MOSFET half bridge switching
This article is written as section 6 on pwm motor control. It is however quite relevant even if you do not follow the rest of the articles.
The 4QD electronics public area has an article which explains briefly how the MOSFETs switch in a half-bridge PWM controller. This article carries on from that, giving more detail and discussing some of the finer points.
At point A, there is no low side MOSFET gate voltage - the MOSFET is off, so its drain voltage is high. Any motor current is now re-circulating through the hiside MOSFET.
At some gate voltage around point A, the loside MOSFETs start to conduct. There may be many MOSFETs in parallel and they may not be exactly balanced, so some may start turning on before others. This is where the positive temperature coefficient of silicon is useful: any MOSFET which turns on earlier will get hotter, so its Rds(on) increases, which tends to slow its turn-on, balancing the currents between the MOSFETs.
Referring to the 2QD circuit, the rate at which the MOSFET gate voltage rises will be controlled (mainly) by the rate at which the 1K pullup to pin 14 charges the two capacitors (C10 and C11).
The whole switch-on period, from A to E, is around 1µS. 'Slowly' (in perhaps 200nS, the time from B to C) the current in the MOSFETs builds up until the MOSFETs is conducting the full motor current. All the time these low side MOSFETs are conducting less than the full motor current, the rest of the motor current must still be going somewhere - it is still recirculating in the top MOSFET. To flow in the top MOSFETs, this must still be forward biased.
Consequently, during this current build up phase (B-C) the drive MOSFETs have the full battery voltage across them. This is therefore a period of enormous dissipation in the MOSFET: Vcc x Im/2 (Supply voltage times average MOSFET current).
When the flywheel MOSFET ceases conducting, the drain voltage is now free to start falling. But now another effect occurs: there is a 'hefty' capacitor internal to the MOSFET, connected between gate and drain (Cgd). This has to discharge as the MOSFETs source drain voltage falls and it can only do this by absorbing the gate drive current. For a BUZ100S, this Cgs is quoted as typically 615 pF. To discharge 615pF by 36V, takes a significant current. There is a resultant 'terrace' in the gate drive waveform clearly visible at this point, between C and D, of perhaps 300nS.
It is an interesting point that, if the gate drive current is limited (as it must be), the terrace period extends. The terrace period is also high dissipation, the bottom MOSFET is carrying full motor current, at an average voltage of half the battery.
It will seem clear from this that MOSFET switching should be done as fast as possible, for best efficiency. Unfortunately there are other tradeoffs, for instance, if you switch fast, you can expect a high level of radiated r.f. harmonic interference.
Also, if you try and switch on the drive MOSFET too fast, you are attempting to remove the flywheel current from the top MOSFET fast - the top MOSFET's gate drive will already have been turned off (hopefully), and the top MOSFET's body diode will be conducting. Now MOSFET body diodes are not the fastest of diodes: the specs for the BUZ100S state a turn off time of 160nS max. If you approach this rate, the bottom MOSFET now has to conduct not only the motor current but also the decaying turnoff current of the top MOSFET before the drain voltage can fall.
At point E, the gate waveform is at its maximum: the MOSFET has long since switched on. There is now full battery voltage across the motor and the motor current will increase from now until the drive MOSFETs turn off. While the drive MOSFETs are off, the motor's inductance keeps the current circulating and the motor current decays slightly. If the motor's inductance is too small, then this current 'ripple' during the cycle becomes significant, reducing the motor efficiency and it is this that determines the minimum operating frequency. Most commercial controllers chose around 20kHz - high enough for any conventional motor, but interestingly, near the minimum that is sensible for the Lynch motor.
At point F, the gate voltage turns off. Now, if you look at the 2QD circuit, the driver base current available for turn off is the whole current that an LM339 can sink (typically 16mA) times the gain of the PNP driver. The gate voltage is removed quickly, maybe 400nS and yes, there is a 'back terrace' of maybe 100nS.
Switchoff of the drive MOSFET is much the reverse of its turn on. It takes time and, as soon as it is no longer conducting the full motor current, the source flywheels upto full supply voltage to dump the balance of the motor current through the flywheel diode/MOSFET, so there is another pair of high dissipation times, initially the voltage across the MOSFETs rise, to allow the flywheel to start conducting, then the Rds decreases further as the current in the drive MOSFET decays and the top MOSFETs conduct in diode mode.
Only after the top MOSFETS are conducting (as diodes) the whole motor current, can they be safely turned of to become resistors rather than diodes.
And of course, the top MOSFET gate drive must be removed before the bottom MOSFETs start to conduct current - or in the worst case (which actually gives best efficiency), the gate drive can be removed from the top MOSFETs just as the current falls to zero.
One of the things that can kill a MOSFET is a high energy noise spike, arriving at the wrong time. Motors have brushes and commutators. Commutators and brushgear arc, especially when old and worn, arcs are good sources of wideband noise. Between motor and controller are wires. Wires act as transmission lines. There is a finite chance of the correct shape of noise waveform being generated to travel along the transmission line that is the motor leads and arrive at the controller with enough energy and at precisely the wrong time. If that happens - well, it's been stated that the FET in MOSFET stands for 'Fire Emitting Transistor'.
There may well be a lot more that should be added on MOSFET switching etc, but the forgoing should have given the reader a good understanding.
Other relevant pages
Part 7 Starts to deal with Full bridge control.
Page Information© 1996-2011 4QD-TEC