Following my previous unsuccessful attempts at building a quadcopter (I have serious issues with the balancing and stability) I've decided that it's likely all this is due to VIBRATIONS and it was time to get into the nitty-gritty of some measurements...
While they are obviously far from complete or scientific, they do show how complex the "vibrations problem" is...
You can find the graphs on my blog (http://trandi.wordpress.com) and do let me know if :
- you would like some more details / numbers
- you have any other suggestions about how to reduce them
thanks,
dan
Comments
It´s very complicate, I agree. But It makes the difference if you want a very smooth fly.
The risk to tune the quad structure is to make a bad tunning and instead of attenuation you´re going to have amplification!! In this case is better to keep the stiffness as high as possible!!
I like the idea to install rubber grommets. I had rubber grommets and spacers separating the sensor+control board and the structure. I had measure the vibration on the sensors and it decreases!!
The important is to keep the rubber grommets free (no tight). You have to tight the bolts against the spacer not against the rubber.
I use this servo rubber grommets: http://www.hobbyking.com/hobbyking/store/uh_viewItem.asp?idProduct=...
They are small for light weight!
Despite all these mechanical stuffs you still must use a LPF on the code (software).
They code I´ve been using makes a mean of several reading of gyros and accels (Bill Premerlani DCM Code)
- Ok, Mauro, you gave us enough academic reasons to convince us it was just a typo! ;)
The LPF makes much more sense to me.
- I know we are entering finite element simulation territory here, and it is not my intention to discuss susch things here.
However, one thing concerns me: we say stiffer== less dampening so wood ok, and metal worse
but CF known for its super stiffness, dampens better and rejects more vibs than alu (quite soft metal)?!
A 2DOF ( Degrees of Freedom) model is better to representing the quad when you have 2 mass connected by a beam (one mass is the motors+propellers, the another mass is the center of the quad and the beam is the the structure)
see http://en.wikipedia.org/wiki/Vibration#Illustration_of_a_multiple_D...
The model that I described on y previous comment is a 1 DOF is not good ...
It´s not a simple issue ... it could be a Master Science work ...
I had a mistake on the follow comment:"- Finally, the code (software) that you´re using to read the accelerometers and gyros signals must have a high-pass filter feature" please read low-pass filter instead of high-pass filter.
- The theory about the quad stiffness is based on its natural frequency. Look at http://en.wikipedia.org/wiki/Mechanical_resonance
- This graph http://en.wikipedia.org/wiki/File:Resonance.PNG shows a typcal response (attenuation or amplification) of a structure like of the quad. The Y axis shows the relation betwwen the intensity of the vibration measured where the sensors are placed and the intensity measured where one of the motors are placed. Note that when this relation is near 1 the motor vibration is been amplified (not desired) and when this relation is above 1.5 the motors vibration is been attenuated (desired). The X axis shows he relation between the motors frequency (WA) and the natural frequency of the quad structure (W0). Note that when WA/W0 equals to 1 there is a big amplification of the motors vibration. This point depends on the quad structure stiffnes (k). If you decrease the stiffnes the W0 decreases too, beyond 1.5 there is a attenuation of the vibration that comes from motors (desired effect). Another parameter that determinate the W0 frequency is the mass (m) on the center of the quad, higher mass lower W0. See the formula f = ..., f in Hz.
- How to find ideal W0? W0 value can be equals to the motors rpm (divide by 60 to transform in Hz) when the quad is a almost taking-off. Above this you need attenuation effect because the quad is flying and bellow this frequency, doesn´t matter because the quad is on the ground ...
- Another important issue is the dampening of the structure: steel and aluminum have low dampening coefficient while carbon-fiber and wood have higher dampening coefficient.
- The structure stiffness is the first barrier against the motors/propellers vibration and few people consider this ...
- I like these IMU units: FreeIMU v.02 see at http://www.varesano.net/projects/hardware/FreeIMU
or
http://www.sparkfun.com/products/10321
they are complete small, light-weight, easy to connect (4 wires) and large range. I have the FreeIMU v.02 that is cheaper and you can connect directly on 5V Arduinos (I2C) with no 3.3V regulator or logic levels converter ...
- Related to the propellers: I had experience only with GWS and APC propellers and I can say that GWS needs a very good balance work before you can use while APC you can use with no balance work ...
- 1024RPM/volt is good
I hope this comment be usefull, See you soon
LMF
That's a typical simple LP filter, some kind of weighted moving average.
LPF: LOW freqs pass, HIGH freqs do not or are attenuated
HPF; HIGH freqs pass, LOW freqs do not or are attenuated
- I am using EMP propellers (do you know anything about their quality?), but it's good to know that not all of them are made equal and APC seem to be better quality
- now with the holes, I had indeed some trouble as the adapters don't fit exactly my motors' shaft
- I use KDA 20-22L 1024kV so it should be ok (but if I were to do it again, I would indeed choose even lower rpms
- I THINK I understand what you mean by the frame "filtering" the vibrations : this means that it should be not too stiff, but stiff enough, to cut out more or less exactly the kind of vibrations that your motors+propellers combination is generating. From what I can see, the stiffer a frame is, the lower the frequency of the amortized vibrations (ie the higher its resonant frequency) and vice versa.
- I'm using some standard 2mm thick foam tape, to no visible effect
For the software I'm not sure:
I'm using some custom code (inspired by Jose Julio's initial code) running on an Arduino.
The sensors are:
- Wii Motion Plus IDG-650 & ISZ-650 for gyroscopes
- ADXL345 for the 3 axis accelerometer (from Sparkfun's 9DOF IMU, that I burnt the MCU of, and now connect to its accelerometers directly through the I2C bus...lol :))
So the code does have filtering for Accelerometers data ONLY, like this :
// Low pass filter on accelerometer data (to filter vibrations)
accelVector[0] = accelVector[0] * 0.4 + _accelX * 0.6; // acc x
But the initial comment said it's a "LOW pass" and to be honest I don't know the difference between low and high pass...
@I.S.
- I don't know what APM code does as I'm not using it...
@Luis
- So super stiff motor-motormount-frame bad idea? better to put some dampening between motor and frame to do lowpass? Could you give some hints for this?
- where can I find gyro tape?
- Does de APM code do high-pass filtering for sensors? gyros may need a high pass to avoid drift, but what about the accels? do they need high pass filtering too?
Hello Dan, I had a lot of experience with vibration I have some clues for you:
- APC propeller are the best, even with no balancing job, they have good balancing,
- Use the original propeller holes, don´t drill it for a larger hole,
- Use motors up to 1000rpm/Volt because the vibration increases with the square of the rpm,
- The stiffness of the quad structure must be tunned to "filter" the vibration like a "low-pass filter (it´s pretty difficult to explain in a few lines). The quad structure must attenuate the vibration that comes from the engines/propellers!
- CSM gyro tape can be used to attenuate even more the vibration,
- Finally, the code (software) that you´re using to read the accelerometers and gyros signals must have a high-pass filter feature
/edit by moderator: it needs to be a low-pass filter
Phil, thanks a lot for your comments and suggestions !
I'll definitely look into the "gyro tape", even though I've already tried simple 2millimeters thick foam tape, with no noticeable effect...
As for the Hyde Mount, wow... looks quite interesting, but hard to put together, especially 4 times ! lol :)
dan
-
1
-
2
of 2 Next