their new Q67 chips. I don't believe they're in any available cameras, Soon Only Not Yet.... ;-)

 

I asked John Gault of Panavision about the two rows per output line on the Genesis, he says they're identical, they don't do any ND or gain tricks, they're just intended to double the vertical Nyquist limit.

 

 

 

-- J.S.

 

Good to know.