Dynamic braking power dissipation

[KrisH]

Hello,

I realize this is likely a fair bit different than a typical build spec, but given the level of knowledge on this forum I was hoping someone could point me in the right direction.

I have a project in mind which requires a high duty cycle (up to 10-15min) variable braking power, in the range of 2-4kW. I believe it is technically possible to use a regen controller and dump that energy into a battery, but I'd prefer to avoid needing a large battery and not having to deal with discharging it as it becomes full. I believe what I'm looking for is a large power resistor, though an active-load based on a bank of FETs strikes me as a very attractive solution as well. The tricky part is that I will also be using the motor as a motor, so I need the power-dissipation circuit to only be active when my motor is acting as a generator. Through my own digging, it seems what I'm looking for is a chopper circuit which can PWM a specified (and variable) current through my load, though I'm open to suggestions if there is a better way to accomplish this.

Is there such a thing as a BLDC motor controller which supports a braking resistor instead of/in addition to regen, or regen onto a separate voltage bus? The best option I've found so far is this: http://kellycontroller.com/kls7230s24v-72v300asinusoidal-brushless-motor-controller-p-1343.html , but it doesn't quite seem to have all of the functionality I'm looking for. From what I've seen, it seems that COTS controllers dump their regen current back onto the main voltage bus, for the BMS to deal with. In the event of an over current/over voltage event, the system shuts off and the motor free-wheels. Freewheeling is a desirable fail-safe state for my application, though I'd like to have a solution which stays working under high load and protects the hardware.

Thanks for your help,
Kris

 

[amberwolf]

What is the specific application? Knowing that might give some other ideas, too.

Note that anything that varies the power to the load also varies the braking force.


The problem you have is similar to that in a thread by Buk in the last few weeks, in the Alternative Energy subforum.

In your case, it is at least relatively simple to engage the load--simply have it on a contactor (relay that's designed to switch under load) that is turned on by the brake handle (or whatever you're using to start regen). A second contactor simultaneously disengages the battery from the controller so it doesn't drain into the load.

The biggest issue you have is that of switching over from battery to load at the controller's input (where it will dump the regen)--if there is insufficient deadtime between the switchover, and regen hasnt' started yet, you drain battery power into the load. If there is excessive deadtime, your controller shuts down from lack of power before it can begin regen braking at all.

(and some controllers will not power up without error unless the motor is not spinning and the throttle/brake/etc inputs are all "at rest" (disengaged/zero). so after it shuts down in the above situation, you'd have to come to a complete stop to power cycle it to get it to work again, if it's one of those types)

The simplest (but most wasteful) way to dump this power without that problem is to not disconnect the battery at all, but that drains the battery as well as the regen power the whole time the load is connected, so that's not really acceptable.

But if you make a short delay (less than a second) after the load is connected before disconnecting the battery, and reconnect the battery before regen drops below battery voltage and disconnecting the load, then while it does waste a bit of power, it guarantees you never have the situation of the controller shutting down from lack of power.




As for a load, depending on voltage of the system you might look into resistive heaters (space heaters, hair dryers, stove elements, toaster ovens, etc). Some are ceramic elements, some nichrome or similar wire, etc. Some may lend themselves to moutning to the vehicle frame or other radiative surfaces for airflow cooling while in motion.



Keep in mind that the motor itself will also heat from the braking power generation, so braking times as long as you're proposing may require active cooling of the motor and possibly the controller. (you may be able to power that cooling from the regen itself).


Also note that there are at least two different systems that are commonly used for electric braking on ebike-type controllers. The oldest is regen (regenerative braking, sometimes called recuperation), which via one method or another takes power from the wheel's rotation and feeds it back as current into the battery at the controller input. The newer type is often called EABS or similar, though its also sometimes called regen, and it actively takes power out of the battery and feeds it to the motor in a way that opposes the direction of motion of the wheel. It's much more effective as a brake, but it heats the motor (and controller and wiring) significantly more than plain regen would, and wastes at least as much power as teh regen would have created (so it's really twice the loss of power).

So make sure the controller you're going to use actually uses regen that gathers power back to teh battery.

 

[MadRhino]

I’d say wear the tire instead of the brakes or power group. Friction roller is probably your best solution if you want to save the weight of a bigger motor, battery, controller...

 

[KrisH]

The specific application I have in mind is a winch for towing paragliders and hanggliders to altitude, as an alternative to foot-launching from a mountain. Towing technology for this application is already relatively well-established, but most solutions tend to revolve around complex hydraulic systems or friction brakes, both of which have their downsides, especially when it comes to [considerable] maintenance. In any case, the basic principle is a controlled amount of line tension applied to the glider which effectively generates "thrust", while paying out anywhere from a few hundred to a couple thousand meters of line. The climb rates that are acceptable to us generally result in the process taking ~10-15 minutes while climbing at an average of 2-3m/s. There are static variations where the tow winch is fixed to the ground and a motor generates all the work required, though the more common setup which gets us higher is to mount the winch on a moving vehicle and simply uses the winch as a tension-regulator.

