Replies to This Discussion

Permalink Reply by afernan on April 2, 2012 at 5:39am

Clarification:

What I want to say for "independent highly twisted lines" is the following:

From the battery I´ve created four independent lines black-red (- & +) driven to each of the ESC´s. Each pair of +/-  cables are highly twisted in order to cancel its own magnetic field by its own. This is done for each of the four pairs of +/- cables .

• ▶ Reply

DeveloperPermalink Reply by Adam Rivera on April 2, 2012 at 8:04am

Great write up! Thanks for sharing this information. I almost wonder if this should be a blog post so more people have eyes on it.... Its good information for everyone.

Thanks again!

• ▶ Reply

Permalink Reply by u4eake on April 2, 2012 at 10:13am

Great stuff afternan !

Did it solve yaw problems you were having ? (if not, what symptoms did it solve ?)

The thing that still bother me is this : with 2.046 till 2.0.55 I didn't have any yaw problems and my wiring was the same as it is now...  

 

Power to my hexa motors is done in a starburst form *.  A central feedline from the battery is split into 6 branches heading into each arm

• ▶ Reply

DeveloperPermalink Reply by Andrew Tridgell on April 2, 2012 at 1:42pm

That's a great result Angel!

• ▶ Reply

Permalink Reply by Jonathan M on April 2, 2012 at 2:51pm

Very cool!  EMI issues are always a pain. Care to post a pictures with twisted pair harness?

• ▶ Reply

DeveloperPermalink Reply by Marco Robustini on April 2, 2012 at 2:54pm

Nice one Angel!
He was the one who flew avoid, but unfortunately it is inevitable ... "come on let's twist again"... :P

• ▶ Reply

Permalink Reply by afernan on April 3, 2012 at 1:44am

Extensive flight test results with new hardware setup

The goal is try to get the best offsets for the better flight. Also to compare the three method we have:

1. Using MP / hand rotation , motors off
2. Using MP / TLOG offline calibration
3. Compass_learn by Tridge

Method 1):

hand rotating (no motors) -43, 30, 25   repeteated again   -30,31,17

Method 2):

Using MP / TLOG offline calibration. I´ve got three long fly records (see attached TLOGs)

TLOG	Thro>10%	Thro>20%	Thro>30%
2012-04-02 20-17-35	-21 31 76	-21 31 76	-21 31 76
2012-04-02 19-54-08	-14 34 64	-14 34 64	-14 35 65
2012-04-02 19-00-45	-17 36 89	-17 36 89	-17 36 89

2012-04-02 20-17-35

2012-04-02 19-54-08

2012-04-02 19-00-45

Method 3):

Compass Learn:

Outdoor flight:

After 2 min flight                   -2,40,-4

After 5 min more                   -19,39,-14

After 10 min more                 -18,52,-11

After 15 min                           -16, 16, -95

CONCLUSSIONS

• Compass_learn gives very different values for "mz_off", but similar values for "mx_off" and "my_off".
• Method 1 (hand rotating, no motors gives also similar values for mx, my but not for mz offsets.
• So we can consider mx_off = 21 , my_off = 31 as good values since in all three methods are similar.
• For mz_off I rely in second method, that is not so far than first method: let say mz_off = 65
• A flight with this values confirms that they are good  (see TLOG graph 2012-04-02 20-17-35)

Looking fordward your comments on this results to try to commint on one method for today until we get another.

Attachments:

• TLOGs.zip, 3.7 MB

• ▶ Reply

DeveloperPermalink Reply by Andrew Tridgell on April 3, 2012 at 2:14pm

Hi Angel,

Thanks for the tests! The 19-54-08 log is very interesting as it shows the learning algorithm doing quite poorly on the Z axis. The Z axis is always the hardest to learn, but this is worse than I expected. I'll see if I can use this log to work out why.

One thing to keep in mind is that what we're aiming for is to get a good heading. So graphing the heading produced with the different offsets gives you some idea of whether the differences matter. Here is a graph of the heading you would have seen with the learned offsets versus the tlog throttle 30 offsets for the 19-54-08 log

there are differences, but they are not huge. That is because the Z offset tends not to change the heading much in level flight (which is also why it is hard to learn!).

Still, I would like to improve the algorithm. That Z offset is further off than I'd like it to be.

Cheers, Tridge

• ▶ Reply

Permalink Reply by Dani Saez on April 15, 2012 at 1:28pm

Hi all,

prior to using 2.5.3 I ha no problems with yaw.

With 2.5.3 yaw would have a will of it's own, wandering about 25º to each side; I believe it must be some kind of process that now is more sensitive to mag interference,

So I followed afernan's advice, high-twisted all power lines and moved the mag a bit further away.

Tested it today, and yaw is back to rock solid.

Thanks! great tip!

• ▶ Reply

Permalink Reply by u4eake on April 15, 2012 at 1:39pm

That's very nice Dani !   Unfortunately twisting powerlines would require a complete dismantling and rewiring of my copter, which I really don't see me doing.

Surely we can't expect all users with yaw problems to rewire their copter ?  It worked fine before (2.0.49) so we should be able to get it working fine again.

• ▶ Reply

Permalink Reply by Ian forbes on April 16, 2012 at 9:44pm

Dying to see some photos! I have been having some issues regarding Random Yaw on my X8.  Would twisting wire always be a good practice in main wiring for future reference? I haven't really studied my logs yet, too busy reppairing and balancing props.  TIA :)

• ▶ Reply