Naze32 - drift after aggressive turning or flips

[presidenten]

I have a full naze32 rev5 board running firmware 2.3.

Like the title says, if I fly aggressively in angle or horizon mode and do a hard turn, it seems like the accelerometer gets offset and the hexacopter starts to drift. After som flips I did today, the drift was quite alot. When this happens, I have to set it down on the ground and stand still for a few seconds. When I fly off again the accelerometers are back to normal. ...until I make another hard turn or flip that is.

I noticed that if I make a flip and it starts to drift, sometimes (but not always) I can get rid of the drift by making a flip in the opposite direction.

Have anyone else encountered this, and is there a way to get around it or make sure it wont happen again. As soon as my fatshark goggles arrive I want to fly fpv, but I see how having to touch down to reset the accelerometer will be a problem.

Is there maybe some way to bind a switch to reset the accelerometer values to what it was when I calibrated it or something?

[Crashpilot1000]

Too much vibration hitting the acc is one reason for that. The mwii IMU part is also part of the problem but holding the copter still in the air for a short time should correct the internal horizon. If only landing is able to correct that your setup has too much vibrations coming through to the acc. There are other IMU algos available but at the end of the day they are not generally safer IMHO (I know BRM will eat me up now).
If you eliminated the vibrations on your acc and the problem persists you will have to mess with the LPF and CMPF stuff.
Cheers Rob

[brm]

Too much vibration hitting the acc is one reason for that. The mwii IMU part is also part of the problem but holding the copter still in the air for a short time should correct the internal horizon. If only landing is able to correct that your setup has too much vibrations coming through to the acc. There are other IMU algos available but at the end of the day they are not generally safer IMHO (I know BRM will eat me up now).
If you eliminated the vibrations on your acc and the problem persists you will have to mess with the LPF and CMPF stuff.
Cheers Rob

no - just no time to do so.

[Plüschi]

mess with the LPF and CMPF stuff

IMHO the LPF may be responsible for a Gravity vector offset after flips.

The LPF does smooth the acc data BEFORE they are rotated, which will give them a good delay of some cycles. Then the CMPF filter will use this delayed acc data to correct a properly rotated, not delayed Gravity vector. I think this is fundamentally flawed. Delayed acc != actual acc in a flipping copter. Check out what happens with a really beefed up (means long) LPF.

Lets throw out the LPF, or better make sure the total delays for gyro and acc data have the same time constant (chip internal LPF included).

@presidenten sure you use the right gyro scaling?

[brm]

mess with the LPF and CMPF stuff

IMHO the LPF may be responsible for a Gravity vector offset after flips.

The LPF does smooth the acc data BEFORE they are rotated, which will give them a good delay of some cycles. Then the CMPF filter will use this delayed acc data to correct a properly rotated, not delayed Gravity vector. I think this is fundamentally flawed. Delayed acc != actual acc in a flipping copter. Check out what happens with a really beefed up (means long) LPF.

Lets throw out the LPF, or better make sure the total delays for gyro and acc data have the same time constant (chip internal LPF included).

@presidenten sure you use the right gyro scaling?

FULL-ACK

[presidenten]

I have not changed any parameters regarding the accelerometer. Just calibrated. I am using stock baseflight with firmware version 2.3.

I have ordered some vibration dampeners to mount the board on. Ill see if that makes a difference. It will take a week or so for them to arrive. Then I will report back here.

mess with the LPF and CMPF stuff

IMHO the LPF may be responsible for a Gravity vector offset after flips.

The LPF does smooth the acc data BEFORE they are rotated, which will give them a good delay of some cycles. Then the CMPF filter will use this delayed acc data to correct a properly rotated, not delayed Gravity vector. I think this is fundamentally flawed. Delayed acc != actual acc in a flipping copter. Check out what happens with a really beefed up (means long) LPF.

Lets throw out the LPF, or better make sure the total delays for gyro and acc data have the same time constant (chip internal LPF included).

@presidenten sure you use the right gyro scaling?

