Replies to This Discussion

DeveloperPermalink Reply by jasonshort on December 15, 2011 at 11:01am

Can you take a macro photo of your board so we can see if you accidentally shorted or knocked off a part?

Jason

• ▶ Reply

Permalink Reply by George Andriyanov on December 16, 2011 at 1:33am

Here they are: imu shield, ardupilot

 I didn't find any damages. 

Please tell, is such gyro behavior is normal? 

• ▶ Reply

Permalink Reply by George Andriyanov on December 17, 2011 at 9:30pm

Up..

There's also scale-like accelerometer date. Is it normal? Can someone try it on his/her IMU?

I'm very sorry that invensense gyros intolerant to shock and cold:( 

• ▶ Reply

Permalink Reply by George Andriyanov on December 20, 2011 at 5:57am

I see noone answers me:(

This is probably because of my bad english. If so, sorry, I'll try to explain it again.

 

The main flaw is that gyroscope data come to the system very slowly - like if there previously was an airball in the water and now this airball is in the honey or melted wax (see upper right chart on the attached image). Also, there's something wrong with accelerometer (or not?) - looks like it sends quantized data, jumping from one value to another (bottom chart).

 

I want someone to make the same test - quickly rotate IMU 90 deg. left or right to see if accel behaves same way, or smoother.

 

I'm not familiar with electronics, but what do you think about this hypothesis: something did broke ADC's clock (does it have one?) and it started to work very slow. It bufferizes gyro data (and therefore it comes stretched to main processor) but does not do it for accelerometer (and its data become quantized). What do you think?

Thre also might be flaw in  Atmega2560's clock, but I didn't see slowing of connection process or loading or something of this kind.

Strongly need help!! 

• ▶ Reply

Permalink Reply by Vernon Barry on December 20, 2011 at 6:48am

something did broke ADC's clock (does it have one?)

No, it wouldn't work at all. In order to quantize the data, it either clocks properly or doesn't work at all.

 

Slow reporting back via telemetry or USB is just a function of the serial bus itself. Have you lowered the baud rate or some other parameter?

That said, you reported that the data coming back continued to rotate after you stopped moving the copter so it definitely sounds like you may have damaged the board

It bufferizes gyro data (and therefore it comes stretched to main processor)

No, that sounds like such erroneous data with outliers being filtered most likely caused by a physically broken sensor or broken solder joint. The only way to fix the solder joints is a reflow oven, and it's also possible you caused enough shock to actually break the tiny parts inside the chip. Showing us a screenshot of the RAW sensor data is much more likely to help us verify the sensor is bad, but ultimately-even if it is bad, you cannot easily fix it without specialized tools and equipement.

I hate to say it but a new APM 2.0 might be in your future.

• ▶ Reply

Permalink Reply by George Andriyanov on December 20, 2011 at 10:19am

Thanks! Now I can either replace the sensor board alone or, knowing that even if the frame survives, electronics may not, buy 10 KK multicopter boards at a price of one Arducopter and have no problems with crashes:)  I will also try to find local fixer who has an oven.

But sorry, what do you mean with raw sensor data? I did the IMU test command in CLI and there was the same thing: the data in all 3 gyro columns was slowly and linearly going to the new expected value. 

 

• ▶ Reply

Permalink Reply by George Andriyanov on December 27, 2011 at 1:13am

Looks like reflowing helped! 

Than it should be amortized enough to prevent such kind of damage in the future.

I'm working on it. 

• ▶ Reply

Permalink Reply by Bill Jowitt on December 20, 2011 at 6:58am

I am having similar problems. Twice now just after take-off my quad has flipped into the deck, bending arms, motor shafts and mounts and breaking props. It then lies there shrieking until I disconnect the battery. It is getting pretty heartbreaking as earlier today I was making great progress.  I attach the log of the last incident. I would be most grateful if somebody could look at them and tell me what is going wrong.

Attachments:

• 2011-12-20 01-50 13.log.gpx, 4 KB
• 2011-12-20 01-50 13.kmz, 6 KB
• 2011-12-20 01-50 13.log, 31 KB

• ▶ Reply

Permalink Reply by Bill Jowitt on December 20, 2011 at 9:30am

Just read another thread which recommended recalibrating the ESCs after a crash. Did that, and it appears to be behaving itself again. So many things to worry about! Anyway, making progress and getting quite good loiter performance - will do some more tuning next time there is daylight and ok weather.

• ▶ Reply

Permalink Reply by George Andriyanov on December 20, 2011 at 10:34am

"So many things to worry about" - the best phrase to describe the project:)

Thanks to Arducopter - I've learned Arduino, C++, I2C, advanced Tx setups, C# basics, got headaches and depression. This is truly new experience. 

• ▶ Reply