CranialRectosis' Twitchity Mini-Tricopter build

cranialrectosis

Faster than a speeding face plant!
Mentor
Yep, looptime will affect the PID's. I see Twitchity has his up at 3500 which is from what I understand too high to really benefit anything (it's faster than the ESC's can update so the FC is wasting effort calculating updates that never get sent to the ESC's)

I'd try dropping that down to 2500 like CM first. PID's are pretty close between the two so that looptime would be what I'd look at first.

Is that right? I thought the higher number was slower on the looptime?

I use 2500 on my Simon K and BlHeli copters and ran 2000 on the Hex when I ran 12A KISS ESCs on them. Was a noticeable difference.
 

Twitchity

Senior Member
Well I tried Codes settings with an update to the looptime at 2000, and motor pwm rate at 490. The tri flies much better (at least inside because it's too dark outside). The tail seems to snap back real fast when I let off of the yaw. I tried to capture this in the video but it was a little on the difficult side. This video was filmed at 120fps from my iPhone so it's not the best of quality. At the 5 sec mark you can see where the rear motor goes back and forth a few times to even the tri out. It seems like its overcompensating for the yaw, but I could just be used to my quad. Outside of that it seems very stable in the air so far.

 

jhitesma

Some guy in the desert
Mentor
Is that right? I thought the higher number was slower on the looptime?

I use 2500 on my Simon K and BlHeli copters and ran 2000 on the Hex when I ran 12A KISS ESCs on them. Was a noticeable difference.

That's what I get for posting more than I intended when I'm in the middle of a frustrating project at work :)

Yes...the smaller the number the faster the loop runs. But lower than 2000 doesn't really have any benefit since 2000us is 500hz which is the fastest any ESC will update.


WolframAlpha is fun....in 2000us light will travel 373 miles in a vacuum, but in 3500us it will go almost twice as far: 652 miles.

2000us is also:
~~3 × time for which a baseball being hit remains in contact with the bat (~~ 0.7 ms )
~~ 40 × sampling time for the highest-frequency tone audible to humans (~~ 0.05/kHz )
~~ 88 × sampling time for 44 kHz digital CD audio ( 0.02273/kHz )

0.002 seconds
500hz

That means at 2000us looptime your FC is updating it's loop 500 times a second.

Kind of gives you new respect for just how much is going on in our FC's.

Thinking in these timescales amazes me...but sure gives me a headache!
 

jhitesma

Some guy in the desert
Mentor

Well, currently when you set the PWM rate higher than 499hz on Naze it goes into brushless mode. So I something would have to change there in the FC code. But a looptime of 1667 would match 600hz if you could do 600hz PWM out.

Note - the ZTW's aren't "running at" 600hz, they can just accept up to a 600hz input signal.

And other than the KISS the big limiting factor on ESC's isn't really the PWM rate they can run at. It's the rate at which they can turn their FET's on and off. I haven't had time to catch up on the experiments with these new ZTW's to see how they stack up - they mention using newer fets but that doesn't necessarily mean they're any faster than what's already in use. But like I said people were still waiting to get their hands on them and test them the last time I had time to read up on them and I've been too busy to catch up on seeing what the final findings are.

The ability to turn the FET's on and off quick is what gives KISS their edge and their ability to do such great regen braking and to take full advantage of oneshot. One of the things the guys testing oneshot on blheli are running into is that most esc's just don't have fast enough fet's or fet drivers to take advantage of oneshot.
 

cranialrectosis

Faster than a speeding face plant!
Mentor
I got a little build time in today, and have the copter assembled. All the motors spin correctly and the servo works. I just need to maiden it and tune it.

The copter is tiny and weighs in at 285 grams. It feels SOLID. It is bizarre how as I move the copter with the lipo connected and in a disarmed state the servo tries to counter my yaw. :)

I have updated the photos at the top of the thread and here is the latest:
P3100003.JPG
 

cranialrectosis

Faster than a speeding face plant!
Mentor
Rats.

No fly today. The number 2 motor magnets seem to be coming unglued and are falling out of the bell...

Danged crapola Sunnyskys...
 

mpbiv

New member
Rats.

No fly today. The number 2 motor magnets seem to be coming unglued and are falling out of the bell...

Danged crapola Sunnyskys...

That's a bummer! Hope you get it fixed soon.

Just curious was that 285g with or without the battery? If that is with the battery, that thing is wicked light.
 

cranialrectosis

Faster than a speeding face plant!
Mentor
ZTWs have no BEC.

