snubbr.com

Sensor Areas? (1 image)
Normally, I play like I'm an expert and try to help beginners. But today, I'm wearing my "Beginner" cap....

Been reading about the D300. Lusting a little....

I've seen this before, but this time it bothered me! Phil (and most other review sites) says that the sensor in the D300 has 13.1 Total MP, 12.3 Effective MP, and 12.2 MP. I finally realized that I didn't understand these 3 numbers. I found nothing on DPR to answer my questions. Yes, I looked at the "Sensor Sizes" for the 1000th time....

I did find an interesting article at:.

Http://www.barrypearson.co.uk/articles/dng/specification.htm#areas.

Which I think answers the questions I had...sortof....

Ignoring the Mask Areas, there are 3 areas on a sensor, per the above specification:.

Image control:Zoom outZoom 100%Zoom inExpand AllOpen in new window.

I hope these 3 areas correspond to what Phil lists in Specifications for cameras..

For those that don't know, Thomas Knoll has a little application called DNG Recover Edges. If you convert a RAW file to his DNG format, then drag the file to the "DNG Recover Edges" on your desktop, it will recover the missing Active Area pixels and store the result in the original file..

My Sony R1 has 156.1K extra pixels that can be recovered this way. The dimensions of the Crop Area on the R1 is 3888 x 2592. When I recover the missing Active Area pixels, it's 3924 x 2608. To keep you from doing the math, that's 36 extra H-pixels and 16 extra V-pixels..

But Phil's review of the R1 says that it has 10.8 MP Total, 10.3 MP effective, and 10 MP. These don't agree with my "understanding" as described above!.

The obvious Crop Size is 3888 x 2592 = 10,077,696. Converting to MP and rounding, I get 10.1 MP, instead of 10.0 MP. When I run DNG Recover Edges, I get 3924 x 2608 = 10,233,792 or 10.2 MP, instead of 10.3 MP. Can anybody explain this?.

My last question is, when DPR and others state that the sensor size is 21.5 x 14.4mm (as with the R1) which area does this describe? I suspect it is either the Total Area or the Crop Area. But which?.

I'd appreciate help with this. Links to the answer?.

Charlie DavisNikon 5700 & Sony R1HomePage: http://www.1derful.infoBridge Blog: http://www.here-ugo.com/BridgeBlog/..

Comments (9)

The total pixels value contains the:- Optical black pxiels- Dummy pixels- Active pixels.

The optical black pixels are physically shielded from light. They are used to sample the average dark current level and as a reference for DC clamping circuits at the input to the ADC..

The dummy pixels, I'm not so sure about. They may be partially shielded, or maybe just not tested or guarrenteed to work within specifications. I'd just trust the datasheet and not use them..

Here's a data sheet for a typical Sony CCD used in dSLRs:http://www.magzero.it/Sony_icx413aq.pdf.

They specify 6 values relevant to this discussion:" Total number of pixels: 3110 (H) 2030 (V) approx. 6.31M pixels" Number of effective pixels: 3040 (H) 2024 (V) approx. 6.15M pixels" Number of active pixels: 3032 (H) 2016 (V) approx. 6.11M pixels" Number of recommended recording pixels: 3000 (H) 2000 (V) approx. 6M pixels" Optical black: Horizontal (H) direction: Front 20 pixels, rear 50 pixelsVertical (V) direction: Front 4 pixels, rear 2 pixels" Number of dummy bits: Horizontal 31Vertical 1 (even fields only).

But even trying to make sense of these numbers, they don't add up. For instance they say there are 70 horizontal optical black pixels, but none of the differences between total, effective, active and recommended horizontal pixels equals 70..

I've designed a few video cameras using Sony chips, and I have always thought their datasheets were lousy..

But anyways, is it worth worrying about? What's the advantage of a few extra pixels anyways?.

Regards,.

Tylerhttp://www.flickr.com/photos/tcoen..

Comment #1

Don't manufacturers usually like to have an illuminated border around the actual imaging area that can be used to start the Bayer interpolation?.

I got the impression that this was responsible for at least some of the 'extra' pixels...

Comment #2

Andrew dB wrote:.

Don't manufacturers usually like to have an illuminated border aroundthe actual imaging area that can be used to start the Bayerinterpolation?.

I got the impression that this was responsible for at least some ofthe 'extra' pixels..

Yes, I think that's correct. I assume that the border outside the Crop Area and inside the Active Area are used for this purpose. But why so many? I can't see 16 extra photosites being needed (as are available on my R1) in the horizontal direction. Also, when I run DNG Recover Edges, I can't see any differences in the demosaicing at the edges. This implies that there is ANOTHER border outside the Active Area that can be used for priming the demosaicing engine..

