AutoQuad Forum

Please, could someone explain how DIMU calibration must be done? (and then, maybe to put into WIKI)

Following other post about DIMU calibration, I´m not able to get good numbers. I´m using command line with cal, sim3 version 2.0. I get

STEP 1: //MAE 0.000596
STEP 2: //MAE 0.00958037, AVG: 9.80531

But STEP 3 I can´t get better than MAE =0.58. I repeated many times, several din files without success.

I continue with STEP4a but command line give bad results (MAE = 0.17). Changing now to QGC (same files) then I obtaing good numbers! (MAE = 0.04) Some explanation?. I get good numbers in 4a, 4b and 5. But not in STEP 6 (related to Step 3, I guess)

Please help to all new DIMUers!

Angel

Hi Angel,

which firmware do you use?

Joerg
aBUGSworstnightmare

r345 (for DIMU).

Anyway, to record data with calipso and then post process files I guess is indiferent. Isn´t it?

I'm about to get to this stage too, as I've received my DIMU and have just mounted it to my AQ, ready to start the static tests. Does it actually matter what version of firmware is used for the static tests?

Regarding firmware, I saw this link which Norbert posted:

ftp://ftp.autoquad.org/3/334693_22529/a ... erimental/

What does the 6.0 and 6.1 refer to, and which of these builds are stable? I've heard that some are unable to store parameter values after a reboot?

Many thanks and sorry for all the questions

David

I'm pretty sure 6.0 vs 6.1 refers to the resistor which allows voltages above 16v to be displayed. Only the first batch of boards sold in 2012 were 6.0, everything sold recently will be 6.1.

Full calibration process is the same, but I found that I get best results with just one static file from a horisontally oriented board.

For Cal-mag step, you need to observe the MAE and AVG and stop calculation at an ideal time.

MAE will drop steadily for a number of loops and then begin to rise. Abort calculation when it begins to rise. Defines will be printed. Let it run too long and the calculation crashes and produces a -nan result.

Observe the AVG, average. You are trying to get as close to 2.0 as possible. I usually can get away with 2 runs that brings the MAE down and gets me within +/- 0.01 of the ideal average of 2.0. If I try a third run, MAE will rise from the beginning, calculation will chrash and produce a -nan (not a number) result.

Otherwise, things are as normal. Look at the raw voltages to look for noise in log files.

I do short runs (20-30 loops) for Sim3 -gyo and -acc (usually need some variance to get -gyo to converge for the first 10-15 loops), then I let the combined -gyo -acc step run until MAE no longer drops.

Thanks for the quick feedback SSS, Jussi, much appreciated

I will crack on with my single static log file, and then mount the board back into my platform. Presumably it doesn't matter which firmware I use for creating the log files?

Many thanks,

David

Hi,

can anybody upload a compiled r345 (for DIMU) .hex-file for others to work with? This would give us the possibility to share our calibration results and to determine best praxis how to do it (since Jussi mentioned he's only using one static file for a leveled (horizontal) board).

Joerg
aBUGSworstnightmare

P.S. Started my old fridge in the basement yesterday. It's running, wasting energy and need something to go in!

Here it is r345, DIMU for both R0 and R1 boards (new boards are R1)
(compiled using Makefile, for v1.01 DIMU corresponding to "board_dimu_v1_1.h" file). I use this version succesfully.

Angel

@Jussi: thanks a lot for your explanations. Anyway calibration with DIMU still very tricky with many bugs. I´d sugest to the team to make an effort to clean the process now we have a very good hardware.

BTW, I´m using DIMU with old params (before instaling DIMU) and flies very well. I only need to do the "simple leveling" before each flight, and go. Again, AltHOLD is awesom!