DRV8350H circuit design failure. what am i doing wrong?

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
Hello, I've been working on this brushless motor controller for a long time now. This is version 2 as I began using a different mosfet driver DRV8350H. After soldering with a heat gun twice with two different DRV's, both don't seem respond at all despite double checking and making sure every connection was correct. I cannot tell what I'm missing here. maybe someone here might have an idea on what I might have done wrong with either my design or layout. I really just copied what the datasheet recommended. I'm truly out of options. maybe also someone might have a recommendation of an easier mosfet driver to deal with, let me know aswell. thank you!




Screenshot 2020-11-23 192332.png


Screenshot 2020-11-23 192420.png
 

michael8

Joined Jan 11, 2015
155
Which chip do you have, your diagram say DRV8350FH but I don't
see that on the data sheet:

product page
https://www.ti.com/product/DRV8350

datasheet
https://www.ti.com/lit/gpn/drv8350

see page 80 for the chip device naming scheme: DRV8350H

DRV83 -> device
5 -> 100 volt
0 -> no sense amps
-> no "R" so no buck regulator inside
H -> hardware interface (alt is SPI)

eval kit
https://www.digikey.com/en/products/detail/texas-instruments/DRV8350H-EVM/9860043

What does "don't seem respond at all" mean?

Is the DRV8350 chip taking power?
Are any other outputs active?
Is the fault output active?
Are other chips on the board working? (processor)
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
Which chip do you have, your diagram say DRV8350FH but I don't
see that on the data sheet:

product page
https://www.ti.com/product/DRV8350

datasheet
https://www.ti.com/lit/gpn/drv8350

see page 80 for the chip device naming scheme: DRV8350H

DRV83 -> device
5 -> 100 volt
0 -> no sense amps
-> no "R" so no buck regulator inside
H -> hardware interface (alt is SPI)

eval kit
https://www.digikey.com/en/products/detail/texas-instruments/DRV8350H-EVM/9860043

What does "don't seem respond at all" mean?

Is the DRV8350 chip taking power?
Are any other outputs active?
Is the fault output active?
Are other chips on the board working? (processor)
So its the drv8350fh. No buck and no sense resistor amps. I have the fault pin pulled high by a 10k pull up resistor so I read 3.3 volts on fault output, which either means its not being pulled low to indicate a fault or its failing to for some reason. All other chips are responding perfectly fine. All three in phase amps are working and so is my microcontroller as well as my magnetic incoder chip. And its not responding as in my microcontroller is sending pwm signals to the inputs of the drv chip, but nothing is coming from any of the gate outputs. I even checked and saw that the c6 charge capacitor is charged to 20+ volts with reference to ground in order to drive the high side mosfets. Whats weird is that when I set the enable pin to low (deactivating the windings), only about 200ma flow (coming from the other chips and power led on the microcontroller board) but when its set high, about 300 to 400ma flow but nothing still happens. Also I currently do not have the outputs to the windings connected to anything. I wonder if that contributes to anything ?
 

michael8

Joined Jan 11, 2015
155
A hard problem, and what's needed is some test of partial function rather
than work/non-work test. Some ideas:

Fault output: The pullup on the fault output is reported as working.
Is there some way to tell if the fault line from the IC is really
connected? (Inspection? Create a NON-DISTRUCTIVE fault?)

The MODE pin is grounded via 47K resistor specifying 3x PWM mode
(datasheet page 30 8.3.1.1.2)

Consider a temporary test in "Independent PWM Mode"
(datasheet page 31 8.3.1.1.4 MODE pin tied to DVDD)

This bypasses some of the "smart" parts and might allow testing of
some of the drv chip. Be careful as this disables some protections
and might conflict with other choices made in the PC board wiring.
It also disables checks on shoot-through.
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
Okay alot of good ideas! So my main issue is that the drv chip only exists as a no leads ic package which makes it very difficult to check if the connections are shorting to each other or simply not connected. I did some continuity checks by following the traces to the nearest passive component solder pad and check to see if they were shorted together, they were not. But I have no way of testing if the connections from the pads to the ic are correctly connected. I can observe that there is definitely solder on the pads but if they connected to the leads of the ic, i do have have an easy way of telling. I kinda squish the ic down while it was being heat gunned so ensure there were connections, maybe thats what messed it up? Any ideas on that would be amazing! I could remove the 47k resistor and route the connection to the power supply to switch it to 6x pwm mode but that turn all the low side mosfets on because there inputs are pulled logic high with pull-ups. I suppose this could be a good test to see. I also am unsure if I fried my second chip or not, or even my first one for that matter. I'm not sure how sensitive these chips are. I only have one more left right now so I'm hesitant to try and solder the last one on if my other chips are even fried.
 

