Motor Driver Chips

Consternation turned to lucidation !!

No, an H-bridge class D amplifier should clamp the spikes into the supply rail capacitors through the FET's intrinsic diodes.

Or using common mode chokes for CM noise.

The inductance of the motor may help its own differential EMI.

And through Rds-on, which is bidirectional and generally lower drop than substrate diodes.

As PWM duty cycle goes up, energy is pumped from the power supply into the motor. As duty cycle goes down, if the motor has stored energy in angular momentum, power flows backwards unto the power supply [1]. That can have hazards.

There shouldn't be much in the way of spikes, except maybe during any PWM dead time. It would be unusual for an H-bridge to deliberately go hi-Z.

[1] Conservation of Energy short-cuts a lot of thinking.

Not sure why you say "No". I'm talking about what motor driver chips do. They DO use the parasitic diodes to clamp the voltage which is dissipative from the motor's perspective. In fact the VNH5019, which we are using to of all things - to control a motor - when the PWM pin is lowered leaves the high side driver on and turns off the low side driver. With the parasitic diodes the current flow is in a loop through the power rail, but not back into the power source, just using the rail as a conductor between the motor ends.

The diodes have significant voltage drop and so dissipate power. If you have low voltage drop transistors, why not use them? The remaining two choices are to simply short the motor which will continue the current flowing with little dissipation or to reverse the connections allowing the current to flow back into the supply (charging up the large caps on the rail). Not sure which is preferred when driving a speaker. Mostly I'd just like to know for sure what the MP6519 is doing.

My goal is to drive a speaker which has much less inductance... I think. No small part of a speaker's behavior is due to it's interaction with the acoustic environment. I've asked the speaker vendor for the Thiele-Small parameters which can be used to calculate the electrical model.

When driven, power goes into the motor through the FETs. When not driven the inertial power does not return to the power source. The inductive power can return to the source when all FETs are turned off through the parasitic diodes. That is probably why motor controllers leave one FET turned on allowing the current to flow in a near short through a power rail and one intrinsic diode. This provides a path for the inductive back EMF which otherwise would be destructive. Although if the FET if not left on there is still a path for the inductive current through two intrinsic diodes, but that *will* feed current back into the power source. This link has some very good images of the waveforms.

I don't see how the motor momentum could feed back into the power source in a PWM system. When connected to the source the inertial back EMF simply opposes the applied voltage and never exceeds it, being proportional to the speed of the motor. There is no way the motor can be connected to the power source to cause the inertial energy to be fed back into the power source. For energy to be transferred current and voltage are both required. Inertial back EMF has to result in current flow to transfer power and there's no way to cause a reverse current to flow into the power source.

It is possible for the load on the motor to cause the motor to spin faster, resulting in a higher back EMF resulting in a reverse current flow. Perhaps that is what Larkin is thinking of. Or maybe he is thinking of variable voltage drive where current can flow both ways.

Yes, but thinking is required to apply conservation of energy properly.

On 2021-03-13 19:55, Rick C wrote: [Snip!]

Evidently, you'll need a boost converter.

Jeroen Belleman

The BEMF has the value of the speed driven constant. So if you drive a motor and turn off the motor drive the BEMF is the same as the drive voltage before turning off

However you have inductance in the loop, so during the dead time the voltage will clamp to either rail through the body diodes

An h-bridge is a boost-buck converter. It's like a DC Variac.

How many kilowatts do you intend to feed into the speaker if you even consider a motor controller ?

What frequency range do you intend to feed into the speaker ?

What is the resonant frequency when mounted into the final enclosure ? The speaker behaves electrically in a nasty way around resonance(s). Also any passive crossovers within the band should be considered.

Do you intend to drive the speaker directly with PWM or through some LPF ?

Less than 1.

Square waves of varying amplitude and 250 to 550 Hz. Harmonics up to 5 kHz are important.

Which enclosure? The one we are using is about two cubic feet by my estimate and has large holes for other purposes (you can put your fist through one) and lots of ventilation holes in the back. Speakers are mounted on the bottom. Yes, I know, far from optimal.

