Undefined behaviour in Position Hold

[rput]

Hi all,

sorry to keep bothering you with problems, I would like to contribute more but am still at the phase where things happen that I cannot explain and resolve myself

Today I was performing a demo with a quad at a different location than where I fly normally. Because of the issues I had before in PH take-off (see other post: ), I took off in manual mode and switched to PH in the air. As soon as I switched to PH it started drifting a bit so I immediately put it back to manual. When I switched again to PH it really went crazy and nearly went 90 degrees. I recovered in manual mode and landed.

Normally I would have stopped there, but I remembered a previous occasion with almost the same symptoms that was resolved after a reboot. So, I rebooted and indeed it flew great in PH this time and I completed a mission without much issues other then some overshoot which I still need to fix.

I reviewed all the log curves, but don't really see any issues to explain the behaviour. I first thought it might have been a GPS initialization issue, but the GPS seems ok. There are some vibs (incidently I did change the props with statically calibrated RC-timer 12" ones since my last post which did not make much difference), but since it flies very well in manual mode and the issue was gone the second time, I doubt this is the cause.

It would be really fantastic if someone could help shed some light on this.... Coupled with the other problems I described in other posts, my confidence in resolving this by myself is getting low. But since the AQ is such a great platform when it works, I am determined to make this work. I look forward to hear from you!

Cheers,
Richard


System:
Quad X, 2.5kg, 6.1 with DIMU, ESC32, Tiger Moter U3, RCTimer 12" Carbon props, Spektrum DX6i

[Jdmagoo]

I haven't looked at your log yet due to been on an ipad but what sort of GPS lock did you have before switching to PH? If your GPS isn't so good then this might explain why your having problems?

[rput]

HDOP was between 1-2, which I think is sufficiently good?

[joebar.rc]

After a quick look I noticed that your GPS accuracy is not really good. Both the hAcc & vAcc start around 1.2m (which isn't really good to start with, but ok) and drift to 2m.
The sensor fusion will correct some of it, but 2m gps accuracy in pos hold is not very precise and will cause drift. This might obviously not the only reason but try to get your GPS accuracy in the lower <1m numbers or better <0.5m.
Maybe it was just a bad day for reception (they exist), your location is blocking reception or the signal is not optimal (antenna position, noise etc)

[JussiH]

Richard,

As Menno points out the GPS HACC and VACC is less than ideal.

The usual suspects for GPS dropouts are connector contact issues, Antenna failure, EMI from some peripheral (radios, power supplies, wiring, cameras (GoPro is a notorious EMI emitter) or interference from outside sources.

Try new antenna, make sure connection from antenna to board is good and tight. Locate peripherals further away, raise antenna up, increase ground plane size, move antenna wire, move location.

GPS quality varies, this tool can help you estimate if you just had a "bad GPS day" or if the problem is something else. But very likely you have a reception problem to look at first.


http://satpredictor.navcomtech.com/

It does take a while for the GPS to arrive at a good estimate. As the solution quality increases, the accuracy will increase. At ground level where there is obstacles in the skyline, you should expect to see lower accuracy, simply because the more sats are visible the more precise the solution will be.

HACC and VACC is the parameters to look at when estimating if you have a good enough signal to do a reliable postion hold. In general you need HACC to be <1m to get a decent position hold, but you should expect to see <0.5m in a good copter with full visibilty of the sky. If its consistantly higher, you likely have a reception or interference problem.

But we need data to help you more here.

Details about setup - pictures of it; details of wiring, antenna placements, peripherals, batteries, frame materials, etc

Video showing the problem and a corresponding logfile is crucial to make us help understand what you mean by "undefined behavior"....the more data you provide the better the chance is that someone will be able to help spot the problem and solve it.

[rput]

Thanks Jussi for the detailed post, that is very helpful.

I noticed that on different flights the HACC and VACC varies quite a lot from 0.5-8m. Would a bad VACC be a reason for a sudden change in altitude? If so, this may very well be the issue, which makes me hopeful. I am using the AQ shield, but am now doing a new build with a CNC-routed PCB and a custom shield made of half a copper-clad board.

We will have to investigate what the cause is of the GPS inaccuracies: EMI, bad connector/wire, other... We do have a GoPro onboard, but are not using the WiFi on it. Still, we will try and zoom in on the different components to see the effect.

I will try to post some pictures later, I have disassembled everything at the moment to put on the new PCB.

Richard

[JussiH]

Absolutely, VACC (Vertical Accuracy) and HACC (Horisontal Accuracy) has direct impact on Postion and Altitude hold under GPS.

But 0.8 should get you a decent altitude hold, provided the ACC and Baro is working correctly and the ACC is properly calibrated. DIMU Tare will not yield a perfectly calibrated ACC, so you might also be seeing some effects of ACC bias. But as the VACC increases, you should see altitude hold improve. But you in some cases also need to tune NAV_ALT PID to get a perfect result.

Bear in mind that the nature of the UKF means that it constantly estimates bias in all the sensors and uses the others to correct it. This means that it can take some time for the filter to converge the position and altitude estimations. GPS reception and the initial state of calibration on the sensors plays a big part there.

You can make a test to determine if the clean altitude hold works properly. Just unplug the antenna and see if you can get a decent altitude hold without GPS.

For the ACC Bias and scale, you can get an idea by looking at the live telemetry and checking that ACC magnitude stays close to 9.806 no matter how the copter is orientated.

Record the ACC Z-axis value with the copter leveled and upside down and tell us. Or post a log that shows it.

[rput]

So, yesterday we had some bad news. After disassembly and re-assembly with the new PCB, the DIMU no longer wants to work. IT gives the same problem as described in this post:

viewtopic.php?f=42&t=3127

with the debug light coming on. So we have decided to quickly try to tune the analog IMU using the latest stable firmware release. Hopefully we can get it all working as we need it ready by tomorrow

Here are some pictures of the system with the new PCB and the custom groundplane installed. IT does not differ much from the previous version except on two points: I was using an AQ groundplane, and 2. the other one has U3 tiger motors. Would love to hear your thoughts.

After the calibration is finished (started last night, but may need to redo the dance as my mag MAE in step 3 stays around 0.4) I will update with some more data.

Cheers,
Richard

[Max]

Looks like a very neat setup, nice work. Do you have insulators between the aluminum standoffs and the AQ board?

-Max

[chschmid]

Very nice setup. The open circuits look dangerous to me. Maybe you can cover them with some insulation. The AQ needs to go on nylon standoffs or at least use Nylonscrews with nylon shims.
Nice landing gear with professional welding

Cheers
Christof