What determines a sensors Base ISO

Started 7 months ago | Discussions thread
dethis2
Forum MemberPosts: 83
Like?
Re: Some corrections/clarification...
In reply to Mark H, 7 months ago

Mark H wrote:

.... Nope - I firmly disagree.

Whilst of course, any part of an electronic system can introduce some noise/error - the ADC alone introduces very little appreciable noise, but mostly just errors, e.g. non-linearities (which are generally very small/insignificant) and by far the most significant error is usually quantisation error averaging 1 LSB or +/0.5 LSB.

Quantisation error can be considered a form of noise though.

Yes, but we aren't talking quantisation error.

It's actually the same thing - simplistically, quantisation error introduces +/- 0.5 LSB of noise.

http://en.wikipedia.org/wiki/Quantisation_error#The_additive_noise_model_for_quantization_error

......

Mark, why do you call the quantization error significant, it is really small even in the absence of read noise (wiki .. stdev 0.28 (rounding model) stdev 0.58 (truncating model).

If we count in some read noise the quantization error becomes negligible.

Of course when the read noise is much lower than the LSB the digitization is setting the limit for DR .. like with DxO's DR results for Nikon 3300 which digitizes at 11bits (12bits file but only half the levels populated even in the linear part).

So D3300 scores at "screenDR" 11.97 stops vs D5300's (14bit) 13.06 stops

-- hide signature --

Ilias

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Heh!New
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow