Unterschiede zwischen den Revisionen 51 und 83 (über 32 Versionen hinweg)
⇤ ← Revision 51 vom 21.09.2009 13:28 →
Größe: 19309
Autor: fredericg
Kommentar:
← Revision 83 vom 17.10.2009 12:25 → ⇥
Größe: 12112
Autor: FredericG
Kommentar: Added changelog
Gelöschter Text ist auf diese Art markiert. Hinzugefügter Text ist auf diese Art markiert.
Zeile 6: Zeile 6:
= Why is it important to reduce vibrations? =
The FlightControl sensors pickup the vibrations and this deteriorates performance; the new FC ME comes with vibration dampers in order to reduce the vibrations transferred to the FC.
Zeile 7: Zeile 10:
In addition, the FlightControl sensors pickup the vibrations and this deteriorates performance; the new FC ME comes with vibration dampers in order to reduce the vibrations transferred to the FC.
Zeile 17: Zeile 19:
[[BR]]
[[Anchor(SampleSignalJpg)]]
Zeile 20: Zeile 23:
[[BR]][[BR]][[BR]]
Zeile 21: Zeile 25:
/!\ ~+'''The version available now (version 0.0) is an alpha version; it was only tested by a limited amount of people. More testers are very welcome!'''+~ /!\ ~+'''The version available now (version 0.9 RC3) is an beta version; it was only tested by a limited amount of people.
[[BR]]More testers/evaluators are very welcome!'''+~
Zeile 23: Zeile 28:
The current implementation does not provide a graphical user interface; it is a "command-line" tool. While this approach is a bit less user-friendly, it allows to experiments more freely, which is important at this stage of the project. Later a more fancy user-interface might be added. At the time of writing, the tool was tested in combination with:
 * PC/Windows XP and Vista
 * FC 1.2 and FC 1.3
Theoretically the tool should also run on MAC and Linux.
Zeile 25: Zeile 33:
Currently the tool was only tested on '''PC/Windows with FC 1.3''' hardware. Theoretically the tool should also run on MAC and Linux. Features/fixes in the pipeline
 * Integrated backup and restore of flight settings
 * Integrated installing and restore of FC code
Zeile 28: Zeile 38:
(i) [[MailTo(frederic AT rc-flight DOT be)]] [[BR]]
{i} [[MailTo(frederic AT rc-flight DOT be)]]
Zeile 30: Zeile 41:

[[BR]][[BR]][[BR]]
[[Anchor(Install)]]
Zeile 33: Zeile 47:
The most logic and clean way to run the tool is to install support for Python on your PC (if you do not have it already). Python is also available for Linux and MAC; the tool should also run on these platforms, but I have never tried it. The most logic way to run the tool is to install support for Python on your PC (if you do not have it already) and the Python libraries the tool depends on.
Python is also available for Linux and MAC; the tool should also run on these platforms, but I have never tried it.
Zeile 40: Zeile 55:
== Option 1: Install Python == == Option 1: Install Python and required libraries ==
Zeile 45: Zeile 60:
Once you have Python support on your PC, unzip [http://www.rc-flight.be/VibrationTest/VibrationTest_0_0.zip VibrationTest_0_0.zip] to a directory on your PC.  * [http://www.wxpython.org/download.php wxPython wxWidgets for Python] Select the the win32-unicode version for Python 2.6
 * [http://sourceforge.net/projects/numpy/ NumPy, the fundamental package needed for scientific computing with Python]
Once you have Python support on your PC, unzip [http://www.rc-flight.be/VibrationTest/VibrationTest_0_9.zip VibrationTest_0_9.zip] to a directory on your PC.
Zeile 48: Zeile 65:
unzip [http://www.rc-flight.be/VibrationTest/VibrationTestExe_0_0.zip VibrationTestExe_0_0.zip] to a directory on your PC unzip [http://www.rc-flight.be/VibrationTest/VibrationTestExe_0_9.zip VibrationTestExe_0_9.zip] to a directory on your PC
Zeile 51: Zeile 68:
[[BR]][[BR]][[BR]]
Zeile 60: Zeile 78:
The VibrationTest tool only works in combination with a special version of the FlightControl software. It must be flashed in the FC board before using the test and you need to reinstall your regular software version afterwards. This just takes a few minutes. Switching the software is done the usual way, using the MKTool. The VibrationTest tool only works in combination with a special version of the FlightControl software. It must be flashed in the FC board before using the test and '''you need to reinstall your regular software version afterwards'''. This just takes a few minutes. Switching the software is done the usual way, using the MKTool.
Zeile 64: Zeile 82:
The LC-Display in the MKTool will look like this: /!\ '''The FC code for the VibrationTest is based 0.74d but is not fit for flying! Before flying, you will need to reinstall your initial FC code. When the version you use for flying is 0.74d, your settings should be preserved. In combination with other versions, it might not be the case. It is a good idea to backup yout settings to PC (via the MKTool) before installing the VibrationTest FC code. '''

