Autotune:0...Ground:2

My first official foray into 3.1 has not gone well, and so far, Gravity - and his mistress, the hard frozen ground - are winning 2:Nil.

Both bird were flying fine on 3.0.1. There have been no physical changes. The Disco even flew ok on 3.1, pre-autotune, although I still can't seem to find the source of the vibes.

The Disco started autotune no problem, and you could physically see roll getting tighter and tighter. Pitched started, and that looked good too. I was getting excited. Right up the moment it totally flipped out of control, Belly bounced off the tarmac, and proceeded to freak out some more and self-destructing. Logs are attached, and apart from less than satisfactory vibes, no idea why it freaked out. All physical connections were tight even after the crash, and there were no loose screws or bolts. The new 3.1 engine readings look weird btw...something in there, maybe?2014-01-12%2008-25.log 2014-01-12%2008-25.log.gpx

So I moved onto the quad, which was even more bizzare. It tried to do a fly away the moment I flicked autotune. So I flicked out, but it now went into failsafe RTL. Broke RTL into loiter, and it freaked out again. Realising this was not a route to safe flight, I went back to stabilise, and discovered I had lost all yaw control. Then I made a mistake - instead of just putting it straight down in stablise, I flicked manual RTL, and boom - all motors stop. Stabilise again, but I was too late. Wish I had the logs for that one, but I think the logs are corrupted - it fails at the same point very shortly after starting the download. As mentioned - It was running sweet in 3.0.1...Compassmot: 0%, Vibes well within +/-5, etc. Only thing changed was the upgrade to 3.1.

2 freakouts in Autotuine on 2 aircraft in the same morning session. Both were flying well previously. Methinks I'll stick to the manual PID method, until my confidence is restored...

Amazingly, both are now phyisically repaired already. Disco ready to fly, and I'm doing a ground up, factory-fresh install on the Quad. The fog has come though, so visibility down to 20m. I had a lot of flight activity planned, but it ended up being a frustrating - and short - weekend's flying :-(

@Crashpilot, so why don't you just go buy DJI then?

At least Euan got his parts back, that doesn't happen when the Naza fails. ;)

Euan: interesting observation about those connectors. I had once hacked some ESC's I had, soldered barrel plugs right to the board as DJI does. But I quickly realized that this is a foolish thing to do. Any stress on the leads results in horrible loading on the solder pads of the ESC, and they are quite likely to break off if they're a cold solder, or tear off the solder pads if they're soldered well.

I have a digital servo tester which allows me to make step-function changes to the output, this would probably be a good way to test it.

My pet theory on what causes the loss of sync on the SimonK with pancake motors, is that the SimonK will instantly apply a load to the motor, and it causes the motor to accelerate faster than the timing system can measure.

Now, if we can just get some purpose built, quadcopter ESCs with better and easier management of timing to better mate them with motor and prop combos. (3DR do you have anything in the works??)

Well, why not just use standard ESC's? They do work just fine. My H8 used the Quattro ESC's with standard firmware, and it was very very stable.

You need to be a member of diydrones to add comments!

Join diydrones

Comments

Calle February 7, 2014 at 6:20am

By the way, I did a loiter endurance test in abit of wind and I hovered for almost 17min! (1000sek)

I had only 7 sats due to bad GPS placement but i didnt touch the sticks for the whole time.. Calculated mah in MP was approx 3000mah in 4 degrees Celcius. Hover amps were around 8-11A.
Calle February 1, 2014 at 6:03am

Hey all,

Ok it took awhile to run the autotune and it worked flawlessly the first time! Well, Im not that happy with the pids but that is a different story. It feels more sluggish now but it was overshooting before so. It still works good and no failures at all. It was not wind free and I had to move the copter back to me alot. That would have affect the settings.

Im running HK 20A Blue ESCs (Mystery) flashed with SimonK (No Becs are used)

HP2212 880KV motors

4S 3700mah

TBS Discovery frame with RCTimer 450 Arms.

APM 2.5 - UbloxGPS

Thank you for the great work that you have put in to this product!!
Roman R. January 17, 2014 at 2:41am

OK-
SimonK FW: awesome!
Arducopter 3.1: awesome!
Thank you for all the developers hard work- do not worry about some of the comments!

