DSLR Sensor technical data

[RobertM]

I thought it would be worth posting links to these two sites as they give fairly objective technical comparisons between DSLR sensors on various cameras over the years:

http://www.dxomark.com/index.php (look under sensors)

http://www.sensorgen.info/ (less information but with a more interesting assessment of read noise and full well depth)

In all they make interesting reading, particularly wrt Canon vs. Nikon.

Anyone know of any other sites with fairly upto date information like these?

Robert

[mickw]

Worthy of a "Sticky" methinks

Very interesting - I need a new camera 

[Rocket Pooch]

EOS1000D very good for the price!

[RobertM]

It doesn't tell the whole story but could be useful.

Note that the sensorgen site data is generated directly from dxomark so is the same data from a slightly different perspective.

Robert

[MarkS]

Interesting site - I'm not familiar with the maths being used to generate these figures so I can't guage how accurate they may be.  However, I have a couple of points to add:

1) Even for modified DSLRs, the QE at the H-alpha end of the spectrum tends to be much worse then for the green wavelength (actually the same is true for many or most astro-CCDS)

2) If 2 DSLRs have the same QE but one has smaller pixels than the other, then the one with smaller pixels will appear noisier because each pixel gets fewer photons because the smaller pixel area is a smaller target for photons.  I belong to the camp that thinks that smaller pixels are therefore a bad idea but another camp says it doesn't matter.  Take your pick.

Mark

[MarkS]

I read through the approach - essentially it consists of taking the SNR graphs and "true ISO" graphs (instead of the inbuilt camera ISO)published on the DxOMark site and then fitting the data to arrive at the camera's gain, read noise, pixel non-uniformity (which is what we correct for using flats) and subsequently calculating the QE (quantum efficiency) by estimating the number of incident photons at a given level of illumination, compared with the number of photons the sensor captures.

There are some figures for other cameras published here by the guy who "invented" the methodology:
http://forums.dpreview.com/forums/readflat.asp?forum=1019&thread=34516718&page=1
including
Canon 350D 26.9%
Canon 450D 30.4%

Note that these figures are for green light.  Personally I think they probably understate the true QE because I'm not sure the model for the pixel non-uniformity completely captures the relatively large variations in pixels (e.g. hot pixels) on the cheaper sensors.

In the same thread, various participants dispute the Nikon numbers because of the problem of obtaining unadulterated raw frames from the Nikon (which makes Nikon images appear cleaner then they otherwise would and results in a higher calculated QE).

I came up with some completely independent figures of my own here:
http://forum.orpington-astronomy.org.uk/index.php?topic=5262.0
where I concluded a QE for Canon 350D as somewhere in the range 33%-43% for the green pixels.
However I really need to repeat the experiment under controlled conditions using the DSLR and calibrated astro-CCD on the same scope.

Mark