Based on the line spool geometries I intend to use and the aerodynamics involved, I'm looking at a motor which can provide at least 20-30Nm of torque at ~800RPM for take off, with peak torques in the 100-200Nm range at considerably slower RPM (at the apex of the tow when the line is pulling in a more downwards direction), which works out to 2-4KW. Note that in the static case the motor would be providing that energy, while in the "pay-out" vehicle mounted case, the motor would be dissipating that work. The ability to reverse the tow and pull the line back onto the spool while under light tension (after the pilot releases from the tow) is also critically important, so the motor needs to be able to act as a motor in that case. In other-words, a "generator-only" configuration won't be sufficient. Other tow systems have gotten around this by using a smaller rewind motor, but I figure if I'm already using a BLDC as a generator, it is less complex to just use that as my rewind motor.

Because there are variables with pilot weight, glider performance, headwind speeds, tow operator proficiency, etc, the actual rotational speed of the motor can vary considerably, but the important aspect is that a specified amount of torque can be set, with a fail-safe being the motor free-spools to dump line tension. The way I had imagined this working with a power dissipation system would be an active/semi-active load which sinks a constant current regardless of bus voltage. I had imagined using a power resistor (or a space heater is a great idea) which is PWM'd and/or subject to closed-loop current control so that the desired current can be drawn regardless of motor velocity. I realize that I can't guarantee that torque at very slow motor velocities, though I can bound a minimum motor speed and/or provide input power to make up the difference in that corner case.

Unfortunately due to the heating issues and finite power supply, I don't think the EABS is the solution I'd want to pursue, though I imagine EABS could be my solution for the low-speed case? I'm happy to go into further details and operational requirements, though I've already written a novel here.

Also note that there are at least two different systems that are commonly used for electric braking on ebike-type controllers. The oldest is regen (regenerative braking, sometimes called recuperation), which via one method or another takes power from the wheel's rotation and feeds it back as current into the battery at the controller input. The newer type is often called EABS or similar, though its also sometimes called regen, and it actively takes power out of the battery and feeds it to the motor in a way that opposes the direction of motion of the wheel. It's much more effective as a brake, but it heats the motor (and controller and wiring) significantly more than plain regen would, and wastes at least as much power as teh regen would have created (so it's really twice the loss of power).

So make sure the controller you're going to use actually uses regen that gathers power back to teh battery.

From what I've gathered, it sounds like some manufacturers use EABS interchangeably with regen braking. Is there any reliable way to know for sure which system is implemented, and/or is possible that modern controllers use true regen and simply supplement with EABS at low rpm?

 

[amberwolf]

AFAICT from the above, you don't need to switch back and forth between motor and braking modes, you're either in payout (braking) mode, or reel-in (motor) mode, right?

If that's the case, it makes it much simpler to switch between the two, as far as power demands and mechanical stuff are concerned, as there's no need for automatic switching back and forth, so you don't have to have contactors.

Between battery and controller positive, you could use a breaker or switch (possibly from solar panel type places) rated for the DC current and voltage of the motor (reel-in) side of things; it doesn't have to handle any of the pay-out load current.

Then load positive connects via a large breaker or switch rated for the DC voltage and current you're dealing with on the load.

(negative of load and battery always connected to controller input)

Turn the load on only when you're in pay-out mode, and only after it's paying out do you shut off the battery.

Remember that if the payout regen voltage drops below the controller's LVC (motor slows down enough), the controller is going to shut down, and it may (probably will) come up in a fault mode because it's still got the brake engaged, when the voltage comes back up above the LVC (motor spins fast enough again). I don't know what effect that will have on the object being paid out, but to fix it you'd have to disable the braking then power cycle the controller, which is going to take at least a few seconds to do the steps.


The only way to avoid the potential for cutout is to always have a battery on there, but if you do, the battery will be draining into the load. There's no setup on a common-bus controller (diodes/etc) that would allow it to remain connected and not still pass current into the load the whole time, even when the braking current is also draining into the load.



As for controlling the amount of braking, there are some controllers with "proportional" braking, where either an analog brake lever or the throttle or some other analog-output control tells the controller how much regen to apply, so the braking force can be varied.


You can't safely use an external PWM control on the load because as it turns off each time, the voltage on the controller will spike, potentially destroying the FETs in the controller. If there's enough capacitance on the battery lines it might work, but I doubt you can fit that much inside the controller, and you'd have to experiment with how much is needed (meaning, keep adding it until FETs stop blowing up, or just add lots and lots more than you think it'd need, and see if it works). Capacitance there causes another problem, which is that the battery has to fill those up every time you connect it if the contorller was unpowered. If there's no precharge circuit then really high capacitance causes high current flow for a relatively long time, and you can get arcing at the connectors as you plug it in or switch it on, even welding the contacts together if it's bad enough. If there is a precharge circuit, it complicates the process of "rebooting" the controller because it now takes up to however long the precharge is setup for to reconnect the battery.



