Folks here have helped me identify a vibration issue with my heli, so I’m hoping I can minimise its impact using notch filters. The problem frequencies identified are 63 and 442Hz so, following the instructions in the Ardupilot how-to, I’ve set notch parameters as shown in the attached screenshots. Am I headed in the right direction? Will this allow my FC to ‘ignore’ those frequencies?
Use the Filter Review Tool to analyze it. That’s what Estimated Post Filter is useful for. With a throttle based notch you need a reference.
It’s clear I don’t understand what I’m doing 
I got those frequencies from the filter review tool – very distinct peaks at 63 and 442Hz before filtering. So are the after-filtering graphs showing what I should expect after I’ve configured INS_HNTCH_FREQ and _BW with the appropriate values, as shown in my screenshots?
Post a link to the .bin log file.
dd you also see some peakes on 189 Hz and / or 315 Hz. the 442 Hz is exactly the 7th harmonic of 63 hz.
Allan,
What size heli is this? Typically any peaks above 4 times your rotor speed don’t have enough energy to cause an issue plus the 20hz low pass filter will help (ie if your rotor speed is 1500 RPM then 4 x 1500/60 =100 would be the highest I use the notch for.). Maybe a tailrotor vibe or if you have a combustion engine then you may need to use the notch for that.
Thanks for the replies. The heli is a TRex 500X with 470mm blades. Rotor speed was tacho’d at 2420rpm. I didn’t notice any peaks on the other harmonics, Jeurgen.
Attached is a link to the .bin file which I got the data from. Here’s the .config file from the same date:
TRex500 ATC_RAT_RLL_FF and _PIT_FF adjusted after first tuning flight.param (17.7 KB)
Before getting further into this I would suggest you log higher resolution data. With an H7 based FC you can set these:
INS_LOG_BAT_MASK,0 (disable)
INS_LOG_BAT_OPT,0 (disable)
INS_RAW_LOG_OPT,9
But here all I see is the mode from Rotor RPM ~37Hz and a small harmonic of that. Anything higher is in the mud (very highly attenuated) from the Gyro low pass filter.
Thank you Dave, I’ll implement your suggestions. Might be a couple of days before I can test-fly again, but I’ll report back then.
I managed to get a couple of 7-minute flights yesterday in gusty weather. Here’s the .bin file from the second of them.
Here’s the config file FYI
TRex500 INS_LOG_BAT_MASK, _LOG_BAT_OPT set to 0 and _RAW_LOG_OPT set to 9.param (18.0 KB)
I hope you can tell me I’m doing the right thing. The heli seemed to fly well.
Hi Allan-You go the configuration right to collect the data! I could suggest a filter configuration but it should be reviewed by people more knowledgeable with Traditional Heli than I. But, this is what I see as working well. It’s possible the 2nd harmonic is not needed.
It’s my understanding that you can’t use a normal throttle based filter on trad-heli because of how collective pitch and the RSC are set up. If there’s no ESC or other RPM data it will need to be a fixed filter. But the good news is the ESC or governor should be holding the RPM pretty solid so the value won’t change much.
I’d try something like this:
2 Likes
Thanks Allister for that info.
Sure! A notch filter removes (or mostly reduces) a narrow band of frequencies while leaving others mostly unaffected. So yes if you understand it as “blocking a specific frequency,” you’re getting it right!
@Allister suggested set up is correct in how to approach reducing noise produced by the rotor system. Since helicopters operate at a constant rotor speed, the noise is at a very narrow band. So if using a governor with out an RPM sensor, the fixed notch is the right way to go. If using a throttle curve, it must be set up to hold approximately the same rotor speed throughout the flight envelope. In this case a wider notch might be used to account for variability. If an RPM sensor is used then the RPM mode should be used for the notch filter and in this case a smaller bandwidth could be used. I have used one as small as 6hz with good results.
Yes a notch filter blocks a narrow range of frequencies and lets others pass.
Thanks Bill, JACK4 and Joy.
Yes, I’m using the ESC’s governor without an RPM sensor. I’ll compare my settings with the ones suggested by Allistair above.
One last question: I understand that the filter is helping the FC to ignore specific frequencies; does that mean that the vibe levels recorded in the log will be reduced, or will they be raw data before filtering?
The log VIBE message uses pre-filtered data because that is what is important when judging the vibration effects on the EKF. In fact it is looking at the linear acceleration measurements instead of the angular rate (gyro) measurements which is used by the control system.
1 Like
Yes a notch filter blocks a nYes — notch filters block (attYes — notch filters block (attenuate) a narrow band of frequencies while letting others pass.