Sensor Tech Generational Improvments

Started 4 months ago | Discussions thread
OP Astrozoid Contributing Member • Posts: 846
Re: Summary So Far

57even wrote:

Astrozoid wrote:

57even wrote:

Astrozoid wrote:

57even wrote:

I've never heard of flats removing DSNU. Are you sure about that?

My apologies. When you said flat frames I read it as dark frames.

No problem.

I'm a little unclear when you say "shorter exposures and higher ISO generate less of both." What is the "both" referring to?

Do you mean FPN and DSNU? Or Thermal signal?

DSNU, PRNU and also thermal signal.

How does a higher ISO result in less DSNU? I know how it helps with FPN.

DSNU is fixed pattern noise in the dark signal. It increases in proportion to dark signal so its proportional to time. Dark shot noise only increases as the square root. You reach a point (usually after many minutes) where DSNU dominates.

This is in addition to hot pixels and other nasties (requiring mapping or median filters).

Shorter exposure times mean smaller signal and dark signal signal = less DSNU (and PRNU) and of course, less dark signal.

Shorter exposures don't reduce thermal signal if you stack to equal a longer one.

If one is not stacking, wouldn't DNSU be proportional to the total integration time just as thermal signal is?

So, stacking 5 x 1min would equal the same amount of thermal signal and DSNU as 1 x 5min.

And the FPN in the DSUN would be 5x a single exposure.

Yes, but you only need 1 much shorter dark frame to remove it

Yes, but for a stack, the dark noise in the dark frame will be added to each individual sub. And when you stack, there isn't any benefit in shorter exposures. The dark signal, and dark noise, will be the same if the total integration time is the same, regardless of exposure time or ISO.

True. My main use of the technique is for handholding images in dim light, where I can use auto-alignment on the best matching frames. I can stack about 20 frames if I use ISO3200 and still get a decent result.

And I assume you already shoot somewhere around unity gain, or ISO 800 or so, to reduce FPN in the read noise ?

This is something I've never heard before.

Could you please explain why unity gain would reduce read noise FPN?

Probably nothing to do with FPN thinking about it. Just that downstream noise which contains most of the FPN isn't amplified. Still additive though.



Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow