H-Bridge blew up 10 MOSFET a day

~~~~~~~~~~~~~~~~~~~ sorry, right side P and left side N

I used two 1uH/14A inductors and many bypass caps you can see from schematic.

L1/L2 (1 uH) are far too small compared with C23/C24 etc. (> 100 uF) At each cycle C23/C24 etc. are completely charged and discharged so about 14 W are dissipated by the MOSFETs P = Frequency * Energy/cycle = 1KHz * 12V * 12V * 100uF = 14.4 W

Either you increase L1/L2 (I would say 1 mH or more) or you decrease C23/C24 etc. (I would say less than 1uF)

ee

Thanks very much for your input. I just updated my schematics, actually I just let the left-P off, left-N on, and switching the right P/N. It doesn't work either. The noisy sound keeps coming out.

for 10A current, I picked a 14A inductor, but I couldn't find >1uH compact size from coilcraft/coiltronics/sumida or digikey. My layout is here, the trace between MCU and MOSFET Driver is very short with vias, is the via (inside the rectangular area) causing the ringing on the edge of my PWM? it's double layer board with joint gound on the bottom. I labeled the current flow as well as the mcu/ driver area.

layout is here:

and my updated schematics is here with corrected pwm signal control

Thanks very much.

ee

I also modify the switching frequency to 200KHz where the LC roll off is far less than fsw. Still toasting.

Thanks very much for your input. I just updated my schematics, actually I just let the left-P off, left-N on, and switching the right P/N. It doesn't work either. The noisy sound keeps coming out.

for 10A current, I picked a 14A inductor, but I couldn't find >1uH compact size from coilcraft/coiltronics/sumida or digikey. My layout is here, the trace between MCU and MOSFET Driver is very short with vias, is the via (inside the rectangular area) causing the ringing on the edge of my PWM? it's double layer board with joint gound on the bottom. I labeled the current flow as well as the mcu/ driver area.

layout is here:

and my updated schematics is here with corrected pwm signal control

Thanks very much.

========================================================================

For one thing, there is not much of a heat sink for the MOSFETs, and they have have something like a 48 to 96 deg C per watt thermal characteristic when mounted on 2 oz copper PCB. If you have 5 amps RMS flowing in the P-channel with 70 mOhm RdsOn, it will dissipate 1.75 watts and will soon reach critical temperature. At 10 amps you will have 4 times the power.

In addition, the capacitors you have connected to the TEC will increase the power dissipation especially at higher PWM frequency, and I don't understand their purpose. They are not bypass capacitors (you do have some of them from power to GND), but instead appear to be an attempt at filtering the DC signal to the TEC, which is probably not really needed.

The series inductors may cause even more problems. What is their purpose?

Paul

WM

/N

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

he

Thanks for the comments.

The original design came from ADN8831 and LTC1923, they also have evaluation boards and application notes. Take a luk at

The cap used heavily is to maintain the current pumping on the power rail due to the switching, refer to ADN8831 application notes. I designed with ADN8831 +/-1.8A/3V without problem. Bypass cap also gets rid of the spikes which are bad to the cooler module.

The inductor used here is for filtering like semi-buck in this full- bridge and then lower the ripple. refer to LTC1923.

Thanks for the comments.

The original design came from ADN8831 and LTC1923, they also have evaluation boards and application notes. Take a luk at

The cap used heavily is to maintain the current pumping on the power rail due to the switching, refer to ADN8831 application notes. I designed with ADN8831 +/-1.8A/3V without problem. Bypass cap also gets rid of the spikes which are bad to the cooler module.

The inductor used here is for filtering like semi-buck in this full- bridge and then lower the ripple. refer to LTC1923.

=========================================================================

The data sheet

There is no capacitor directly across the TEC. The inductors are shown as 10 uH, but the part number SUMIDA CDRH6D2B-220NC sounds like 22 uH. They are bypassed to GND with 22 uF X7R ceramic capacitors. The transistors are shown to be Si9801

The application note

The PWM frequency appears to be in the order of 50 kHz. And the power supply seems to be 5 VDC.

I don't know the characteristics of the TEC, but I don't think these circuits are designed for any more than about 1 or 2 amps.

Paul

nt

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

ey

ic

e?

=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D

NC

You are right. Those circuits are for 5A/5V use. I use 1uH/14A, the part# is wrong, I made BOM separately.

TEC acts as a power resistor, you adjust either the voltage or the current, it is a semiconductor cooling device, it is sensitive to spikes.

When I do simple analysis I just use buck converter to do simple calculation, I didn't start firmware yet, thus the digital PID is not involved.

Today I ran at 4A it got hotter and hotter, at 2A so far so good. I'll update this post tomorrow.

After looking at the PDF on the driver, I really think you should revisit the input of the gates.

I see that you have them both tied together. The PDF states these are AND gates with an inverted input on the B pin of each gate. In theory, I can see these doing nothing but fighting with each other. If you want to turn on AND gate 1 for example, A input sets to 5V and B input goes to common or below the logic threshold that specifies 0 logic for that chip.

The way I see it working now, A input gets turned on while B input is off because the input is inverted.

That's just an observation, maybe i've over looked something.

That's 4469. I use 4468 without invert. They give a logic wiring diagram with two pwm, one direction control with shoot through protection? I didnt' take a further look. I use MCU to get pwm with dead time inserted. The driving pwm 0-12V looks okay without ringing now.

Well that explains it. The PDF only showed the inverted gate model in the circuit examples.

You may have removed the ringing but have you resolved the destruction problem? Because it's obvious that uncontrolled signals on the gates are going to cause issues in this circuit.

P.S. I've made circuits similar to that but I didn't use any inductors at the bridge output because I was driving a non movement resistive load, I just placed a snubber across the load point.

tion

ors

a lit bit different than the motor driver. right now I got: a nice pwm signal on Right-gate, Right N-gate, and a bad switching on the drain where connected to the inductor. If they are turned on/off nicely the waveform on the N-drain should be a pulse width but it looks like:

____ __ _______ |_| |________| .................

multiple switching on each period! the multiple switching varies as I adjust the dead time of PWM. no clue so far.

Not sure about your load type you're using "TEC", but if you can place a resistive load in place of the TEC for now, you can better determine the issue . It's possible the TEC is causing some issues. This is assuming that you are taking these readings from the actual bridge output ? That being a low voltage unit, I'd use something simple to handle the current like an automotive incandescent lamp. This also gives you a nice visual view of things while you're working on the unit.