After insatlling the FC code for the VibrationTest, the LC-Display in the MKTool will look like this:
Zeile 68: Zeile 88:
== The command line ==
The current implementation of the tool does not have a graphical user interface. The simplest way to use the tool is from the command prompt.		 == Starting the VibrationTest tool ==
In the directory where you unzipped the tool, dubbleclick VibrationTestGui.py (in case of the Python version) or VibrationTestGui.exe
Zeile 71: Zeile 91:
On Windows XP open a command shell by selecting Start>Run and typing "cmd". == Using the VibrationTest tool ==
[[BR]]
[[Vimeo(6948282)]]
[[BR]]
Zeile 73: Zeile 96:
On Windows Vista, click Start and type "cmd.exe" in the "Start Search" box. == Checking the reliability/accuracy of the results ==
Before you start, it is good to check the reliability of the test and setup.
Zeile 75: Zeile 99:
Now, if you have for example unzipped the tool (the Pyhton version or the compiled version) in directory "D:\MK\VibrationToolExe" you should go to that directory by typing: You should alway start with a sweep and look at different sensors. If the vibrations is rather constant over motor-speed, the combination might be well balanced or you might not be measuring what you are hoping to measure. In this case, sticking some tape to the prop should should unbalance it and you should see a difference.
Zeile 77: Zeile 101:
{{{
D:
cd \MK\VibrationToolExe
}}}
Now, the tool can be started by just typing "VibrationTest.py" if you have installed Python or "VibrationTest" if you use the compiled version. Without any parameters, it tool should display a description of the required parameters.		 You can have a look at the recorded signals, the filtered signals and perhaps the spectrum. If this seems ok, you can repeat the sweep multiple times, using one of the sensors. Each time you should get similar results. Results will never be 100% the same.
Zeile 83: Zeile 103:
attachment:cmdLine.jpg Another test you should do, is let the tool measure multiple times at the same speed; for example "10*100,10*200". The measurement at the same speed should report the about the same amount of vibrations.
Zeile 85: Zeile 105:
== The command line parameters ==
The tool expects to be started with a few parameters. When the tool is started without or invalid parameters, a brief description of the expected parameters is displayed:		 The figure below shows the result of such test with a "Roxxy 2824-34 / EPP1045 / 3s / Standard ESC" combination. All but the green line are from the Nick sensor. The black line are vibrations measured at speeds 130 and 185. The other lines are sweeps. This does not look too bad.
Zeile 88: Zeile 107:
{{{
VibrationTest.py COMPORT MOTORS SPEEDS CHANNELS [-m MINSPEED] [-s NBSAMPLES] [-n NAME] [-d FILENAME] [-v]
 COMPORT Serial port to use. e.g. COM4
 MOTORS Motors to activate during test. Multiple motors can be used at the same time. e.g. 1,2,3,4
 SPEEDS Indicates at what speeds the motors need to be tested.
                   Format 1: e.g. 50,110,140 Tests at speeds 50, 110 and 140
                   Format 2: e.g. 100-200:50 Tests at speeds 100, 150 and 200
 CHANNELS Channels to monitor. e.g. 5,6,7
                   Channel 0 = GyroYaw
                   Channel 1 = GyroRoll
                   Channel 2 = GyroNick
                   Channel 3 = Pressure
                   Channel 4 = Batt
                   Channel 5 = AccTop
                   Channel 6 = AccRoll
                   Channel 7 = AccNick
 -m MINSPEED Minimum speed of the motor(s)
 -s NBSAMPLES Number of samples
 -n NAME Name of the test
 -d FILENAME File to which the measured values will be logged in
 -v Verbose
}}}
4 parameters are mandatory: COMPORT MOTORS SPPEDS and CHANNELS

