For my EE friends...
Damn I am sick of this already... but here is what I have come up with... and you'll probably find it interesting if you like this sort of thing.
In short:
I thought that the zener trick would for sure be the cure... it eventually was proven not to.
I resorted to trying what was suggested to me by a coworker, which was to shunt the transient at the power supply and at the mosfet, amazingly this didnt work either.
And alas, I did what I knew would have to work.... add a capacitor of sufficient size... and YES, that worked.
So..... WHY then does just shunting, not appear to work?
I think the answer can be illustrated thus-
In the above circuit, L3 and L2 represent wire inductance from the power supply, with R5 and R6 being the wire resistance (not much). The mosfet is not shown, but it would be between R3 (the load) and L4. L1 and L4 are representitive of on-board inductance.
The key point to note is that the induced voltage is fundamentally related to the rate of current change in the circuit by the relationship "V = L * di/dt". With no math being done, what is important here is that di/dt in this circuit is NOT effected. The very forceful method of shunting the arc necessitates that the device doing the suppression be VERY fast. Also....and more importantly, Yes, this solves the issue of what I will term "Upstream induction", i.e. the induction of those wires connecting to the power supply, however, the inductances and circuitry represented to the right (R1, R2, L1, L4, R3) are further down the transimission line yet left to deal with the di/dt that surge suppressor does not address. Thus, I believe that you would find the necessity to have surge suppessors placed nearly everywhere which a transient could dammage the circuit.
I tried adding surge suppression right across the mosfet, and I dont know exactly how, but it STILL blew. I gave up, on the recognition that a surge suppressor does NOT address the fundamental CAUSE of the spike, but merely tries to forcefully arrest it.
So... the capacitor idea:
This is a much more elegant solution basically, but with some other draw backs. First, you need sufficent capacitance, and in this case a few thousand uF are needed to begin to bring things in check (i know that the circuit shows 100uF ;) . As a result, the cap is big and bulky... and NOT cheep either. But... it is the best solution, and here is why:
When the switch is opened, the inductive voltage, again generated by a di/dt term, would be present... except that the capacitor C1 now supplies the power needed to momentarily run the motor. The motor draws 30 amps, so I will note that indeed the time it runs IS short... but it need only be long enough to reduce di/dt to a 'managable' level. The current supplied by the capacitor now keeps the current in the wires flowing in the same direction. What DOES change, is the di/dt on the capacitor wires. So clearly, the leads of the cpacitor need to be kept VERY SHORT.
As for the zener idea I was working on in the beginning. Yes, the idea fundamentally should work. and it certainly showed indications of vastly improving the problem. But as with the shunt, di/dt remains unchanged. I strongly suspect that the spike caused problems in the gate drive circuit wich affected the main mosfet, but the most likly cause I think is that if you have a very very short turn on and turn off event, the transient rize time is much faster than the mosfet can turn on to supress it--- keep in mind, in all of these circuits, the problem I am trying to solve is NOT the mosfet turning on and off the circuit, it is due to the circuit being broken, and jittered rapidly, while the mosfet is already on, or in an unknown state of on or off. All of this asside, however, the idea was scrapped. Still, I have included a pic of the basic idea below for your geeky consumption.
Also, creating a vast improvement was to add a capacitor across the 'freewheel' diode (the cap I am talking about is not shown). I discovered that the diode is too slow to suppress the entirety of the transients created by the load induction.
I was in the basement all day today working. Mostly oblivious to the world around me.. *sigh* I feel like the world is passing by me while I work on this... anyway.. hope you enjoy this. I look forward to much geeky commentary.
sorry- no spell check- too tired. Sound it out, you'll be fine.
In short:
I thought that the zener trick would for sure be the cure... it eventually was proven not to.
I resorted to trying what was suggested to me by a coworker, which was to shunt the transient at the power supply and at the mosfet, amazingly this didnt work either.
And alas, I did what I knew would have to work.... add a capacitor of sufficient size... and YES, that worked.
So..... WHY then does just shunting, not appear to work?
I think the answer can be illustrated thus-
In the above circuit, L3 and L2 represent wire inductance from the power supply, with R5 and R6 being the wire resistance (not much). The mosfet is not shown, but it would be between R3 (the load) and L4. L1 and L4 are representitive of on-board inductance.
The key point to note is that the induced voltage is fundamentally related to the rate of current change in the circuit by the relationship "V = L * di/dt". With no math being done, what is important here is that di/dt in this circuit is NOT effected. The very forceful method of shunting the arc necessitates that the device doing the suppression be VERY fast. Also....and more importantly, Yes, this solves the issue of what I will term "Upstream induction", i.e. the induction of those wires connecting to the power supply, however, the inductances and circuitry represented to the right (R1, R2, L1, L4, R3) are further down the transimission line yet left to deal with the di/dt that surge suppressor does not address. Thus, I believe that you would find the necessity to have surge suppessors placed nearly everywhere which a transient could dammage the circuit.
I tried adding surge suppression right across the mosfet, and I dont know exactly how, but it STILL blew. I gave up, on the recognition that a surge suppressor does NOT address the fundamental CAUSE of the spike, but merely tries to forcefully arrest it.
So... the capacitor idea:
This is a much more elegant solution basically, but with some other draw backs. First, you need sufficent capacitance, and in this case a few thousand uF are needed to begin to bring things in check (i know that the circuit shows 100uF ;) . As a result, the cap is big and bulky... and NOT cheep either. But... it is the best solution, and here is why:
When the switch is opened, the inductive voltage, again generated by a di/dt term, would be present... except that the capacitor C1 now supplies the power needed to momentarily run the motor. The motor draws 30 amps, so I will note that indeed the time it runs IS short... but it need only be long enough to reduce di/dt to a 'managable' level. The current supplied by the capacitor now keeps the current in the wires flowing in the same direction. What DOES change, is the di/dt on the capacitor wires. So clearly, the leads of the cpacitor need to be kept VERY SHORT.
As for the zener idea I was working on in the beginning. Yes, the idea fundamentally should work. and it certainly showed indications of vastly improving the problem. But as with the shunt, di/dt remains unchanged. I strongly suspect that the spike caused problems in the gate drive circuit wich affected the main mosfet, but the most likly cause I think is that if you have a very very short turn on and turn off event, the transient rize time is much faster than the mosfet can turn on to supress it--- keep in mind, in all of these circuits, the problem I am trying to solve is NOT the mosfet turning on and off the circuit, it is due to the circuit being broken, and jittered rapidly, while the mosfet is already on, or in an unknown state of on or off. All of this asside, however, the idea was scrapped. Still, I have included a pic of the basic idea below for your geeky consumption.
Also, creating a vast improvement was to add a capacitor across the 'freewheel' diode (the cap I am talking about is not shown). I discovered that the diode is too slow to suppress the entirety of the transients created by the load induction.
I was in the basement all day today working. Mostly oblivious to the world around me.. *sigh* I feel like the world is passing by me while I work on this... anyway.. hope you enjoy this. I look forward to much geeky commentary.
sorry- no spell check- too tired. Sound it out, you'll be fine.