MarkB
I did another test with v3.3. With this one the ND Filter was in use and focus was set to MF, started off with 150 images as a 5sec interval, followed by 150 at 4sec, then 150 at 3sec and finishing off at 150 as 2sec interval.
This is quite interesting. The random jumps in exposure are all gone. However, the script did not attempt to use the ND filter - probably because the shots were later in the day and the scene darker so it did not need to use it. I'm thinking I need to do more testing of the ND filter use.
Other than underexposing at the shorter interval (but that could be in part due to the conditions) the exposures seem fairly consistent.
You are seeing the "filter factor" at work there. The filtering is so high that the exposure changes lag the actual scene illumination changes as evening approaches. I have an improve version of the filter I'm working on to remove a small offset error at high filter factors. But you will probably want to Medium or even Low settings for the filter factor in any case based on what I see here.
In spite of script set to MF it appears to change from shot to shot. I am assuming some of my focus issues have to do with the camera itself and how it reacts to the script. Could this me a case of not the right tool for the job? Has there been anyone who had done a fair amount of timelapse with the S series camera (S90,S100,S110,S120) since they do have manual focus?
There have been many thousands of CHDK timelapses with all kinds of cameras. I personally have used my G, S, IXUS, A and SX series cameras. However, the vast majority of all those sessions have been with the camera in autofocus mode or with a simple AFL set at the start of the script. On many of those cameras, up until recently, the AFL really did nothing.
In the last eight month, there has been a lot of work done on using MF & AFL for manual focus ( a.k.a Subject Distance override ) in CHDK 1.3.0. The rather long forum thread is here :
Setting focus from scripts or menus
I thought we had made pretty good progress as we seemed to be able to make most cameras work in MF mode. And those that did not would focus in AFL mode instead. At least based on the test script we used they all seemed to work.
However, I've come to realize that we actually only evaluated the ability to set focus distances between 0.4 meters and 1.0 meters. We reported success based on the camera's reported subject distance versus the focus distance requested. While I checked the resulting images from my cameras, the majority of the test results did not, so the camera reported focus distance could be wrong too.
So now we seem to have a few cameras that passed the tests but that do not manually focus well out to infinity ("soft images"). Even though reports from owners of the same camera models (including mine) do not have that problem when running the same tests.
And we seem to have cases where the MF setting gets lost after initially working. Like in your tests.
And even I have had issues with my
A1200 :
https://chdk.setepontos.com/d/11078/383 that have never been resolved.
So for now I continue to collect test results, write new tests, study what we are seeing, and report it here. Somewhat frustrating as I have not been able to reproduce the problems on my six different camera models.
tl;dr note : I would like to test the way the kap_uav.lua script handles the set focus at infinity part. Are you are up for another test round? Currently the script enables MF and/or AFL mode when it starts but it tries to set_focus before each shot as well That should be fine but I'd like to see if that is actually causing issues so I've attached a modified version. Run a few hundred shots from your balcony in MF and AFL modes and see what happens to the reported focus setting ?
Thanks - progress here will only be made here with the help of people like you !