I dont know if it helps, but Bazaki here on ES (and others I can't recall) have windsurfing/wakeboarding winch setups posted around somewhere.

 

[KrisH]

It certainly isn't a requirement to fluidly switch between modes, but it would definitely be a plus if reasonably possible. But yes, I could live with a non-automated switch if the alternative is overly complicated.

Your comments on the PWM were something I feared. Fortunately space is not at a premium so I can certainly add in extra capacitance, though since I'm looking at primarily COTS controllers I would likely end up with an external switching circuit. Ideally I could find a controller in which the "brains" were powered with an auxiliary power supply that could be always on, but I'm not sure such a setup exists. This brings me back to an active FET load; would operating a bank of high power FETs with a good heat sink and using them in a voltage-controlled current source circuit solve the arcing problem? Is it wildly unrealistic to expect to be able to dissipate these kinds of loads?

In order to deal with the battery on/off issue, would it be in the realm of possibility to use a hall-effect current sensor and disconnect the battery/connect the active load when a sufficiently large back-current is measured? I'm envisioning a flow that works as such:

1) Battery is connected as normal
2) Regen current exceeds some threshold, activating a hysteresis circuit
3) Hysteresis circuit toggles Relay1, connecting active load to the common bus
4) Some finite but non zero time later, Relay 2 disconnects battery from common bus
5) Eventually hysteresis circuit drops out
6) Relay2 reconnects battery to common bus and Relay1 removes active load (possibly with a time delay, or different hysteresis threshold?)

 

[MadRhino]

I believe that your specific application would be best served by the design of a custom controller. It could even be made automatic.

 

[amberwolf]

Ideally I could find a controller in which the "brains" were powered with an auxiliary power supply that could be always on, but I'm not sure such a setup exists.

It didn't occur to me until just now, but actually most ebike type controllers *do* have this ability--the "ignition" or "keyswitch" wire is the battery supply to the low-voltage-power-supply that feeds the MCU and FET gate controls.

The main battery wires go to the FETs and capacitors.

So...you could switch just hte main battery positive input wire off from the battery itself, and leave the "ignition / keyswitch" wire connected to the battery positive.

This brings me back to an active FET load; would operating a bank of high power FETs with a good heat sink and using them in a voltage-controlled current source circuit solve the arcing problem? Is it wildly unrealistic to expect to be able to dissipate these kinds of loads?

No, it's been done with various kinds of cell dischargers and test equipment, so it's certainly possible, but getting it right might cost at least a few sets of FETs (and possibly gate drivers, PCBs, etc) in development.

Keep in mind that using them linearly will require serious heatsinking and active cooling, just like high-power amplifiers.

If you're PWMing them, it's different, but at that point it gets complex enough you might as well use existing controller technology (powerstage development is often the hardest part--there's a number of controller development threads that might help with that if you want to go that route.

For simpler non-PWM'd stuff, you might look at Fechter's precharge / switch device thread (should be on the first page of whichever forum it's in as it's presently active) for stuff that might help see some development issues.

Methods also had a precharge / switch thread, though it's been a few years since it was active so you'd hve to look thru his threads for it. I don't recall if it had info you'd be able to use though.

In order to deal with the battery on/off issue, would it be in the realm of possibility to use a hall-effect current sensor and disconnect the battery/connect the active load when a sufficiently large back-current is measured? I'm envisioning a flow that works as such:

I wouldn't even bother getting that complex. Since you have to switch to braking mode anyway via the ebrake input to start regen in the first place, just use that signal as the trigger for your battery-disconnect/reconnect contactors, rather than a current sensor.

 

[KrisH]

I'm looking into all of options and trying to evaluate if I can use one (or several) large power resistors to do the same job instead of an active load. However, I'm a little unclear on what will happen with the bus voltage under variable regen conditions. Nominally I would expect the back EMF voltage to be linearly proportional to the motor speed, but my reading seems to suggest that the motor will act as a boost amplifier and bring the bus voltage to some higher value.

If a power resistor is connected across the DC terminals (with no battery in the loop other than to power the controller), will the motor try to push the demanded regen current through and the bus voltage will be dictated by ohms law, or will it be a relatively steady-state bus voltage at a given RPM and the power resistor will end up sinking a speed-dependent amount of current, regardless of the demanded regen? I'm hoping that the former case is true and thus I can get by with an appropriately specified power resistor power and resistance rating, with the compromise that I may not be able to generate the specified current under some small RPM.