RE: DRV8825: DRV8825 driver failure

Hey Sarath, 

Overall the design looks good.  A few small design things:

• Add a 1-MΩ resistor between CVP and VM.  Is there a reason this isn't in there?  
• 10K is a large series resistors for the inputs (R6/R7/R8/R9/R10) - consider changing those to 1K resistors.  
  • Note - the way you have it should be fine.  The I_IH (input high current) level is 100uA, and your setup is 3.3V / 10K = 330uA.  This should be plenty above the threshold, but since you're having issues suggest changing to 1K.  
• R12 and R13 as pull up resistors along with being a MCU controlled pin is unusual.  Recommend either controlling those pins from the MCU or the pull-up resistor and not both.   
• On the board layout, are the filtering capacitors located close to the driver as recommended in 11.2 Layout Example? 

A few questions on the failures:

• Do the failures happen more often for the boards powered by a power supply (Main mode) or by a battery?  
• Is the motor driver the only thing that is failing? 
• Do you have any temperature data of the boards that have been failing?  Are they indoors in the air conditioning?  
• What environment are the boards in - high humidity, very dusty, etc?  I'm trying to see if we can rule out any external factors that might cause the boards to fail.  

Regards,

Jacob

Dear Jacob,

As you given suggestion will be update in my design. 

I will add 1-MΩ resistor  between CVP and VM. Filtering capacitors placed as mentioned in layout.

• The failures happen both power supply (Main mode) and a battery mode.
• Yes, Motor driver only failure in that board.
• We are use the system indoor air conditioning. The system is related to health care product, Which is mainly used in ICU (Intensive Care Unit).
• The System can operate both room temperature and air conditioner environment.

Regards,

Sarath

Hey Sarath,

Thanks for answering my questions.  Given the operating conditions I doubt it's a thermal issue or humidity issue.  

Let me ask my team whether they think any of the notes I commented previously would case this type of failure.  Give me 24-48 hours for them to get back to me.  

In the meantime, could you give me specs on the motors these chips are driving?  Specifically the winding resistance and/or the inductance of the motor.  

Regards,

Jacob

Hey Sarath,

Can you give me a picture of the top of the chip?  I'd like to check the lot trace to make sure the chips aren't counterfeit.  

What is MCU voltage - 3.3V?  5V?  Typically overvoltage event causes damage to the outputs not the input, it is very odd that we are seeing failures on the logic-level input pins on the chip.  

Cheers,

Jacob T

Hey Jacob,
We would like to have your advice on the similar kind of problem we're facing with the same driver DRV8825.
We are using this driver for the stepper motor which has a rating of 1A/phase and 1/32 stepping mode and using 0.27ohm as Rsense and for limiting maximum current through the winding to approx 800mA, we designed Vref according to the calculation in the datasheet to be 1.1V and thus the voltage divider of 60kohm (misprint in the schematic, actually we're using 60.4kohm as R7) and 30kohm.
Right now we're using it in mixed decay configuration and for the bulk capacitor part, 100uF electrolytic is there on PCB.
The layout is also as very similar to the one given in the datasheet.
The problem we're facing is that the motor takes near about 500-600mA only and after some trials, the Vref pin somehow gets short and the voltage drops there to 0V or 0.01V even though the voltage divider is connected with it. We even took out those resistors (R7 & R9) and verified, that they're not short or open and working properly when not connected to the Vref pin.
A mistake was found in the previous designs in which we had mistakenly put a 1kohm resistor in place of a 1Mohm between VCP and VM.
Do you think that could've been a major reason for the demise of the pin?
Also, the Rsense resistors are of 1/8W capacity.
Does that affect the performance too?
We've rectified that and are afraid to continue as we don't know what is exactly damaging the driver so that its Vref doesn't become what it is designed to be. We've already damaged multiple drivers unknowingly, so waiting for your expertise on this.
Also, if you could put some light on some internal working of Vref and other parts.

