1
\$\begingroup\$

There were some good posts regarding choosing the PWM switching frequencies considering the needs of the motor control. Here I am focusing rather on understanding of how to decrease losses in the power bridges of the controller. Mainly analyzing two main components of these losses: the conductive losses (power MOSFETS + inductor + resistive load) versus switching losses in power MOSFETs (Miller effect between full open and fully closed state).

Consider example of 3000RPM motor , 50Hz outer loop frequency, having 4 pairs of poles whence 200Hz outer loop motor frequency. According to the accepted "rule of thumb" the required PWM switch frequency starts from 10 x 200Hz = 2kHz This number is thought of been able recreating the "natural sinusoidal" EMF of the motor coils needed for voltage control of the coils current.

Let's assume that I have other means to change the voltage applied to the half bridges MOSFETs in order to control the speed of the motor (in this case to maintain steady 3000RPM). Is it correct to assume that I can reduce switching losses in the power inverter if I just apply the same switching frequency 2kHz thus not recreating sinusoidal waveform at all but only switching polarity of motor phases to make motor run with constant speed ? (Again, assume for a moment that issues of the close loop, position sensing and fine tuning are passed to another block - the voltage controller , adjusted for a permanent speed.) Or are the conductive losses increase so dramatically that saving on switching losses will not help at all? What kind of losses prevail within say 100Hz to 10kHz switching range ? Would conductive losses increase dramatically in the lower end of this range so that eliminating switching losses don't matter ? Or is it more beneficial to use upper end of the range to have minimal total losses?

Yet one more way to put this question: Is PWM switching frequency chosen only to recreate sinusoidal voltage control of the motor or is it also helping to decrease overall energy losses in the inverter compared to the long period of the ON state conductive losses in the MOSFETs?

Anybody who had heavy practical experience with motor control please share your thoughts. JonRB please? Edit: To clarify the expression "to recreate sinusoidal voltage control" is only one purpose of using PWM, another one - more frequently used for - is to control voltage level below of the provided fixed voltage source. In any case, this post is not about voltage control but about power losses in the inverter : the balance between conductive losses vs switching losses.

\$\endgroup\$
  • \$\begingroup\$ I'm sure there should be a simple question in here somewhere. What does "whence 200Hz outer loop motor frequency" mean? \$\endgroup\$ – Andy aka Jan 25 '18 at 17:18
  • \$\begingroup\$ 8 pole motor requires 200Hz to run at 3000rpm. \$\endgroup\$ – Brian Drummond Jan 25 '18 at 17:26
  • \$\begingroup\$ I feel that your question may have a number of misconceptions built into it. Switching frequency is not really related to motor electrical frequency, except that it should be somewhat higher than the highest frequency required by the motor. In your case, that means much faster tha 200 Hz. The effective voltage applied by the controller is simply VDC * duty cycle, irrespective of frequency. Very often, BLDC motors are not controlled by a recreated sine wave, but with a square wave. For sure, switching frequency effects efficiency. \$\endgroup\$ – mkeith Jan 26 '18 at 5:20
  • \$\begingroup\$ It will probably work fine to use a 2kHz switching frequency if that is what you want to do. However, 2kHz is audible, and if people are around, they may find the 2kHz tone irritating. One of the more common ways to control a BLDC motor is using 6 step commutation. In this form of commutation, the effective voltage applied to the motor is not sinusoidal at all. It is more like a square wave. If you search for "6-step commutation" I am sure you will find some interesting information. \$\endgroup\$ – mkeith Jan 26 '18 at 5:26
  • 1
    \$\begingroup\$ because every time a fet gate is charged and discharged, that is energy that is lost forever. Likewise, every time a FET transitions from on to off, a certain amount of power is dissipated internally in the FET as heat, and that energy cannot be recovered. \$\endgroup\$ – mkeith Jan 28 '18 at 1:27
1
\$\begingroup\$

According to the accepted "rule of thumb" the required PWM switch frequency starts from 10 x 200Hz = 2kHz This number is thought of been able recreating the "natural sinusoidal" EMF of the motor coils needed for voltage control of the coils current.

The PWM frequency must be much higher than the commutation frequency to get a reasonable representation of a sine wave, but also to reduce current ripple. High current ripple causes greater loss due to the increased rms current (which heats the controller and motor) relative to average current (which produces torque).

Current is smoothed out by the action of winding inductance, so low inductance motors need higher PWM frequency. Slotted iron-cored BLDC motors are typically run at 8kHz, while slotless and ironless motors may need 32kHz or higher. Another reason for using >20kHz is to reduce audible noise.

Is it correct to assume that I can reduce switching losses in the power inverter if I just apply the same switching frequency 2kHz thus not recreating sinusoidal waveform at all but only switching polarity of motor phases to make motor run with constant speed ?

Controllers that use '6 step' commutation rather than 3 phase sine waves can have reduced switching losses because PWM is only applied to each MOSFET for 2 of the 6 steps. At full power there is no PWM so switching losses are further reduced.

Or are the conductive losses increase so dramatically that saving on switching losses will not help at all? What kind of losses prevail within say 100Hz to 10kHz switching range ?

Most BLDC controllers use FETs which have low switching loss below 10KHz, so I doubt that it is significant. However current ripple increases as frequency is reduced, so the controller should suffer higher conduction loss at lower frequency. All the controllers I have tested ran at 8kHz or higher, and even at that frequency there was high current ripple. Recirculation currents are another factor. Most controllers rely on the FET body diodes to carry the flyback current when the FETs are turned off. This results in higher conduction loss below 100% PWM.

