Well I feel better now. I'm always a little nervous to fly a new ship until the first crash. You never know how she'll deal with adversity.
I was flying FPV in loiter at ~15m when it suddenly pulled hard right and back making a nice arc into the ground at about 20kph according to the telemetry log. The current jumped from the ~30A normal to over 60A in the few seconds of the death plunge. When I got there the motors were spinning and wouldn't react to transmitter commands including disarm. All the props were broken of course.
Looked through the logs a bit, but I can't tell what happened. If anybody cares to take a glance and give me some insight, I would be very grateful. I don't mind when things go terrible if I can at least understand why it happened.
The damage seems to be limited to a couple of cracked arms and a few other frame bits. The GoPro was fine thanks to the waterproof housing. Not even a broken antenna on the 3dr radio or video Tx! I'm taking it all apart tonight and checking the electronics, they got covered in snow, but no physical damage that I can see so far. I'll upload some pics once Iphoto stops crashing every time I try to get them off my camera. There must be weird solar weather or something going on today, I've never had a problem getting pics off of my camera before, really strange...
I'll have to upload my tlogs from the windows machine, just a minute.
: )
Comments
@Randy, there is now spike in x and y magnetometer. Looks like there is change in value corresponding to rotation of vehicle.
I have had many issues with the compass calibration procedure in the past. I always test the cardinal points(0,90,180,270) on the bench after calibration. It is very important you rotate around the axes of the magnetometer as the magnetic field a few inches away can be different. Then I elevate one side of the quad into a 45 deg bank and retest again if i continue to get errors I change the values manual to get consistent indications. I have had a bank create a 30-40 deg change in magnetic heading after a auto calibration. The auto calibration seems to be more consistent now but I always double check.
OK I was just looking at the log playback vs the FPV video. Looks like a large compass error. as you cross the road in loiter the log indicates a heading of 293 after the slight yaw turn. The FPV video shows you looking between the roads that should be ~180. Is it just a compass calibration?
How about a picture of your APM instillation. How is the power distributed. I am working on a quad that I am not using a power distribution board. The issue Randy mention did cross my mind. A wire caring current does create a magnetic field around it. An increase in the current = increase in magnetic field. If this field effected the magnetometer it could have created a feedback into the code that caused the crash.
I am using the power module which should be able to supply the gimball servos. Next time I will use the BEC rail to power the servos
Joshua,
Ah right. The nav_bearing is the direction towards the waypoint so that's the direction it thinks it needs to go in to get to the target point. So really it's recognising that the direciton of the target is changing...but it really wasn't successful getting there.
The GPS lag is about 0.5sec on the ublox and about 1sec on the mediatek. We don't use the gps for heading in anyway but it would be involved in the nav_bearing in the graph.
My guess is that it got confused about it's heading because the sudden throttle up of the motors interferred with the compass. We've got some code on the way, probably for 2.9.1 that will help compensate for that.
Randy, I think the heading deviation I was commenting about is just normal GPS lag after researching a little, is this correct?
Randy, I was just re-reading your comments, you say that the loiter point was 5m away is that at ~640 point on the time axis in the graph? At 622 it looks like the nav roll is starting to go against my inputs.
Thanks for all feedback folks it's the most vital component in our endeavors, in my opinion.
Randy,
The first thing that jumps out to me after looking at the KMZ file you posted, is the orientation of the plane graphic. It's 90degrees clockwise from the actual orientation of the craft starting in the middle the little circle around that structure on the ground. On a similar note- I've always assumed that in the mission planner ground station map, the copter icon is in a + be default. I'm pretty sure that I have my config set up as X as that's how all the hardware is set up for sure (would it even fly in the wrong config?)
I've been going through the logs of the previous flights( I've flown this guy about 5 flights with mavlink connected plus a few short hovers in the backyard(I need to pull the logs of the sd card for those)). In the past flight I'm seeing that differential in the outputs of diagonal pairs, the crazy thing is that 1&2 are higher in some flights. It's looking like maybe I have some kind of mechanical issue, which is crazy because it seemed to be flying so well. I've been mostly focused on the gimbal setup and pretty much just assumed the flying components were fine, I didn't do much tuning or looking at logs until now.
Leo,
I measured the total current draw for the APM and servos while wildly moving the frame around so that both servos were being worked hard and I was seeing ~.325A max. The gimbal mechanics are very friendly to the servos, there is never any leverage against them, they just spin the camera while it remains in a very balanced position. I checked the battery that powers flight and camera control and it was at ~60% full, it's a 3s and was well above 5V for sure (I have a standalone voltage alarm on it, but only have voltage and current measurement capability for the big motor battery). The video shows that the gimbal was still under useful power during the incident so I assume APM was well powered also.
Quadguy,
That looks interesting, this seems to support the anomalous heading in the KMZ file Randy posted above. I'm powering the input and output rails with a 20A v-reg and the jp1 disconnected.
Richard,
iMovie does have stabilization and even a rolling shutter motion reduction, but I never really mess with them as I'm trying to eliminate those problems at the source so I want to see all of the bad stuff.
I'm running (3) batteries when using the videoTx, (2) otherwise.
Emile,
Glad you saved it! I wonder what happened there?
Here is a video of my last glitch that was similar but did not end up on the ground. I was in loiter trying to do a slow decent down. A review of the OSD shows the APM went into stab mode with a hard roll then back.( not my selection) If it was a loss of signal It should have gone into RTL.
-
1
-
2
-
3
-
4
of 4 Next