 * '''COMPORT''': The serial port that is connected to the MK e.g. "COM4"
 * '''MOTORS''': List of motors (comma-separated) that need to run during the test. Typically only one motor will be selected. It is also possible to activate all motors to evaluate the global vibrations before and after calibration for example. e.g. "1" or "1,2,3,4"
 * '''SPEEDS''': List of speeds at which will be tested. The "speed" is the I2C value that will be sent to controller and is a value between 0 and 255. Typically interesting values for vibration-testing vary between 100 and 200. [[BR]]There are 2 formats possible. First there is the simple list of speeds, e.g. "100,150,190,200". The second format specifies the minimum, maximum speed and step, e.g. "100-200:20" will test at 100,120,140 ... 200
 * '''CHANNELS''': List of "channels" that will be monitored. A channel is in fact one of the MK analog sensors. Measuring the pressure or battery does not make much sense in this context, but is it possible. It is my experience that channel 6 and 7 produce best vibration signals, but YMMV.

[[BR]] In addition, these optional parameters can be provided

 * '''-m MINSPEED''': The motor(s) will first be started at "idle" speed before being spooled up to the speed at which the measurement will take place. By default, this speed is 25 but it can be changed with this option. Your motors need to run smootly and start reliably at this speed. In combination with the TP converted ESC for example, the default value of 25 will be too low.
 * '''-s NBSAMPLES''': During a measurement, the FlightControl board will monitor a given channel and store 1000 samples in memory. Afterwards the VibrationTool will read these samples. These samples can be dumped in a file (see the "-d" option) and will be used to find the amplitude of the signal (the difference between the minimum and maximum measured value). [[BR]]It is important to read enough samples to cover a few periods. By default 400 samples are read but with this option, the number of samples can be modified (with a maximum of 1000).
 * '''-n NAME''': An indication of what you are testing. This string will be added to the dumps (see "-d" option) and should not contain spaces e.g. "-n AfterBalancing"
 * '''-d FILENAME''': With this option, all samples will be dumped in a text-file. This file can later be used for further analysis. It can for example be inported in MS Excel or OpenOffice Calc to make graphs of the vibration signal. When the file does not exist yet it will be created, otherwise the new data will be added. All measurements dumped in the same file should have the same number of samples (see the "-s" option) e.g. "-d motor1.txt"
 * '''-v''': With this option, the tool will be much more verbose; it will indicate what it is doing. Also when an error occurs, the tool will provide more technical details (the callstack)

== The output ==
=== The results on the screen ===
Without the "-v" option, only the results of the test is displayed:
{{{
C:\DATA\temp\VibrationTestExe>VibrationTest COM5 1 100,150 6 -n Demo
      Demo Speed=100 U=12.1V Channel=AccRoll Min=497 Max=506 pp= 9 *
      Demo Speed=150 U=11.9V Channel=AccRoll Min=498 Max=508 pp= 10 **
}}}
 * "'''Demo'''" This is the name of the test as we provided with the "-n" option.
 * "'''Speed=100'''" Indicates the motor speed for this measurement
 * "'''U=12.1V'''" The measured battery voltage when the measurement was done
 * "'''Channel=AccRoll'''" The measured channel
 * "'''Min=497'''" Lowest value of all samples
 * "'''Max=506'''" Highest vale of all samples
 * "'''pp=9'''" Difference between Min and Max, this is the actual result
 * "'''**'''" The number of stars gives a quick indication of the pp-value


With the "-v" option more details are provided:
{{{
C:\DATA\temp\VibrationTestExe>VibrationTest COM5 1 100,150 6 -n Demo -v
comPort = COM5
motors = ['1']
minSpeed = 25
speeds = ['100', '150']
channels = ['6']
nbSamples= 400
fileName = None
testName = Demo
Opening comPort...
Version: 0.74
Voltage: 12.1V
Minimum Voltage: 10.5V
Starting motor(s) (speed=25)... OK
Setting speed to 100 ...
Getting data... Demo Speed=100 U=12.1V Channel=AccRoll Min=496 Max=508 pp= 12 **
Setting speed to 150 ...
Getting data... Demo Speed=150 U=11.9V Channel=AccRoll Min=497 Max=508 pp= 11 **
}}}

