I am dumb struck!!
I have a Bixler 1 and have been flying it for three to four months with no problems at all.
Have a bigger motor in and FPV gear on it. Last Thursday went to fly and something very strange happened. The Bixler nose-dived on its own, in stabilize mode. I did manage to save the bird, but 1 min later it did it again. But this time I could not save it as it was too low. I also did not have any telemetry connected.
Rebuilt it and went to the field Saturday morning. This time I had telemetry connected. I took her up and flew in manual mode for 3min, then switched to stab. It was not even 1min and it happened again. Saved her but the second time I could not. Crashed again !! I was busy turning (with the ail and elev) the nose just went down uncontrollably. I looked at the tlogs with Graham Dyer, what we could see is that the vcc was not very consistent at the time of the crash. Tlog attached named Saturday.
Rebuilt the Bixler AGAIN !! I swapped the ESC too and reloaded the APM firmware (v2.68). Calibrated everything and tested it on the table. Sunday morning went out to the field. Once again took her up in manual and flew around for a while. Switched to stab and it went well for a couple of min’s. Then it happened again. Nose down 100% and whatever I did, did not help. CRASH NUMBER THREE !!!
We looked at the tlog and the power this time was 100%. But at the point of the crash there was nearly no input from ch2 but 100% output on ch2. Tlog is attached below, named Sunday
Rebuilt Bixler AGAIN !!!! Took out the APM 2 and swapped it for another one I had in my quadcopter. I went this morning and flew the Bixler again. It was much better but found something interesting, when I bank full ail the Bixler will tip over and dive, but this time it recovered after a second or two. Still very stressful to see. I tested this a couple of times, it loses control but gets it back again. Still not 100%.
Thursday I flew with 2.66 the rest was done with 2.68. I also loaded the default bixler.para from this site every time.
Please can someone look at the tlogs and give me some advice.
Hi Luis, i am very sorry for your crashes,
Ihave like many others a bixler, so will try to point some reasons, most of them you may already checked.
- I see you still have the magnets on the canopy.. it was my No 1 reason for eratic behavior. specialy with the magnito.
- You have a bigger motor, is your esc efficient enough? maybe you have experienced tiny brownouts. do you share the same bat, for the fpv gear? try putting an external bec?
-is your mount secure ? many times the spaggeti and battery underneath the APM mount push and displace the APM.
-Now some silly ones, did you make radio calibration after uploading the params?
-interference from fpv and xbee cables?
from your tlog, something caused APM to 100%elev, i will look your logs to see if can find something
ok i made some checks, thanks for your answers,
so we can narrow it down to the canopy magnets and the ESC bec, both really mess with APM 2.0
and rc in-out @ 466 (just before the fall)
-another strange think is that your mags are all zero! (canopy magnets maybe?)
-Also there is a small drop out at vcc just before the strange behavior. I had 4/5 unsuccesfull nav flights using the esc bec, and magnets. 120/120 succefull flights with external low noise bec and magnets off.
-3rd minor prob, your GPS alt is a little jumpy near the fall, common issue with dirty power. from the log, it looks like you gain 5-7meters rapidly, and AP tries to keep it down.
- One last think, did you erase-reset after firm update?
Again all those are just speculations, iam very familiar with the bixler, but not the best guy arrount for log analysis and debugging.
hope i helped
1.Always use a separate voltage source for the autopilot and servos
2.vibration can shake gyroscopes
Do you think you realy realy need the magnitometer? I'm flying UDBs and AUAVx and MatrixPilot and never used a magnitometer.....no serious crashes at all....pls think about. On APM you can disable the magnitometer.
I've been working with Louis to try get this sorted.
1) It's not the magnetometer, as Louis said he's been flying for 3 to 4 months with the magnet in the same place, I have THREE magnets, (two ahead of the APM and one behind) on my plane canopy and have not had this behaviour, the magnetometer software learns which is magnet flux and which is north pole flux and compensates for this. Any magnetometer misbehaviour would also not cause the extreme pitch down that Louis experienced.
2) The one log (using the old ESC) shows the voltage at around 4.6V dropping to 3.9V at one stage which definately caused a brownout so the ESC was changed for a brand new one and the log of Sunday and Monday's crash clearly shows the APM's voltage was stable at around 4.95V for the entire flight so it's not a power issue (see image below).
As James' pictures show we need to figure out WHY the gyro suddenly thinks the plane is pointing straight up. At that point it commands the servo full down as shown with the green trace in his pic.
Looks like your flights were a bit more exciting than they should have been!
