Broken PCB Trace

[donavan]

Hello,

i have an diy H-bridge with 4 Mosfets that switches 50V, 15A at 1kHz to 16kHz. The copper thickness should be 105um.

Now my problem is that the copper trace right after the mosfet pad keeps breaking at the red line you can see in the first attached picture.
I dont have a picture of this but the copper is bend upwards and the solderstop mask is broken. What might be the problem here, is 105um copper thickness not enough for 15A?

the second picture shows one case where a big section of the copper layer just evaporated, what could cause this?
My theory would be, first the trace broke and didnt make good or any contact, then the voltage on that trace got so high it just sparked across and evaporated the copper plane.
 

[capt bullshot]

I'm just guessing:

Usually a too high current caused by a dead short would cause damage seen in the second picture. Rather not simple thermal overload by having too much current heating the trace too much. Maybe arcing, but I'd suspect 50V and 15A wouldn't be enough to make this damage by thermal overload alone.

For upwards bent copper, my first thought is mechanical stress that broke the trace. One could be able to see if the damage is caused by mechanical stress or electrical overload by closely inspecting the broken trace (look for signs of heat, an overheated trace looks different as a mechanically broken trace). A too high current could cause the symptoms you describe, but in my opinion the layout should take the 15A without overheating. So I'd suspect that a failure resulting in a significantly larger current than 15A caused both kinds of damage.

[Neilm]

This is a DIY H-bridge - have you checked that you don't have shoot-through? That would cause the second picture

[M4trix]

Maybe it's not a 105um copper clad in the first place !? 

[suicidaleggroll]

This is a DIY H-bridge - have you checked that you don't have shoot-through? That would cause the second picture

Agreed.  I'd take a close look at your H-bridge driver timing.  Do you have a current monitor anywhere on the system (high side or low side, input or output of H-bridge)?  If so, watch that with a scope to see if you're getting current spikes at the switch points.  If not, you can try monitoring the rail voltage right at the input to the H-bridge to make sure it's relatively flat, if you're getting shoot through you'd probably have large voltage dips at the switch points.

What is this H-bridge powering?  If it's an inductive load, do you have freewheel/flyback diodes on it?

[donavan]

I'm just guessing:

Usually a too high current caused by a dead short would cause damage seen in the second picture. Rather not simple thermal overload by having too much current heating the trace too much. Maybe arcing, but I'd suspect 50V and 15A wouldn't be enough to make this damage by thermal overload alone.

For upwards bent copper, my first thought is mechanical stress that broke the trace. One could be able to see if the damage is caused by mechanical stress or electrical overload by closely inspecting the broken trace (look for signs of heat, an overheated trace looks different as a mechanically broken trace).

external mechanical stress shouldnt occure in this particular place, can the mosfet vibrate while switching and cause this?
i will inspect the broken trace next time it happens, i guess thermal stress will cause a black color on the soldermask?

A too high current could cause the symptoms you describe, but in my opinion the layout should take the 15A without overheating

Yes i was pretty shure the layout should handle the current thats why i ask for a second opinion to be shure

Maybe it's not a 105um copper clad in the first place !? 

Yes that was my thought too, i am currently checking what thickness the copper layer has

This is a DIY H-bridge - have you checked that you don't have shoot-through? That would cause the second picture

The gate drivers are two UCC20520 (http://www.ti.com/lit/ds/symlink/ucc20520.pdf). I guess that could happen when the control signals are not synchronised i will check that.

Agreed.  I'd take a close look at your H-bridge driver timing.  Do you have a current monitor anywhere on the system (high side or low side, input or output of H-bridge)?  If so, watch that with a scope to see if you're getting current spikes at the switch points.  If not, you can try monitoring the rail voltage right at the input to the H-bridge to make sure it's relatively flat, if you're getting shoot through you'd probably have large voltage dips at the switch points.

What is this H-bridge powering?  If it's an inductive load, do you have freewheel/flyback diodes on it?

