#Probably a bad ICM42688P Gyro but...
1 messages · Page 1 of 1 (latest)
4.4.0 and tried 4.3.1
I've heard there are improvements on the latest version of BF for the ICM42688P's. I'm not sure if those changes have made their way into and release version yet though so you may need to try a dev build.
the only thing I could think of is the gyro hardware lpf setting but that shouldn't be necessary
i'll try a nightly
just haven't made it that far yet
I thought this was a bug specifically for the ICM42688P that newer builds aimed to fix.
I have quads with this gyro that fly fine on 4.3, it's prolly a bad gyro or pour board design
@zealous torrent The latest nightly won't even pick up the gyro, nice
oh, damn. ok. womp.
4.4.1 no dice just yaws on the setup tab like a freakin owls head, lol. I hate this gyro. Even though when the board is right they are really good. Just too sensitive!
the fix is in 4.4.1 and later and fixes a bug that caused the cut off of the gyro's internal filter to be set way to high.
The gyro not being detected in 4.5 is likely a cloud build issue
I do remember that the ICM-42688P can pick up a static offset if it is moved slightly when powered up. Not sure if this is a BF thing though or if its just that specific IMU.
but that offset was small, and not back and forth.
Can you make a log?
Yeah I will make one here in a little while
4.4.1 doesn't fix the issue. I'll post a BB log.
@light shell @frosty crag here is a GD link with a video of the behavior. It "seems" that it only starts doing this after a battery has been connected and it has taken flight for any time at all. After it sits for a bit un powered then plugged in via USB the model in BF is solid as a rock.
Oh, it's causing RTH to fly off in the wrong direction and you can forget a locked in acc trim.
@frosty crag Here is a log from this mornings test flight.
Just so I'm sure: was the quad sitting stationary for the entire duration of that video?
Yes
đ
Definitely use 4.5 Zulu with this gyro. I don't know exactly when, but the filtering was optimised for this gyro relatively late during 4.4 development, perhaps 4.4.1, not sure.
The log looks alright. There is a little too much high frequency noise, but not a massive problem.
D is relatively high and Dmax in particular very high, leading to lag.
There is some 100hz resonance most likely props/frame exaggerated by D.
D filter is higher than normal.
Gyro filter is higher than normal.
The gyro does not appear faulty.
These gyros have much less internal filtering than the MPU6000, with less delay as a benefit, but do let more noise through at the high end.
Even so I wouldn't say it's outrageously bad.
To get rid of the really high frequency noise you could make gyro lowpass 2 into a PT2, and put all filter sliders in the default position.
The resonance at 100hz is difficult to remove, the best way is to run less D. You could lower the lower limit of the dynamic notches to 80hz so that one would focus on it, but they introduce delay and distortion.
I would use more feedforward, less P and less D on this build.
I don't see any evidence that the yaw axis is bad. I do see RC link issues, worse on yaw, but the yaw PIDs are doing what you request reasonably well.
In this image you can see abrupt setpoint steps leading to abrupt feedforward blips.
However the gyro for yaw shows that it is following those steps quite well.
Most likely you haven't turned the ADC off in EdgeTx. You need to do that.
Overall the gyro is perfectly fine, I think, just that the tune is a little hot and the filters a little too high for it.
The icm-42688 fix is in 4.4.1.
But that bug shouldn't cause the quad to yaw back and forth in configurator like it does in the video.
Since the battery being connected seems to make a difference I suspect some kind of weird interaction with the GPS, but I'm not sure about the details.
I didn't spot any magnetometer data in the log, so I assume that there isn't one on the quad. Otherwise that could explain some odd yaw behavior.
If your problem relates purely to issues in Angle mode, please flash 4.5Zulu and try that.
It's possible that there is a pure accelerometer issue affecting angle mode, that I don't know, and I don't really know how to test that.
The GPS rescue has pitch blips every second so it is likely an M10 and the GPS is not working properly.