michael8

Joined Jan 11, 2015
155
> So my main issue is that the drv chip only exists as a no leads ic
> package which makes it very difficult to check if the connections are
> shorting to each other or simply not connected.

Understood.

> I did some continuity checks by following the traces to the nearest
> passive component solder pad and check to see if they were shorted
> together, they were not. But I have no way of testing if the connections
> from the pads to the ic are correctly connected. I can observe that there
> is definitely solder on the pads but if they connected to the leads of
> the ic, i do have have an easy way of telling.

So you've done a visual inspection (as much as possible). Good.

> I could remove the 47k resistor and route the connection to the power
> supply to switch it to 6x pwm mode

MODE pin 47K resistor? Why do you have to remove the resistor? It won't
take that much current at +5v (DVDD value not the power supply.)

Oh, what's your VCC voltage?

The datasheet says MODE high gets 1x PWM Mode (8.3.1.1.4 page 31)

> but that turn all the low side mosfets on because there inputs are pulled
> logic high with pull-ups. I suppose this could be a good test to see. I
> also am unsure if I fried my second chip or not, or even my first one
> for that matter.

I'd be careful to check out all the implications of the mode switch before
trying it -- there might be conflicts between the board wiring/FETs etc
like shootthrough to avoid damage.

> I'm not sure how sensitive these chips are. I only have one more left
> right now so I'm hesitant to try and solder the last one on if my other
> chips are even fried.

I would be hesitant to try the last chip too. The current chip has
at least the boost supply and enable pins doing something.

You said enable to the drv chip increased the board current by 100 to
200 mA. Another approach would be to chase down where this current is
going. Is it staying in the drv chip? How much power is that -- does
the chip temperature change? Or is it going elsewhere? Anything else
getting warm? Can you trace the current via PC board trace voltage drops?
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
Well now setting the enable pin high doesn't result in any current increase anymore. Again I'm not sure why. The chip does not noticeably heat up at all even when there was current flowing. I haven't tested changing the mode to independent pwm mode as I do not know what will happen with all my current connections. I haven't looked at the datasheet in a bit cuz I'm currently at work so I can read it again later tonight. So currently I do not know how changing the mode to independent pwm mode disables safety features? I really just focused on 3x pwm and 6x pwm but 3x is my preferred one for the code I currently have written. Also the voltage supply along is 12v. I do plan on having it at 24v but the current voltage regulator can't handle that much wasted heat so ill have to change it to a buck converter set up in a later version. But right now I can't even get the drv chip to work.
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
okay so I've ran a few tests. first i bypassed the 47k resistor straight to the 3.3 supply in order to change the mode to independent mode. made no difference. no voltage on any gate pins of the mosfets. at this point the motor wires are connected to complete the circuit.
then i noticed that the amp connected to phase c was outputting .4 v instead of 1.5v (offset voltage) and the others were somewhere at 1.3 to 1.4v indicating that there was some kind of current flowing even though my power supply isn't detecting it. so i changed it back to 3x pwm mode and it still was the same. so i disconnected the motor wires and saw that the output was still the same. i assume that's because i some how have a faulty amp on phase c because no current should be able to flow through the sens resistor with the motor disconnected. even after setting the enable pin to low did not change the amp output.

then i checked the voltage of the charge cap c6 after every change and noticed that it stays at 12v now no matter what.

another weird thing is that when i checked every gate of the mosfets i noticed that MHA 's gate was at .7v once i disconnected the motor and no other mosfet is getting that, just 0v.