There will be EMI filtering. The PWM is around 250 kHz so I'm thinking the speaker cone can take care of integrating the pulses.

The reason for using a motor controller is because every amplifier device I find is either large and messy ($$) to use or won't work on 7.5V. There are a number of small motor controllers in small packages that would do a good job. Only a few of them allow for measurement of current to provide for detecting a failure in the speaker, open or short. The MP6519 generates its own PWM output to control the current and so provides a signal from the current. It is not clear exactly how it switched the outputs in a PWM cycle. I've exchanged emails with a factory engineer and he keeps answering the wrong questions. He even had someone work on a simulation, but none of the data is very useful to answer my question. They have their own simulation tool and model. Maybe if I fire up that tool it will give me answers.

The square wave is supposed to sound like a woodwind. Maybe I should add a bellows, reeds and some valves on a tube?

I'm sure I will someday, but not for this circuit.

I did see a similar design using a buck converter as a single ended driver complete with current detection. But it used a $7 LT part.

...

Usually PWM motor controllers use a high enough rate that the spike you refer to lasts the entire duration of the half-cycle.

During that time the current still flows in the same direction through the motor and can supply energy to the mechanical load by transferring it from the inductive store of energy.

Also it is normal for this current to pass through the active device rather than the body diodes or any explicit diodes in the circuit (they are useful for catching the voltage during the dead time used to avoid current passing directly through both devices from power to ground.) Even bipolar transistors will pass current with the collector in the opposite to normal direction though not as well as FETs. IGBTs can't pass reverse current so they require diodes.

At the high frequency of operation current flows continuously the motor inductance can act as as boost converter to transfer energy from the motor inertia into the supply to slow the motor down if needed. (as JL mentioned).

Pushing power back into the supply can cause problems in some system. EVs use it for regeneration (as they have a battery to dump energy) but others may not. A few decades ago I read about a large HP disk drive that used the power from the motor slowing down to power the electronics for a graceful shutdown in the case of power failure, not sure if they do the same on small disk drives.

If you use a low frequency excitation the system will be less efficient and more power will be dissipated in the motor resistance (dissipated as heat). Can also cause audible noise or vibration.

The system acts pretty much the same as if there was a separate buck/boost converter powering the motor with DC (except there is no smoothing cap).

It is also the same if you drive a speaker - although the speaker inductance may not be enough to maintain continuous conduction and may require adding additional inductance in series. The back-EMF from a speaker is still there as with a motor although because they are not as efficient as motors will not be as significant.

kw

Burst of white noise into a feedback comb filter with the resonance turned up just below self-oscillation makes a very uncanny flute sound with very little DSP overhead:

That would be the "brake" mode in motor controller documentation. All transistors open is what they call free wheeling. As you say, different rates tap into different effects.

Sorry, not following that. How does the inductance of the rotor tap into the rotational energy of the rotor? That would be the inertial back EMF which is simply the motor acting as a generator producing the same voltage as the applied voltage, but opposite current. The inductive effect is to maintain the existing current at whatever voltage is required in either polarity. I don't see how the two interact and in fact, here are images that show the distinct effects.

As you say between the high speed PWM pulses the inductive EMF is apparent. Then when the PWM is stopped the inductive EMF creates a large negative spike and only after that attenuates does the inertial EMF become apparent maintaining a steady level. How would the inertial effect pump current back into the supply?

The inertial back EMF can act as a brake if current flows. Yes, EVs have regenerative braking that charges the battery, but that is done on the circuit boards controlling the motors. The motor by itself does not create a larger voltage required to charge the battery or feed the power supply. At least I don't see it.

So in normal PWM operation the motor controller switches between providing power to the load and switching to "brake" mode where both low side or both high side switches are turned on to allow the inductive current to continue to flow?

Does a class D amp do the same thing or does that reverse the applied voltage when PWM?

I'm preparing a simulation of the TI DRV8876 which might be a better part and I noticed they call for a "VM" rated capacitor for the charge pump for the gate driver. I can't find anything on this. I assume VM stands for voltage multiplier. What would be special about a cap used in a voltage multiplier? Is it the ripple current?

