I finally did the test of hard fixing the APM case to the frame, thus not having two separate masses with vibration dampening materials in between, as before (i.e. : frame/props on one side and APM on the other side).
This is explain by Forrest Frantz in a post I made earlier about moongel use here:
and in more details here:
I have to say that I felt on my bottom when I saw the results : TWICE less vibrations with this method WITHOUT any moongel nor any other vibration dampening techniques. JUST ONE SINGLE mass (APM fixed on frame directly).
Look at this graph that shows ZACC within plus or minus 0,1g which is twice less than the best +-0,2g I had with moongel combined with rubber bobbins and/or silicon gel. It is measured on the exact same aircraft (except moongel was removed and rubber bobbins were replaced by hard pylons). APM is stcked to the plate with NON foam double sided tape (hard double sided tape, as I did not want to drill screw holes in the APM plastic case).
Not only have I less vibration but I will also get better control on the craft since I get rid of the lag effect of dampening materials.
I confirm thus the theory of Forrest Frantz. It works for my frame.
It would be nice if others could confirm this also on their frame.
I publish this UPDATE after the information about 200hz dangerous vibrations came out thanks to the debate initiated of this post: the APM and its integrated sensor's hardware have a limit (due to low processing power limiting sampling frequency - that should be solved with the more powerful PX4 hardware) in such a manner that all frequencies of 200hz (and multiple of) will impact the APM control (accel) and will go through the 20hz software filter, undetected (thus noise id processed instead of valid signal, possibly creating catastrophies). As there is no way to analyze in a log AccelZ graph if you get these 200 Hz frequencies on your arducopter, it is necessary as a precaution to use dampening technique(s) to eliminate these high frequencies (even if the AccelZ log seems to show low vibrations by direct fixing APM on the frame).
Replies
So let's find out ... and agree to the following test.
I'm proposing a test that we do and verify.
If we do this and share the results, then a lot of good will have come from this complex topic.
Background: I think we all agree on the following:
o Balance the props.
o Balance the props on the hubs so both blades rotate within the same plane.
o Ensure that the frame is as stiff as possible.
o Aliasing might be a problem on some ships that can cause erratic and dangerous flight behavior.
o All dampeners and isolators disconnect if only slightly the APM from the attitude of the ship. But because of aliasing, there may be an optimal net trade even if that trade means slightly increasing the accel-xyz noise input to the APM.
Hypothesis (please help me edit this so it can be tested): Adding [TBD Gary or someone] under the APM decreases the amplitudes of 100 - 400 Hz frequencies.
I think I can test this with some help. Procedure:
o hard mount a light platform to a speaker box.
o hard mount an analog accelerometer to the platform of the same mass as the APM 2.5.
o subject the accelerometer to the following sound waves to vibrate the board at a repeatable dB: 100Hz, 150Hz, 200Hz, 300Hz, 400Hz.
o record the amplitudes for each frequency.
o dampen the accelerometer using the perscribed dampening method(s).
o retest
o compare and report
Accept the hypothesis if the dampened amplitudes are significantly less than the undampened. Reject the hypothesis otherwise. [standard deviations will have to be tightly controlled]
Mull this over. If enough of the community agree that this test would help, I'll do the test and fully document each step. Also, more than one of us can perform the test (I'll pay for the sensors and can mail them around to those that have the test equipment).
I'm also willing to bet a six pack of local micro-brew beer to any takers. We have pretty good beer here in Hood River, OR, one of the UAV and kite surfing capitals of the world. Makes for a fun combination.
P.S. The H and V copter both have more free-air time than test-stand time. But, I confess, not much. I'm not easily satisfied with results. For me, a 30 minute hover is not good enough. I'm now onto Version 3, a really low KV octa.
I
I do not use moogel or gums, or anything like that, in my case the fill box apm foil, those monitors inside and foam as insulation, screw the box with nuts and bolt. excellent result. A good balance of the propellers and everything securely tightened is the most important.
2013-07-24 15-23 1.log
Forrest
Forrest said: "On the 50Hz (it actually varies, but that's a knit). The wiki recommends 30 seconds of stable hover because it is a sample (if you are curious as to why sampling works, I can point you to many science journals on this topic). I recommend 40 seconds."
I don't wish to get into an argument about this but it is clear you do not understand what is going on and so we wish you would stop preaching flawed gospel that contradicts the truth we are trying to convey for everyone's safety/enjoyment. Nyquist theorem is BASIC knowledge in this topic, stating that one must sample at twice the frequency of the signal - hence if we're sampling at 50Hz then we can only observe up to 25Hz of frequency. So, in answer to your response, yes there is plenty of information to tell us that sampling does not work in this case because our sampling rate isn't high enough (a trade off - the IMU can sample at 200Hz but that would use significant amounts of flash for the logs - if we can even get the throughput - I don't have the information on that to hand).