And I still don't know which of these photosites are included in the physical measurement....

Thanks for your input. It's appreciated..

Charlie DavisNikon 5700 & Sony R1HomePage: http://www.1derful.infoBridge Blog: http://www.here-ugo.com/BridgeBlog/..

Comment #3

That's a good point (avoiding boundary conditions for the bayer interpolation)..

Do you have any info on the algorithms used for bayer interpolation? What size of window they use? I assume a simple bilinear combination of the 4 nearest like-coloured pixels would be too simplistic for photography?.

I found this paper from microsoft which uses 5x5 windows:http://research.microsoft.com/.../users/lhe/papers/icassp04.demosaicing.pdf.

Interesting stuff. The bayer interpolation probably has a big impact on image quality, but I haven't heard it discussed much. I wonder if there is some optimal method the major camera manufacturers have all settled on..

Tylerhttp://www.flickr.com/photos/tcoen..

Comment #4

Tcoen wrote:.

The total pixels value contains the:- Optical black pxiels- Dummy pixels- Active pixels.

The optical black pixels are physically shielded from light. They areused to sample the average dark current level and as a reference forDC clamping circuits at the input to the ADC..

Yep. I understand what the black pixels are used for. I had no question about them, so ignored them....

The dummy pixels, I'm not so sure about..

They call them "dummy bits", so I assume they are something in the serial data stream, but not necessarily associated with a photosite?.

They may be partiallyshielded, or maybe just not tested or guarrenteed to work withinspecifications. I'd just trust the datasheet and not use them..

Here's a data sheet for a typical Sony CCD used in dSLRs:http://www.magzero.it/Sony_icx413aq.pdf.

They specify 6 values relevant to this discussion:" Total number of pixels: 3110 (H) 2030 (V) approx. 6.31M pixels" Number of effective pixels: 3040 (H) 2024 (V) approx. 6.15M pixels" Number of active pixels: 3032 (H) 2016 (V) approx. 6.11M pixels" Number of recommended recording pixels: 3000 (H) 2000 (V) approx.6M pixels" Optical black: Horizontal (H) direction: Front 20 pixels, rear 50pixelsVertical (V) direction: Front 4 pixels, rear 2 pixels" Number of dummy bits: Horizontal 31Vertical 1 (even fields only).

Yes, that's informative. I need to get a SOny datasheet for the sensor in the R1. I think the part number is "IMX0xxAQ", but don't know what the "xx" numbers are..

But even trying to make sense of these numbers, they don't add up.For instance they say there are 70 horizontal optical black pixels,but none of the differences between total, effective, active andrecommended horizontal pixels equals 70..

Yes, that's one of my issues. The data I read here on DPR doesn't agree with what I measure at home on my computer..

I've designed a few video cameras using Sony chips, and I have alwaysthought their datasheets were lousy..

Why pick out their datasheets for scorn?.

But anyways, is it worth worrying about? What's the advantage of afew extra pixels anyways?.

I'm not trying to get extra pixels. I'm trying to understand the sensors..

Thanks for your help..

Charlie DavisNikon 5700 & Sony R1HomePage: http://www.1derful.infoBridge Blog: http://www.here-ugo.com/BridgeBlog/..

Comment #5

Tcoen wrote:.

That's a good point (avoiding boundary conditions for the bayerinterpolation)..

Actually, I started this quest thinking that the width of the "border" might tell us something about the interpolation algorithm..

Do you have any info on the algorithms used for bayer interpolation?.

Very little. There is lots of info about algorithms used in PP, but the camera manufacturers seem to hide every detail. I've never seen anything about it. If there are small quality issues with a camera's JPEG output, everybody seems to attribute it to JPEG issues, completely ignoring Bayer demosaicing!.

What size of window they use? I assume a simple bilinear combinationof the 4 nearest like-coloured pixels would be too simplistic forphotography?.

I was, unfortunately, involved in a recent discussion here:.

Http://forums.dpreview.com/...forums/read.asp?forum=1002&message=24654101.

It started out with a question about how to reduce noise, perhaps by reducing the resolution in PP. Then quickly degraded... But there is some good info there. Pretty basic stuff, but we really don't know exactly how various manufacturers do this. My assumption is that the choice of demosiacing algorithm is heavily influenced by the available computing HP...that one reason expensive cameras have good IQ is that they have the HP to use a BIG window and many surrounding photosites. One of the sources I read said that some cheap, low-HP cameras actually only use ONE green photosite, to save time!!!.

I found this paper from microsoft which uses 5x5 windows:http://research.microsoft.com/.../users/lhe/papers/icassp04.demosaicing.pdf.