How much less than 1 kW if you intend to drive it from 7.5 Vdc ? :-)

Apparently you meant below 1 W.

Near the low end of the speaker response the impedance varies between inductive and capacitive so you may consider this in your interface modeling. If the intended frequency range is well above resonance, the impedance curve is more stable.

So this is essentially a baffle construction. At 250 Hz the wavelength is 1.3 m, so the distance between the front side of the speaker cone around the baffle to the back side to the cone should be at least 0.7 m avoid acoustic short circuit between the front and back side of the cone, causing strong attenuation of the lower tones.

At least at 1000 Hz a few watts of power into a speaker into a small room is a torture method due to the standing waves in the room :-).

I do not know about 250 Hz but the audio levels should not be too high, not at least for a longer periods of time. Of course good for alarm tones, but that should be easily acknowledged and silenced.

Is the energy consumption an issue to justify class D (PWM)

With 7.5 Vdc a full H-bridge can deliver about 6.5 V peak into 4 ohms (10 W) or 4 Vrms sine 4 W which should be more than enough.

Un bel giorno Rick C digitò:

VM is the power supply for the driver. You need to use a VM rated capacitor for the charge pump because when the low-side switch is ON, the source voltage of the high-side switch will be near 0V, therefore VCP will be typically 5V and (VM-VCP) will be VM-5V.

I think class-D audio amps don't have to worry about back-charging the power supply. The signal frequency is relatively high and speakers don't store much energy.

The question is what "VM rated" means. A google search on VM rated capacitors turned up nothing.

They are using the term like is used for a class Y capacitor. If they are just talking about the voltage on the cap, that's not a problem.

On Saturday, 13 March 2021 at 21:17:42 UTC-8, snipped-for-privacy@gmail.com wrote: ...

That is not just used during braking - it is also used to control the effective source voltage during sustained operation at less than full speed. For example if the desired motor speed is 50% of the full speed obtained when the full battery voltage is applied across the motor then a duty cycle of 50% would be required (ignoring losses in motor resistance etc). The motor would think it is being fed from a low-impedance source at 50% of the battery voltage.

Stopping the PWM allows the back-EMF to be visible at the motor terminals. This can be used to provide negative feedback to control the motor speed in the face of varying loads.

There is an interesting paper that gets into the technical details here:

I did a controller using similar principles recently. It worked fairly well although as described in the paper it is difficult to make a good PID controller with the limited sample rate that is limited by the inductance of the motor.

...

The motor inductance is used as the inductor in a boost converter to increase the back-emf voltage to the level required for current to flow into the battery - there is no magic occurring in the electronics apart from the bridge devices and controlling the duty cycle of the PWM.

When the lower switch is on current will flow from the back-emf and store energy in the inductance, during the second part of the cycle the switch is opened and the upper switch closed. Since the current through the inductance remains (approximately) constant the voltage at the junction of the switches rises until the upper catch diode (and the upper switch when it closes) allow current to flow to the battery. Then the cycle repeats.

Since this process removes energy from the back-emf it causes mechanical torque that slows down the motor (and anything attached to it).

The voltage multiplication may be up to 10-20 to one to allow regeneration in EVs to be maintained down to a handful of MPH

...

... Yes - although I the class D amps I have read about exclusively use a full-bridge and either apply positive or negative battery voltage to the output - in that case 50% duty cycle would result in zero volt output. Greater than 50% would result in a positive output, less than 50% a negative output.

Motor controllers usually switch only one pair of the switches (for example the lower ones) at a higher rate while the others ones would be statically switched one way for forward and the other for reverse. This approach results in lower losses in the less than perfect motor inductance and lower peak-peak current ripple and voltage swings on the wiring at the expense of a potential discontinuity when going through zero. This isn't important for motors but is for speaker outputs.

Similarly TEC drivers often operate in the cooling or heating mode with the circuit being reconfigured to operate in the two modes, only one half of the devices operate at the high-speed PWM.

kw