Outstanding explanation! I was really scratching my nugget on this one for a while. The control reversal phenomenon explains chasing my tail on RC channel and output config. And I simply misdiagnosed the throttle output by looking at RCOUT instead of ThrOut (which is a much better indicator of autopilot behavior).
Hopefully my mistakes here helped someone other than me learn something.
Thanks again guys!
if you have never flown before then you need simulator time, this is a free Android game that should get you familiar with the basics. it’s not the most realistic but it’s just to get your thumbs going the right way without thinking about it.
ardupilot just isn’t designed for flying a plane without an experienced pilot to set it up first , it can be done but its a very steep learning curve for somone can’t fly.
ARSPD_USE is set to 0, so that means the plane is going to run the trim throttle setting, in this case 45. Basically, that AIRSPEED_CRUISE parameter doesn’t do anything in this configuration. You need to set your trim throttle by testing and tuning the airplane.
I see, thanks! Based on the log, a trim throttle setting of 45 is fine, and probably corresponds to a cruising airspeed of around 16 m/s. The issue is that the autopilot increased the throttle above 60% at 07:31, and then to 94% at 07:41.
When you switched to RTL the plane is going to try to climb to 100m and turn towards the home location. Both will cause an increase throttle setting (without calibrated airspeed) as a means to reduce stall risk, as risk of stall increases naturally in a turn.
It was already above 100 m, so the plane was actually slowly descending the whole time when the autopilot was increasing the throttle. But the plane was turning to circle the home point, so maybe that explains the throttle increase. Still, it was already above AIRSPEED_MAX and throttle was above TRIM_THROTTLE, so I’m puzzled why Ardupilot wanted to floor it.
I have an open-ended question for you: in your best judgment, based on the logs, what parameters would you have changed (and to what values) to prevent the crashes, and what parameters would you have changed (and to what values) to make autotune succeed?
@Rolf @dkemxr @Allister @geofrancis I’m not ignoring your advice to start with an easier plane (or with a simulator, which I’m already doing), but I still want to figure out the exact causes of these two crashes and how to avoid them in the future.
Yes, that’s correct. I over looked that. But it doesn’t change the fact the plane was trying to go to higher pitch and high bank angle. Like I said, the throttle will be increased to reduce the risk of stall. The plane is commanding a 50 degree bank angle, and that nearly doubles the required amount of lift that the wing needs to produce.
Like any accident/incident, there isn’t one smoking gun. A skilled fixed-wing RC pilot maybe could have flown it in manual mode. We haven’t addressed if the C of G was correct. The TECS tuning wasn’t even started so going to an auto-throttle is a gamble. Are the servos mechanically trimmed? SERVO_AUTO_TRIM wasn’t turned on.
Going forward, start with something slower and work through the wiki step by step. Start with a more traditional setup, that’s why I suggested the Swordfish or something like that. (there’s lots out there, that’s just the first one that comes to mind) Flying wings (elevons) can be tricky to get tuned. Make sure the control throws are set properly and in the correct direction for both manual and FBWA modes. Don’t assume RTL or FBWB/cruise will work properly the first time and be ready to switch back to manual or FBWA. Once the plane PIDs are tuned, then work on the TECS tuning.
I do have my Nano Goblin flying just fine on Arduplane, although mine is built as a sub-250g plane (246g takeoff weight with HD camera). Top airspeed is ~120kph / 75mph / 33m/s, minimum airspeed is around 45kph / 28mph / 13m/s. It has quickly became one of my favorite planes for doing stuff planes were not meant to do.
This was on iNav:
Arduplane: flies better.
General tips regarding this plane:
I could post the PID parameters of my nanogoblin. Using someone else’s PID values is a can of worms but in this case (little experience and light/agile plane for which the defaults are not great) it might be a better starting point than the defaults.
Many thanks DaBit, and great flights! I’ve reduced the weight from 470 g to 330 g by using a 3200 mAh battery, and also put the clear tape that came with the kit onto the elevons. How much tape did you put? Did you tape the whole elevon (or even the whole wing), or did you just use the tape that came with the kit to cover half the elevons?
After those changes, I had my first successful flight! Flight log here: first_successful.bin - Google Drive
Autotune seemed to work, though I can’t tell for sure because I’ve never flown a perfectly tuned plane before. The plane still enters a rapid roll whenever it exceeds an airspeed of 25 m/s or so, which it often does in RTL mode, but I can recover from the roll by idling the throttle, letting FBWA stabilization do its magic, and bringing the throttle back to 10% after two seconds. For the next flight I’ll decrease both THR_MAX and TRIM_THROTTLE, because the motor and propeller are overpowered for the lighter craft.