I will check the timing and the voltage/current at the output, thanks.
The load is inductive, i look into if freewheeling diodes make sense there. 

[suicidaleggroll]

The load is inductive, i look into if freewheeling diodes make sense there.

They absolutely will, when your FET goes to shut off you're going to get a huge reverse voltage spike from the load, which if it doesn't kill the FET, could arc across the pads and could cause damage.
See here for wiring:
https://forum.arduino.cc/index.php?topic=50931.0

[capt bullshot]

I'd expect to find hints of melted copper at the broken trace, not necessarily blackened solder mask. If the trace would be broken mechanically, I'd expect the edges to look rather sharp. And I wouldn't expect mechanical stress (vibration) due to normal operation.

Cross conduction wouldn't be a concern with your driver chip, as long as you used it as intended. Switching 15A and 50V into an inductive load causes the body diodes of the MOSFETs to conduct. So one would always drive the MOSFETS in synchronous recitfying mode (either high or low side turned on, with minimum possible dead time) to minimize the body conduction time. Switching speed of the MOSFETs is kind of critical, too fast will cause heavy ringing than can lead to various kinds of destruction of the H-Bridge components. Usually one of the MOSFETs goes into a short, causing exploded packages or vaporised traces like the second picture.
Finding the best combination of switching speed and dead time is an iterative process that may include damages as seen here, fastest possible switching often isn't the best solution.
Using real fast (Schottky) diodes as freewheeling diodes may help, but usually not done for a MOSFET based H-Bridge.

[donavan]

i cut a layer out and measured it, but its very hard to get rid of the solderstop mask. With the mask on it has a thickness of 160um. I scratched the mask off with a knife and some heat and it had about 60um. So either i scratched off too much or the copper is actually just 70um thick. I tried acetone but it didnt do anything, i read potassium hydroxide works.

[Neilm]

i cut a layer out and measured it, but its very hard to get rid of the solderstop mask. With the mask on it has a thickness of 160um. I scratched the mask off with a knife and some heat and it had about 60um. So either i scratched off too much or the copper is actually just 70um thick. I tried acetone but it didnt do anything, i read potassium hydroxide works.

Try measuring the resistance of a piece of track and compare it to the theoretical resistance. You will need a 4 wire measurement for this but the difference between the two values should be obvious.

I see that the dead time is programmable so check that the correct resistors have been fitted. Also, check how long it is taking for the FETs to turn off. I had to diagnose a blown H-Bridge when I found the FETs were turning off much slower than I expected.

[donavan]

The load is inductive, i look into if freewheeling diodes make sense there.

They absolutely will, when your FET goes to shut off you're going to get a huge reverse voltage spike from the load, which if it doesn't kill the FET, could arc across the pads and could cause damage.
See here for wiring:
https://forum.arduino.cc/index.php?topic=50931.0

wouldnt the body diode act as freewheeling diode? Are additional diodes faster or more robust than the body diode?
Can i use schottky diodes with a fixed breakdown voltage of about 80V to safe the mosfet from destruction when the mosfet is closed and Vds gets too high?
Also, can i use a varistor at the H-brdige output to protect the mosfets until the fuse blows?

I looked at some Eval boards like http://www.ti.com/tool/TIDA-00436#Technical%20Documents and they didnt seem to use additional diodes.

[SVFeingold]

wouldnt the body diode act as freewheeling diode? Are additional diodes faster or more robust than the body diode?

Maybe to both. You'd have to check the FET datasheet for the freewheeling diode properties. Depends on the size of your inductive load as well.

Also, can i use a varistor at the H-brdige output to protect the mosfets until the fuse blows?

Varistor operation is still a little bit of a mystery to me, but I think you would want to carefully check the specs to make sure it will actually protect the MOSFET. AFAIK they are intended to suppress voltage surges, not necessarily current surges. So if it is rated for the normal operating voltage and you get shoot-through of your MOSFETs...well, the voltage hasn't increased so the varistor won't really do anything to protect them.