Thread Tools
Sep 05, 2013, 09:41 PM
E-Copter
Bricobrac's Avatar
Quote:
Originally Posted by R_Lefebvre
Ok, so few new things.

First, I'm finding that every time it boots, the response of the "PID" is slightly different. I booted it up, using the same setting that buzzed before, and the buzz was gone. Plugged in the USB, which makes it reboot, and it started buzzing. Wondering if it was the USB causing the problem, I rebooted a few more times, with and without USB, and basically got different results every time.

Second, I did a 6-axis calibration. The results aren't great. After step one, flat level, the GUI shows it's perfectly level. Great. Then I used a machining gauge block (ie: perfectly square) to help align it while doing the other axes. However, none of the axes show straight before or after calibration, they are all off significantly. And when I was done, the "flat level" position is no longer level. I had to redo it.

Third, I did some basic slow yawing of the copter by hand. I can see the horizon going off. What's going on?

Any ideas?
Hi,

i'll answer technically about the 3 previous post you did write, just objectively and commercialy talking.

1) PID : as far as i know, a PID is Proportional, INtegral and Derivative gain tuning. True that a perfect PID system is setup based not on mechanical guessings and tuning as we do on brushless gimbals, but shall i write that to make a really accurate PID setting, it is needed first of all to have the following elements in your hands :
- Motor torque measurement
- Motor angular speed measurement
- Motor acceleration measurement
- Exact weight of the load on the gimbal
- Mathematical model of the entire assembly and " asservissement" model (compensation)
- advanced fusion algorythm knoledge and some othe things i won't write down because competitors would use it to search what they are missing in reproducing some Alexmos featurs


All of this is supposed to make give a regulation of a system based on measurement of physical units of an industrial system.

This is valid only if you know everything , mechanicaly and thoricaly about that system and that you are able to quantify or measure accurately everything about it , but it also has to be designed around perfect measurement of angles, accelerations (velocity) , and force / unit values..

In our gimbals, we have tons of unknown things :

2) The static and dynamic balancing precision,meaning that you don't need same torque and gains on the roll for example depending if your roll right or left becauwe we are unabel to set it as a standard except making it like: DJI Zenmuse (who can put basic working parameters because they have a knowledge of the mass, inertia, balance of the system and also knowledge of the motors ( maxon) exact performances charts and measurements, and because you can only put the go pro on it and nothing else, no accessory.. Using the Alexmos for "Universal" gimbals sometimes correclty mounted, sometimes not, sometime with good balance, sometimes,not, leads to only one thing : you need to work on the PID"s by yourself and find something that works + setup the power of the motors depending on the motors and load you have hooked to it.

3) MOtors precise datas : i can see weights, i can see dimensions, but i can't see any velocity / torque / acceleration / speed exact measurement, because nobody takes care to make it, everybody sells gimbal motors ( we made it for our motors on our side...)

4) Data fusion and filters : as you are int he Arducopter developer team, you should be used to the words " filters", " complimentary", " linear" and "non linear". And also Fusion. Alexmos system is based around the multiwii code and has been since fully rewritten and modified to remain closed source but ti stays not far away in term of working pattern of this. With the actual filters used in this system, it is not possible to setup a PID model even with a fully solid mathematical model of a gimbal because the datas can not be processed accurately enough, which is normal on this kind of product range, and which is already very enough for 90% of applications. I will not detail why but i am sure that you will know about it from waht you said in your previous posts so let's say that there are various way to process the datas of an IMU and correct the errors due to linear accelerations for example. Sure a Kalman will never be able to correct redings of an IMU on a gimbal that is flown in a copter.. But that's anothe story..

5) AHRS : yes, it is a flase statement and i agree on that one. AHRS is the combination of Accelerometers, Gyroscopes but also magnetometer. Don't know why there is something called AHRS and who wrote it but yes, it's false. It's a Acelerometer / Gyroscope sensor onboard, maybe a magnetometer will come later on...