Elevon throw should be 6mm each direction or so. With Emax ES9052 servo’s I am using the hole closest to the shaft for the linkage.
I have elevon throw maxed out (so that the spar almost hits the wing at SERVO_MIN and SERVO_MAX). What are the benefits of limiting elevon throw? Also, I’m using the hole farthest from the shift. What are the pros and cons of using the different holes?
With these small and light planes, follow the ‘Tuning light, agile aircraft’ recommendations: Automatic Tuning with AUTOTUNE — Plane documentation
I changed SCHED_LOOP_RATE, ONESHOT_MASK, INS_GYRO_FILTER, PTCH_RATE_FLTT, and RLL_RATE_FLTT as suggested. Did you use a dynamic throttle notch filter or in-flight FFT? My F405 WMN doesn’t support in flight FFT, and I haven’t enabled raw IMU logging to set up the notch filter yet.
I could post the PID parameters of my nanogoblin.
That would be great! You can see mine from the log. I’m very curious how the two compare.
Why so much battery? I am using a 1300mAh 3SHV pack when flying without HD cam or 1100mAh 3SHV when flying with HD cam (as said before: this is my <250g plane). Flight time is 10-15 minutes which is plenty for the small gap seeker. And that is with leftover drone stuff (1407 Emax ECO motor, 4" prop)
I am using regular clear packing tape. You know, the same stuff one uses to close cardboard boxes. I usually cover every square inch of foam using long continuous pieces of tape. On the elevons for example I start with a strip of tape on the bottom side and fold it over to the top so it adds a lot of torsional stiffness. If you taped the top and bottom separately it would not add as much rigidity.
Control surface throws: it depends on the plane how much is still effective. Too much deflection and the airstream detaches from the control surface, turning it into a turbulence generator instead of a control surface. Plank wings such as the nanogoblin just won’t act nice with too much throw. Start out reasonably small on a new plane build and increase if you feel the need to do so. That’s why you have multiple holes in the servo arms; easy to move the control rods a notch.
I will pull the parameters from the nanogoblin when I get near to it, and shoot a picture or two. However, that won’t be today I think.
Not a nano goblin but another plane I’m building. Just to show that I cover everything. It makes a HUGE difference in airframe stiffness and an even biger diference in crashability. Which happens when 5" racing/freestyle drone pilots start flying fixed wing. Must-Hit-That-Gap…
Going around the corners is the hardest part, but you will figure out where to put a slit once you start doing it.
(and to be honest; this is a clear laminate experiment and not packing tape, but the principe and appearance is the same)
Great to hear you got it flying!
Planes don’t ride on their motors the same as multi-rotors do, so the filters don’t have the same fundamental impact. Setting up a notch filter certainly won’t hurt but it’s not going to be as dramatic as you might be used to. (quad planes, different story). If you want a basic throttle filter that will do the trick and not suck up too may resources on the F405.
Planes don’t ride on their motors the same as multi-rotors do, so the filters don’t have the same fundamental impact. Setting up a notch filter certainly won’t hurt but it’s not going to be as dramatic as you might be used to. (quad planes, different story). If you want a basic throttle filter that will do the trick and not suck up too may resources on the F405.
Great, so I can just ignore the notch filter for now? I’m always happy to have one fewer thing to worry about!
Personally I wouldn’t get too worried about a filter on a normal, single motor electric plane. There just shouldn’t be the vibration to worry about. If there is, you’ve got other mechanical problems to deal with. A plane just isn’t change the motor speed as much as a multi-rotor.
I followed your lead and covered the wings and elevons with packing tape, adding 11 g to the weight. That plus glue for the various crashes brought the total weight to 349 g. I also limited the servo throws to 12 mm. I know you did 6 mm, but the Nano Goblin instructions said 10-12 mm, so I thought I’d try that first. I also reduced THR_MAX to 50% to avoid overspeeding, and did some tests to calibrate the TECS parameters. I set PTCH_TRIM_DEG to 4 degrees, TECS_CLMB_MAX to 3 m/s, TECS_SINK_MIN to 2.2 m/s, and TECS_SINK_MAX to 9 m/s. Then I flew again and did an autotune.
