View Full Version : Flipping!
:) At least it landed on the right side. I had put on some long screws to protect the PCB if landed upsidedown...
09-28-2010, 05:20 PM
Maybe you should try a complete roll next time instead of counteract ;)
09-28-2010, 05:57 PM
I stand by my word - PM me your address and I´ll send you an APC set :D
PS: Offer not available to other forum members due to limited stock ;-)
Hehe. Well, I think the greatest mistake I made was that I didn't increase throttle enough on the way up, when passing 180 degrees. But hey, it was worth a try! Gonna practice more later. Too bad since I just got stable mode working :) It was windy though...
By the way lokling, where do you get your APC's from? The Swedish dealer takes so much time delivering them! At least the last time...
PM sent :)
09-28-2010, 06:34 PM
I ordered a huge batch from the US directly - that arrived in about a weeks time, but you need a big batch to get the shipping and tax down to a reasonable price. So I have 5 prop sets here of each size. Thought I´d go through them in a week, but they are harder to break than I thought.
The german cnc-modelshop scammed me for a set, paid upfront using VISA and never heard from them. http://micronrc.co.uk/ has good service, I´ll probably buy from there next time. Then again, maybe its an idea to get a big order directly from the US and split it up here between us? :-)
09-28-2010, 06:53 PM
On the topic of practice and saving props: Here´s an idea I´ve had: A simple simulator that would take logs from the accelerometers and the motors/RC commands to learn how a specifc quad reacts. The physics simulation is quite straight forward as theres only one object, the quad, to simulate - and with actual logged data I think it would have a chance to feel quite realistic? With a little effort we could have this connected to the RC controls and have our own little trainer. If you agree this sounds doable I´m willing to give it a try. I code better than I fly ;-)
Ok, I get it. Well, yeah I would really like to split a big order since practicing and coding new stuff really eats props :P
Well, what do you think about, like a graphical simulator? I was thinking in terms of coding an auto-flip function! That regulated throttle and roll throughout the flip to make it awesome! I think that I was just too slow on increasing throttle in time, but an auto-flip function in the code would read off the actual angle (absolute) and make the flip very fast and effective. Since we have barometers we can also check that we are at a good height for doing a flip before activating :)
09-28-2010, 07:24 PM
Main effort in getting the simulation quite accurate by learning from actual logs. The physics dont change of course, but how it reacts to throttle could be derived from the logs. Graphically it would be something quite basic in OpenGL 3d.
Main value of something like this would be experimenting with manouvers. Both for yourself and any logic we would write - I guess we in theory could have the actual arduino/aeroquad software hooked up to it aswell, with fake sensors and such. I´ve got my IDE up here now to do some first work on it.
Hmm, yes you're on to something! Real environment logs are really useful of course. Start a thread about it and more people can help!
Have you checked out Processing by the way? It is what Arduino IDE is built upon, and it's used for this kind of logging and graphical use. I've never worked much in it but I guess it's a little like OpenGL, but more "Arduino adapted".
09-28-2010, 07:57 PM
Yes, I´ve been playing with processing. But as I´m working with Java for a living processings simplifications just make me frustrated :-) Great project though, bringing a tool like that to the masses. I´m playing with implementing this in Scala here now - (scala-lang.org ). Partly because I want to, need some training in it. Partly because it is well suited for creating a "domain specific language" so its accessible for non-coders. Ie create code that actually reads;
quad throttleTo 50%
quad yaw 10deg left
quad roll 90deg left
and so on..
Honk, looks a bit like my latest ArduCopter crash.. Ok I whacked it from 3 meters with full throttle to concrete due were doing some testings.. Point is that do not to loops or loop tests close to ground if something goes bad :)
I was doing autoleveling tests in form of throwing quad up in sky with 0 throttle. After upwards motion was stopped, throttle up and see if it levels properly.. Well it did level properly 9 times out of 10.. It would have do it 10 out of then 10 but ground was just too close so result was WHACK.. Broke 2 props, 6 nylon screws, one motor shaft (cut in half) and hole on ground. 10 mins later it flew again :P
09-29-2010, 08:05 PM
...Too bad since I just got stable mode working :)...Sounds good. Can we see some vids of stable mode next time?
lokling, I checked Scala and read some about it. A guy I know also work with Java and told me about Scala too. I must say I was a little disappointed looking at the syntax and structure but, it's probably because it's a new thing to me. Tell me when you get somewhere with it! I'll try to log some useful flight data when I'm back up in the air and send it over.
JP, Hehe, but why full throttle into concrete? I have noticed with quads that they like throttle to level out. When I was trying to learn to recover from low throttle I think it like inherently stabilized itself just because of added throttle. I have also learnt not to be so close to ground, but actually it was kinda hard to push down and up the throttle stick in time, especially since I have your silicon tube idea on the throttle stick making it slow :)
RCvertt, Absolutely! I'm dying to get it "proved"! :) I was amazed how much vibration did harm the process of autoleveling, when I tried it 2 days ago with lower vibration props it did autolevel fine in my hand. Quite slow but certainly enough, and then I could push it up into the air. And the barometer is implemented in my flight version, but I found some bugs with it when I was flying before, might have been the strong wind though.
So I can pretty confidently say that the Nintendo Nunchuk unit (alleged ADXL335 with more than 10bit AD) can handle angle estimation good enough for autoleveling in flight in certain vibration conditions. And I don't think it was a one-timer. I posted my values for everything in the "Low cost..." thread.
Powered by vBulletin® Version 4.2.2 Copyright © 2014 vBulletin Solutions, Inc. All rights reserved.