General ESC Theory Discussion
 

I thought I'd start a thread on exactly how ESCs work in general. Some of this is to clear things up for me, and some to help others understand how they operate.

From what I understand;

• ESCs use PWM to send varying levels of power to the motor coils. Greater pulse width means more power.
  
  
• ESCs use FETs for a few reasons:
  
  
  1. FETs can easily be paralleled because they have a positive temperature coefficient (which means they conduct less as they heat up) preventing any one FETs in a parallel set going into thermal runaway - a sort of self-regulation. Regular bipolar transistors have a negative temperature coefficient, which means that if several are paralleled, one will conduct more than the others, which heats it up more, which makes it conduct more (etc) until it blows leaving you with one less device to handle the current.
     
     
  2. FETs have a lower "on" resistance than transistors, which means they drop less voltage, which means less heat generated. And paralleling FETs further reduces total "on" resistance for less heat.
     
     
  3. FETs drive inputs are high impedance, so you don't have to worry about current gain and any loading down the drive signal at high currents like you would with plain bipolar transistors. This makes FETs ideal to drive with logic signals.
  
  
• Heat generated by an ESC is caused by two things:
  
  
  1. The voltage drop caused by the on resistance (Rdson) value of the FET.
     
     
  2. Slew rate, which is the time it takes to go from 0v to max voltage. Faster slew rate means that it will not have a voltage drop for as long of a time (if that makes any sense).
  
  
• By looking at several different ESCs, they always use a multiple of 6 FETs (2 FETs times 3 phases). So, does that mean these use the classic H bridge configuration? If so, then an ESC like the MM with 36 FETs means that power is flowing through 12 FETs at once - battery power through 6 FETs and ground through 6 FETs. Correct?

One of the things I don't understand is why using the motor for braking is hard on controllers and batteries. For the induced voltage from the motor to even get to the battery, that would mean all the phases get turned on at once, right? So, why not simply short all phases to ground instead? The controller would still have to be able to shunt that power, but the motor would be taking the majority of the abuse.

So, feel free to comment or correct any of above statements, or add more information. :smile:

Not all controllers brake in the same manner. Castle controllers use the motor as a generator and recharge the battery to slow the vehicle down. It generates less heat than trying to slow the motor down in other ways.

From what I know, a motor always generates some sort of energy when it is used as a brake. If johnrobholmes is correct, it seems that Castle takes the energy and puts it back into the battery (which I actually don't think is right, because this on a freshly charged Lipo might damage it - I think it just puts it into the caps or something), then it means the ESC takes the grunt to save the motor. Other controllers might make the motor take the grunt to save the ESC (by shorting out all the motor leads).

Brain, regardless of how many Fets take the current under braking, if the braking causes voltage to increase X amount over the current input voltage from the battery, and X is too much for the Fets, the Fets pop. Having 1000 Fets each rated for 25V, all 1000 will pop if you put 30V through them (I'm saying this for exemplary purposes).

Personally, I don't think it would be a bad idea to have some sort of super capacitor attached to the ESC that would absorb all the energy created under braking. This energy can then be put back into the ESC when throttle comes back on. Obviously, going 40-10mph is going to generate a good bit of energy, much more than a few ~300-400uf capacitors will take. Maybe if they had a few 10000uf capacitors in parallel (even if they were a bit large compared to the average cap) I'm sure it would help quite a bit.

One thing I DON'T understand about ESC's is why adding capacitors help with keeping temps down. From what I understand, it just helps keep the input voltage more constant (higher, when you are on the throttle). Wouldn't this cause the ESC to work harder, and thus run hotter?

Looking at the power board of the MM (in my hand presently, it does look like it is using 12 fets per leg, wether it uses-switches them all at once, or 6 at a time alternating I'm not sure... they give the resistance for the esc, couldn't we reverse the math to figure out how many fet's are in parallel?

I think the work done by the esc when braking is that it is powering the coils out of phase to slow the motor, the inductive current isn't enough to really slow down the motor (shorted the three wires on my 5700kvMM motor, put on a 20t gear and spun it up with compressed air, not much different with the wires loose).

Not claiming to know anything for sure... I got this MM thing, made sure it worked, then tore it apart befor it ever saw the car it will end up in.

I guess the esc could just be charging the coils and keeping them like that but how would it know the car had stopped, like it does in the forward/brake/reverse setup? back emi?

First of all, I want to thank you BrianG for opening up this thread. And second, I hope this does not turn into a babbling mess like other thread on the subject! :wink:

I think we should take this slow, step by step... and Artur should definitely be here!

