Quote:
Originally Posted by gregbradley
I am not sure that Craig Starks article is conclusive. IO response is not linear very often with DSLRs. You can get a drop in sensitivity by going to a nonstandard ISO setting (ie not a standard multiple of 1600, 3200, 6400 etc. That varies by camera maker and model I believe. For example the Nikon D800 does not drop dynamic range between ISO's but other models do.
|
I think the dark noise investigations are pretty conclusive (pages 3-5 of the article). Based on how imaging sensors work, we know that dark noise only increases over time - thermal noise doesn't subtract from the well counts. However, when extracting data from the "raw" files it clearly does...
Quote:
|
Of course it may be a secret routine is being applied but it smacks of conspiracy theory - in this case against Canon.
|
I disagree that it's a conspiracy theory - the methods he used are clearly stated, his findings match what I've experienced, the theory for what he's saying appears legit, and he provides very compelling reasons why this behaviour (scaling) would be highly desirable for normal shooting.
The main point is that the process is being performed before the data is being written to the CR2 raw file - hence there's no way to undo the rescaling. CR2s are a lot more "raw" than JPEGs but they're obviously not equivalent to FITS files from astro CCDs either.
I should point out that this process works extremely well. I can fiddle around with equivalent Exposure Values, e.g. 30 sec @ ISO 25600 gives me a histogram with the same peak location as 8 mins @ ISO 1600. Likewise shooting timelapse sequences on manual doesn't change the location of the histogram peak (that I've noticed) even as the sensor warms up substantially.
In fact, it worked so well that before getting into astro I didn't even realise that there was such a thing as dark current
