With the recent release of APM 2.9x for Arducopter, reducing vibration has become a top top priority. This is because inertial navigation is being implemented, starting with an inertial PID loop in altitude hold. Vibration problems show up in altitude hold and any mode that requires the use of the altitude controller. I was having problems with my quad climbing uncontrollably in forward flight (while in altitude hold) and then descending again when forward motion stopped. I thought it had something to do with transnational lift or air blowing into my APM case, but Randy (developer) helped me determine that the root cause of the problem was in fact vibration.
By enabling RAW logging I was able to see the acceleration forces on the quad in flight. The current thought is that during normal flight (not just a hover - you need to include data from high power settings) the Z acceleration should generally stay between -15 and -5. As shown by the green line on the graph below I was well outside that. I've got a motor that really whistles while flying, so I would guess it has a bent shaft. All my props are meticulously balanced. The red line shows the current vibration level the APM feels after I made a simple mod to the mounting of the APM. By no means was this my idea, but it works great and I highly recommend anyone having trouble with altitude give something like this a shot. Here is the wiki on Vibration Control. By importing the log data to Excel I was able to determine that the standard deviation of the red line is 1.4. I learned how to analyze flight data here. The second video discusses using Excel.
I previously had my APM mounted on a high quality gyro mounting tape, the same kind that the xBee is mounted on in the second picture. The ONLY modification I made to achieve the lower vibration to the APM was to switch to supporting the APM, receiver, xBee and a 42 gram plate with four really squishy earplugs. The diameter of the holes I drilled for the earplugs are a .25". The distance between the plate and frame is 7/32". I admit, 42 grams is A LOT of weight to just add to the quad, but the results are excellent. The all up weight of the supported structure is 111 grams. I first saw this mounting method here. The earplugs hold very well, I can actually pick up the whole quad by the APM and not have it all come disconnected. I messed around with using a little CA glue to hold the ear plugs in more permanently and found that thick CA and some accelerator did not attack the foam at all.
I hope this can help someone who has had trouble switching to 2.9.1.
Comments
Hi Hugues,
ill post them to thingiverse together with some text at the beginning of next week (found some more possibility to shrink the frame -> making it smaller)
ill post an update on this blog once its done, please stay tuned.
cheers christoph
Hello Christoph,
This looks excellent ! Z accel graph results are below the +1/-1 (tenth of G).
I am inetrested in your case. Do you have 3D files so that I can have it printed ?
Time for an update, the parts a printed and i have my first logs to share.
so here is the foam-setup i started with
first i came to try to shrink my initial design a little bit further which you can see in the above picture.
Once i hade everything assembled i glued in the earplugs (bought them at the pharmacy and spent 6 euro for that box) using 2k-epoxy just to find out that they might be a little bit to "hard" but i ran a quick test anyway which just proved my feeling.
after that i put in a new set on which i cut them in half making the suspension way softer.
next i added weight to it (first "only" 20gramms lead then a total of 60gramms extra, not including the 16 grams of the center-part of my suspension)
and here are the test-results so far
and here a flightlog of today with all 3 axis
Cheers Christoph
Ive been keeping close eye on the effects on the ear plugs and so far they look good as new. I am deciding if I want to stick with plugs long term, if I do I am going to just run a bead of thick CA around the earplug, then it will never break. I found that the thick does not penetrate the earplug, so it stays squishy. I may still go the moongel route however, the earplugs impinge on the folding characteristics of my frame.
Thanks Ted - I'm in the early stages of building so I follow this thread closely. I bought some Moon Gel to try when I get that far. I plan to make my plate from acrylic sheet. I have 1/4" so I'll start there. Just FYI - I found the following material weights per cubic inch: Aluminum = .10, Acrylic = .04 and G10= .072. It's easy enough from that to pretty closely estimate the weight of a particular proposed plate.
With the earplugs pulled through the hole in the plate do you worry about the vibration eventually cutting through the plug material? Should there be a secondary attachment?
Ken,
My plate is actually 1/8" fiberglass (g10), my dimensions are 4.5" x 2.6". The only reason I used those dimensions were to accommodate all gear to be mounted on the plate. The weight of the plate is not relevant what is relevant is the weight of the total supported structure: APM+RX+TELEMETRY+PLATE. My total weight is 111 grams (IMO insanely heavy for a little quad). I also cut out a 1/16" plate I am going to try.
The whole idea is simply to increase the mass from only the mass of the APM to something higher. More mass=better VR reduction...to a point. You don't want the mass to be so high that you decouple the movement of the APM from the frame (ie. the frame rolls but APM stays put). You could use 1/4" plywood to start and some earplugs. I found in my testing that most earplugs are tolerant of CA, so you can easily do testing that way.
This topic needs more research and development, but for now just increasing the mass of the object your are trying to isolate (the APM on the plate) to the neighborhood of 40-120 grams seems to be a great start. The dimensions probably are not super critical, but my gut tells me the greater the distance between the earplugs the better, but that's just a gut feeling.
Its also worth looking into moongel (any store that sells acoustic guitars). Apparently people have had great success with that.
Also, propeller choice makes a big difference. Bigger diameter props (lower RPM) will tend to produce less vibration at the source. The 12" APC SloFly series is easy to balance and very effective. However, if you plan on flying aggressively they are not an ideal choice as they do flex and will induce vibration when they do.
Ted - First Thanks for posting. If you don't mind, here are a few questions: What are the dimensions (L x W x D) of the aluminum plate you're using? Is the 42 gram mass of the plate critical in this anti-vibration setup or might it be more or less effective with more or less mass? Is their a reason for choosing aluminum for the plate vs say acrylic or polycarb? Is there a ready source for the aluminum? I couldn't find anything I thought might work at either Home Depot or Lowes.
kb
@Axel Why not roll back to 2.8 if you don't like 2.9?
Hi,
@Ted: yes they will be 3D printed (rep-strap)
@Axel: i come from mwc and previous had some trys with megapirate (only partly succesfull) and now am useing apm2.5+... i tuned all rate parameters from zero and i realy like the CH6 for tuning option and now its flying quite smooth (with adjusted d-term) (have you adjusted the lowpass-filter as stated i the 2.9 anouncement?)
given... mwc manages impressive stability with gyro only on allmost any frame and default settings.
I just gave up on APM2 because of 2.9.x problems.
With same frame on MWC (multiwii), no need to use accelerometer to have a stable flight.
With APM2 i tried everything, seems this card doesn't want to flight :/