I few days ago I took this short video of my stock DJI Phantom and stock 3DR Hexacopter on position hold side by side. What I want to share is with the new 3.0 beta firmware Arducopter now performs very similar to DJI on the part where DJI until now has been the best alone.
Being user of both APM and DJI flight controllers it is very inspiring to follow how APM Arducopter gets better and closer to DJI performance for every new firmware release. This new beta release is a milestone for the Arducopter project and the Developers have done a great work with improving the firmware.
From video you can see 3DR Hexacopter performes better than DJI Phantom. Phantom flies good as always. But compared to 3DR Hexa I noticed it moved nervous up and down and around the spot the whole flight. At last Phantom was drifting under Hexa and I ran out and grabbed it to avoid collision (not in video). I think difference in performance may be model spesific, but this show Arducopter now also is very good on position hold!
DJI Phantom from video is complete Stock With the newest firmware release.
Arducopter from video is 3DR Hexa with 880 motors and 11x5 Carbon prop (RCTimer), 3DR ublox GPS, no sonar.
APM 2.5 with Firmware 3.0.0-rc6. All PIDs default from rc6 release.
Comments
Bob Thank you for your feedback about the Phantom behave. I was thinking maybe my Phantom was faulty before you shared your experience. I also agree Naza on F-models behave better compared to my Phantom.
I have a pair of Phantoms, an F330, F450 and F550. The Phantoms, F330 and F550 have V1 Naza with the 3,12 (V2) firmware and the F450 is V2 hardware and firmware. They all have the GPS module. All of the F series are more stable in hover than the Phantom, and side by side comparisons produce similar results to your video.
I've noticed that flying weight, gimbals/landing gear, and propeller size all have effect on stability in hover. I've also noticed that advanced IMU calibration every few flights adds a lot to stability with all Naza based DJI aircraft, even when the Assistant app says it isn't needed. It's interesting to compare the various brands and configurations though. It looks like the 3D Robotics unit you built is pretty rock steady. Way to go!
Yesterday I did two repeated auto mission With Automatic landing. I did mark the spots and distance was 30 cm apart. I may be Lucky With my setup. But I also want to share Picture of my Stack.
With all the talk about separating Compass from Electric noise please not follow advice from 3DR manual to put APM on bottom deck!
Put the biggest spacers 30mm in bottom, and Place APM on the second deck. Receiver and 3DR radio to bottom deck. You now need 4 more 30mm spacers to the Next deck and they are unfortunately not supplied in the kit.
You need more 30mm spacers, or combine the 18mm included spacer with the 8mm spacer supplied for the Power distribution Board to get enought distance.
On top deck put the GPS With doubleside tape.
For my model I have not done compassdance or compassmot. Only a short flight with "compasslearn" enabled and then turn off again.
Nicely done. Appreciate the effort in making the video and demonstrating the effectiveness of both systems.
Well, first of all nice work on the arducopterside! I myself flew an naza 1 unit with gps on my bigger Quadcopter (55cm motor to motor diagonal with 1,3KG) and never experienced that altitude up/down wobble seen in the video. Like mentioned by F1P something heavier is always more stable. Besides that i had Hacker motors and perfectly balanced props on. So that setup might exceed the Phantom specifications. The gps was also very solid working. On the other hand naza became too limited and boring after a time. I also had to overcome some issues with naza that were caused by EMI and solved by an alu box. I read some reports, that the current naza FW3.12 can cause severe problems and ground your copter/put you out of control. Well I put the naza out of order with some 2.X Version on it, don't know what actual problems are based on. It waits in the box for future usage.
Cheers Rob
Different class and weight.
But DJi behavior is very strange in this footage.
One difference I see between Naza and Arducopter is Naza is easier to fly (for beginners). This is because of softer RC setup on Naza. And the same reason I prefer to fly Arducopter myself :)
Arducopter is setup default from experts to experts. If we asume all unknown new pilots is a beginner on default setup we can make Arducopter easier and safer to learn. And then let the experts take a active choice to make settings more responsive.
I think we need two important parameters to do this:
- Maximum tilt angle. Now is 60degrees. I compile with 45 degrees for my kid.
- Rate of RC input. To reduce RC input ("servotravel") without also reducing stability PIDs. This must be done on the FC and not transmitter.
It is so good to see these comparisons being made so favorably. I suspect any perceived shortcoming of the DJI stuff is their choice of conservative or softer tuning or simply bad luck with regards to GPS reception.
We still have a lot to learn from DJI when it comes to ease of setup and tuning. However I am personally more interested in dynamic flight performance comparisons. By this I mean how DJI handles fast corners, decelerations, fast descents ect. Is there more we can do to improve the controllers, if so where do we need to look?
Thank you for the post.
Excellent demonstration.
-=Doug