Hey Raj,

I'm glad you were able to find the issue with the 1K/1M resistors on the previous board!

Here's a note about current regulation, specifically setting VREF with resistors is the Resistive Network Voltage Divider section:  https://www.ti.com/lit/an/sloa170b/sloa170b.pdf#page=3

The Rsense resistor of 1/8W capacity is a little on the small side.  If it's up to 1A motor, the max power through the resistors is I²R = (1*1)*0.25 = 0.25W = 1/4W.  So 1/8W won't be able to handle it.  Even with 600mA it's 0.09W which is close to 1/8W (0.125W).  I'm not how this would affect performance, but definitely would check if this resistor is getting too hot or not.  

Just to clarify:  When you say the motor is only using 500-600mA, is that the desired current amount?  Or would you rather run it at a higher current?  The Vref value you have of 1V will be a pretty low current limit, it might be good to increase the Rsense values as detailed in this post.  

Interestingly enough I did find another post discussing blown VREF pins.  Another post discusses some manufacturing issues with the DRV8825, maybe read that thread and closely look at your board and see if similar is happening?  https://e2e.ti.com/support/motor-drivers-group/motor-drivers/f/motor-drivers-forum/578047/drv8825-powerpad-via-treatment  This would explain the short.  

Regards,

Jacob

First of all, thank you Jacob for responding so promptly.
Can you put a little light on how that 1k instead of 1M resistor would be affecting the circuit behavior so we get a better understanding of the process?
And will change the resistor to higher wattage or lower value and set Vref accordingly to the calculation of current and voltage divider.
And till now, we've observed that Rsense resistors were not getting heated up but irrespective of that, should it affect the circuit so much that the Vref pin may get damaged?
And 500-600mA is not desired current, we wish to run it at 700mA as max. rating is 1A/phase of our winding.
Also, the forum you shared has shown problems regarding soldering, layout, and size of IC, we're totally fine with that as we're soldering the IC manually and checking every pin and corner before testing and there is no issue with that, I can take full guarantee of that part.

Hey Raj,

RAJ LADHA said:

Can you put a little light on how that 1k instead of 1M resistor would be affecting the circuit behavior so we get a better understanding of the process?

My understanding is that the 1M resistor on the charge pump discharges the charge pump during low-power sleep mode or when the device is powered down, and I know that not including it can cause the charge pump to stay charged and find a different discharge path through the device.  With a 1K instead of a 1M resistor, I would guess that the device would be nearly-instantly absorbing all the power from charge pump.  I would expect this to make the charge pump not really function.  

RAJ LADHA said:

And till now, we've observed that Rsense resistors were not getting heated up but irrespective of that, should it affect the circuit so much that the Vref pin may get damaged?

I don't think so, I wouldn't expect the Vref pin to be damaged from this.  Interesting that they weren't heating up, I guess because of just 600mA instead of 1A.  

RAJ LADHA said:

Also, the forum you shared has shown problems regarding soldering, layout, and size of IC, we're totally fine with that as we're soldering the IC manually and checking every pin and corner before testing and there is no issue with that, I can take full guarantee of that part.

Okay good!    

Could you post an oscilloscope picture of your board running the motor, preferably near max load?  I'd like to see a phase voltage and current along with the VM voltage.  I'd like to see if anything stands out to us as being unusual in normal operation.  

Also, can you give a picture of the top of some chips so I can check for counterfeit?  

Regards,

Jacob

Hello Jacob,
Thank you for explaining how the 1M resistor part works.

I'm attaching the pictures you demanded.

The driver IC we're using, ordered from mouser.

Setup: Microcontroller is STM32G070RBT6 and at the bottom left corner is the driver and whole assembly, the big capacitor of 470uF is connected at VM. 

