Beta Testers requested to fix Compass Issue

[sar104]

Sar104, agree with your sentiment and testing method. The real fix should have the initial compass readings right on target without any adaption and testing the original CL is the best way to see that. I will say though that your results are different to mine. I need to go out again this weekend and do some more thorough CL tests.

I plan to do the same today, and I'll include resetting CL after some initial passes. That is why the CL response is so important - it should work fine with or without declination correction, and will only be wrong if there is a bad attempt at declination correction. I guess from your reply that we really don't have any direct information from DJI that the Naza has declination information, but unless there is a flaw in my reasoning then it seems that it has to. I'd really like to see some definitive information from them that shows either why we are mistaken in our understanding or that they do realize that the problem is not a learning issue.

 

[ianwood]

I don't think they're going to be that transparent. Something is adjusting for declination above and beyond in-flight learning but it certainly doesn't seem to be as simple or as accurate as the lookup table solution.

 

[John Shaw]

I don't think the current fix is just faster learning. The TBE is essentially gone at takeoff. I takeoff to a hover and there is very little motion before it settles in. I don't think there is enough motion for learning.

Obviously the baro altitude is relative as we never enter the ground elevation or the baro setting as you do in an airplane. In flying an airplane you are required to reset the baro setting periodically. I expect even though we are not going far that in just the 25 minutes we have on a battery there could be barometric changes.

As Ian mentioned since our altimeter is always operating in the prop down wash it is going to change with power settings and the effect will be different at altitude compared to close to the ground. Airflow close to the ground is different than at altitude (ground effect).

The vertical error in GPS is generally 1.5 times the horizontal error and I don't think would be good enough for altitude holding. Even using GPS for relative altitude would be hard as that error constantly changes with satellite geometry - which is constantly changing.

For me, Course Lock error is the biggest problem I noticed but it does seem to be less than the previous twice declination error. When I switch to CL initially it heads out directly on the wrong track but it does not yaw. Very soon I can reset CL in flight and it's right on.

It would be great if DJI would at least give us a vague idea of what they are changing. If we knew we could help more. I think this is a complicated enough control system that any effect we see could come from various causes.

 

[RemE]

Last comment on the barometer, I'm pretty sure that it's inside the Naza case, not on the mainboard as Naza V2's of which this is a variant, don't use mainboards in other models like Flamewheels.

I agree with Ian, I bet that the initial power up and time to settle is not long enough when the altitude is first zero'd. I have found that powering down the model and powering back up when I do have a larger error seems to give me a better, closer to zero number.

As others tested, altitude is not derived from GPS, it's a relative number based on your starting point each flight (or set home location). That does remind me, I have for sure tested and verified that setting a home location while airborne, sets the home altitude as well. If you set home at 30ft high, the model will go 60ft above that when you execute a RTH.

Next time out I will manually set a home location when I have an altitude bias and see if it resets the number to zero.

 

[AnselA]

That does remind me, I have for sure tested and verified that setting a home location while airborne, sets the home altitude as well. If you set home at 30ft high, the model will go 60ft above that when you execute a RTH.

In 3.05 or some other versions as well?

 

[RemE]

That does remind me, I have for sure tested and verified that setting a home location while airborne, sets the home altitude as well. If you set home at 30ft high, the model will go 60ft above that when you execute a RTH.

In 3.05 or some other versions as well?

All versions!

 

[AnselA]

That does remind me, I have for sure tested and verified that setting a home location while airborne, sets the home altitude as well. If you set home at 30ft high, the model will go 60ft above that when you execute a RTH.

In 3.05 or some other versions as well?

All versions!

P2V, firmware 2.0, resetting home location does not reset altitude.

 

[Xrover]

DJI has taken on our observations from the first beta and are preparing a new version for us to test that they believe will give us better results. Stay tuned...

Great news Ian. Thx for the update.

 

[landonkk]

Not sure if the beta is going to open up to anymore users, but in case it does: I have three P2's I can test in a -12 declination and was the fifth person to respond I had the problem in this thread. Thats a pretty good resume if you ask me...

 

[RemE]

P2V, firmware 2.0, resetting home location does not reset altitude.

Hmmm, well since I don't have 3.05 and have only tested previous versions I have to decide if I believe you, or my lying eyes :shock:

Several of us on RCG tested setting home locations (in Naza Mode, toggling S2 between CL and HL until the green LEDs flicker) at different altitudes and locations. Activating RTH caused the model to go 20m above that spot and begin to land in that new location column. If the location is "virtual" up high, the model will descend in that column, thru that point, until it touches ground or is cancelled out of RTH.

 

[Navman]

Oh, think I'm getting it, P2 uses ground touch rather than barometric to auto land?

Wondered how it knew where ground was.

 

[N017RW]

Oh, think I'm getting it, P2 uses ground touch rather than barometric to auto land?

Wondered how it knew where ground was.

The [auto] descent continues until there is no longer an increase in barometric pressure.
If this this equates to 'touching the ground' by your definition then OK but there is no tactile feedback.

 

[RemE]

Oh, think I'm getting it, P2 uses ground touch rather than barometric to auto land?

Wondered how it knew where ground was.

It detects "landed" using the barometric sensor, when there is no longer a descent detected and the motors are at lower throttle setting, then the motors idle down.

