Impossible LR

[jihlein]

Well come to find out the impossible tilt for the Baby Tri (https://www.thingiverse.com/thing:2869393) is a direct fit for the Tri LR. I retfited mine this morning. Curious to see if it helps tame the tail any, making it easier to tune, etc. All my other tri’s are outfit with variations of this unit, so off we go. Here’s a few photos of the mod.

Attachments:

1. 

   IMG_1907.jpg
   

2. 

   IMG_1908.jpg
   

3. 

   IMG_1909.jpg
   

4. 

   IMG_1910.jpg
   

[jihlein]

Well FWIW, adding the impossible tilt was a great change. I went back to the dRonin firmware (the latest version), and reset everything up from scratch. Used the stick cmds to set the servo angles, worked like a champ. Servo speed ~250 DPS.

But the the better news is the autotune worked very well. I did the autotune, came back inside and applied the new gains, zeroing the yaw D gain as I typically do. Back outside, it flew nicely. No odd tail drift behavior or oscillations that I could tell. This was all LOS flying.

Completed the in flight servo cal on the first attempt (further proof the yaw is behaving correctly). Motor thrust factor in my configuration computed to be 12.9.

Continued flying around the backyard in hover and slow forward flight. Acro, level, and horizon flew as expected. Altitude hold needs some tuning (and some closed cell foam over the BMP280). Did not try position hold. I suspect that after tuning alt hold and pos hold, RTH just might work correctly.

After 12 or so minutes of flight, landed to check motor and ESC temps. All good, not really much above ambient, so I was happy with that.

The impossible tilt is a worthwhile mod in my opinion.

[yumemi5k]

That seems promising! Do you think there is a significant improvement to yaw performance? My tail is either sloppy or wobbly and I’ve almost learned to fly with a sloppy tail.

[jihlein]

Without a lot more experimenting, I can’t say for sure, but with 12 minutes of flight after tuning it sure seems better.

My yaw P/I gains are 110/408, directly from the autotune performed with the impossible tilt, and I observed no oscillations or sloppy heading control. I believe this ratio is backwards from what most are flying, which is an I term significantly smaller than the P term.

I could not use the autotuned gains computed with the standard yaw tilt mechanism, I had to play with them manually to get something flyable, but I was never happy with the results.

Hope someone else eventually tries this too and reports on there results/observations.

[Kevin_Erik]

@Jihlein,

What your seeing is the Mass of the Motor / Propeller as well as their Gyroscopic-Mass (GM) is being held closer to the Servos pivot point. This reduces the amount of induced Roll that results from the copters Yaw.

Case-in-point: Your copter Yaws Left, thus bringing its Tilt-Mech Mass / GM Right, inducing Right Roll that the copter must now compensate for. The compensated Roll, induces both Yaw and Pitch changes that also must be compensated for and so on and so on. While some of this interplay can be compensated for or dampened by the Flight Controller, this doesn’t always work due to the Servos slow response time.

The same is also true for the Mass / CG being too far above or below the copters true CG. However not to the extent seen from the Mass / GM being held far from the Servos pivot point. Now by using the shorter Tilt-Mech, the amount of interplay between each axis is drastically reduced. The closer the Mass / GM is held to the Servos rotational axis the smoother / faster the Tail-Mech can respond. (This can be seen in the tail calibration results)

The Gyroscopic mass is based on the propeller length / mass. Which is why smaller Tricopters with lighter / smaller propellers appear to behave better. This IMO is one of the short comings of dRonin managing very small or very large copters due to the differences in propeller GM. If you don’t believe me, try running a large copter that has tuning issues using smaller propellers. The difference is night and day…

Beyond all that, I’m still a fan of the “inverted” Tail-Mechanism. Simply because the copter Rolls in the direction of Yaw. Making the overall flight feel smoother and more predicable. Attached is a Step-Response from a 3″ inverted Tricopter I’ve been working on. (125-130 grams with 4S LiPo) While there is a little bit of P overshoot, the rest of the plot is really smooth. Keep in mind that I’ve yet to tune this copter and am currently using the BF (4.06) default PID’s. (except for Yaw) Previously, i built up a non-inverted version of this copter and the Step response wasn’t nearly as clean.

Cheers!

Kevin

Attachments:

1. 

   fusion-360-image.png
   

2. 

   LOG00028BF-Defaults-Oct-9-2019-4-minute-flight-with-63-percent-battery-left_0_response.png
   

3. 

   3-inch-inverted-Tricopter-WIP.png
   

4. 

   Wiring-Layout.jpg
   

[yumemi5k]

Well, yaw control in a tri is a circus compared with a quad. With un-inverted tail, you get unwanted yaw (from mass displacement), roll (from turning servo mechanism) and pitch (because spinning tail rotor has angular momentum), all before yawing at all.

