Originally posted by BigMackCam If you go ahead with that, I'd definitely be interested to see the results. There's absolutely no doubt that noise increases with sensor temperature.
The big question remains doe IBIS further increase it and with a first set of tests done last night out in the cool garage so far it looks to be a big nope. As you stated it seems that overall sensor heat is a bigger factor but that seems to mostly just show as an overall brightening of the image that reaches a peak and then drops and levels off. That said I am going to do a similar shoot tonight but with the SR settings inverted for each shot. Last night was 7 lens cap shots done the following way:
Camera had been in the cool garage all day (about 50 degrees)
All shots were 30 second shots at ISO 51,200 and were pictures of the inside of a lens cap for a 17mm SMC fish-eye Takumar
All shots were done on a tripod fully extended
The focal length was manually set to be 300mm.
For shots where SR was turned on the tripod was continuously tapped moderately hard with a wrench to introduce a fair amount of vibration that needed compensation.
The camera used was a K-3
High ISO noise reduction was turned off
ISO51200 was chose because it seems likely to be most likely to show thermal variations.
1 - SR off camera was powered on initially here and picture taken immediately. afterwards SR was turned off as we the camera
2 - SR on camera was powered on after a 10 minute wait and picture was taken immediately
3 - SR off right after previous picture
4 - SR off right after previous picture
5 - SR on right after previous picture
6 - SR on right after previous picture
7 - SR off right after previous picture
Now tonight I plan on doing the same set of tests (the camera is out in the garage again) but with SR settings inverted. I should be able to have some results Sunday or at the latest Monday night to provide. I have way too much going on it seems other wise would say I would post results tonight. The problem with this test is I don't have access to another camera with a similar sensor and of similar age that lacks IBIS so here the only meaningful results would be IBIS on vs. off. Since we are dealing with noise some worthwhile measurements to look at would be:
RGB average value
RGB median value
RGB STD deviation
per channel average value (might be interesting so why not)
per channel median value (might be interesting so why not)
per channel STD deviation (might be interesting so why not)
LZW tiff file size (this gets to the concept of entropy in information theory, a high entropy data does not compress much so this would be a good quick data point)
Maybe figure out some way to feed the image data into a tool like the
diehard or
dieharder battery of tests and get the test results.
Maybe also do some transformations on the data and subtract the first frame from all others and provide some data on new noise introduced. This may work for adjoining frames as well where the SR setting was changed so that may be worth looking at too.