Serious hardware issues with GY-521 MPU6050 modules

[SergioMForster]

I have an automotive development using MPU-6050 mounted on GY521 modules.

I am driving the MPU-6050 using a PIC24EP512GU806 Pic micro at 3.3v.

For this reason I am bridging in the GY521 modules, the 5v to 3.3v regulator, since I have my own 3.3v regulator that powers my processor & MPU6050.

They were very useful for this project the Jeff Rowberg libraries and examples, so I want to thank Jeff for this great work.

I am using the MPU6050 DMP, with a sampling rate of 100 ms, which interrupts my processor every 100 ms to update Accel, Gyro, Quaternions, Euler angles, etc.

Now I tell you my problem, I have very serious hardware problems with the GY-521 modules.

I bought 1500 units of gy-521 modules from two different Chinese suppliers, The price of gy-521 was very good, less than one dollar each per 1000 units.

The appearance of these modules are identical to all I've seen in forum photos, each one in its bag, with the two connectors one straight and the other at 90 degrees.

They all seem to be from the same manufacturer, then sold by different vendors, they are even identical to the ones I bought locally during development.

I have built a test tool, which connects to the gy-521 board with a clamp, does a calibration, and test with controlled movements in all axes, accel, gyros and DMP output from MPU6050.

When I receive the modules, the first thing I do, is to remove the 5 to 3.3v LDO regulator, and make a bridge between pins 1 and 5 since my power supply is 3.3v.

The second thing I do is test them with my tool, during this test I am getting a rejection of 10% of the modules (they are new modules taken out from their bag).

The main reasons for rejection are: the DMP does not start, or any of the axes is fixed at a 2G value, no matter how the module moves.

Then with the 90% approved modules, I assemble my equipment, test them and encapsulate them in epoxy resin.

This equipment is then installed and calibrated in trucks, where they work 24 hours a day, measuring certain acceleration and incline parameters continuously, in an unattended environment.


It is good to clarify that I have experience in both design of automotive equipment and unattended equipment, with adequate power supplies and intelligence and watchdog in the processor to recover from any software failure, with the capacity of a total to partial power off, and automatic restart.

The problem I am having is that after approximately 2 months of continuous good operation, the MPU6050 suddenly dies and stops working.

I cannot access to measure electrically since they are encapsulated in epoxy, but I see with the monitor leds, my processor trying to power off and restart continuously the MPU6050, I see also with a led that shows the status of the interrupt pin of the MPU 6050 that the DMP is not starting.

You know someone who has had similar problems, it will be counterfeit modules or mounted with scrap Invensense chips?

Do you know any reliable vendor for these modules?

Thanks in advance for your help.

[Ttelmah]

Buying modules that are as cheap as the bare chip is usually a fairly
sure sign that they are using chips that may well be copies or have
failed testing. Probably explains your silly initial failure rate...

Now you say 'epoxy resin'. Is this a proper potting compound or just
a normal resin?. Proper potting compounds have altered chemistry to
ensure the hardener does not use a chemistry that will adversely affect
the electronics. Conventional epoxies use chemistries that can cause
issues. Same applies with silicone materials. Potting with a standard
silicone sealant is a sure way of having parts fail after a short time. The
acids used as hardeners in these will cause component damage.
Also before potting are you cleaning and drying the board?. Otherwise
fluxes may well cause issues at the component legs. This has also
become more of a problem in recent years. Historically PCB's were
typically cleaned using materials like carbon tetrachloride. However
modern practices have encouraged use of water or similar cleaning.
These can leave boards 'wet' and if potting these need to be dried
before encapsulating.

Then are you cleaning the assembled board before potting?.
Again modern practice may involve ultrasonics. These should not be used
on MEMS devices.

MEMS devices are also delicate mechanically. Use of really rigid
encapsulants can cause issues with these with failues due to thermal
expansion/contraction.

[PCM programmer]

Counterfeit MPU6050 chips from China:
https://www.eevblog.com/forum/projects/counterfeit-mpu6050-chips-from-china/
It shows photos of the modules, and test results on the WHOAMI register.

[Ttelmah]

Though as you see, he had one that gave the correct 'WhoAmI' result.

There was a huge operation a few years ago, with people raiding the line
rejects, encapsulating these and selling as real chip...

If we want a reliable product you have to buy from a supplier who warrants
their source.

[temtronic]

wow, what a PAIN. As pointed out it PAYS to buy from 'big name' distributors like Digikey...
3 decades ago, place where I bought communications boards from used a regular household dishwasher to clean then dry every module. He found out if they survived that 'test', they'd work in the Real World.
I just can't believe anyone can make money selling fake chips..

[Ttelmah]

Don't get me wrong. Though 'fake chip' was my first thought, it is still
possible that the later failures are because of the encapsulation used.

Really rigid epoxies can very easily lead to mechanical failures at
the legs. Boards do need to be carefully prepared before potting. Using
the wrong potting material can lead to lots of problems. Also the MEMS
device is sensitive to high G forces. If solidly potted rather than resiliently
potted this could also cause problems. A rigid epoxy could actually result
in higher G's than for an unpotted device...
Choosing the right compound is quite important.

The modern 'water cleaning' of PCB's is also a thing I have seen a lot
of problems with... :(

[SergioMForster]

Thank you very much for your comments.

Ttelmah, I am using a flexible epoxy resin, specially formulated to encapsulate electronic circuits.

I am not doing any previous cleaning before encapsulating them, what method would you recommend?

I have already successfully used another type of modules mounted in China at a lower price than the individual components, and I assumed that they could reach that price because they manufactured millions of modules.

But in the case of the GY-521 it seems to be of very poor quality, it struck me that I did not find comments on forums about these modules.

Thanks PCM Programmer for the Counterfeit MPU6050 chips from China link.

I am evaluating redoing my PCB to directly mount the original MPU6050.

[Ttelmah]

On PCB's, the problem is 'ionic contamination'. Comes from 3 sources:
The etching chemicals. This should be cleaned by the PCB house, but
problem is that on new boards they tend to use water based cleaning.
Then water. Conventional fibreglass PCB's contain significant moisture.
This unfortunately is made much worse by the water based cleaning.
Then flux.
This contamination can encourage dendritic growth and corrossion.
We have found we have to dry boards before use for reliability. A warm
dehumidified chamber. Then alcohol clean before soldering and again
afterwards. Isopropanol.