Yes they have three wires coming off them, mystifying to me too. These are ROHS compliant ESCs and use leadless solder. First thing I do with them is strip off all the wires and ALL the solder from the motor pads, power and ground pads and the signal pad and re-tin with the solder I use to get a better connection.

Not that I don't like leadless solder (I don't, it's hard to melt), but that mixing solders provides a poor connection. I do this for any ESC leadless solder or no so I may have the best connection possible.

With all the wires stripped off, I am free to use my own Arduino wires in the colors and lengths I choose to provide signal to my ESCs. I only connect signal wires reducing the wiring substantially.

I like the Pololu voltage regulators. They are much smaller than a dedicated BEC as they don't have that clunky capacitor. Because I am powering a servo, I am using the 1A, 5V output step down voltage regulator which is connected directly to the 12v power harness. I run smaller servo wires as 5v power and ground output from the Pololu up into a 3 pin Dupont, add the signal wire for ESC1 and connect the 3 pin Dupont to the Naze at M1.

The toughest, most time consuming part of the process is removing all the solder and all the hot glue from the ESCs and not zapping them with ESD or overheating them while soldering directly to them.

I should have taken more photos of the wiring details and I may yet. I have everything except the Pololu under the Naze and wired correctly but with a little more effort, I think I can get the Pololu up under the Naze as well and make for a cleaner build.

It's -28 outside right now. It's a better day for soldering than flying for sure. :)
 
twitchy, i came here to track down tricopter naze issues.... i've set the servo_pwm to 100 (digital), haven't reflown yet....
will let you know how it goes. everything else is stock pids.

Twitchty, what software do you use for pcb design?
 

Twitchity

Senior Member
twitchy, i came here to track down tricopter naze issues.... i've set the servo_pwm to 100 (digital), haven't reflown yet....
will let you know how it goes. everything else is stock pids.

Twitchty, what software do you use for pcb design?

I've been using DipTrace for all of my PCB designs http://diptrace.com/download-diptrace/. The free version should be more than enough to create anything you could need on a simple board for multirotors.

The completed boards are ordered from OSH Park https://oshpark.com/
 

cranialrectosis

Faster than a speeding face plant!
Mentor
I was able to get outside and maiden the tricopter today. The yard is still full of snow and ice with some muddy patches showing through so I maidened over the concrete (also icy as heck) patio. At first, the copter spun out of control. Mustang pointed out the yaw reversal on Baseflight in the servo tab. Checked that and the spinning is done.

Note to self, it is easier on rotors to test yaw direction with the props off and in your hand in the shop. Yaw in any direction with the motors spinning, the copter should push the motor to counter. If it pushes the motor to augment your yaw movement, don't fly it over concrete. :)

I then reduced yaw P to 6 and tried again. The copter toilet bowl wobbled and ascended out of control and was again dumped on the concrete. I LOVE zip tie landing gear. Yes it bounces but nothing broke.

Now with P on all three axes cut in half again, the copter flies. I need serious tuning but don't really want to go until the ice melts off the lawn and I can crash in something that isn't snow or concrete.

As it is, it feels light in the air and responsive to throttle (not heavy). I just have some drifting issues. I haven't tried horizon mode yet, just acro mode, but the copter flies well enough to tune once I get a break in the weather. :)
 

cranialrectosis

Faster than a speeding face plant!
Mentor
I am still working on tuning this little monster. It has power to spare and is pretty light so flying in acro mode with the tuning all out of whack has me bouncing a lot.

The copter is yawing to the right. It is a slow, steady yaw that I am not wanting to trim out in the transmitter but would prefer to work out in the build. I am spinning both front motors CW and the tail rotor CCW. My tail rotor seems to want to center tilted so as to spin the copter to the right. This is likely a centering issue or a Naze tuning issue that I haven't quite figured out yet.

So far my bouncy zip tie landing gear has kept the nastiest crashes from hurting much and I now have a muddy patch in the front 'lawn' in the middle of the icy patches over which I may fly and dump the copter when I make it unflyable in tuning.

Most of the oscillation is tuned out, I just need to figure out this yaw thing and this little copter will finally be buzzing the squirrels.
 

Twitchity

Senior Member
I'll need to adjust the mid point on the servo tab for my tri as well as it has a slight left yaw.

I'm surprised at how zippy this little tri is running 1806's and 6045 props. Hopefully the weather will get nicer for you soon. It's in the 70's and sunny here in Florida :p This is the only time of the year I enjoy living in Florida.