6) I2C errors: here, well, the only error is to use a board from one manufacturer with it's IMU, which is an I2C device, and another manufacturer 3rd axis board that uses also I2C. reason being very simple: I2C is a simple thing but is more complicated to get working fine than it looks. The bus must stay " charged" with capacitive charges and avoid getting bothered by noises, and to work properly, when desinging the I2C interface between master / slave elements, it's needed to take into consideration the lenght of the cables, the resistance of these, and determine the right values of the resistors and route all this little world correctly depending on th various elements compsing the system. So sure thing is, if the boards are not mached to work together on the I2C bus, you will have hard time getting them work without I2C errors. A USB simple cable carries 500 mah and needs a ferrite ring in some cases to avoid bothering the I2C readings with it's own noise, having wires routed on alu parts also does not help, and adding ferrites ( right out of the Alemxos controller ) for the motors and the I2C plug can help a lot especially with strong motors. People reading the 100 and more pages of the Phillips documentation about I2C design can find out how to design a proper I2C bus. But Serial is still used in many designs because it's very strong and safe compared to I2C in some cases..

7) Your PID head scratching : Giving 50 on power, and 0 proportional gain, 0 integral and .1 derivative gain only lead to one thing : you allow 50 / 255 power range ( in fact, we could replace 0 to 255 by voltage instead which would be more accurate but would put even more the customer in doubt..). So ok, you feed your motor. But you give 0 proportional gain : so you don't allow the system to give a " consigne" power of any kind to ask the motor to position the gimbal. And 0 on I means, you don't give it anye consignee of speed also. So the motor will " lock" it's position due to the way it's driven and come at any random position or even wag in some cases.

8) Commercial approach. I think i wrote that Open source was a fantastic thing. BUt you forget the difference between open source and closed source products, in Open source everybody brings some of this time and talents to put his hands on some code or electronics and make it move forward. This takes sometimes years.. Alexmos product has been designed early this year and i think well supported by the different affiliated partners, and there has been a continous development since that time on software and hardware will follow. It brought at 110 euro a stabilisation system working pretty well compared to many with a new standard in stabiisation itself. It might not be perfect, but let's be honnest, at that price, what do you expect that is not already given ? You got a defective board ( it can happen ) and you can get a replacement asap. YOu got a DJI wookong GPS problem, you wait 4 weeks for a 1000 euro system that grounds you for a defective GPS / compass device and willget no support from a world sized company..

8) have to setup the gimbal. There is no magic board that will magically setup PID's on a gimbal composed of various manufacturers products ( motors, gimbal, camera, wires, video linik maybe, possible wrong balancing or unacurate symmetry). Some adaptive PID technologies can work out, but still need to have very precise informations on the system, dimensions, power, inertia.. A solid Math model again... So yes, tu use this board on a gimbal that is not exaclty modelisd and already setup in factory, tweaking is needed.

9) Calibration : When you do the 6 axis calibration, make sure that :

- You start witht he flat IMU and you put a small adhesive tape under the IMU especially if you are going to put it on a metal part ( level you talked about) and eve to avoid making a contact with fingers under the IMU because there are 2 spots, if you touch them wit a finger, you kill instantly the I2C reading

- You must not press" write" at each calibration. You press write only at the end ( written in manual).

- Before the calibration, it is necessary to setup the direction of the sensor., save ( write) and then perform the calibration

- Between each " calib acc" button press, wait for the convergence to finish on the right graphs ( the small white tabs must get bac to 0 around and you must be able to see some readng on the real time tab graph).

10) Jitter on USB pluging : you throw 500 mah when plugging the USB inthe board and it can sometimes cause a small hicup /. jitter due tot he capacitors if they re charged or discharged. But once in the GUI and connected, everything is ok.

So if you are part of this or that Arducopter (tht i find really awesome) project that is a that level of almost military specs ( that neds hundreds of pages of manuals documentation, aeronautical test and certification that costs around 150 000 euro for a single flight control board) , maybe one of the guys handling IMU, fusion algorythms and filters will be able to confirm the following: :

Saying that you need to scratch your head on PID tuning on a system built of various components and that none of these components has any mathematical models, precises quantifications and measurements, i'm sorry but ... it's normal . MOst users who are not used with PID's at all could manage to get the board working with not too much time and when they have difficulties they come on Rcg and ask and people answer if we are not able to answer before.