=== The dumped signal ===
The "-d" option allows to dump the measured signal in a text-file. This is a "comma-separated" text file that can easily be imported in MS Excel or OpenOffice Calc.

attachment:calc.jpg

1000 samples are acoputed in about 90ms (0.09 s), the time between samples is about 0.9 ms (0.00009 s).
Suppose the prop is spinning at 5000 RPM, this means 83 rotations per second. In 0.09s we have captured 7.5 rotations. We we take 400 of the 1000 samples we should see 7.5/1000*400 = 3 rotations of the prop.


== Proposed procedure to balance prop/motor assembly ==
First we need to find what what sensor provides the "best" footage of the vibrations. The largest signal is probably the best. Candidates are Gyro and Acc signals. My experience is that AccNick (channel 7) and especially AccRoll (channel 6) provide the nicest signal, but YMMV. The vibration will typically peak at a certain motor speed. So, first we perform a sweep between speeds 100 to 200 and look at channels 6 and 7 and choose a channel and speed we will concentrate on.

Next I check if the prop can be balanced a bit better. I stick small strips of tape to the prop, close to the center. One strip to the side where the prop-dimension is mentioned I denote as "-1", one strip on the other side, I denote "+1". I test different configurations and check what gives the best results for my chosen channel and speed.

Now that the prop is better balanced, I start to rotate the prop in relation to the motor. As names for the test I use "XsA" where "X" is the number of strips and "A" is the angle; e.g. "+1s60" means 2 strips (opposite to the prop-marking) and rotated by 60 degrees (compared to the starting position). Again I try to find the best position.

Once we have found the best rotation, we can rebalance the prop in the same way as before.

After all this we should have found the best combination. We can perform the swipe again to compare with the starting point.


=== Example 1 ===
This is an example with standard Roxxy 2824-34 motor with EPP 1045 prop.
[[BR]]

Finding what sensor provides best information at what speed:

{{{
VibrationTest.py COM5 3 100-200:20 6,7
      Speed=100 U=11.3V Channel=AccRoll Min=492 Max=505 pp= 13 **
      Speed=100 U=11.3V Channel=AccNick Min=498 Max=507 pp= 9 *
      Speed=120 U=11.3V Channel=AccRoll Min=489 Max=503 pp= 14 **
      Speed=120 U=11.3V Channel=AccNick Min=497 Max=507 pp= 10 **
      Speed=140 U=11.2V Channel=AccRoll Min=469 Max=519 pp= 50 ********** <<=======
      Speed=140 U=11.2V Channel=AccNick Min=495 Max=508 pp= 13 **
      Speed=160 U=11.1V Channel=AccRoll Min=478 Max=517 pp= 39 *******
      Speed=160 U=11.1V Channel=AccNick Min=493 Max=511 pp= 18 ***
      Speed=180 U=10.9V Channel=AccRoll Min=479 Max=511 pp= 32 ******
      Speed=180 U=10.9V Channel=AccNick Min=493 Max=514 pp= 21 ****
      Speed=200 U=10.7V Channel=AccRoll Min=479 Max=515 pp= 36 *******
      Speed=200 U=10.7V Channel=AccNick Min=493 Max=513 pp= 20 ****

VibrationTest.py COM5 3 130-160:10 6
      Speed=130 U=11.1V Channel=AccRoll Min=484 Max=511 pp= 27 *****
      Speed=140 U=11.1V Channel=AccRoll Min=469 Max=524 pp= 55 ***********
      Speed=150 U=11.1V Channel=AccRoll Min=475 Max=520 pp= 45 *********
      Speed=160 U=11.1V Channel=AccRoll Min=479 Max=518 pp= 39 *******
}}}
So, lets concentrate on channel 6 (AccRoll) at speed 140.


Now, the preliminary balancing of the prop:

{{{
VibrationTest.py COM5 3 140 6 -d motor3.txt -n 0s0
  0s0 Speed=140 U=11.0V Channel=AccRoll Min=470 Max=521 pp= 51 **********

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s0
  1s0 Speed=140 U=11.0V Channel=AccRoll Min=482 Max=513 pp= 31 ******

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 2s0
  2s0 Speed=140 U=11.0V Channel=AccRoll Min=475 Max=520 pp= 45 *********
}}}
For now, 1 strip seems best.