I've been having a look at your logs, and as Graham has said, it's not a magnetometer problem (you had the magnetometer disabled anyway). It looks to me like you went into a spin because of a too sharp turn at low speed. I'll go through the key graphs below.
When analyzing a set a logs like this what I tend to do is extract out just the portion of the log near the crash to make it easier to graph the key variables. I've attached two extracted logs below, one is the Sunday crash, and the other is the first big dive on the Saturday.
Let's start with the crash on Sunday. First off, the altitude and airspeed:
The red line is the altitude in meters with the scale on the left. You can see where it drops rapidly as trouble begins. At that time the airspeed was quit low - about 8.1 m/s. That sort of speed would normally not be a problem for a bixler unless it was very heavily loaded. It is worth noting that the problem started at minimum airspeed though.
Next let's look at what the APM thought the attitude was - the roll and pitch in degrees. Note that you don't start by looking at the gyro values - the gyro tells you your rate of rotation about each of the 3 axes, not your attitude. What we're interested in initially is the attitude. Here is the roll and pitch of the plane in degrees, with the altitude in meters to show us what stage of the crash we're at:
notice that the roll was very high (peaking at 82 degrees) when the trouble starts? A plane like the bixler likes to fly fairly flat - a roll of 60 degrees is used in a sharp turn. A roll of 82 degrees is way beyond what it can reasonably handle, especially when flying slowly. The amount of vertical lift that the wings give is proportional to the cosine of the roll. At 82 degrees it is getting less than 14% of the lift that it would get when flat. So unless it is travelling very fast it just won't have enough lift to keep it in the air, and it will stall.
Next we need to check if the APMs idea of the attitude is accurate. There is some pretty complex maths that the APM does to work out which way is up, and we need to check if it was getting it right. The way I normally check that is to do some much simpler maths that gives an approximate attitude based solely on the accelerometers and the centripetal forces from the airspeed and turn rate. It doesn't use the X and Y gyros at all, which means it isn't using the main sensors that normally are used for attitude. Here is the APMs idea of attitude against the roll estimate:
That shows that the APMs idea of the roll matches the roll estimate quite well. It breaks down a bit as the plane nears the ground, but that is not surprising. The key is that it matches well when things start to go wrong, so I think we can trust the roll estimate.
That also means the gyros were working correctly. So now we can look at what the gyros were telling us. As I mentioned above, the gyros tell you your rate of rotation about each axis. The xgyro is the roll rate, the y gyro is the pitch rate and the z gyro is the yaw rate - all 3 rates are in body frame.
Here is what the gyros were showing:
The key result here is the xgyro, which shows that when the trouble starts the plane starts rotating very rapidly about the X axis, peaking at -256 degrees/second. That is a very fast roll. It is also rotating quite fast about the z axis. What it looks like to me is that the low airspeed combined with high bank angle led the plane to stall, and essentially fall out of the sky.
That's enough graphs for now, it may be more useful to suggest what you can do to avoid this in the future!
First off, I'd recommend you fly in FBWA mode instead of STABILIZE mode. In FBWA mode the APM will prevent you from rolling over too much. The config file you have loaded now has a LIM_ROLL_CD of 4500, so it won't roll past 45 degrees. That is a conservative setting, and you'll need to use some rudder to make a tighter turn, but it is unlikely to stall unless you pitch it up hard at low speed. It is generally much easier to fly in FBWA than STABILIZE mode.
In STABILIZE mode the plane will add your control inputs to whatever it thinks is needed to level the plane. That means if you use too much aileron then the plane will roll over too far, and can stall and fall out of the sky (which is what happened to you). In FBWA mode it won't roll past the limits you have set in the config file.
Another factor was probably the weight of the larger motor and FPV gear. The bixler is a small plane, and isn't really meant to carry a lot of weight. You can load it up, but if you do then you need to fly it much more conservatively. If it has the standard motor and no extra gear then it is very light, and doesn't need much airspeed to keep it in the air. As you add weight it needs a lot more lift to compensate, and that means you need to either fly it faster, or fly it flatter. In the case of your crashes it was flying slow, and at high bank angle. The small part of the wings that was still generating lift at 82 degrees roll wasn't enough.
I hope this helps!
Thank you for the detailed expert analysis, Tridge, it's much better than watching the program "Seconds from Disaster".
While I was at the field on for the first two crashes I didn't actually see them so didn't think that a stall might be the cause, but it makes sense especially as the weight of Loius' Bixler (v1) is higher with a bigger motor, bigger battery, 40A ESC, APM & peripherals, Spektrum telemetry and the extra glue and tape used to fix it after the crashes.