Hello everybody, hello Paul, I am French, new on this forum, I subscribed only to answer to this post and congratulate you for your work
I am an electronics developer (hardware/software), and I have some experience in controller design, and I plan to build a commercial universal controller with a price as low as possible because of the lack of reliable controller at a reasonable price. In France it's the beginning of EV conversion, I personally convert a car at this moment, and it's difficult to find good controller for high power (brush/brushless or AC) without pay a lot of money.
I read all this topic and I also seen your impressive work on the DC controller. I see actually you are maybe testing your "low voltage" 144V power stage based on mosfets. I maybe have some advise to give to you, concerning some vicious things.
The first thing is that at beginning (and me the first!!) we think it's better to commute mosfet or igbt as fast as possible to have a better efficiency. It's an error ! Let's imagine a couple of high side + low side mosfet which send current in a phase. When the high side mosfet is ON, current flow from BAT+ to the phase through the high side mosfet. When the pwm cycle is finished, we switch high side mosfet OFF, wait a dead time, switch the low side mosfet ON (synchronous rectification). Then we switch the low side mosfet OFF, wait dead time and switch the high side mosfet ON. But when we switch the high side mosfet ON, current actually flow from BAT- to the phase through the internal diode of the low side mosfet. And this diode cannot become blocked instantaneously when the high side switch ON and apply BAT+ on the phase. It's called the reverse recovery time. During this time the current can flow in "reverse direction" in the diode, and we have a "shoot through" condition during the reverse recovery time. The lower the Rgate value, the higher the switch on speed but the higher the shoot through current during reverse recovery time of the opposed free wheeling diode. The reverse recovery is generally very fast, but with high speed commutation the current peak can be very very high.
In my first designs I smoked some mosfet for this reason, they were stressed by this very high current peak. After I discovered this, I understood that it's not alway better for efficiency to commute mosfet ON as quick as possible, because of this high current peak during reverse recovery time. I was interrogating myself why the mosfet became almost hot with a very small motor and small current. It was the reason.
Thus to solve this problem we increase the Rgate value, but now a new problem appear ! Let's imagine the low side mosfet is OFF and we commute ON the high side mosfet. The potential on the low side mosfet drain will increase quickly, and because of the parasitic miller capacitance between drain and gate, when potential increase quickly on drain, the gate potential increase too (as if it were sucked up by the drain potential). And because we increased the Rgate value, gate potential can increase higher, and low side mosfet begin to commute, and we have another shoot through condition which decrease efficiency and can stress mosfets too (or igbt of course).
This is typically called autocommutation, and the higher the bus voltage (higher dv/dt on the drain potential), ang the higher the Rgate value, the higher the mosfet autocommutation ! It's really easy to see this phenomenons by looking the mosfet Vgs potential on the scope. When the opposite mosfet commute, you will see what happen on the gate...
To solve this problem, on simple design, we can use a different Rgate to switch the mosfet OFF, with the help of a schottky diode. When bus voltage increase, some time we need to completely remove this Rgate, just use the schottky to be sure to lock correctly the gate to gnd. Of course it needs a powerfull gate driver to absorb the current which flow out of the gate and tend to increase its potential.
The good new is that is not a big problem to have a low Rgate when switching off mosfet, because it's not a big problem to swith off quickly. On the contrary, it's better for the efficiency.
But sometime with high DC bus voltage and high dv/dt it's difficult to solve problem even with no switching off Rgate (because the parasistic Rgate of the mosfet himself) thus solutions are to slow down the dv/dt rate thus increase the switch ON Rgate (but decrease the efficiency), or to use negative voltage to control the gate (for example +15V/-5V). The gate potential will also be sucked up but will not reach the commutation threshold. It complicates the design but sometime it's mandatory.
Anyway, with igbt, it's often mandatory to control them with negative voltage to increase the switch off speed.
To conclude this long (and soporific !?
) message, I will say that the theory of operation of a brush or 3 phase power stage is alway easy to understand, but the practice is really another story !!
When you begin the development of a controller you alway begin by the theory, and these vicious things I explained are difficult to find in application note. I discover all this things by myself with experiments, and only after that and long hours of searches I find some small explanations on some articles on Internet which confirmed all my personal experiments conclusions. But I only find these articles because I knew what I was looking for !!
Regards,
Cyril
Today
Other popular topics in this forum...