Now, lets rotate the prop:

{{{
VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s0
  1s0 Speed=140 U=10.9V Channel=AccRoll Min=481 Max=513 pp= 32 ******

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s60
 1s60 Speed=140 U=10.9V Channel=AccRoll Min=480 Max=512 pp= 32 ******

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s120
1s120 Speed=140 U=10.9V Channel=AccRoll Min=477 Max=514 pp= 37 *******

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s180
1s180 Speed=140 U=10.9V Channel=AccRoll Min=482 Max=509 pp= 27 *****

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s240
1s240 Speed=140 U=10.9V Channel=AccRoll Min=482 Max=510 pp= 28 *****

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s300
1s300 Speed=140 U=10.9V Channel=AccRoll Min=482 Max=514 pp= 32 ******

VibrationTest.py COM5 3 140 6,6,6 -d motor3.txt -n 1s360
1s360 Speed=140 U=10.9V Channel=AccRoll Min=482 Max=515 pp= 33 ******
}}}
Rotating the prop by 180 degrees produces best results, now lets try to balance the prop again:

{{{
VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s180
1s180 Speed=140 U=10.9V Channel=AccRoll Min=483 Max=509 pp= 26 *****

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 0s180
0s180 Speed=140 U=10.9V Channel=AccRoll Min=472 Max=519 pp= 47 *********

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 1s180
1s180 Speed=140 U=10.9V Channel=AccRoll Min=483 Max=510 pp= 27 *****

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 2s180
2s180 Speed=140 U=10.9V Channel=AccRoll Min=489 Max=508 pp= 19 ***

VibrationTest.py COM5 3 140 6 -d motor3.txt -n 3s180
3s180 Speed=140 U=10.9V Channel=AccRoll Min=483 Max=512 pp= 29 *****
}}}
So, 2 stickers is best, let's verify again:

{{{
VibrationTest.py COM5 3 140 6 -d motor3.txt -n 2s180
2s180 Speed=140 U=10.9V Channel=AccRoll Min=488 Max=508 pp= 20 ****

VibrationTest.py COM5 3 140 6 -d motor3.txt -n end
  end Speed=140 U=10.8V Channel=AccRoll Min=487 Max=510 pp= 23 ****
}}}		 attachment:acuracy1.jpg
Zeile 283: Zeile 110:
Now, we can perform a sweep in order to compare with the starting-point: When the results are disappointing, one can experiment with the way the MK is attached to the desk. It is also possible that the filter parameters might not be optimal for your setup. In the figure below I have almost disabled the filter and switched to "pp" measurement:
Zeile 285: Zeile 112:
{{{
VibrationTest.py COM5 3 100-200:20 6,7
       Speed=100 U=11.1V Channel=AccRoll Min=493 Max=501 pp= 8 *
       Speed=100 U=11.1V Channel=AccNick Min=499 Max=507 pp= 8 *
       Speed=120 U=11.1V Channel=AccRoll Min=489 Max=507 pp= 18 ***
       Speed=120 U=11.1V Channel=AccNick Min=497 Max=508 pp= 11 **
       Speed=140 U=11.0V Channel=AccRoll Min=489 Max=507 pp= 18 ***
       Speed=140 U=11.0V Channel=AccNick Min=498 Max=509 pp= 11 **
       Speed=160 U=10.9V Channel=AccRoll Min=482 Max=508 pp= 26 *****
       Speed=160 U=10.9V Channel=AccNick Min=496 Max=510 pp= 14 **
       Speed=180 U=10.7V Channel=AccRoll Min=489 Max=505 pp= 16 ***
       Speed=180 U=10.7V Channel=AccNick Min=495 Max=508 pp= 13 **
       Speed=200 U=10.6V Channel=AccRoll Min=482 Max=508 pp= 26 *****
       Speed=200 U=10.6V Channel=AccNick Min=491 Max=509 pp= 18 ***
}}}		 attachment:acuracy2.jpg

[[BR]]
This is another example of such a test (Turnigy 2836-750 / APC 12x3.8 / 4s / TowerPro ESC):

attachment:acuracy3.jpg

[[BR]][[BR]]
= Examples =

== Balancing example: Roxxy 2824-34 / EPP1045 / 3s / Standard ESC ==
[[Vimeo(7011506)]]

[[BR]]
== Effect of prop position, Roxxy 2824-34 / EPP1045 3s / Standard ESC ==
The importance of the position of the propeller relative to the motor was first reported by RCGroups member "Old Man Mike", see [http://www.rcgroups.com/forums/showpost.php?p=12832508&postcount=228 this post]. My tool also shows this effect. However, I observed that sometimes rotating the prop did not produce significant variations...

