I will resume my problem and the solution:

My board is 2.5.2 (not original) and with firmware 2.9 works fine (or it appears to be but it is in danger).

After trying to update firmware from 2.9 to 3.0 or 3.0.1 the problem is the bad gyro health caused because of 5V on 3.3 regulator.

I am engineer so I unsold the 3.3 regulator and checked it. It was ok, but the 3.3v rail is at 5v, even without regulator soldered

Then I tried to replace the regulator for another one, same problem...

I tried to downgrade to firmware 2.9, the board works OK, so I tried to fly and it flies ok (or it apears to be OK, but all the sensors are overvoltage).

After 30-40 mins fly time the quad started to do some suddenly movements (I dont knew whats happening then, now I know it was the sensors at 5V instead of 3.3...)

But now I understand everything:

There is a problem in the board layout, there are a level shifter from 5V of the ATMEGA to 3.3v of the flash memory.  The 3.3 and 5v references are inverted. Thats all the problem, nothing from regulators.

SOLUTION: Very hard and complex soldering skills. You must cut wires from the dataflash level shifter and interchange 3.3 and 5v. Cut the blue lines and replace with cable the red lines.

OR maybe you are not interested in use the dataflash but you want to repair your board: It is only necessary to cut the 2 wires in blue and the board will work perfect but the dataflash will appear as disconected.

It is incredible that all the sensors worked at 5v for so long time ago and didn't crashed :o. Also, the dataflash is working at 5V without problems... Incredible!!!

I didnt try to fly yet but the board loads the 3.0.2 firmware properly and passes all the tests.



Mine is working with this change but the problem was the data lines and not the power lines of the level translator. You can check it in the datasheet of the chip.

So if you want to solve this the first thing you have to checkout it is if your pcb has the bad level routing on the dataflash. Definitive solution is very complex to routing but like me you can repair it. 

Thanks everyone. 

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –


  • GRACIAS!!!! mi APM chino Volvió a la vida cortando las lineas indicadas!!!! tuve que reemplazar mi regulador 3v3 puesto que ya se había dañado.


  • I Think that this problem (and others) have been solved by chinese cloners. Check this items!



    • +1

  • Developer

    Hi All, please see the conversation at http://www.diydrones.com/forum/topics/ac3-1-rc5-spi-speed-problem

    there is a new solution there for the 3.3v issue. It can be replicated, and the level shifter is vindicated :)

  • The data lines PG0, PA6, PJ2, PJ1 from the AT2560 link to the A port of the TXB0104 level converter. This is the LOW voltage side. THe output lines from this chip - the HIGH voltage side B port - go to the AT45DB161 dataflash chip which is powered from the 3.3V rail.

    Check the schematic. No need for an extensive investigation and not what I would have described as a technicality.

    This is not a 'genuine' versus 'clone' issue. This is a design error.

    Yes, the AT45DB161 is ~ 6V tolerant, so maybe one can get away with it but it is the root of the problem for al that.

    • Developer

      Simon, we all know this, it is nothing new. It has been the same since the APM 2.5 came out in the first place.

      It is a technicality, as it DOES NOT cause any flight issues.

      there is NO unacceptable risk of flying with this. it DOES not cause any other issues.

      we have all checked the schematic. what is your point? we know that the level shifter is around the wrong way. and it has been PROVEN not to be the cause of the 5v on the 3.3v rail.

      If this was the cause, it would affect the 3DR boards, and the clones the same, it does not.

      Again, to everyone.


      • You are focused on 5V on the rail but what about short circuit on 3v3 comming from a failure of the buffer used the wrong way ! (this was the case on my board)

        In this case the result is the same, no more infos from imu, baro...

        So we need the proof not only of the 5V on the rail but also any other failure of the board linked to this buffer. Please write a clear note to the users and not only posts.

        Concerning the failure of the boards, there are a lot of boards failed every where and it is very difficult to demonstrate that they all failed the same.

        In France there have been failures with crashed quadris.

        The users do not have the capacity to investigate the exact reason of the crash. Some have even lost there quadris.

        If you can provide an analysis showing that this component can permanently sustain reverse voltage, please do so.

        Concerning the behaviour of the 3DR and clones, it could be explained by the difference in the failure mode of the regulator linked to a stress generated by the buffer. If the regulator can sustain the stress then it does not generate 5V if it can't then the story could be different.

        In my club the 3DR board has strangely burned when we have intensively used the data logging.

        Furthermore, I can't understand that 3DR has not changed the schematics and the layout of the board, knowing the issue for 6months !


  • Hello,

    I have the same probleme and same conclusion, I have removed the buffer because you can't change only the +3v3 and +5V because this component must have B lower than A it can't be reversed.

    here is an extract of datasheet : 1.2 V to 3.6 V on A Port and 1.65 V to 5.5 V on • Latch-Up Performance Exceeds 100 mA Per
    B Port (VCCA ≤ VCCB)

    See my post on : http://www.modelisme.com/forum/aero-vol-en-immersion/189120-bug-des... (in French !)

    You must intervert lines from ATMEGA and flash !

    I think we should gather all the cases to send a common request to 3DR because it is really a danger for all and the way 3DR manage the thing is not acceptable at all. They know since november this problem and have done nothing !


    • 100% of all boards, both 3dr and clones, have a design fault that causes this chip to operate out of spec. With that said, there is no concrete evidence that links the 3v3 reg failures to this problem. Nobody really knows the impact of this issue. Philip will be testing boards extensively over the next few days to investigate.

      Although it is true that somebody messed up the traces which was an honest mistake, I cannot blame them. In fact I have nothing but appreciation for the hard work and effort that they put into this project. Without them and all other donors of their expertise, none of us would be here.

      I am confident APM will emerge a better product after all of this.
      • Developer
        Great reply :)
        It is under investigation, and we will post a full report
This reply was deleted.


DIY Drones via Twitter
Using ArduRover with an RTK GPS https://ift.tt/2N9I3RO
18 hours ago
DIY Drones via Twitter
DIY Robocars via Twitter
DIY Robocars via Twitter
DIY Robocars via Twitter
DIY Drones via Twitter
DIY Robocars via Twitter
RT @Heavy02011: @diyrobocars : A Home-brew computer club* for Connected Autonomous Driving on Jan 23rd, 2021 https://www.meetup.com/Connected-Autonomous-Driving/events/275728684/ #Meetu…
DIY Robocars via Twitter
David Hori liked Isabella Domi's profile
Jan 12
DIY Robocars via Twitter
RT @Heavy02011: ⁦@diyrobocars⁩ Autonomous Driving Assembly at #rC3. join us at https://rc3.world/rc3/assembly/diyrobocars-f1tenth/ ⁦@f1tenth⁩ ⁦@DAVGtech⁩ ⁦@DWalmroth⁩…
Jan 11
DIY Robocars via Twitter
RT @chr1sa: New car designs coming for our next @DIYRobocars @donkey_car virtual race on the 23rd. Choose any one you want at race time Le…
Jan 11
DIY Robocars via Twitter
RT @RoboticMasters: Thanks to @EllerbachMaxime and the Sydney Uni Capstone Students the @donkey_car @diyrobocars simulator is getting a ma…
Jan 11
DIY Robocars via Twitter
Jan 6
DIY Robocars via Twitter
Dec 28, 2020
DIY Robocars via Twitter
An interesting line-following simulator to use with with your robocars: https://github.com/ron-grant/LFS
Dec 23, 2020
DIY Robocars via Twitter
Dec 23, 2020