And I've two quantities of this board, one has IC installed in which the Vref pin is damaged and another is working(but has a different setting now).
I've made little adjustments as I said earlier in one of the comments, I didn't have 0.1ohm SMD so, I put two 0.27ohm on each other and made them into the parallel combination and hence solving both things of lowering the resistance and the heating problem. So, my Rsense now is 0.135ohm and I've set voltage divider of 60.4k & 10k and thus Vref is 0.468V and I checked in DSO, it measures 0.454V so same. and according to the calculation, the current limit is 0.694A.

Ch.1(yellow): phase current(using the current sensor) and Ch.2(blue): VM voltage and in the image below you can see average values respectively.

DC Power supply: This current is not the actual current flowing through the winding, I learned recently that DRV8825 steps down the voltage and steps up the current. So, I have a new question, how much should we consider as losses while calculating the actual current flowing through the motor winding when this much current is flown through the supply? Generally, the output voltage across windings is near about 1.8V.

Yellow: Phase current, Blue: Phase voltage(between ground & any winding terminal) (attaching probe's both sides on a winding was not letting motor start)

zoomed IN

Now, I'm attaching waveforms of the Board in which the Vref pin is damaged. Rsense is 0.27ohm and voltage divider 60.4k & 30k and rest calculation accordingly.

Yellow: Phase current, Blue: VM voltage

Yellow: Phase current and Blue: Phase Voltage(measured similarly as above)

Zoomed in


Thank you.
Waiting for your response.

Hey Raj,

I tried to look up that part in our lot trace tool but it didn't come back with anything correct.  The best I can tell from the picture is DRV8825 / 19TG5 / CHR7 or CHRZ.  I looked up both CHR7/CHRZ and neither had a hit, can you look and tell me if those codes are right on the chip? 

Even if they are, it's weird that the markings appear to be etched and not printed on.  See my pic below of the chip on my EVM board to compare.  I recommend you request samples for the chip here:  https://www.ti.com/product/DRV8825#order-quality Just click Request Sample and log in to your account and you should be able to request up to 50(?) for free, or you can just order them.  

The signal in this shot is confusing me, it doesn't look like a normal stepper motor control sequence:

Let's replace your chip with chips ordered on TI's website and then see if that solves your problems, because those chips are 100% real and Mouser doesn't appear to have stock for the DRV8825 for another year.  

Regards,

Jacob

Hey Jacob,

Thank you for responding!

I looked at the etching on the IC is DRV8825 / 19TG5 / CHP7.
And I've requested samples as you mentioned.
Also, I see the routing for the IC on the EVM board is very different from then in the layout recommendation in the datasheet.

And the signal in which you're confused, even I'm not sure what I had to measure and thus I measured phase voltage as across the winding. and it's a zoomed version of the signal above it.

Regards,
Raj

Hey Raj,

That etching actually does come back as a valid lot code for DRV8825PWP, so perhaps counterfeit isn't our issue like we expected.  Still am interested in if the samples perform differently.  

Sometimes our EVM's are designed well before the datasheet is finalized and detailed, so I would recommend referring to the datasheet not the EVM as a layout reference/guide.  

Can you take some more oscilloscope shots for me, see if you can clarify where the aforementioned signal is?  

Regards,

Jacob

Hey Jacob,

Good day and I hope you are fine.

I get your point regarding counterfeit and I have been declined to provide free samples.

I have designed my board layout with reference to the layout given in the datasheet and not in any other way.

I feel that rather than taking the same signals pictures, can we discuss some small-level testing for better understanding?

Regards,
Raj

Raj,

It is a US holiday today.  Please expect a response tomorrow.

Regards,

Ryan

Hey Raj,

Sorry to hear that you were declined free samples, I'll see if I can help with that.  

Can you provide me with an oscilloscope waveform similar to the image below, and a zoomed in version of the same on any problem-areas?  Showing the phase current, AOUT1, AOUT2, and nFAULT signal.  

Also can you post or directly send me the full circuit schematic (at least for the power input/regulation and motor drive sections), and maybe your PCB layout as well if you don't mind?

Regards,

Jacob

Hello Jacob,

Good day and hope you are fine.