I discovered that when I first balance the prop as good as I can using the VibrationTest, rotating the prop produced very little difference.

I come to the conclusion that I need to deliberately unbalance the prop a bit in order to find the "sweet spot". I also observed that when I unbalance the prop by sticking some tape on one side of the prop the vibrations peak while they go trough a minimum when I stick some tape to the other side.
I think this supports the theory that the slightly unbalanced prop can compensate for unbalances in the motor.
[[BR]][[BR]]
attachment:StudyRotation.jpg
[[BR]][[BR]]
On the X-asis the rotation of the prop in steps of 60 degrees. The red line is with a prop that I balanced before (needs some tape opposite to the prop size-markers). The green line shows vibrations when I attach an additional strip of tape (so unbalancing it). The blue line shows what happens when I attach tape to the other side (where the markers are).

== Balancing example: Turnigy 2836-750 / APC 12x3.8 / 4s / TowerPro ESC ==

/!\ For this setup I have found that setting the LP Filter cutoff frequency to 200 Hz
[[BR]]
attachment:Mk2BalancingExample1.jpg
[[BR]]
The red line is the initial balancing. The prop seemed best balanced with no additional tape.[[BR]]
The green line are the measurement while rotating the prop (steps of 45 degrees)[[BR]]
The blue line is the re-balancing, due to the repositioning of the prop the results are better. The first 3 measurements showed that no tape produced the best results. In the last 3 measurements of the blue line, I experimented with smaller strips and this produced the best result.

[[BR]][[BR]]
[[Anchor(Changelog)]]
= Changelog =
||'''Version'''||'''Date'''||'''Changes'''||
||0.9 RC3||17/10/2009||- Removed limitation of 4 motors[[BR]]- More parameter checking and clear error reporting[[BR]]- Support to connect to NC[[BR]]- Fixed exception at start of measurement||

[[BR]][[BR]]
Zeile 301: Zeile 156:

=== VibrationTest related ===
 * [http://svn.mikrokopter.de/mikrowebsvn/listing.php?repname=Projects&path=/VibrationTest/#_VibrationTest_ VibrationTest Project in Subversion]
Zeile 302: Zeile 160:
 * [http://svn.mikrokopter.de/mikrowebsvn/listing.php?repname=Projects&path=/VibrationTest/#_VibrationTest_ VibrationTest Project in Subversion]  * Mail FredericG: [[MailTo(frederic AT rc-flight DOT be)]]

=== Related Threads ===
 * [http://forum.mikrokopter.de/topic-7794-1.html Thread: Vibrationsdämpfung der Antriebsmotoren]
 * [http://forum.mikrokopter.de/topic-post106888.html Thread: Tilt servo play => Vibrations]
 * [http://www.mk-fr.info/forum/index.php/topic,812.msg12038.html Thread: Equilibrage des Hélices RC (French)]
 * [http://www.vimeo.com/4257366 Video: Is the Optical Image Stabilization of the FX35 causing problems?]
 * [http://www.rcgroups.com/forums/showthread.php?t=807347 Thread: Multikopter (mikrokopter, UAVP, x-ufo-3rd, etc) AP pictures]

=== Related Projects ===
Zeile 304: Zeile 171:
 * Mail FredericG: [[MailTo(frederic AT rc-flight DOT be)]]
=== Used Libraries and Tools ===
 * [http://www.python.org/ Python]
 * [http://www.wxpython.org/ wxPython]
 * [http://wxglade.sourceforge.net/ wxGlade]
 * [http://numpy.scipy.org/ NumPy]
 * [http://www.py2exe.org/ py2exe]

=== Technical references ===
 * [http://en.wikipedia.org/wiki/Sampling_(signal_processing) Sampling]
 * [http://en.wikipedia.org/wiki/Fourier_analysis Fourier Analysis]
Zeile 310: Zeile 188:
 . KategorieTools  . KategorieTools KategorieProjekte KategorieEnglish KategorieAnleitung

TableOfContents(2)

What is the VibrationTest Tool?