Thank you for the log break down. I'm not the best at reading these yet but I get by just fine for tuning. This is Dave C's factory tune for this quad. I haven't tuned it yet. ADC filter is certainly off, just went and checked, when you said that, it messed with my head for a sec. Running Edge TX 2.8 and ELRS 150Hz, v3.2.0. with the correct preset, made by you actually. Yeah this BNF quad is pretty much straight out of the box from Flywoo. I put thread locker on the motor and bottom frame screws. Created an EMI shield from copper foil tape around gps leads and grounded it. Then wrapped the EMI shield in cloth tape, for protection and all. I put the arm braces on, checked it for potential mechanical resonance issues, made sure the stack screws were just right then put it in the air for a few flights. My first and second GPS RTH test went fine but my 3rd flew off and I had to take control, come home and power cycle the drone. It did it again after a few more RTH tests. I brought it in and noticed the model doing the twist in the setup tab. Other than that it flies fine but I'm 100% sure the ACC is drifting pretty bad. Probably why it wants to fly the wrong direction upon initiating RTH sometimes and trimming the ACC to a consistent state is impossible. I did try 4.5.Zulu finally and it seemed better but not fixed. Thanks, I would like to know what might cause the abrupt setpoint steps though. Would a very loose receiver cause this? @light shell
Sounds like a very good build
I'm going to rip it apart and go over it with a fine tooth comb if flywoo doesn't decide this acc is bad.
complete rebuild
The issue with random gps weirdness isnât an error on your part. Under certain conditions the initialisation of the GPS hardware isnât quite right and the whole flight it just behaves weirdly. This affects about 1% of my tests. Typically it shows the home icon but then puts home in a weird place some distance away. If I return and power cycle it, the problem goes away, but disarm/arm cycles wont fix it. This appears to be an issue with the gps initialisation code and people are working on that as we speak. That problem has been there for many years.
Ok i've seen this a few times in testing. I'm in the tester s group.
wish I had more free time though
There is a second gps problem that we do have a workaround for, there everything is perfect from takeoff but if the quad is drifting hard and the climb is >4-6 seconds, then during the climb the quad loses orientation and heads off the wrong way. This is also a long-standing problem but I only really figured out exactly what it was just now. So there is a PR in 4.5 that handles this much better than before. You can flash it with 4.5 Zulu and youâll find that if the IMU is confused, the quad will initially head off the wrong way (this is inevitable, we cannot stop it), but then it will turn for home more effectively. Read the PR for details.
If these things happen occasionally, it doesnât mean your accelerometer is faulty. They are ânormalâ for 4.0/4.1.
What I donât understand is what you say is wrong with yaw? I donât see anything in the log that seems wrong.
What exactly is the yaw problem you are experiencing? What flight mode? Is it reproducible? Did it happen in the log?
Did you watch the video in on the goggle drive. I've never seen this behavior in the configurator from a quad sitting still
What do you mean by âdoing the twistâ in the setup tab?
Ok I looked. Was the radio on?
The radio was off and the indicator is on angle in the modes tab
I thought the yawing model and the issue were related
Ok so maybe repeat and watch the modes tab and the receiver tab. Check what mode it thinks itâs in.
with the radio off?
Yes
ok it's up now
Iâm suspicious that the Rx is a bit weird
I can fix that
I think this is down to the newness of the ICM42688P at this point and it's not a deal breaker. 4.5.zulu is definitely better
Youâll know if there is random signal in the Rx panel. If the Rx panel is silent, then thereâs no setpoint change.
I think the quad image on the start screen comes simply from accelerometer values.
it does
all is still and everything looks good with movement
If you tap the quad youâll see movement on gyro and acc. Check that acc traces are stable.
they stop moving when the quad stops
Yeah leave it alone and watch.
so far nothing
If the acc traces, especially yaw, are steadfast on zero, then check the setup screen
I got rid of that funky proprietary ELRS on the receiver earlier and updated to 3.2.0 BTW
good and stable running 4.5.Z
It could be that there are transients on the acc that are not showing up. Set it to the maximum fastest time scale, 10ms I think. Thatâs still only one sample every 40 samples so you may miss single sample transients.
