sim3 --acc crashing sim3.exe - solved

[Max]

[Max]

Here's my dynamic log plot, FWIW.

-Max

[kinderkram]

Hmmm ... just wild guessing ... but could you please try the following:
- start the taskmanager, select "processes" tab, sort by name, scroll down to the processes starting with "s"
- run the simulation as usual
- as soon as sim3.exe is coming up in the taskmanager, right-click, select "Priority" -> "Realtime" and watch it munge the resources.

[Paul0711]

i've got the same problem with this step, i try to use variance set to 1e-7, seems to work...... i'm in loop 60

[Max]

[Max]

[Max]

[Max]

Well I managed to get through step 4b. I ran ~300 loops with variance set to 1e-8, stopped it and c/p the values to the params file. Then set the variance to 1e-7 and continued running the step. After ~2500(!) loops total (I stopped/started it a few times so don't have the exact number) my MAE was still dropping by 1e-6 on each loop before I aborted it. I tried a few times to resume the calcs with the default variance (var. checkbox unchecked) but the MAE would start to climb and then sim3.exe would crash, every time.

So I'm running step 5 now, and I still have to set variance to 1e-7 otherwise the MAE starts to climb dramatically.

It sure would be great to have a way to determine "good" data from "poor" data, if that is the underlying problem here. Although I'm not sure how I could do the Calibso any steadier than I already did. The heavier the 'copter the more error prone the process is. I'm thinking some kind of mechanical device to do all the tilting/spinning would be cool.

-Max

[Max]

Just realized I should update this thread with the eventual resolution... d'oh!

The underlying problem on this whole calibration run was, I believe, due to using an older version of cal.exe to do the first 3 steps (due to the "nan" problem with v1.0.2 viewtopic.php?f=32&t=814). Once I found and eliminated the one static log file which was causing the NAN issues, all the steps ran smoothly, with no need to set the variance manually, and MUCH lower MAEs (was in the tenths, now in the hundredths). This was not just a matter of changing the spelling of "ALN" to "ALGN," it seems like the whole set of data generated by the older cal.exe was just poor, or somehow incompatible with the newer sim3.exe.

The saga got continued in another thread, which is why I neglected to update this one. viewtopic.php?f=35&t=807&start=10#p3981

-Max

[LPR]

max

Very good to know.

Larry