Anyway my parts should arrive in the mail in a few days, I’ll try it out asap.

[yumemi5k]

I don’t own a 3D printer so I had it printed by a service and it’s in the mail. The SLA printed epoxy is gorgeous, but the servo spline was unsurprisingly messed up. I plan to use hot glue (which should form around the spline but still allow removal), loctite and a tension washer to keep it from decoupling mid-air. Still, it could be interesting to modify the design to accept servo arms that come with the servo instead of servo spline itself, which should make precision requirement much looser.
I hope to get it set-up in a few days and try it out.

[PeterPankras]

This is interesting.

I might have to give this a try 🙂

I was thinking about a new tiltmechnism since this shop is closing.
I work at a manufacturing company, and I might be able to do one later in aluminium.

I’ve not been able to do more testing to get rid of the tail wagging because of the weather here.

So nice time to print one of these.
I’ll download and print one.

[yumemi5k]

# Editing the post made it disappear for me yet still trigger duplicate warning…

I got it fitted and it was immediately obvious that the tail was much less inclined to oscillate even with aggressive PIDs on the ground. I then took a test flight indoors, taking cover behind furniture when arming. It’s hard to say about flight characteristics from the 1-min test hover, but it flew, hovered with same current draw than before and showed no erratic behaviour, so everything looked fine.

The retrofit created a few side effects: fore props (gemfan 8045) no longer readily contact the coils of aft motor when folded, eliminating a minor worry that the sharp prop might damage rear motor coils during transportation. Servo wire strain relief needed to be removed due to prop clearance. Servo wire also no longer lie flush with tail arm as it’s not long enough now. It was THE ONE wire that I cut to exact length in this whole build… Servo mounting tab did clear props, but with pretty tight margin so I trimmed it.

I was initially convinced that the servo spline had too little material and I needed to use glue, turned out that I had the opposite problem. The spline was still there though, and by carving with a pointy tip I was able to make them mate. The SLA printed resin felt tough but probably also brittle as I was able to remove material from it pretty easily. I also just learned that it will continue to become more brittle with UV exposure so I gave it a coat of black acrylic paint. Exact material properties will change with your choice of raw material dramatically – I was talking of my Chinese $10 incl. expedited shipping SLA job. I’d say that a well-tuned FDM probably is better suited to this application than SLA in terms of structural soundness.

Anyway I’m glad it worked till now, gonna take a proper test flight someday and I’ll report back.

• This reply was modified 7 months, 1 week ago by yumemi5k.

Attachments:

1. 

   2s-1.jpeg
   

2. 

   1s-1.jpeg
   

3. 

   3s.jpeg
   

4. 

   2s-2.jpeg
   

5. 

   3s-1.jpeg
   

6. 

   1s-2.jpeg
   

[Kevin_Erik]

Just my opinion mind you but you may run into problems with the prints being fragile around the splines contact area. One good hit and while the servo maybe be fine, the tilt mechanism will be drastically off center from it.

SLS Nylon and injection molded parts are by far the most resilient.

[PeterPankras]

I would not fly with an SLA part. Too brittle.

Better to use FDM with at least PETG or ABS if possible.

A friend of mine has PLA filled with Carbon lying around. I might ask him to print one in that to see how well it holds up.

Or have one printed in Nylon

[ME-tan]

I printed mine in SLA and it seemed fairly strong, wasn’t aware of brittleness issues with that kind of print so I’ll take that into consideration. I can destructively test a piece and compare possibly as I have access to both kinds of printer.

I’m still planning the rest of my build and haven’t bought components yet so I’ll see how this thread develops. Out of interest, is the testing for this happening on the Kakute or the Matek FCs? I see people talking about using both and haven’t decided on which one yet.

[yumemi5k]

In my test flight today the rear holder screw backed out, and the servo-side screw held the whole thing for god-knows-how-long. Adding a nut between motor cradle and back holder should help, but for now I just replaced the embedded nut with a nyloc and hope that it at least doesn’t back out mid-flight. I also wish to add that the SLA epoxy motor cradle held together with tail screw missing, if it failed I’d have had a messy and mysterious crash…

Can’t say for flight performance as it’s a bit windy and I struggled with tailtune, but I think at least vibration is now a more pressing issue for me than yaw holding. I might look into soft mounts for camera later, as it’s seemingly not common for a 8″ commercial drone to have hard-mounted camera and I might learn from them.

As for flight controller I’m using a Matek F405 CTR, they are somewhat different in feature set and you can pick one you like.

[ME-tan]

Get some blue threadlock on there. Essential for anything that screws into anything that might vibrate, but the nylock shoiuld be fine for now as long as it’s a fresh one.

[jihlein]

I simply use nylock nuts. Never had a problem with them coming loose. I also print the parts from PLA+.

The above has worked fine on V4, Mini, Baby, and BiCopter vehicles.

[Author]

Posts