If the quad spins on yaw in the main screen and the acc and gyro traces are dead flat well thatâs something very peculiar indeed
It's not yawing right now
it may do it after i go out and fly it
but
i have 4.5.z on now I'll test it tomorrow and see if this weirdness is gone
whatever it is it only rears its head after flight
on 4.5.z i can get a consistent acc trim inside, i couldn't do that before
If youâre game, flash Pr 12738 the gps IMU fix one. Set your max return angle to 40 degrees see how it goes. Donât attempt a rescue if the wind is around 50kph or higher if itâs critical, but otherwise by all means test in high wind to find how high a wind makes it fail.
I wonder if itâs an electrical issue.
But then if you see it on usb only then probably not. Very strange indeed.
yes but it's nit picky and you said yourself sometimes gps rth acts up
Intermittent faults are a pain. Often the only certainty comes when a part is replaced and book itâs all good.
I'm thinking of ordering a replacement FC ith a BMI270 and calling it a day
Yeah but in that Pr the gps will be far more likely to be good in situations where it would have been bad.
I'll test that PR tomorrow
Thanks
Thank you
I think the mpu6000 is available again and itâs been the most reliable and durable for me
for sure, i like the BMI270 as well
50% or more 42688P give me crap
Well I'll let you know what shakes loose, have a good night.
@frosty crag that PR doesn't detect the gyro, lul
?
its based on master, should be no different from any other flash. Did you include the correct gyro when specifying the build config?
Just another sanity check: have you been doing full chip erases when flashing and tried with just the minimal settings (I suspect you have since you must have flashed the board at least half a dozen times by know, but never hurts to make sure)
I downloaded the assets folder from the PR like we used to do and extracted the f405.hex file. It won't give me options for cloud build when I load locally. I'm sure I'm going about it the wrong way though. Something else, I loaded the hex file from flywoo, 4.3.1. It didn't detect the gyro. After I loaded the CLI DUMP they provide for 4.3.1 it did. I don't know of a cli command to detect a gyro. Right now it has the latest development release of 4.5.Zulu from the latest nightly configurator. I'll test that this morning. How do I use assets directly from the PR you specified with the new cloud build system.
10-4
These days you build the code in the cloud. Use Configurator in expert mode. Choose the settings when you build.
Ok, that's what I've done for the testing with the latest development build of 4.5.Zulu in expert mode which TMK is correct. I'll test that this morning and see how it goes.
Packs are charging now!
đ”âđ«
@frosty crag I loaded the latest development release in expert mode for 4.5.Zulu. Configured the quad from scratch, all looks fine. The 4.5.Zulu test went terrible, RTH would have put me somewhere in the woods a hundred feet away had I let it. When I landed and disarmed then tried to re arm I got RPM FILTER from OSD warning with no ability to arm until power cycle and my VTX broadcast a black screen to my goggles 50% of the power cycles. Everything else seemed to work as usual. Cool huh? All is working very well with 4.4.1 now. I'm getting RTH landings within 3 feet of home point on 4.4.1 but this could be a good GPS day. All is working well on 4.4.1 for the time being.
Looks like 4.5 still needs some fine tuning.
4.5 is good. Lots of people flying it with no problem. Must have been something very strange about that flash I think. None the less, stick with what works.
@frosty crag I will, I think this board is poorly designed, I don't have trouble with any of my other 42 quads besides an odd ball or two. It is what it is, thanks!
Good luck! I too have had boards that I just gave up on.
@frosty crag Rebuilt that quad from the ground up and moved the receiver t the rear and all issues cleared up. I had a feeling the Fllywoo ELRS receiver was causing some sort of interference as it was injecting RF noise into my camera feed being right behind the camera, she built right now. No more BNF quirks.
Is it flying ok now? I hope so
@frosty crag Yes on 4.4.2-Zulu just fine and all the way back to 4.4.1
Thanks again she just needed an overhaul
You would think these manufactures would know better
Thatâs incredible, really.
Itâs the second recent build related âgyro failureâ that weâve had here recently.
So glad itâs flying well!
That's interesting. I'll have to add "blast a ICM42688 with wifi" on my want to do list