All you can do is to provide values based on most used hardware, and i will even go further, if you press the " use defaults" button on the general tabb, it will bring the I and the D at 5 and P at 10, and pwoer at 50 (from memory) which helps with starting with something. So the values you had were values that were not in a vlid profile or maybe they were erased after firmware upgrade and needed a " use default" to kick back with regular values to start with.

Last of all, in the case one day the Alexmos could be delivered with a preset that is already optimised and working, then it needs to be either for a very specific gimbal and camera that is reproduced 99% same all times, either some profiles are shared between the various users for close configurations.

But the head scratcihing is part of the deal as everybody has different equipent and installation and in that case only the user can tune the PID to get his harware driven and stabilised decently. The PID are a bit biased from what would be a real PID tuning but this is only because of the way the motors are used, and driven, so it's not matching trully a real PID system ( on that, i do agree) but well it's close enough for anybody knowing how to tune it.

I had published a video, some other did also, there are some documentations everywhere, some posts with presets and profiles, and on the first page of the thread i think there is enough dcumentations to start safe, and finaly.. In the manual, it's also written ( www.simplebgc.com)

So to make it short :

POwer : 0 to 255 : this is the max voltage you throw in the coils

P gain : it works alsmost as a Proportional gain tuning but acts also on the motor coils power in some ways. When you rise it, it starts wagging large, then faster and then you finaly get a position lock with or without compensation. For sall go pro gimbal you would set it for pitch at 7 and roll at 14 for a start

I : simply, how fast the motor is spinned to get back to position after position is altered from an external influence, also the speed for manual control, a bit dizzy on the real integral as we drive directly a motor and not realy a mechanical assmebly so it's a bit crap to setup... But with a 5 as a start it works. Too high I wil indeed act liek a real I reaction when it's too high : fast oscillation that never ends

D : will be used to stop oscillations after a fast " stop" for example or to keep the position, simply. Too high I will generate the motor buzzing, and D can nothing about it, you need to drop I or power, and then have a decent behaviour, the better the gimbal is setup, the lower the D will be. Also as many motors are used beyound their real abilities the parameters sometime can end up with some reallys trange values. The fun of the hobby lol !

LPF : will allow to smoften start and stop by applying a simple low pass filter, but the higher the value is, the weird the correction will be and bigger the correction delay also it will be...

Acceleration compensation : works only with an external FC connected and after calibration + right tuning in the FC gimbal setup software. But performances will depend on the FC you are using...

OK, i hope i answered all the questions in a decent way, i wanted to write all in one post and sorry for the headache lol !

Best regards,

Fabien
Sign up now
to remove ads between posts
Sep 05, 2013, 09:43 PM
E-Copter
Bricobrac's Avatar
Quote:
Originally Posted by flashervn
Hi Fabien, it would be nice if in the GUI, the realtime data tab can be moved right next to the Basic tab, and the "Write" button in basic tab is right next to the PID too. I know they are small detail of a really great GUI already but personally I think it d make tuning much faster and effective. Cheers
Hi,

i asked ALeksey about that already long time ago ( i will ask again) as it was a pain in the a.. to make the video and explain how each parameter changed the behaviour using the real time thing.. I asked also for another modification that would help setting up easier and will see what he says.

For the Write button maybe just adding a small write button on the PID's in addition could be helpful when tuning like ageek, a agress that woudl be helpful

BEst regards,
Fabien
Sep 05, 2013, 10:57 PM
Registered User
flytahiti's Avatar
I have alexmos and 3axis extension board from viacopter and i would like to know two things
1st do i need to flash firmware with extention instaled. because i have just plug the extension and it innitialise only when connected to pc if i don't only the main board initalize .