i really cant make sense of anything that is going on. the only other thing i can do is desolder the chip and check everything again. if all is good then i might have to use my last drv chip. its not the end of the world if i have to but its just a waste even if i cant figure what's causing all of this.
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
Yeah its not super obvious if the thermal pad is supposed to soldered down or not but this is an interesting idea. When I first soldered the ic on my first run it kinda floated and moved in the wrong place when the solder paste reflowed, forcing me to re solder it. Then I had to take it off and clean all connections because of too much solder. After using a new chip, I only slightly tinned all pads and then resoldered the chip by pressing down gently on it. So mabe something might not be connected properly but again I have very little way of checking that.
 

michael8

Joined Jan 11, 2015
155
> bypassed the 47k resistor straight to the 3.3 supply in order to change
> the mode to independent mode

See the datasheet page 12 SEVEN-LEVEL H/W INPUTS (IDRIVE, VDS)
which specifies "Tied to DVDD". Is that 3.3 V or 5V from
drv pin 29 5V internal regulator output?

So, possibly that wasn't independent mode.

> then i checked the voltage of the charge cap c6 after every change and
> noticed that it stays at 12v now no matter what.

Chip not enabled? or?

> i really cant make sense of anything that is going on. the only other
> thing i can do is desolder the chip and check everything again. if all
> is good then i might have to use my last drv chip.

I'd worry about using the last chip (or order more?) before finding some
error which would have damaged the previous chips. Perhaps they are all
working just fine considering their environment. Or are you thinking
of static or thermal damage?

How expensive (money? time?) would it be to make a *small* "breakout" PCB
for the drv chip to test it in with the minimal prewired connections?

This would have larger connection points for each pin of the IC.
I picture this including the bypass caps and boost cap plus
pullups/pulldowns which don't interfere with configuration.

Possibly the FET gate drive would connect to dummy loads.
Something like 2 resisters in series between VCC and gnd with the midpoint
pretending to be the FET gate, something like 1k to 10k resistors so
the current would be minimal but a lot more than a 10M input voltmeter.

Also do you have a "better" PCB trace picture (gerbers?) or something
so that the traces could be followed.
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
okay so i see what you mean by pin 29 being a 5v regulator and that the pin mode is determined by the 5v power and not the 3.3v. sorry i am not exactly an expert at this stuff lol (that's why i have you guys) so thank you again.

also i attached the gerber files here so you can get a better look at it.

now when i tried to measure pin 29 from the cap pad and i measured no voltage even after enabling the chip. so i must not have the connections correct. perhaps any ideas on how i can solder this thing correctly?
 

Attachments

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
okay some news. so i desoldered the chip and resoldered it back on and noticed some interesting effects.

when in 3x pwm mode phase a and b both connected to ground (so the mosfets work) but phase c was connected to the 12v(somehow the high side mosfet was turned on), resetting didn't change this but setting it to independent mode did fix this and connected phase c to ground. at which point i set it back to 3x pwm mode.

but when i tried to turn on all three high side mosfets only phase a and b went high but not c, it stayed low. which at first doesn't really make sense to me until i sent 50% duty cycle to all three phases.
when this happened phase a worked perfectly and even the high side mosfet was receiving 24v, but phase b stayed low and phase c was barely pulsing at very low peaks like 5v and i noticed this voltage at the high side gate of phase c's mosfet but the low side pulsed just fine.
this has led me to believe that the charge cap is failing to properly charge all three high side mosfets before the voltage drops too low and the safety features of the chip are kicking in, because if i let it run like this for too long sometimes the fault pin will go low.
so maybe i need a bigger charge cap for c6? in the example schematics on the datasheet show a 47nf but maybe with my mosfets i need a bigger one?
any thoughts? should i use a bigger one and see what happens?
 

michael8

Joined Jan 11, 2015
155
> i desoldered the chip and resoldered it back on and noticed some
> interesting effects

So probably there was a soldering problem which is now somewhat or
completely better. (Or something else, floating CMOS inputs or bad
connections elsewhere can change from touching the board...)

> when in 3x pwm mode phase a and b both connected to ground (so the
> mosfets work) but phase c was connected to the 12v(somehow the high side
> mosfet was turned on),

And the INLx INHx were set to?

I'm looking at the 3x PWM Mode Truth Table and am really confused.
I thought the SHx pin was an input but the table has it as Hi-Z, L, or H.
Any ideas what this means? I must be missing something...

> resetting didn't change this but setting it to
> independent mode did fix this and connected phase c to ground. at which
> point i set it back to 3x pwm mode.