As you demanded, I have captured some oscilloscope waveforms for you but I have only 2 channels so have arranged them accordingly.
Rsense right now is 0.135ohm and I've set voltage divider of 60.4k & 10k and thus Vref is 0.468V and I checked in DSO, it measures 0.454V so same. and according to the calculation, the current limit is 0.694A. And I'm giving 12V by attaching two wires below 470uF through the DC power supply.



Another 3.3V is for the STM microcontroller.

A+ voltage and A+ current: (current was AC coupled and Voltage was DC coupled in DSO settings)

zoomed:

zoomed more:

My colleague said the noise seen in the current can be because of the current sensor we are using and can be ignored.

A+ voltage & A- voltage:

zoomed:

zoomed more:

zoomed more:

A+ voltage with nFault signal:

My colleague also said that we are already using nFault signal to stop the code whenever it shifts from high to low and it never did. So, there is no nFault fault in the randomness.
Also, I saw signals similarly for B+ & B- winding and those were the same ones and so didn't attach them here.

The Schematic is the same one I've shown you before and the power supply is the DC power supply(shown above):

And R16 is not installed, B1 and B2 are not connected.

Layout:

The upper two white arrows show the supply for VM = 12V; the lower circle denotes all signals coming from the microcontroller.
(I can only share this much of the design while blurring other parts of PCB and don't worry about them, they are not even installed yet in PCB.)

Will be waiting for your insight.

Best Regards,
Raj

Raj,

Jacob will reply shortly.  He is taking a vacation day now.

Regards,

Ryan

Hey Raj, 

Thanks for all the additional information, please give me 48 hours to consult with coworkers on the waveworms and put together a response. 

Can you confirm what microstep setting you used for the waveform capture?

Regards,

Jacob

Hey Jacob,

Yes please, take your time to answer.
We used 1/32 micro-stepping for the motor. You can evaluate the same from the schematic.

Best regards,
Raj

Hey Raj, 

It seems like our main issue is to answer "How is the VREF pin being damaged".  It's going to be hard to answer that with the current setup because of several board changes along the way and swapping out the motor driver chips.  There could be other unrelated issues with this board or other components that are making it confusing to try to debug. 

I think it will be best if you can use a brand-new, clean PCB that you can solder down a new, legit part down with the PowerPad as well.  Use the properly sized sense resistors and the 1M charge pump instead of 1K.  With this new, clean setup with new board and components see if the problem can be reproduced.  

We'll also continue trying to get you an EVM one way or another so you can compare oscilloscope waveforms with the EVM vs your own board. 

Regards,

Jacob

Hi All,

I think PCB layout may need some improvements. Taking a look at picture above (I assume PCB is 2-layer one) I noticed a few errors like: 

- thermal vias under DRV8825 have thermal reliefs,

- C5 grounding trace is long (high inductance), C5 could be grounded directly to ground plane,

- I can not find any grounding vias connecting top ground layer with bottom one.

I think it would be good to redesign PCB layout focusing on thermal performance and current return paths (grounding).

Regards,

Grzegorz

Hey Jacob,

I will try that new board and fresh setup but it will take time. 
Also, regarding free samples or other things, I guess my profile was not updated clearly, it had some glitches. So, I have updated it properly this time. You can try one more time I guess.

Thank you so much!

Best regards,
Raj

Hi Grzegorz,

Good day.

I must say you have a sharp eye to find those thermal reliefs at the bottom plane. Yes, I accept my mistake. I forgot to overlap polygon to the same netlabels which is why that grounding of C5 got far and similarly for thermal reliefs. Although, the top and bottom layers are connected by multiple places as Bat- pad of C1 and some more through-hole pads on the left side which I had to blur because of company policy.

Thank you for your suggestion.

Best regards,
Raj

Hey Raj,

Let us know if you would like us to review an updated schematic and layout with all previously discussed changes.

Regards,

Jacob

Hey Jacob,

I guess it will take some weeks as multiple projects going on. but will let you know here once done or updating something.

Regards,
Raj