IRIS Run Amuk

Posted by StevenK on June 5, 2015 at 12:46pm

Hello all,

  I am not new to APM or arducopter, but I wanted something easy so I bought an Iris+. I have flown about 10 batteries and it was flying great until this afternoon.

  I had one good flight, landed and put a new pack in and took off in Alt hold mode (since I am using the TX as it was setup). I Switched to loiter mode and the IRIS hovered fine for a minute, and then started acting unusual. It started to not respond to my inputs, so I switched to RTL and the IRIS started to climb and fly away. It was like I had very little input into what it was doing. I switched back into Alt Hold mode with no change. I managed to get it to come back to me and got all the way down to a few inches off the ground, put throttle to min, and it proceeded to accelerate sideways and slammed into the side of my house.

   This is quote unnerving and I really need to get to the bottom of it before I fly again. I am running Copter 3.2.1. I have attached my log, if anyone has any insight into what the heck was going on I would greatly appreciate it.

 -Steve

30.log

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –
Follow

Replies

  • Rob_Lefebvre June 9, 2015 at 7:57am

    I had a look at the log, and don't really agree with the statement that this was a vibration problem.  The vibration levels actually look pretty good to me, and and I definitely don't see a connection between high throttle and vibration.  I think what Craig is seeing is actually movement in the Z-axis accelerometer levels which are actually due to the copter moving around and accelerating.  That's signal, not noise.  There are a few areas with noise, but they seem more correlated with GPS speed than throttle.

    So I called up Randy and we had a look at this together.  The problem appears to be more likely a gyro problem.  I noticed the GyroY values diverge between IMU1 and 2, just before the problems start.  It's not sudden, and not extreme, but it's enough to cause DCM pitch to diverge from the EKF pitch.  This is what then leads to the uncommanded climb rate, the copter thinks it's falling when it's really not.  The strange thing is, the Gyro divergence gets better, then diverges a bit later on, then the other axis goes off a bit too.
    Randy noticed that the IMU1 (MPU6000) has a HUGE Y-Gyro offset on boot up, 14 deg/sec.  This seems very large. It's not clear if it was because it was moved on booting, or perhaps an indication that the MPU hardware is not healthy.  We should investigate this further.  Perhaps review of other logs from this machine might reveal similar high offsets, and we could consider a health check for offsets being reasonable.
    The yaw estimation also goes off during the uncontrolled climb, and the EKF yaw estimate can actually be seen to reset, while the DCM does not.  If we assume the EKF reset was correct, we can then assume that the DCM yaw remains wrong, and that would then explain the general navigational errors that lead to it hitting the house.
    We're not saying for sure this wasn't a vibration problem, as it does have many of the symptoms of that.  But the data just doesn't seem to show that at first glance.
    Steven, can you please send us some earlier logs from this machine?  We'd like to see if the IMU1 GyroY offset is consistent between boots.
    We also noticed that it's quite hot where you are, the barometer is registering 56°C.  This might be affecting things, don't know yet.  Was it particularly hot on this flight, or similar to your other flights?
    Rob
    • StevenK > Rob_Lefebvre June 9, 2015 at 12:08pm

      Hi Rob,

       Thanks for taking a look at my issue, I have included two more logs. Log 28 I believe is from a flight where I first experienced strange flight behavior. I convinced myself that it must have been loss of link, but now I believe if was a less severe manifestation of this issue. Log 29 is from the flight immediately before the first log I posted (the crash). It flew fine, I landed changed the battery, plugged it in (with the copter stationary on level ground, and then took off. It was warm out, probably around 90 degrees and the sun was out, but it wasn't extremely hot or anything. Thanks for your help.

       -Steve

      28.log.zip

      29.log.zip

      https://storage.ning.com/topology/rest/1.0/file/get/3702572192?profile=original
    • Rob_Lefebvre > StevenK June 9, 2015 at 4:24pm

      Interesting, both these logs also have a very large Yaxis MPU6000 offset.  28 also has the same problem with the MPU6000 diverging, then coming back, and includes navigational errors as well.  29 is OK.

      Can we go further back?  Trying to find out if this chip was "born with it".

      The chip has a zero offset of about 16 deg/sec, and the spec is 20 deg/sec, so it's "in spec" but it seems to be outside the normally observed range which is usually very small.

    • Andy Piper > Rob_Lefebvre June 10, 2015 at 2:59am

      Why do the EKF learned offsets oscillate in log 28? Seems very weird.

      3702789278?profile=original

    • Rob_Lefebvre > Andy Piper June 10, 2015 at 5:59am

      I don't know, but I don't think it's a good thing.

    • Jim Phelps > Rob_Lefebvre June 9, 2015 at 10:35pm

      I had a recent crash with exact same Yaxis offset.  The vehicle was at loiter for about 5.5 minutes at 10 meters, then started drifting upward.  There did not seem to be response to the sticks.  It rose to about 28 meters then headed in a westerly course for about 200 meters then took nose dive.  My partner is uploading the logs. 

      Jim

    • DavidFolts > Rob_Lefebvre June 9, 2015 at 6:29pm

      Rob

      How do I convert Steven's a.txt files into a .tlog or a mavlink file?

      tks

      Dave

    • Developer
      Craig Elder > DavidFolts June 10, 2015 at 12:58pm

      The don't convert.  Please read the wiki on log file types.

    • StevenK > Rob_Lefebvre June 9, 2015 at 4:33pm

      I am afraid I am no longer at home, and won't be for some time. I should have pulled all the logs, I just didn't think about it.

       I just got a response from 3DR, here it is, please let me know what you think:

      Hello Steven,

      Thank you for being patient. I checked the logs and noticed that your Iris+ may be acting weird because of bad startup. That’s why I would recommend you to do a compass calibration while being outdoors so there would be no interference, you can follow the steps on the next video: https://www.youtube.com/watch?v=oD9Z9IR2TNU

      After you do the compass calibration follow the steps from the next link to arm the Iris+ correctly, to avoid any miscalibration: http://copter.ardupilot.com/wiki/flying-arducopter/arming_the_motors/

      Let me know how it goes,

      Ruben Jacobo
      [Tech Support Engineer]

      YouTube
    • DavidFolts > Rob_Lefebvre June 9, 2015 at 9:02am

      Rob-

      I had a similar event on a clear hot morning (Texas... its always like that).  The flyaway event was caused by a "gyro problem" per 3DR analysis.

      At any rate, on previous flights as the day wore on, I would get a "GPS Glitch" warning on the Tower app. 

      I don't have any temp readings, but I suspect the black bodied Iris, with just a few air holes, is a heat sink and under the top where the Pixhawk is mounted, there may be some overtemperatures causing hardware glitch. 

      The other working theory I have related to heat is that there is something in firmware 3.2.1 causing some heavy CPU loading, at times, which is also over temping the CPU whilst in warm ambient environments.  Its seems slightly more than coincidental that there are a more occurrences of flyaway now that the weather has warmed up.

This reply was deleted.