[timecop]

need actual version reported by "version" from CLI

[presidenten]

need actual version reported by "version" from CLI

This is from the log topside:
22:31:57 -- Running - OS: MacOS, Chrome: 36.0.1985.143, Configurator: 0.49
22:31:59 -- Serial port successfully opened with ID: 1
22:31:59 -- Unique device ID received
22:31:59 -- Firmware Version: 2.3

And here is the output if I type "version" in the cli tab:
Afro32 CLI version 2.2 Jun 1 2014 / 22:14:50

[timecop]

So, upgrade firmware.

[presidenten]

So, upgrade firmware.

Oh. Thats embarrassing... I looked at the date 1 June 2014 and thought I had the latest firmware.

Well, thanks for the help anyways everyone.

[presidenten]

Ok ... I flashed firmware. Now it says:

Afro32 CLI version 2.2 Aug 3 2014 / 21:51:54

Is this the latest version? Should this work? Im going to try it out tonight or tomorrow, as soon as I get some time on my hands.

[timecop]

Yes, latest.

[presidenten]

Ok, I can now confirm that I still have the same problem, even with the new firmware.

The rubber vibration dampers should show up on friday or early next week. Ill report back to see if they fix the issue.

[Shifty-PK]

Any news/succes with the new bootloader?

[presidenten]

It was the vibrations that were causing it.

I got the vibration dampeners this week, so I padded the naze with them and went out end flew really agressivly. Like 4 batteries of only hard turns, flips and rolls. No more drift. Before I got drift after every single flip. So problem solved

So the conclusion is:

Make sure you have the latest firmware and put vibration dampeners under the board. I used these guys:
http://www.ebay.com.au/itm/100pcs-Black ... OU:US:1120
I cut away the radially smaller part to have 4 rings and put one between each distancer and the carbon fiber plate. Then, because I really have no clue how much is needed the get rid of the vibrations, I also put a dampener under each corner of the naze. Now it works great, so I'll just leave it like that.

Thanks for the help guys!

[marcelokron]

It was the vibrations that were causing it.

I got the vibration dampeners this week, so I padded the naze with them and went out end flew really agressivly. Like 4 batteries of only hard turns, flips and rolls. No more drift. Before I got drift after every single flip. So problem solved

So the conclusion is:

Make sure you have the latest firmware and put vibration dampeners under the board. I used these guys:
http://www.ebay.com.au/itm/100pcs-Black ... OU:US:1120
I cut away the radially smaller part to have 4 rings and put one between each distancer and the carbon fiber plate. Then, because I really have no clue how much is needed the get rid of the vibrations, I also put a dampener under each corner of the naze. Now it works great, so I'll just leave it like that.

Thanks for the help guys!

Hey mate, do you have some photos just to check how you made it?
Thank you!!

[presidenten]

Hey mate, do you have some photos just to check how you made it?
Thank you!!

Sure!


Full rubber niplet between board and spacer. Small rubber ring (half a niplet) on bottom between spacer and body.

[nebbian]

I can't see how those rubber nipples would go any way towards reducing vibration, but if it works it works.

I noticed the drift in acro mode after a quick flip, or doing the following:
Fast forward flight, then give a hard yaw through 180 degrees so that it's flying backwards. The craft would always then have a drift towards the forward direction, which I needed to trim out with a couple of clicks on the Tx.

I have been flying this frame for over 6 months with a KK2 (2.0 and 2.1) and never noticed this issue with the KK boards.

[presidenten]

I can't see how those rubber nipples would go any way towards reducing vibration, but if it works it works.

I noticed the drift in acro mode after a quick flip, or doing the following:
Fast forward flight, then give a hard yaw through 180 degrees so that it's flying backwards. The craft would always then have a drift towards the forward direction, which I needed to trim out with a couple of clicks on the Tx.

I have been flying this frame for over 6 months with a KK2 (2.0 and 2.1) and never noticed this issue with the KK boards.

Maybe its enough for the naze to think its vibration dampened hahaha. I am almost always in horizon mode these days doing flips and what not. I have run through like 16 to 20 packs since I modded it with around 10 to 11 min per pack (turnigy nano-tech 3000mah 25/50c thus handling 75A continous for those interested to know).
No drift any more. So it works for me. And the rubber niplets were really cheap. So its all good