The tool evaluates the vibration generated by the motors which allows balancing motor/prop assemblies and experimenting with different motor-mounts. The test uses the MK build-in sensors, so no extra hardware is required.

Why is it important to reduce vibrations?

The FlightControl sensors pickup the vibrations and this deteriorates performance; the new FC ME comes with vibration dampers in order to reduce the vibrations transferred to the FC.

Reducing vibration is very important, especially for Aerial Photography or Video. Vibrations result in blurry footage. Some claim that high frequency vibrations can be picked up by the Optical Image Stabilization lens assembly which is incorporated in almost all recent cameras. The OIS is not designed for these vibrations and makes thing worse, even when you switch the OIS feature off in the camera menu.

Propellers can be balanced the traditional way. However it was discovered that also the motors need balancing. One way to accomplish this is let a small unbalance in the prop compensate the motor unbalance, so the orientation of the prop is important (see [http://www.rcgroups.com/forums/showpost.php?p=12832508&postcount=228 this post]). Looking at the vibraions of a running motor allows balancing the prop/motor assembly as a whole.

How does it work?

The standard FlightControl program samples all sensors periodically and provides a command to read the latest sample from all sensors. The MKTool uses this command to generate graphs of the analog values. In a first approach the VibrationTest used this standard command to monitor vibrations. The problem is that this mechanism only allows to sample the values 30 times a second. The actual shape of the signals cannot be determined and one can only hope to catch the peak values. In order to get reliable results the values need to be monitored for a long time (a minute for example).

This new approach is based on dedicated FlightControl software providing a new command. The new command instructs the FC to sample one of the sensors as fast as possible and store the info in its internal memory. Afterwards the recorded data can be read to analysis. This allows to grab a reliable signal in a very short time. This approach can capture about 11000 samples a second, 500 times more than the MKTool can.

This is an example of the signal measured by the Roll-ACC sensor: BR Anchor(SampleSignalJpg) attachment:test.jpg

BRBRBR

Current Project status

/!\ The version available now (version 0.9 RC3) is an beta version; it was only tested by a limited amount of people. BRMore testers/evaluators are very welcome! BRBR At the time of writing, the tool was tested in combination with:

  • PC/Windows XP and Vista
  • FC 1.2 and FC 1.3

Theoretically the tool should also run on MAC and Linux. BRBR Features/fixes in the pipeline

  • Integrated backup and restore of flight settings
  • Integrated installing and restore of FC code

BRBR {i} [http://forum.mikrokopter.de/topic-11446.html MK Forum thread for discussions] BR {i} MailTo(frederic AT rc-flight DOT be)

BRBRBR Anchor(Install)

How do I install the tool?

The tool is written in programming language called [http://en.wikipedia.org/wiki/Python_(programming_language) Python].

The most logic way to run the tool is to install support for Python on your PC (if you do not have it already) and the Python libraries the tool depends on. Python is also available for Linux and MAC; the tool should also run on these platforms, but I have never tried it.

If you do not feel like installing Python, a version that is compiled to a windows executable is provided.

So, you have two options.

Anchor(InstallPython)

Option 1: Install Python and required libraries

For a Windows machine I propose you install:

Once you have Python support on your PC, unzip [http://www.rc-flight.be/VibrationTest/VibrationTest_0_9.zip VibrationTest_0_9.zip] to a directory on your PC.

Option 2: Get the compiled version

unzip [http://www.rc-flight.be/VibrationTest/VibrationTestExe_0_9.zip VibrationTestExe_0_9.zip] to a directory on your PC

BRBRBR

How do I use the tool?

Prepare your MK

Secure your MK

I use luggage straps to secure my MK to the table while it is still able to vibrate:

attachment:mk1InTest.jpg

Installing the FlightControl software

The VibrationTest tool only works in combination with a special version of the FlightControl software. It must be flashed in the FC board before using the test and you need to reinstall your regular software version afterwards. This just takes a few minutes. Switching the software is done the usual way, using the MKTool.

You will find the required hex file in the same directory with the VibrationTest tool.

