VStabi Pioneer of Flybarless

Since the launch of VBar EVO, we got reports from customers about tail glitches and other unwanted behaviour on some helis. The effects appear to get worse at higher rpms and with aggressive inputs. In this article, I recap what we found, and how we found a solution:

Since we did not see bad behaviour on various test helis, I started to develop an special test device at the lab, which is able to produce high and violent vibrations. This allows me to quantify when those effects happen, and compare the results to NEOs. This special shaker is able to produce vibrations up to 100 G and up to 2 kHz on a single axis. Operating the shaker is a bit dangerous, since it will immediately cause hearing damage if operated without ear protection.

Analysis

Comparing EVO with NEO gives a bit unexpected results: both do have limits up to which they operate normally, NEO up to 40 G @ 1 kHz, EVO up to 60 G @ 1 kHz. I can compare both very well, since they have both the same highrange sensor on board, which makes it comparable. For comparison, I use a tail servo connected to the unit during shaking, set to 333 Hz drive rate and the FFT from Vibration Analysis Pro. 1 kHz was choosen, because it is the range where the sensors in general are most sensible to effects. Below 500 Hz, I was not able to produce any problems up to 100 G (limit of the shaker).

NEO starts showing increasing servo jitter at 40 Gs. This increases more and more until 80 Gs are reached. At 80 Gs, the servo goes to full deflection and does not react on movements anymore (along with out of range log file entries).

EVO starts showing more and more random movements at 60 Gs, and up to 85 Gs. Above this, it goes to full deflection as well. The difference is, that the random movements do not cancel out to a valid average position like on NEO.

Since the effect seen on EVO is coherent with reports from our customers, I assume that this is what is critical here: vibrations at very high frequencies (1 kHz) and very high intensity (> 40 G).

An interesting consequence: if swapping NEO for EVO shows problems, this means that the tail performance of NEO was at least already degraded!

Sensor Specs

Unfortunately, there is no definitive spec for the allowed vibration loads, for the sensors. This applies to almost all available products on the market as well. I have seen a spec for this only for military grade devices, which are not available to us. There are some specs that describe cross-sensitivity of the gyros to accelerations, but they do not extend to the level we have seen on some helis. Not even close. So we are forced to do our own measurements, an hope that the manufacturer provides consistent quality for our use.

The manufacturer specifies a G limit of 10,000 Gs to internal desctruction if unpowered.

Frequencies

If i do a sweep with the shaker from 10 Hz to 2 kHz, there are some quite sharp peaks, which seem to have most influence. Depending on the sensor, they range from 700 Hz to 1.1 kHz on EVOs, and 500 Hz to 1 kHz on NEOs. The dangererous zone is only super narrow, higher frequencies are usually not problematic, and lower frequencies are never leading to any effects. It's stunning how well the sensors cope with extreme Vibration levels up to a few 100s Hz. At these frequencies tests are only possible if all cables are firmly fixed, otherwise any connector will disconnect in seconds. If you touch the device during these tests, it hurts!

Frequencies up to this range are the signals we need to measure and which the software sees and needs to see. The problematic frequencies are way higher, and are not seen as real measurements at the sensors output. Running a test at 1 kHz, the shaker does not move visibly, feels quiet to the touch, but produces enourmes noise. You dont see it, but you can hear it down the street.

Solutions

The very first and most simple solution that comes to mind is, just do not let these vibrations develop and propagate. As long a the high freqency vibration stays blow 40 G, all is fine, no negative effects appear. Of course our customers cannot do much here, since this is something that has do be adressed in heli construction. I hope it will get adressed by the designers of future helis, so we dont see these problems in the future.

The next idea is, use thicker tape to absorb the vibrations. This works to some degree, but it is very hard to find a pad that really absobs the vibrations, and does not just shift it somwhere else and cause new problems. So some try and error can indeed help. In reality, this solution is of limited effect, it dampens a bit, but you still get a lot of unwanted signal through. If you do the math, you need very thick pads and very weak mounting. This is not feasible if you want to throw around the heli in hard 3D flying.

Using the software to filter out the vibrations is a tempting idea that may just need a software update. Unfortunately it is simply not possible. Software does not see the vibrations, they do not appear on the output as direct signal. The measured signal is a sideeffect of the MEMS structure of the sensor in a direction that it is not designed for. The effect is highly non-linear and not consistent, so it is impossible to compensate for it in software.

After some research and testing, it turned out that the sensitivity at these frequencies can be shifted by changing masses of the board and the case. Loading a small mass on the sensor does help in reducing high frequency effects.

So immediately the idea developed to make modifications to the case that hold the board more firmly, to have the same effect. The case was modified, production samples were made, but it showed no significant improvement on the values in the lab. It was not tested in real life which we may change in the future.

Further experiments showed that the maximum effect can be reached by glueing the sensor into the case. This shifts the resonance frequency down very much, so it comes into a range where the sensor is almost immune. Tests with this modification in the reality showed striking success. So this is the preferred solution, which will be used for future devices as well.

Lessons learned

Effects of frequencies in the Range of >= 1kHz are difficult to predict. They behave not as common sense tells you. So I learned very much about the propagation of these vibrations. We did not expect to see this level of vibration in some helis, it is still not 100 % clear where they originate, but it can be assumed that its mainly harmonics that develop if moving parts strike together with lower frequencies, amplified by frame designs and materials and mainfested at the mounting position of the unit.

Conclusion

The method seems simple, but the road to a solution was bumpy. We are happy that there is a solution, that unleashs the full potential of the VBar EVO also on more critical yet very popular helis. My wish as developer is still to get solutions that avoid these effects from bottom up; this will enable a lot of other possibilities that can be done by an IMU in a heli.