To the gent who asked why the vibrations are so catastrophic - this is why. We use measurements from the accelerometer, if the accelerometer says acceleration of 1g for 0.5 second, then we integrate that into velocity, and then integrate again to get position. Noise disrupts this hugely - keep in mind, we're integrating twice the original values, meaning noise is amplified significantly.
We're saying, reduce the vibrations over all - this is likely to reduce vibrations across the spectrum. Bolting to the frame is not advised at all and very strongly discouraged - it easily transmits vibrations across the frequency spectrum.
Vibrations can be dispersed as heat (this is what dampening actually does). But it is also dispersed as mouvement. So if we can somehow guide the waves to a part of the frame that is purposelu made ro move, would not this reduce vibrations on APM?
I thought about that by seeing my hand made aluminium landing gear, based on openbeams, vibrating a lot during flight but not affecting at all my video nor the zacc measurements.
Am i totally off path here or is it worth drilling deeper ?
of course moongel does not work if you only glu APM to it,you have to make sandwich construction and press it.....also can you explain how do you plan to get rid of electric noise comming from LiPos when APM is attached???
@ Forrest,
Your comments:
What I don't know, but should find out. The accelerometers need good signal to noise, electrically and mechanically. So does moving the accelerometer chip diagonally away from CG help or hurt the processing of the signal? Accelerometers measure distance per second per second in x, y, and z. Take the case where the ship tilts forward. Accelerometers located dead center would see nothing (no change in x, y, or z (the gyros would). But if the chip was located forward to the right of CG, z would show a drop, x would show backward movement (the chip via rotation is moving under the ship toward the CG, and y no change. So is that good or bad? I'd guess that would be a false signal so put the chip as close to CG as possible.
WRT the bold part of your statement - This is not true. Accelerometers do not only measure acceleration due to their own movement, or rather rate of movement, but also measure gravity, obviously. So rotating an accelerometer around its own axis will measure a sinusoidal rate of change in the gravitational field it is immersed in. In ALL cases, it is better to have ALL inertial sensors as close to the CG as possible, that way attitudes are more easily computed, since the best signal to noise measurements are obtained this way,; rates are the lowest for such installations.
Joe
The Nampilot
I would expect this result for a fixed wing aircraft, since the aircraft and its interaction with a viscous atmosphere is itself the best damping device you could get (it has high mass, large surface area and is NOT a rigid body, so it can absorb high frequency energy by flexing)...
... I wouldn't naturally expect this for a copter though, particularly as we try to make them rigid, they have a relatively smaller surface area and generally a lower mass than our fixed wing counterparts...
@Hughes: What I would be interested to see is that rather than displaying the time-based signal, you can display the spectral density for each of the signals... have you truly reduced noise over all input frequencies, or simply the lower frequencies... it's hard to see this from your picture.
Good to see others have had similar results to myself.
Though admittedly I have just never liked moongel to begin with. Most of these vibration absorbing gels are bloody shams to be honest. I have tried moongel for a number of things - especially camera vibration dampening and I finished with some pretty disheartening results compared to high quality silicon vibration dampeners. NOT the stuff you buy from fox tech and NOT the stuff that comes with a vibration dampening "kit".
All of the "designed for RC" dampening systems are really gimmicky and I would advise you look elsewhere.
In regards to the APM or flight controllers the trick seems to be to get something fixed, but not quite completely rigid - which is why I would advise against hard mounting in most cases honestly.I know for me what worked the best was hard mounting the apm and simply using a tiny sliver of highly compressed wire insulation around the screws. (I have the old non cased apm 2.5) The wire insulation was very compressed though, and if you tried to wiggle the APM around you would certainly not be able to tell the wire insulation was there.
Thanks for the informative post - I really wish we could get a dedicated thread where we just listed vibration dampening techniques and results for both the APM and camera systems.
There is a ton of info out there but it is outdated and spread through hundreds of rcgroups threads each with hundreds of pages.
Thank you so much for taking the time to do that test.
You are absolutely correct. Will any other test this? The score is now 0 to 2. The underdogs lead.
From personal experience I've learned that when someone has a lot of time invested in something they built, it can be hard to face the reality that all their work and all their inventiveness was all for not. For me, an inventor, I was really hoping to find a solution that improved vibration over doing nothing but tying the APM securely to required mass (the batteries). So much for hope. But at least now I can focus on other more rewarding issues.
Well done Hughes. One of the brave. One of the few. Thanks for doing the test.
P.S. Something I haven't tested yet, since we tie the APM indirectly to the mass of the battery, is if multiple straps of Velcro is adequate to secure the battery to the electronics platform or if the battery needs to be more firmly secured. And if so, how to do that using only a few grams.
P.P.S. Your z beats mine!
-
1
-
2
of 2 Next