My big(15 lb) quadcopter project

[dirty_d]

I haven't been on this forum In a couple years but I thought I'd share a project I've been working on for a couple months. Its pretty much finished it last night and did a quick 1 or 2 second hover test in the living room

here are the specs:
4x 5055 size 1600W 700kv motors
4x 60-70A ESCs (these are $16, amazing!)
4x 14.8V 5Ah LiPo
4x 15"x4" APC propellers
spectrum DX6i transmitter/reciever
control hardware and software I designed with an atxmega microcontroller and MPU-6050 16-bit accelerometer+gyroscope
the frame is 0.75"x0.062" 6061 square tube and 0.062" sheet 6061 which is very much overkill and one of the reasons it's heavier than I planned
I haven't weighed it but I'd guess it's around 15 pounds which is higher than the 8-10 I planned on

I finished putting it all together and added some simple fiberglass landing gear which unfortunately almost flatten under the weight. I did a very nervous hover test and it did manage a very stable hover for 1-2 seconds at about 40% throttle before I wimped out and let it land(fall). The landing gear did nothing and it bottomed out on the floor, but it's ok. I think I'll just double up on the fiberglass rods and that will be ok.

here are some pics and a video of the first time I tried to hover a few inches at about 20% throttle without the batteries attached.

Video
http://www.youtube.com/watch?v=-1Qf4pqjAqY

[mrtc]

This looks awesome! I'm trying to build my own right now but it's definitely not as big as that.

Any idea what sort of payload capacity it has?

[liveforphysics]

Nice job! I like it!

[dirty_d]

Thanks. I tried my first flight today, I think I need to adjust my PID, lol. The payload should be somewhere around 15 pounds, I planned for 20 but i don't think its going to do it. Right now it weighs about 15, and it hovers at about 40% throttle.

Video:
http://www.youtube.com/watch?v=cBCbTRqPMTY

[adrian_sm]

Pretty cool to see it running. Looks like a bit of a handful to control at the moment.

What can you tell us about the control system. Are the ESCs stock, or have your modified the firmware to speed up the response, like this: http://airhacks.org/?p=445

[dirty_d]

Yea its uncontrollable right now. I had it up about a foot off of video for a couple seconds and it started oscillating back and forth very fast. From the physics simulation I wrote before building it, I can tell that that's either from too much P, or an ok amount of P and not enough -D. The integral term is 0 for now. Ill have to play with it for a while to figure out the right tune. The xmega is running a 400Hz PID control loop. The ESCs are the stock HK SS 60-70A models. They have an atmega8L micro on them, so if i have to reprogram them I can. At 400Hz throttle input they are pretty snappy though. Due to the size and weight of the quad, I think the throttle response from the stock ESCs at 400Hz will be fine. I think as the size of the quad gets smaller and lighter the response time of the ESCs becomes more of a problem because the quad can be accelerated a lot faster by any destabilizing forces. I have a field programming interface on the control board for adjusting all of the PID values etc using an led, 6 switches and a potentiometer. That should make tuning it a little easier, although the external programming board is all implemented on a breadboard for now.

[adrian_sm]

Do you know if the ESC actually responds to all the 400Hz signals? I thought there was some sort of low pass filter.

Check out the first video in this thread http://aeroquad.com/showthread.php?328- ... -for-speed. It shows the difference one guy saw before and after tweaking the firmware.

Oh and nice work on developing your own control board. I would be lazy and just buy one like this: http://www.hobbyking.com/hobbyking/stor ... duct=19534.

[dirty_d]

No, it doesn't respond to the input signal at the same 400Hz that its given. There is a digital impulse response low-pass filter in the ESC. But The filter is designed to work with a 50Hz input signal. When you feed it 400Hz it shifts the cutoff point upward because its based on the time between pulses. So where the cutoff freq for a 50Hz input might be 5Hz the cutoff freq for the 400Hz might be 40Hz (I see a huge difference in throttle response between hooking the ESC straight to the reciever or to my control board). People have made much smaller quadcopters with standard ESCs at 400hz so I doubt I'll have a problem getting this to stabilize. I just use a wild ass guess for the pid values, p=800, i=0, d=-50. I think I'll try p=400, i=0, d=-160. With the i term being 0 the quad would not be able to hover level if i.e. there's more weight to one side. I'll worry about the i term after it can fly relatively stable.