So INHC had to be 0 to cause that.

> but when i tried to turn on all three high side mosfets only phase a and
> b went high but not c, it stayed low. which at first doesn't really make
> sense to me until i sent 50% duty cycle to all three phases.

> when this happened phase a worked perfectly and even the high side mosfet
> was receiving 24v, but phase b stayed low and phase c was barely pulsing
> at very low peaks like 5v and i noticed this voltage at the high side
> gate of phase c's mosfet but the low side pulsed just fine.

Could there be a capacitive "connection" instead of a real connection
on some of the phases so that pulses make it through (barely) but
DC doesn't? Or two fets in the drv IC fighting each other or
something else on the PCB?

How are you measuring the voltages? Is there some load resistance
so that a very weak capactive coupling won't show as a signal?

> this has led me to believe that the charge cap is failing to properly
> charge all three high side mosfets before the voltage drops too low and
> the safety features of the chip are kicking in, because if i let it run
> like this for too long sometimes the fault pin will go low.

Can you watch the voltage on the VCP pin (charge pump output)?
I would think loss of charge pump voltage would be measuable.

> so maybe i need a bigger charge cap for c6? in the example schematics on
> the datasheet show a 47nf but maybe with my mosfets i need a bigger one?
> any thoughts? should i use a bigger one and see what happens?

C6? I'd think more like C7. Or there's too much of a load on one or
more of the GHx pins.

Would it be safe with all power off on the PCB to measure the resistance
between GHx and SHx for each phase? I'd only do this with a DMM which
put out less than a diode drop voltage for resistance measurements.
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
>And the INLx INHx were set to?

>I'm looking at the 3x PWM Mode Truth Table and am really confused.
>I thought the SHx pin was an input but the table has it as Hi-Z, L, or H.
>Any ideas what this means? I must be missing something...

remember that all INLx pins are pulled high by my circuit so no matter what the low side mosfets are turned on by default but only when i set any INHx pin high the low side mosfet turns off and the high side should turn on.

>Could there be a capacitive "connection" instead of a real connection
>on some of the phases so that pulses make it through (barely) but
>DC doesn't? Or two fets in the drv IC fighting each other or
>something else on the PCB?

interesting, i would like to know but i do not have a good way of checking that.

>How are you measuring the voltages? Is there some load resistance
>so that a very weak capacitive coupling won't show as a signal?

i have a simple volt meter and a small hand held oscilloscope to see the signals but its pretty crude.

>Can you watch the voltage on the VCP pin (charge pump output)?
>I would think loss of charge pump voltage would be measuable.

i see a steady 22.4v at VCP when its pulsing as i described.
 

kaindub

Joined Oct 28, 2019
81
You're trying to troubleshoot a pretty complex circuit
At least you will need an oscilloscope with 20Mhz bandwidth and minimum two channels.
Its not just a case of seeing that there is an output at the chip. You need to see the phase relationship between a and b and c.
Also get the datasheet and measure and confirm the voltages at all chip pins. Its never going to work if the inputs are missing/incorrect. Find test instruments that are small enough to access the pins. Other wise your just guessing.
 

Minesh

Joined Oct 1, 2013
1
Hello, I've been working on this brushless motor controller for a long time now. This is version 2 as I began using a different mosfet driver DRV8350H. After soldering with a heat gun twice with two different DRV's, both don't seem respond at all despite double checking and making sure every connection was correct. I cannot tell what I'm missing here. maybe someone here might have an idea on what I might have done wrong with either my design or layout. I really just copied what the datasheet recommended. I'm truly out of options. maybe also someone might have a recommendation of an easier mosfet driver to deal with, let me know aswell. thank you!




View attachment 223227


View attachment 223228
Hi Have you looked at the TI Forum DRV8350 there are several people who have designed with this chip and have similar problems. Most common problems are DRV8350 gate output pins shorting to Gnd causing the phase output to be stuck.
 

Thread Starter

PaperBackJack

Joined Jan 22, 2019
15
Hi Have you looked at the TI Forum DRV8350 there are several people who have designed with this chip and have similar problems. Most common problems are DRV8350 gate output pins shorting to Gnd causing the phase output to be stuck.
I will have to take a look.
 
Top