With all of those changes, the plane flies great! It responds smoothly to inputs, cruises at level altitude at trim throttle, and circles beautifully in RTL after coming back to me. Because THR_MAX is 50%, it never exceeds an airspeed of 22 m/s, and therefore never enters into an uncontrolled roll. I flew it for 82 minutes (65 km) today before landing it successfully. I was tempted to try going into manual mode and gunning the throttle to see if taping the wings and elevons made the plane sufficiently rigid to avoid control reversal, but decided I didn’t want to risk crashing this time because I wanted to do a flight time test. Maybe next time, I’ll do the experiment.
My biggest problem now is hand launching the plane. Using TAKEOFF mode, I’ve only successfully taken off about half the times I’ve tried. The other half of the time, it hits the ground with the motor spinning, usually causing the motor mount to snap–which I then have to superglue together. Hopefully now that I’ve autotuned the PID parameters, my success rate will go up. Do you have any tips for launching the Nano Goblin?
With any plane, throw it only slightly nose up, maybe 5 degrees up. It’s more flat than you think. It needs to accelerate first, then climb.
There is also a dedicated takeoff max throttle parameter so you could use higher throttle for takeoff.
TKOFF_THR_MAX
Sorry for the delay, life has been busy.
As promised, here is my parameter file of the nano goblin: Nextcloud
This is not tuned to the bone; once it did what I wanted it to do I stopped messing and started flying.
Regarding your questions:
6mm in each direction (6mm up, 6mm down) is 12mm total, which matches your elevon throw nicely?
With the wings and elevons stiffened with tape speed should not be an issue anymore. Although your 2207 motor with that high pitch prop is still really, really overkill. It destroys your ability to easily get the CG right, puts too much torque on the foam and in a crash it is too heavy and breaks.
I am running a 1407 with 4" (4x2.5 I think) prop because I wanted 5mm shaft instead of T-mount and that is already excessive.
Launching the nano goblin: In my opinion it is a very easy to launch plane. I have Arduplane setup to start the motor after the launch (I managed to cut a few fingers to the bone by handlaunching a pusher plane. Blood, hospital. So I am careful with spinnng props nowadays). A gentle horizonal throw is all that is needed to get the pane airborne. It doesn’t even have to be against the wind.
In your case it is weight and that excessive amount of torque from the large prop and motor which makes launching the plane more difficult.
Do yourself a favor, and get the stuff needed to operate the plane in it’s intended operating envelope. 1306-1508 motor, 3-4" prop, 1000-2000mAh 2-3S Lipo, takeoff weight <300g. That will provide the most smiles per mile and allows almost quad-like freestyle capabilities.
However, if you want a long range FPV cruiser, get a larger plane. That class of planes is not entirely my cup of tea so others can probably provide better advice, but I had an ARWing Pro once which could fit an insane amount of battery. Not a beginner plane either though, which I found out the hard way.
Thank you again DaBit! I’m in the process of building a second Nano Goblin with the “lightweight & efficient” kit, which will be about 210 g. I might add a FPV system to it, but it’ll still be around 250 g. I’m looking forward to seeing how differently it flies from my first plane.
But speaking of my first plane…after taping the wings, it flies perfectly even at full throttle! I made it do full throttle cruises, climbs, and dives. No problems whatsoever. It reached an airspeed of 39 m/s (87 mph) without a hint of instability.
For takeoffs, I’m also using the automatic takeoff mode. I’ve gotten my launch success rate up from <50% to >75% by setting TKOFF_THR_MAX to 100, and TKOFF_THR_SLEW to -1. The latter was causing lots of problems because the default is 100, meaning it takes a whole second to go from zero throttle to full throttle. An object in freefall from 1.5 m reaches the ground in 0.55 s without air resistance. Of course a plane has plenty of air resistance, but one second was still much too long. I’m thinking of setting TKOFF_THR_MINSPD to 0 as well (and maybe bumping TKOFF_THR_DELAY from 2 to 3 to compensate); I set it to 4 m/s as an additional safety feature, but the GPS velocity estimate is noisy, so sometimes the motor starts too late. What are your automatic takeoff settings?
- 6mm in each direction (6mm up, 6mm down) is 12mm total, which matches your elevon throw nicely?
I have 12 mm of travel in each direction. I should mention that 12 mm is with both max pitch and max roll. If I just do max pitch or max roll, the travel is 6 mm.
Thanks, I’m trying. Is it OK for the pitch to be ~20 degrees during the acceleration phase of the throw, if it leaves my hand at 0-5 degrees pitch up?
Yes, it’s common with wings to have to do that. But you’ll find your timing needs to be really good to get it level by the time it leaves your hand…