Interesting stuff. The bayer interpolation probably has a big impacton image quality, but I haven't heard it discussed much. I wonder ifthere is some optimal method the major camera manufacturers have allsettled on..

I wish we knew! Phil...are you listening?.

Charlie DavisNikon 5700 & Sony R1HomePage: http://www.1derful.infoBridge Blog: http://www.here-ugo.com/BridgeBlog/..

Comment #6

Not quite every detail. Unless I have got it wrong this secret scientists business Nikon site seems to imply the algorithm they use, at least for those cameras:http://www.microscopyu.com/articles/digitalimaging/colorbalance.html...

Comment #7

Les Olson wrote:.

Not quite every detail. Unless I have got it wrong this secretscientists business Nikon site seems to imply the algorithm they use,at least for those cameras:http://www.microscopyu.com/articles/digitalimaging/colorbalance.html..

Hmmm....

I didn't see anything about the issue of sensor areas. Bayer was mentioned once. Demosaic = zilch..

Charlie DavisNikon 5700 & Sony R1HomePage: http://www.1derful.infoBridge Blog: http://www.here-ugo.com/BridgeBlog/..

Comment #8

Sony datasheets aren't necessarily bad, they are just very limited..

I think Sony probably has much better in-house documentation, but it is likely in Japanese..

Sony datasheets tend to have the absolute minimum information required. They also use non-standards measurements of noise and performance. However, on the plus side they do rigourously define the conditions used for those measurements..

I think it is their market philosophy. Sony only wants to deal with big fish that will buy tens of thousands or more parts. For that reason the datasheets for most Sony sensors are not publicly available. And for those that are, they don't really bend over backwards, like other manufacturers do, to make designing with them easy..

You can find datasheets for some lower resolution (VGA and ~1MP) Sony sensors here:http://www.framos.de/www.dir/en/start_en.html.

Compare the information in a Sony CCD datasheet to that in Kodak CCD datasheet:http://www.kodak.com/...US/en/dpq/site/SENSORS/name/ISSInterlineProductFamily.

The Kodak datasheets are much more oriented to encouraging third-parties to develop with them. Also, for Kodak chips you can get full schematics and samples of reference designs to speed up design. Same with Micron (for CMOS sensors)..

You can also easily find datasheets for Dalsa, Micron and Cypress sensors. These however aren't players in dSLR sensors. I've never seen a datasheet for Canon or Panasonic dSLR sensors. I'd be interested to see them if you've ever come across any links..

Sorry to get off topic..

The dummy pixels I think do have photodiodes and readout circuitry. I think they are just not fully specified (for defects, non-uniformity, etc) and Sony and other manufacturers don't recommend using them. They may be partially obscured as well. Kodak calls these buffer pixels..

Ya, I just don't know. I think there is nothing special about dummy pixels. They have no function, other than possibly as a spacer between the optically shielded pixels and the active area..

Tylerhttp://www.flickr.com/photos/tcoen..

Comment #9

Click Here to View All...

Sponsored Amazon Deals:

1. Get big savings on Amazon warehouse deals.
2. Save up to 70% on Amazon Products.


This question was taken from a support group/message board and re-posted here so others can learn from it.

 

Categories: Home | Diet & Weight Management | Vitamins & Supplements | Herbs & Cleansing |

Sexual Health | Medifast Support | Nutrisystem Support | Medifast Questions |

Web Hosting | Web Hosts | Website Hosting | Hosting |

Web Hosting | GoDaddy | Digital Cameras | Best WebHosts |

Web Hosting FAQ | Web Hosts FAQ | Hosting FAQ | Hosting Group |

Hosting Questions | Camera Tips | Best Cameras To Buy | Best Cameras This Year |

Camera Q-A | Digital Cameras Q-A | Camera Forum | Nov 2010 - Cameras |

Oct 2010 - Cameras | Oct 2010 - DSLRs | Oct 2010 - Camera Tips | Sep 2010 - Cameras |

Sep 2010 - DSLRS | Sep 2010 - Camera Tips | Aug 2010 - Cameras | Aug 2010 - DSLR Tips |

Aug 2010 - Camera Tips | July 2010 - Cameras | July 2010 - Nikon Cameras | July 2010 - Canon Cameras |

July 2010 - Pentax Cameras | Medifast Recipes | Medifast Recipes Tips | Medifast Recipes Strategies |

Medifast Recipes Experiences | Medifast Recipes Group | Medifast Recipes Forum | Medifast Support Strategies |

Medifast Support Experiences |

 

(C) Copyright 2010 All rights reserved.