[adrian_sm]

No, it doesn't respond to the input signal at the same 400Hz that its given. There is a digital impulse response low-pass filter in the ESC. But The filter is designed to work with a 50Hz input signal. When you feed it 400Hz it shifts the cutoff point upward because its based on the time between pulses. So where the cutoff freq for a 50Hz input might be 5Hz the cutoff freq for the 400Hz might be 40Hz (I see a huge difference in throttle response between hooking the ESC straight to the reciever or to my control board).

Cool. Thanks for the low pas filter explanation, that saves me some time and effort. I might just have to try out 400hz for my throttle interface and see what happens.

People have made much smaller quadcopters with standard ESCs at 400hz so I doubt I'll have a problem getting this to stabilize. I just use a wild ass guess for the pid values, p=800, i=0, d=-50. I think I'll try p=400, i=0, d=-160. With the i term being 0 the quad would not be able to hover level if i.e. there's more weight to one side. I'll worry about the i term after it can fly relatively stable.

I usually get away with just PI no D. If you do use D make sure you have a pretty clean signal. As the D term is based on the derivative, it will be a lot more sensitive to noise in the angle readings from your gyros.

- Adrian

[dirty_d]

You used only PI on a quadcopter or some other project? I wrote a pretty accurate simulation in c++ before I began building anything and i was unable to get the thing to stabilize without a D term. the quad would oscillate itself out of control and flip over, with a D term it was rock solid stable.

I did as much PID tuning as i could in my living room, hopefully the next outdoor test will be better.

[Hillhater]

Isnt it amazing what technology can do ...
I use PID control to make my morning coffee !!

[adrian_sm]

You used only PI on a quadcopter or some other project?

No very different applications. Looks like you have got it covered.
It has all been done before so you should have heaps of references like this to help you out.
http://aeroquad.com/content.php?126-PID ... r-AeroQuad

Anyway, I'll shut up now since I am not contributing anything.

Keep up the great work.

[dirty_d]

Hmm i just read that, it said to reduce -D to reduce oscillations in stable mode. That's strange to me. Unless by reduce they mean make it more negative, but they say reduce it by -0.5, so id think they mean increase it towards 0. I'm going to try to alter my simulation a bit to try and simulate some of the delays involved better, like the ESC's IIR and see what happens.

[dirty_d]

While working on the C++ simulation I noticed that I made a mistake in my PID code on the micro-controller.

The statement for calculating the D term was basically:
d_term = (pv - last_pv) * dt

Where it should have been:
d_term = (pv - last_pv) / dt * d_weight

Sure enough after changing the corresponding code in the simulation to the incorrect code on the micro-controller the quad would oscillate out of control and flip over. Basically I've been trying to fly with no D term. Hopefully after making these changes it will fly stably.

Also, I added an IIR filter to the ESC class in the C++ simulation, It does increase the oscillations, but its brought into control by more -D. For the curious, the IIR filter was pretty simple to implement Its simply "throttle_filter = 0.95*throttle_filter + 0.05*throttle;" in my code where throttle_filter is an accumulator an the output of the filter, and throttle the instantaneous throttle input. I guess this is a single pole IIR filter. I'm curious to find out what a multiple pole filter would look like.

[dirty_d]

Well I think i have the PID code all fixed now. I could tell with my indoor test that the -D term is definitely kicking in now. I set it way too high though. Just the vibrations alone were amplified and apparent on the motor output which made it shaky and took 75% throttle to hover, it did seem to stay level though. A lower -D term and maybe some filtering on the derivative should fix everything.

[dirty_d]

I've been real lazy since getting back from Vegas. I need to get this thing outside one of these days and get it tuned right.

[Xanda2260]

That thing looks awesome! Just got a microquad (mQX) and I love it. May have to try something like this soon...
Keep up the good work!

[bzhwindtalker]

I am moving away from quads because of the propellers price. Hard to find quality props pairs too. I am braking way too much of them when fooling around with my small quad! I will probably build a big tri instead.
Btw, your quad looks like a really nice death machine, beware when testing those can be uber dangerous

[grindz145]

Just needs a grapple hook and an NFC pay device, and it will your personal milk-buying robot.