Accelerometer drifting out of calibration during flight

[JacobGH111]

I'm using a Multiwii Pro from Hobbyking - this one: http://www.hobbyking.com/hobbyking/store/__26588__MultiWii_PRO_Flight_Controller_w_MTK_GPS_Module.html - and I'm having a strange issue where the accelerometer will gradually drift out of calibration during flight. I'm flying a hexacopter.

So I set the drone on the ground, turn it on, and take off. The thing flies straight and level immediately, but after a little time, maybe after 10 seconds, I have to start holding the pitch/roll stick a little off center to keep the drone level, and the amount that I need to move the stick from center to keep it level will gradually increase over time. After a minute or two it's substantial.

I've got a bluetooth dongle attached, so if I land the drone and connect with my phone (I use the Multiwii EZ-GUI app), I can see that the accelerometers are way off. Sitting on flat ground I've seen it measure the pitch/yaw as high as 22 degrees off of level. But sometimes it's only a few degrees off of level. No matter what, I can run the accelerometer calibration from my phone, take off again, and enjoy the same few seconds of level flight before I have to start manually compensating.

What could be causing this and how do I fix it?
-Jacob

[waltr]

What is the starting temperature and what is the outdoor air temperature?
This could be caused by thermal drift in the Acc chip.

If the persists after the FC has come to a steady temperature then it could be the Acc chip is bad and the FC board needs to be replaced.

Are you flying in Angle mode? or have some other flight mode enabled?
How does it fly in Acro made? If it drifts in this made then it will be the Gryo that is drifting which can also be temperature sensitive.
Or the Gryo could be bad.

[brewski]

As I have said many times it is disgusting & dishonest that HK continue to sell these flight controllers with last generation technology.
If that is not bad enough the quality control of the PCBs is well below par. Search this forum as well as RCGroups using this board description to see the issues that builders have with this board/GPS. Several guys at my local model club bought these & after months trying to get them either just to work (but not reliably) or worse total loss of control returned them for refund.
For a few $ more you can buy A Crius AIOP V2 or 2.1 using latest sensor technology & with Ublox 6M GPS , have consistent flight experience. This board is also the benchmark for running APM branch for non APM spec FCs Megapirates due to it's reliability & ease of configuration. Unfortunately MW cannot yet log to the 8Mb of flash RAM installed.

[FengShuiDrone]

.

[kjriisne]

Just want to say that I have the EXACT same problem as Jacob here, I get it flying level and fine, and after 20-30 seconds I can't get the thing to stay level cause my stick is maxed out and I have to go back to Accro mode to recover it.

And to you FengShui, we have all seen your video and know that the board has the possibility to fly fine. This could be related to production batches, of the accelerometer maybe? I don't know but maybe there was a bad batch of accelerometers installed on some of the HK boards, or they might have started installing cheaper parts. All I know is that this problem has nothing to do with if the thing was level when you calibrated the ACC, if it was a degree or two off when you calibrated it, the driftangle should not increase during flight and at least not to the proportions that you cant even fly that way. I am going to do a few more tests on mine and then I'm scrapping the board and going for something better. I don't have the time, will or knowledge to troubleshoot this thing for hours on end.

[FengShuiDrone]

.

[JacobGH111]

Hey guys,
For some reason multiwii.com stopped emailing me about thread updates, and I put the hobby on hold for the winter so I haven't checked back so I just read this conversation again now. Anyways, I can comfirm waltr's initial suggestion that the problem is caused by temperature. There were a few day this winter where I got to fly inside a space heated at room temperature and during those flights I had no issue with accelerometer drift. The drone actually flew really nicely and stayed level. As soon as I took it outside to the cold again (around freezing temperature) the issue came back. So that rules out general calibration issues related to not having a flat surface and also issues with the transmitter. It's definitely the temperature.

I didn't realize that that the Multiwii uses such crappy hardware. I guess you get what you pay for. I think I'm going to get either Crius AIOP V2 or a Whitespy, as suggested above. Thanks, guys!

[waltr]

Thanks for the update. It sure sounded like a thermal issue.

I remember looking at the 6050 data sheet for temperature drift specs and there are none. So it doesn't surprise me that these do drift.

I've been flying during most of this past winter but mostly fly Acro mode where the accelerometer is not used.
I did go into Angle mode a few times to tune LEVEL PIDs and did need to do ACC TRIM stick commands and then Trim them again when testing in my basement. Didn't think much about his until seeing your post update.

My suggestion when flying in colder weather is to allow some time for the FC board (acc chip) to reach ambient temperature then land and do the ACC TRIM commands so it flies level.
The trim is saved to EEPROM so when you change battery the new trim values will be used. You will then need to re-trim if flying indoors or in warmer weather.

[JacobGH111]

Just wanted to give everyone an update. I purchased the Hobbyking CRIUS AIOP knockoff board (http://www.hobbyking.com/hobbyking/stor ... _V2_0.html) and it works much better. I can fly in cold weather without it drifting way out of calibration. So, confirmed. The Multiwii Pro board from Hobbyking is made with shoddy hardware and doesn't work well in cold weather. Thanks for al the help, guys. My hex flies beautifully now!

[waltr]

Ok. thanks for the update.
Am really wondering if China is counterfeiting the Gyro/acc chips or just used seconds.

[billyboy]

I also have the same problem. Being a newby to the MultiWii this has had me going round in circles for many days trying to workout why the Quad sets off OK but over 1 minute decides that the horizon is no longer where it should be! I can re-calibrate once on the ground and get another 30 seconds of good flight but after that is starts to go horribly wrong again. The ACC values seem to move from 0 on the ground to between 15 and 40 once in use and never go back to zero once back on the ground so the boards view of the world changes. I got this from Banggood some months ago but have now given up on the MultiWii in favour of the 3CCD board which seem to be much better and is configured with relative ease using with LibrePilot.

BTW: I don't think my issues are related to temperature as it seems to go wrong in the same way when used indoors.

[Kbev5709]

I bought one of those red boards from HK. Mine works OK now but at first was kind of drifty. They are obsolete using an obsolete baro sensor. They have the same MPU6050 ACC/gyro combo as the Crius. The mag it uses is still used on later generation MW boards. The only thing really obsolete on that board is the BMP085 baro that has been replaced on better boards with the MS561101BA. Maybe I got lucky with my board because I've never had any major problems with it or maybe I got one with genuine parts not made in China (highly unlikely because I think they make all of these there now, don't they? Just depends on who manufactured the chip.) Mine even does Altitude hold OK (not as good as my other boards that use the MS561101BA) and can even do waypoint missions. It was my first board though and at first I had problems similar to those described in this thread. I really think the secret to getting one of these to work correctly is to have everything as balanced as possible. This board seems to be ultra sensitive to vibration. You need to get rid of as much vibration as possible. Balanced props and motor bells are a must for this board. Even after everything is smoothed down there is still enough commotion at the FC to make it necessary to enable the low pass filter in the config.h for either 98Hz or 42Hz. For mine the magic setting was 42Hz.

[gregd72002]

I've seen the same problem on mine MPU9150, not using stable mode much so not a big deal but could this be related to:
viewtopic.php?f=8&t=7323