The so called "cogging" problem is not new and using leading edge technology CAN cause problems I guess because so many factors influence the outcome...

To test your specific Motor/ ESC FW/ Prop/ Battery combination Simon K gave me the following tip:
- mount the combination you want to test on a table
- connect a receiver
- programm a switch on your remote to the throttle channel: 20% thr pos 1; 100% thr pos 2

Now with the flick of a switch you can simulate the strongest FC reaction you can imagine
;-)

This is a pretty harsh and somewhat dangerous test- so do not blame me if something fails! Wear goggles because this obviously can shatter your props!!!
Especially with comp_pwm- impressive forces ;-)

Btw.: I run MT 2216 with HK F30 ESCs, Comp_PWM FW (removed BEMF filter caps) and 11" Props on a "heavy home brew cat" without any issues- just lowered the values after auto tune somewhat for smoother flying... Flies like on rails with loads of power!

And- keep up the great work! You guys are awesome!
Luis Morales January 14, 2014 at 2:47pm

thanks for pointing that out Marco, such good information about this there.

i just added my 2 cents to that conversation.
Developer

Marco Robustini January 14, 2014 at 2:26pm

Hi Euan, i'm sorry for your double crash!
Read here: this is what I wrote a few days ago, follow the "drone-discuss" list, you can save your quad... ;-)

Cheers, Marco
Rob_Lefebvre January 14, 2014 at 2:01pm

One interesting possibility is if the acceleration limiter that I'm working on would prevent the problem. If the problem is that at the moment of switching for AT to Stab, the Stab controllers gives a massive spike on the output, then if we implement the acceleration limit such that it softens this, it would help.

Still, seems like it's only avoid the problem rather than solving it.
Euan Ramsay January 14, 2014 at 1:42pm

Are t-motor 2216-11's still considered pancake motors? That's what was attached to these RCtimer ESC's.

I run some MN3508-29's (which I definitely consider to be "pancake") on the quad, but I've never opted for simonk on these, for exactly the reasons you state.
Rob_Lefebvre January 14, 2014 at 1:34pm

@Crashpilot, so why don't you just go buy DJI then?

At least Euan got his parts back, that doesn't happen when the Naza fails. ;)

Euan: interesting observation about those connectors. I had once hacked some ESC's I had, soldered barrel plugs right to the board as DJI does. But I quickly realized that this is a foolish thing to do. Any stress on the leads results in horrible loading on the solder pads of the ESC, and they are quite likely to break off if they're a cold solder, or tear off the solder pads if they're soldered well.

I have a digital servo tester which allows me to make step-function changes to the output, this would probably be a good way to test it.

My pet theory on what causes the loss of sync on the SimonK with pancake motors, is that the SimonK will instantly apply a load to the motor, and it causes the motor to accelerate faster than the timing system can measure.

Now, if we can just get some purpose built, quadcopter ESCs with better and easier management of timing to better mate them with motor and prop combos. (3DR do you have anything in the works??)

Well, why not just use standard ESC's? They do work just fine. My H8 used the Quattro ESC's with standard firmware, and it was very very stable.
Developer

leonardthall January 13, 2014 at 10:50pm

yeh, I am sure the repeated tests done by autotune combined with the higher stab gains work the ESC's quiet hard and is directly responsible for taking them over the edge.

It is surprising though because the two tests autotune does two tests. It asks for 20 degrees per second rate and then it asks for a step change in angle of 20 degrees. So the tests are pretty small and reasonable. So all I can conclude is that autotune is hitting a sweet spot of bad. I also don't see the problem early in Autotune so I suspect that heat is building up or some other thing getting gradually worse each test and the esc doesn't have time to recover.

It would be nice to find a way to avoid this problem but I can't see how this is possible as the tests are already very conservative.
Euan Ramsay January 13, 2014 at 10:09pm

So short version: my RCtimers are probably not the required quality level needed for AT, and I have retired them to the parts bin for use in a future project.

of 6

This reply was deleted.