Regarding startup altitude;
I just did a quick test on the bench and outside. Power up, altitude read 6ft, as it warmed up the altitude climbed to 20ft. I powered the model down, power back up, altitude 1ft. I took it outside, altitude showing 4ft. Toggled S2 to reset home, lights flickered but the altitude did not reset to zero. Power cycled the model, alt then zero'd again.

So warm up causes some altitude drift it seems. I do know that setting an airborne Home location will be used for RTH, it will climb 60ft above that set point, however, the telemetry display always displays the home altitude that was locked in at power up.

 

[nzvideoguys]

DJI has taken on our observations from the first beta and are preparing a new version for us to test that they believe will give us better results. Stay tuned...

My second days test results.

Christchurch, New Zealand.
Running 3.05 firmware

Perfect day. Zero wind and sunny.
Powered up, took off straight up using only throttle. No TBE, as before. All good.
Forward stick applied, P2 went straight ahead, negligible hook at the end.
Back and forth 3 or 4 times. By the last run there was an increase of hook, about 4 or 5mtrs, but direction was maintained very well.

Tried IOC. Forward stick applied and it immediately started drifting off about 30 degrees to the left.
Tried HL and got similar results. Wouldn't maintain a straight path.

In GPS mode, when I line up the P2 at a target, and flying FPV with goggles, there seems to be a slight sideways yaw during flight.
During RTH there is a similar slight yaw (to the left) as it comes home.
Accuracy of GPS RTH landings varies a bit. Sometimes 2mtrs. Sometimes as much as 7-10mtrs away.

Cheers
Paul

 

[John Shaw]

DJI has taken on our observations from the first beta and are preparing a new version for us to test that they believe will give us better results. Stay tuned...

My second days test results.

Christchurch, New Zealand.
Running 3.05 firmware

Perfect day. Zero wind and sunny.
Powered up, took off straight up using only throttle. No TBE, as before. All good.
Forward stick applied, P2 went straight ahead, negligible hook at the end.
Back and forth 3 or 4 times. By the last run there was an increase of hook, about 4 or 5mtrs, but direction was maintained very well.

Tried IOC. Forward stick applied and it immediately started drifting off about 30 degrees to the left.
Tried HL and got similar results. Wouldn't maintain a straight path.

In GPS mode, when I line up the P2 at a target, and flying FPV with goggles, there seems to be a slight sideways yaw during flight.
During RTH there is a similar slight yaw (to the left) as it comes home.
Accuracy of GPS RTH landings varies a bit. Sometimes 2mtrs. Sometimes as much as 7-10mtrs away.

Cheers
Paul

Just checking...
The term "slight sideways yaw" isn't clear to me. Are you referring to a sideways motion (which for a quad is due to roll) or a true yaw motion which is a rotation of the vehicle about the vertical axis so it's looking in a different direction? Or a combination of the two?

 

[Xrover]

I also tried a couple RTH with 3.05 Both times it landed over 15feet from Home Point. I had the same experiences with 3.04 but It landed within 8 feet of Home Point. I don't know how that feature plays into Firmware programming but it would be great to have a RTH that is more accurate.

So far, I do feel the Beta 3.05 is a vast improvement in the right direction.

Thanks Ian
Thanks to all the Beta Testers
And thanks to DJI for working on the issues.

 

[ianwood]

I did two flights this morning after the Presidential TFR ended in my area. Neither were controlled tests and instead were more of my downtown LA filming. I used course lock on both flights. It wasn't 100% perpendicular to my target each time which could be caused by the firmware but also could be caused by how I lined it up and/or a slight magnetic deviation given I was taking off a few inches above the pavement.

Point being the course lock was still much better and totally usable by comparison to before. I am pleased as it has opened up a world of new tracking shots.

Still some un-commanded yaw which is annoying.

I may take my Phantom to Chicago this week to see how the beta does in low declination areas. It's a PITA as I still don't have a case.

 

[d4ddyo]

progress i guess but seems so far!

 

[nzvideoguys]

Just checking...
The term "slight sideways yaw" isn't clear to me. Are you referring to a sideways motion (which for a quad is due to roll) or a true yaw motion which is a rotation of the vehicle about the vertical axis so it's looking in a different direction? Or a combination of the two?[/quote]

To clarify, the P2 is yawed slightly to the left. It is moving in a crablike way, where the direct heading is not centered, but off to one side by about 10 - 20degrees. It's as if you were flying with IOC turned on, the craft knows where it is heading, but it is rotated slightly.
There is no roll. Just yaw.

 

[tom3holer]

The weather was much better today and I was able to get 2 flights in.

Phantom2 with H3-3D
FPV equipped
weight 1350
Dec -15deg

Flying with 3.05 this is the fifth flight with the beta s/w.

The TBE which before the upgrade was very noticeable was not observed on the first takeoff. In GPS mode I set out and observed, as before, a 20deg or so off heading track. I had at that time 12 sats. Flew the first pack down to 40% and noticed the tracking was perhaps somewhat better but still there. I did not notice any J-Hooking at all.
The alt as read on the IOSD mini was accurate as to Takeoff as 0 and upon landing the same.

Overall its a big improvement but still needs some work.

Tom