Hello. I'm an amateur having a hard time wrapping my head around designing a snubber for a buck converter LED driver powered by a bicycle dynamo.

Some background: Bicycle dynamos reach a saturation current (typically around 0.5-0.6A) at a fairly low speed. The way I understand it, the voltage across the dynamo is load-dependent. At higher speeds there is more potential power available (in the form of higher voltage), but it cannot be gotten without adjusting the load. Getting more current out of the dynamo can be achieved using a buck converter by lowering the duty cycle at higher speeds. So, I built a digitally controlled synchronous buck converter using an Arduino 32kHz PWM channel to drive a MOSFET driver, pair of N-fets, inductor, etc . It works as expected (more current at lower duty cycle) and is reasonably efficient on the breadboard (70-80%).

The problem is I was zapping the high side MOSFET (Q1), presumably by exceeding the max Vds during the off-cycle from the inductance in the windings in the dynamo. My first solution was to try to protect the MOSFET with C3 and D2 (D1 is a Vgs protection diode that I put in before I realized that the likely problem was transient voltage from the dynamo and D3 is another attempt at limiting the voltage out of the dynamo).

This seems to work; I'm no longer frying the high side FET (but haven't tried duty cycles below 50%). Although I still see lots of transient spikes on the scope, they seem to be clipped to about 20V or so, presumably from the 19V TVS diode.

However, if I disconnect the LEDs I fry the high side MOSFET, despite my crude efforts to save it. The more I've read about it, the more it seems like I need a proper RC snubber. My question to you guys is where should this go? If left to my own devices I'd put it between the drain and source of Q1. Are there any pitfalls to doing that? My second questions is what format should it be? A simple RC snubber? Any suggestions on how to calculate values would also help.

Comments and suggestions encouraged!

 

I would think that substantially increasing the size of C3 and adding a Schottky diode in parallel with Q2 could help a great deal in alleviating the problem. Using a MOSFET that has an avalanche rated body diode would also be a big help.

C3 provides a storage place for the generator output, and increases the time required for the voltage to rise on Q1's drain. Adding a Schottky diode across Q2 will "buy more time" for Q2's body diode to turn on/off, and for Q2 itself to turn on. If Q2's drain is not going negative, you won't need to worry about adding this diode - however, it'll be tough to tell if that is occurring without an O-scope.

 

Thank you. That's a good start. So if I add a resistor between C3 and ground that would create an RC snubber that would dissipate the energy during Q1's off cycle. Is that correct?

I need to see what's going on at the node between Q1 and Q2 with my scope.

I should add that I don't fully understand why a load disconnect is frying the high side mosfet, when it is fine when actively switching. The 28V zener diode should be limiting the output of the dynamo to a voltage below the max Vds (which is 30V).

 

In general, your dynamo is a current generator, rather than a voltage source. This means that, it will push a certain current whether or no. In other words, when unloaded, or lightly loaded, the output voltage will be very high, which may be what is zapping your FET.

The trick is to provide a path for this current when no load is attached. The avalanche rated body diode mentioned above is the right idea. There is such a thing as a power zener diode that is essentially the same thing. The shunt resistor would also be OK I think, as it would draw more current/power as the dynamo's voltage increased with less load!

In the old days, you simply mechanically removed the dynamo's drive wheel from your bicycle tire to shut the thing off....

 

Most bicycle "dynamos" are in fact alternators, designed to have high leakage inductance so that the rising frequency as the bicycle moves faster results in a larger and larger series reactance restraining the current from increasing.

The resulting output is therefore usually AC. Are you sure that a rectifier is not required at the input to your system?

 

Most bicycle "dynamos" are in fact alternators, designed to have high leakage inductance so that the rising frequency as the bicycle moves faster results in a larger and larger series reactance restraining the current from increasing.

The resulting output is therefore usually AC. Are you sure that a rectifier is not required at the input to your system?

Sorry, I should have clarified that Vdyn is, in fact, rectified DC from the dynamo, not AC.

 

In general, your dynamo is a current generator, rather than a voltage source. This means that, it will push a certain current whether or no. In other words, when unloaded, or lightly loaded, the output voltage will be very high, which may be what is zapping your FET.

I'm sure that was what was happening originally, but D2 should be taking care of that voltage spike when Q1 is off. Maybe it doesn't conduct fast enough and there is a very fast transient spike before D3 starts to conduct?

The trick is to provide a path for this current when no load is attached. The avalanche rated body diode mentioned above is the right idea. There is such a thing as a power zener diode that is essentially the same thing. The shunt resistor would also be OK I think, as it would draw more current/power as the dynamo's voltage increased with less load!

In the old days, you simply mechanically removed the dynamo's drive wheel from your bicycle tire to shut the thing off....

I'll give this a try, thanks!

 

And just to recap, it seems like neither of you are suggesting a traditional RC snubber, is that right?

 

And just to recap, it seems like neither of you are suggesting a traditional RC snubber, is that right?

A snubber at the input to L1 probably is not going to do much unless there is enough dead time in the cycle with neither transistor conducting to allow significant overshoots to occur. I do not know enough about your driver IC to know if this is likely. Checking with an oscilloscope might be best, if you have access to one.

SgtWookie's advice about increasing the value of C3 however sounds to be a very good idea. Unless there is a big capacitance here (thousands of μF may be best), there will be a danger of the input voltage periodically collapsing during the "dynamo" AC cycle. This may lead to reverse current flow from the output filter and perhaps odd things happening with the bootstrap cap.

A large C3 will also help to suppress any spikes from the dynamo coil, although the diode you have added will help with this - one important difference being that any current conducted by a Zener represents wasted energy, whereas charge stored in a capacitor does not.

Edit: If you are thinking of putting a resistor in parallel with C3, note that this would likely to be consuming a lot of power if it was made low enough to help much with overshoots. If C3 is made large enough, such a resistor would be redundant, provided that you have a suitably rated zener to hold the peak voltage down.

 

A snubber at the input to L1 probably is not going to do much unless there is enough dead time in the cycle with neither transistor conducting to allow significant overshoots to occur. I do not know enough about your driver IC to know if this is likely. Checking with an oscilloscope might be best, if you have access to one.

SgtWookie's advice about increasing the value of C3 however sounds to be a very good idea. Unless there is a big capacitance here (thousands of μF may be best), there will be a danger of the input voltage periodically collapsing during the "dynamo" AC cycle. This may lead to reverse current flow from the output filter and perhaps odd things happening with the bootstrap cap.

A large C3 will also help to suppress any spikes from the dynamo coil, although the diode you have added will help with this - one important difference being that any current conducted by a Zener represents wasted energy, whereas charge stored in a capacitor does not.

Edit: If you are thinking of putting a resistor in parallel with C3, note that this would likely to be consuming a lot of power if it was made low enough to help much with overshoots. If C3 is made large enough, such a resistor would be redundant, provided that you have a suitably rated zener to hold the peak voltage down.

Thank you for the advice. The consensus seems to be to increase C3 by a lot. I'll give that a try!

 

With a very large C3, be careful not to exceed the peak current ratings of Q1 when it turns on. C3 will be storing 1/2 CV^2 energy and can dump it very quickly without any series resistance to limit current.