\$\endgroup\$
  • \$\begingroup\$ Thank you Bruce for bringing up the specifics of increase of the conduction losses due to the flyback diode currents - current ripple effect. I am not sure about the following though: whether we control voltage (and hence output current) by using PWM or we control voltage otherwise and provide just 6 step rectangular commutation, the certain Irms over the motor winding will produce the needed torque and speed. And the total conductive losses from both Ron and from flyback diode losses per given Irms will be the same in both cases. Do you agree? \$\endgroup\$ – VladBlanshey Jan 28 '18 at 18:54
  • \$\begingroup\$ If you control motor speed by varying supply voltage and just apply 6 step commutation then flyback only occurs for a brief period after each commutation, and the rms flyback current is low. If you use PWM there is a flyback after every PWM pulse, so the rms flyback current is higher. At 50% PWM average flyback current = supply current (assuming PWM frequency is high enough to maintain continuous current flow through the stator windings). rms current is even higher because the diode current waveform has a high crest factor. \$\endgroup\$ – Bruce Abbott Jan 28 '18 at 19:57
  • \$\begingroup\$ that means that compared to fundamental 6 step commutation, adding PWM for voltage control not only increases losses due to FETs switching but also adds more of flyback current spikes. That makes total power dissipation in the FETs even higher in comparison than I previously thought! Thank you. \$\endgroup\$ – VladBlanshey Jan 29 '18 at 1:54
  • \$\begingroup\$ If you can edit your answer to include this information about conductive loss increase and switching loss added when PWM used to control voltage -I'll mark it as problem solving as it fully answers this non-trivial question. \$\endgroup\$ – VladBlanshey Jan 29 '18 at 2:04
2
\$\begingroup\$

You are absolutely correct, motor speed can be controlled without the use of PWM by changing the voltage. There are a couple of considerations to this approach.

Brushless motors are not all wound for sinusoidal drive. You would want a motor with a more trapezoidal back EMF for this approach. You mention two functions of PWM in your post. One is obviously speed control, and the second is to match the back EMF of the motor to the drive waveform. When you go without PWM, the drive mismatch from the EMF means that the instantaneous current requirement will change ("current ripple") and so will the instantaneous torque. If you have a large inertial load, you will have a lot of current ripple as the motor goes from a position where there is high torque (the back EMF voltage is low with respect to the drive) to a a position where there is less torque, but your inertia will keep the speed relatively constant. If you don't have a lot of inertia you may actually have some "cogginess" in your output speed.

Your real question is then concerning your switching losses. If you have a driver with a lot of stray inductance and are switching at high PWM rate, your losses can be reduced by your approach. We run downhole motors in this manner. You will still need PWM to keep your current from spiking during start-up. Get the best motor wind you can, but you will have to put up with current ripple of 15-20%.

\$\endgroup\$
  • \$\begingroup\$ Good points you brought to consider. But I'd like to arrive at the explicit answer. To be specific: if you consider the assumptions in my post, then can't I reduce or even eliminate those current ripples by externally controlling the voltage otherwise instead of PWM? The same applies to start-up: why would I need PWM with it's switching losses if I control voltage otherwise for the start-up? Do you agree or not that, if only considering losses in FETs, the elimination of PWM would only lead to the total reduction of those losses? \$\endgroup\$ – VladBlanshey Jan 28 '18 at 18:11
  • \$\begingroup\$ You would eliminate the switching losses related to PWM if you don't use PWM; there would still be some small losses associated with switching between phases. You do not have a shoot-through situation since you will never be turning the lower half bridge on immediately after switching the upper one on (and vice versa). However, controlling the voltage is problematic, because during start-up voltage is very low. You need some mechanism to control the input voltage, and it usually comes back to a PWM somewhere. \$\endgroup\$ – John Birckhead Jan 29 '18 at 14:44
0
\$\begingroup\$

The main reason for operating the FET's in a PWM mode at all is to alter the effective applied voltage. The effective voltage applied to the motor is simply VDC * duty cycle. VDC is the battery or DC bus voltage. Very often, this effective voltage is not modulated to approximate a sine wave. Instead, it follows a simple six-step commutation scheme. If you are not familiar with six-step commutation, please look it up using a search engine. Please don't be offended if you already know all about this. (I have no way of knowing what you know or don't know).

If you are able to control VDC rapidly using feedback, there is no real benefit to using PWM control at all. Losses will be lower if you simply turn the FET's on and off fully once per commutation state. There are many, many control systems out there that do not try to approximate a sinusoidal waveform. They are only using PWM to match the effective applied voltage to the back emf.

You mentioned conductive vs switching losses. The problem is that during PWM, the conductive loss (due to Rds(on)) is always present, because the motor current always flows through a FET. Each time the FET's switch, additional losses are incurred due to charging and discharging FET gates, and also there are resistive losses in the FET channel due to transition from low resistance to high or vice-verse. So to a certain extent, there is no trade-off here. Lower frequency is more efficient in this special case where VBUS is controlled.

Please note that when you apply a voltage to the motor which is very different from the back EMF voltage, it may result in very high current and torque. Essentially, setting the voltage sets the speed of the motor. Acceleration is the derivative of speed. If you change speed rapidly, it can lead to violent action. So if you rely on DC bus voltage to set speed, you need to vary it from very low (near zero) up the maximum, and if the motor speed changes due to sudden load change, you need to respond very quickly by changing the DC bus voltage to match. It may be worthwhile to add some form of current sensing so that you can treat sudden over-current as a fault condition and possibly respond in some fashion.

Finally, note that there may be conditions where you end up with regeneration current being delivered from the motor to the DC bus. The DC bus needs to have some way of accommodating this without blowing up. Regeneration can occur if you need to stop the motor fast, or if an external force (like another motor, or a falling mass) is driving the motor and you attempt to reduce speed. In short, there are a lot of system-level issues here that need to be considered, but are probably outside the scope of what you asked.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.