/!\ The FC code for the VibrationTest is based 0.74d but is not fit for flying! Before flying, you will need to reinstall your initial FC code. When the version you use for flying is 0.74d, your settings should be preserved. In combination with other versions, it might not be the case. It is a good idea to backup yout settings to PC (via the MKTool) before installing the VibrationTest FC code.

After insatlling the FC code for the VibrationTest, the LC-Display in the MKTool will look like this:

attachment:LCDisplay.jpg

Starting the VibrationTest tool

In the directory where you unzipped the tool, dubbleclick VibrationTestGui.py (in case of the Python version) or VibrationTestGui.exe

Using the VibrationTest tool

BR Vimeo(6948282) BR

Checking the reliability/accuracy of the results

Before you start, it is good to check the reliability of the test and setup.

You should alway start with a sweep and look at different sensors. If the vibrations is rather constant over motor-speed, the combination might be well balanced or you might not be measuring what you are hoping to measure. In this case, sticking some tape to the prop should should unbalance it and you should see a difference.

You can have a look at the recorded signals, the filtered signals and perhaps the spectrum. If this seems ok, you can repeat the sweep multiple times, using one of the sensors. Each time you should get similar results. Results will never be 100% the same.

Another test you should do, is let the tool measure multiple times at the same speed; for example "10*100,10*200". The measurement at the same speed should report the about the same amount of vibrations.

The figure below shows the result of such test with a "Roxxy 2824-34 / EPP1045 / 3s / Standard ESC" combination. All but the green line are from the Nick sensor. The black line are vibrations measured at speeds 130 and 185. The other lines are sweeps. This does not look too bad.

attachment:acuracy1.jpg

BR When the results are disappointing, one can experiment with the way the MK is attached to the desk. It is also possible that the filter parameters might not be optimal for your setup. In the figure below I have almost disabled the filter and switched to "pp" measurement:

attachment:acuracy2.jpg

BR This is another example of such a test (Turnigy 2836-750 / APC 12x3.8 / 4s / TowerPro ESC):

attachment:acuracy3.jpg

BRBR

Examples

Balancing example: Roxxy 2824-34 / EPP1045 / 3s / Standard ESC

Vimeo(7011506)

BR

Effect of prop position, Roxxy 2824-34 / EPP1045 3s / Standard ESC

The importance of the position of the propeller relative to the motor was first reported by RCGroups member "Old Man Mike", see [http://www.rcgroups.com/forums/showpost.php?p=12832508&postcount=228 this post]. My tool also shows this effect. However, I observed that sometimes rotating the prop did not produce significant variations...

I discovered that when I first balance the prop as good as I can using the VibrationTest, rotating the prop produced very little difference.

I come to the conclusion that I need to deliberately unbalance the prop a bit in order to find the "sweet spot". I also observed that when I unbalance the prop by sticking some tape on one side of the prop the vibrations peak while they go trough a minimum when I stick some tape to the other side. I think this supports the theory that the slightly unbalanced prop can compensate for unbalances in the motor. BRBR attachment:StudyRotation.jpg BRBR On the X-asis the rotation of the prop in steps of 60 degrees. The red line is with a prop that I balanced before (needs some tape opposite to the prop size-markers). The green line shows vibrations when I attach an additional strip of tape (so unbalancing it). The blue line shows what happens when I attach tape to the other side (where the markers are).

Balancing example: Turnigy 2836-750 / APC 12x3.8 / 4s / TowerPro ESC

/!\ For this setup I have found that setting the LP Filter cutoff frequency to 200 Hz BR attachment:Mk2BalancingExample1.jpg BR The red line is the initial balancing. The prop seemed best balanced with no additional tape.BR The green line are the measurement while rotating the prop (steps of 45 degrees)BR The blue line is the re-balancing, due to the repositioning of the prop the results are better. The first 3 measurements showed that no tape produced the best results. In the last 3 measurements of the blue line, I experimented with smaller strips and this produced the best result.

BRBR Anchor(Changelog)

Changelog

Version

Date

Changes

0.9 RC3

17/10/2009

- Removed limitation of 4 motorsBR- More parameter checking and clear error reportingBR- Support to connect to NCBR- Fixed exception at start of measurement

BRBR

Links

Used Libraries and Tools

Technical references

BR