2nd i would like to control roll, tilt, yaw and function from RC but there is only roll and tilt plug ?
Sep 05, 2013, 11:18 PM
SkyHi Aerial
Daloaf's Avatar
Quote:
Originally Posted by flytahiti
I have alexmos and 3axis extension board from viacopter and i would like to know two things
1st do i need to flash firmware with extention instaled. because i have just plug the extension and it innitialise only when connected to pc if i don't only the main board initalize .

2nd i would like to control roll, tilt, yaw and function from RC but there is only roll and tilt plug ?
No firmware upgrades are needed for the 3rd axis expansion boards and you can use the FC Pitch or FC Roll inputs to control the yaw axis if you want. In the software RC Mapping section, you can assign EXT_Pitch or EXT_Roll for the yaw channel. You can then use the other channel if you want to assign a switch to change between profiles (such as heading lock or follow mode). Basically you can assign any of the 4 RC and FC control plugs to do other functions then they do by default.
Sep 05, 2013, 11:47 PM
Registered User
flytahiti's Avatar
do i need the 2,2 i am in 2,04b
Sep 05, 2013, 11:50 PM
Registered User
flytahiti's Avatar
sorry 2-03b
Sep 06, 2013, 02:29 AM
Registered User
Quote:
Originally Posted by flashervn
Anyone has to use "i2c pullups" to stop the errors?
Anyone? Whats the downsize of using pullups apart from potential sensor death?
Sep 06, 2013, 03:10 AM
Registered User
Quote:
Originally Posted by flashervn
Anyone? Whats the downsize of using pullups apart from potential sensor death?
Yes I use it on my largest gimbal which has very long imu wires. No problems so far and it fixed the errors that remained after careful routing of the wires and fitting ferrite rings...

Malcolm
Sep 06, 2013, 03:51 AM
Registered User
Hello
I have try to connect com port on my computer but it cannot connect to comport
I try on Win7,8 it not work but XP is working ,
Anyone use Win7,8
Please advice me
Sep 06, 2013, 04:00 AM
Registered User
pardoez's Avatar
Does anyone have a electric diagram of how to do such a tilt button on should close?
http://www.ebay.com/itm/New-PS2-Joys...item4d10d93c2a

I added some else it than just this button?
Sep 06, 2013, 05:07 AM
E-Copter
Bricobrac's Avatar
Quote:
Originally Posted by golfhaha
Hello
I have try to connect com port on my computer but it cannot connect to comport
I try on Win7,8 it not work but XP is working ,
Anyone use Win7,8
Please advice me
Hi,

you need to install the Win 7 compatible drivers v. 6.6.1. from Silab:

http://www.silabs.com/Support%20Docu...CP_Windows.zip

Launch the setup as an administrator ( right click on the setup to get the menu)

Once installed you should be able to connect then, check also firewalls and antivirus that they do not lock the applications.

Hope this helps,

Best regards,

Fabien
Sep 06, 2013, 05:20 AM
Registered User
Quote:
Originally Posted by Bricobrac
Hi,

you need to install the Win 7 compatible drivers v. 6.6.1. from Silab:

http://www.silabs.com/Support%20Docu...CP_Windows.zip

Launch the setup as an administrator ( right click on the setup to get the menu)

Once installed you should be able to connect then, check also firewalls and antivirus that they do not lock the applications.

Hope this helps,

Best regards,

Fabien
Thanks for advice.I will try
Sep 06, 2013, 08:11 AM
ZGA
ZGA
Registered User

test my 3 axis on my electric bike


http://youtu.be/9XvJ3S0EzDg
Sep 06, 2013, 09:30 AM
Registered User
http://item.taobao.com/item.htm?id=12467119761


Quick Reply
Message:

Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Mini-HowTo Super Simple Gimbal - SSG - KK2.1 & OpenPilot support crezzee Aerial Photography 1675 Sep 05, 2017 02:57 AM
Yippee! Simple Brushless Gimbal Controller--FPV MODEL fpvmodels Multirotor Drone Talk 3396 Dec 26, 2014 07:23 PM
Discussion A very simple camera-gimbal for Gaui 330xs babanan Multirotor Drone Talk 2 Oct 09, 2013 08:46 AM