PWM and frequency response

[DigiTan]

I spent the last few days experimenting with PWM speed control with DC motors and got me wondering: is there an optimum PWM rate? I know in terms of average power, there is normally no frequency preference with resistive devices, but I don't really expect this apply to reactive components like solenoids or motors. But mostly I am looking for a minimum rate to free up the most processor time possible.

So is there an optimum frequency? How can I determine it? :?

 

[ljcox]

Some years ago, I designed a DC motor speed controller for a friend's model train.

I used the EMF generated by the motor (ie. when the pulse was off) as a feedback signal. So the circuit only generated a pulse when the EMF fell below the threshold set by the throttle potentiometer.

So it was not driven by an oscillator. The motor was pulsed only when it needed a pulse.

My friend said it was that best controller (particularly at low speed) he had seen.

Len

 

[DigiTan]

That sounds a lot like what I'm looking for. How exactly did you isolate the voltmeter from the speed controller

 

[Styx]

Generally the rule of thumb with power electronics is the higher the switching frequency the better.

IF you could switch 100kW at 1GHz it wouls eliminate soo many problems, torque ripple, resplonse time, EMF,EMI

However, at the moment it is impossible to switch a power-electroinics switch with any decent power behind it at 1GHz for more that one swithc (burns out after first)


The problem is is switching times. It takes a finite time for the volts across teh device to collapse and the amps to rise in turning ON (equally amps to collapse and voltage to rise at turn-OFF)

THe problem is you have to fully switch the device off/on before turning it on/off otherwise the device will sit in its active region for too long

equally if you are in a H-bridge you have to ensure that you do not turn an upper device on at the same time as a lower device, be it hard switch (ie command both on) or during switching and the devices are in their active region, thus rule of thumb is an interlock of twice the stitching time (+headroom), this is also true for minimum pulsewidth/max

Some 3rd gen IGBT can turn on and off (at full rating's) in 700nS which is bloody fast. From this you could then say:

Say 1us to give some headroom

From all this it is saying we could switch this device at: a period of 4us, which is the same as: 250kHz safely.
However, this is at a fixed 50% duty and is of no use to man or beast.

Give a 10-to-1 control the switching frequecy needs to be dropped to 25kHz to be able to have control over the drive. That works and would be the APSOLUTE maximum such a device would be concidered to be switched at.




However, there is always a however.
It might be advisable to swtich as fast as possible in all cases, there are other factors that will force you to switch slower. With increase switching frequecy, the current ripple and such might be signifcantly reduced. BUT switching losses go up with switching frequency.

Thus depending on your cooling (that 25kHz was suming holding the substrate at room temerature!!!) that switching frequecy might have to be dropped to 15kHz.

Now to drive the switch you will need a gateboard with a burn resistor on the gate, to switch the device fast you have to have a low gate resistor, low gate reisitor will give rise to high current flow though that resistor. With a high switching frequency that could lead you to a situation where 15kHz is too high for the power avaible (and what it can dissipate) on the gateboard, so the switching frequency could then drop to 12kHz.

As it turns out 10kHz to 15kHz is pretty much hte standard for a 20kW drive.



All in all it really depends on your output power and the type of switch you are using, in your case you could push it upto 100kHz

 

[ljcox]

That sounds a lot like what I'm looking for. How exactly did you isolate the voltmeter from the speed controller

Its a long time ago so the details are vague. I think I used an analogue switch to connect the feedback path when the pulse was off. I'll find the circuit and post it tomorrow.

Len

 

[ljcox]

Here it is.

Len