About motor braking.... on the one thread I made a while back (can't remember which one) we were discussing motor braking. My theory back then was it is a "PWM controlled motor short"... if that makes any sense? The FET switches between each phase, effectively shorting out each phase as it does this, and ultimately shorting out the motor... and we all know what happens when you short the motor phases and try and turn the rotor; it 'resists' moving. Obviously, this resistance to movement is because it is acting like the generator that it is, equal but opposite reaction in the coils (as voltage is generated) makes the motor slow down when this happens. So, now emagine this in your BL Revo truck, for instance, you are going at speed, you slam on the brakes, and the truck slows very fast.... and depending how much reverse throttle you gave it, it slowed at a variable rate. The esc is basically controlling the braking action by the PWM ('Pulse Width Modulation', for those that don't know) the FET's are conducting between each phase... the wider the pulse, the more braking effect you have. So, at what frequency does this PWM operate at? I am not totally sure, but I would have reasonable assumption that it is the same frquency the esc generally operates at (40KHz for the Quark, example).

As for the regenerative part, we have to get Artur on this, but he said it has something to do with the voltage being higher than the diode.. then ultimately higher than the battery voltage.


EDIT: I forgot to mention... that in the act of motor braking, most, if not all the energy is dissipated by the motor as HEAT.

First off, I would like to say Welcome back JohnRobHolmes! I always enjoyed discussing things with you. I'm glad to see you back.

From what I gather, there are two types of speed controller motor braking.

1. Regenerative Braking. The AC (Brushless Drive Motor) Forklifts at work use this technology. It does give a short recharge burst back to the battery thus increasing runtime.

2. Shunted Braking. We have two type of forklifts. The DC trucks shunt the energy coming back from the motor while coasting.

One statement on the group of 36 FETS. That would be 2/3 the total in use at any given time. 24 would be energized at once. Two legs out of the 3 are always hot under power.

My 2 cents.

The Castle controllers use regenerative braking, even with lipo packs. The recharge rate is about 15% effective, so whatever amp hour you use to accelerate the vehicle will only be recovered 15% on a full stop. This doesn't put a lipo pack in danger. This info is straight from Pat's mouth, I talked with him last week for about 30 minutes on the subject. In his opinion it is the most effective and efficient way to brake a motor, as the battery takes the load from the motor and the ESC doesn't really have to do much work.

that slowing down/braking is discussed quite some tim ago. I wrote that it used the low internal resistance of the batteries to slow the car down. I remember i didn't made friends with that post.

almost everybody seemed to disagree.

another thing about sensorless controllers; the way they start varries;

I read an article on a DIY brushless controller that the controller starts at a lower set sampling frequency in order to start the motor, once it sences its EMF it changes the sampling frequency and takes it to another level. Interesting stuff; the sampling frequency would affect the timing?

A theory of myself;

Some controllers just put out a certain RPM (fixed) to start the motor (MGM/Mambamaxx) there is no way the motor will turn bellow a certain RPM; they simple use this to reduce cogging. It doesn't look at EMF, it just starts at ~120rpm or so.

i think this is done, because the EMF is hard to measure for controllers on lower RPM, and the electric magnetic field might jump over the actual rotorspeed (this results into cogging)

There is another thing that affects the heat of the esc Brian;

the quality of the sign-wave. If this is an incorrect signal, with saw-tooth effect on high mark of the signal, this results into heat as well. The FETT tries to follow the steering signal (which is obviously amplified as well)

Regarding braking, different ESCs will likely apply different techniques, but I'm still not convinced any RC ESCs on the market now are capable of recharging the main pack during braking.

I also should mention that theory and debate is great if you're proverbially Greek, but the Roman side of me would really like to see experimentation and proof behind some of the theories. ;)

Watching with interest...

Wow, more responses than I would have thought!

So, it looks like the two braking methods are shunting all windings to ground (or to each other - same thing really), or to apply the back-EMF pulse to the battery to help recharge the cells somewhat. I can understand both braking theories, but don't agree on how they are implemented. Regenerative braking works fine for NiMH (or lead-acid in things like forklifts, cars, etc) or maybe even A123 cells, but lipos are more sensitive to overvoltage. It would be nice to have the ESC smart enough to automatically disable regenerative braking if a lipo mode and/or a LVC value selected since those settings are usually set only for lipos.

@BP-Revo: Yeah, I know that an FET will pop if subjected to too much current. What I was trying to get across is that one FET won't "current-hog" when paralleled like bipolar transistors do because of the way they conduct.

Well, this is something I'm not sure about. In a 36 FET ESC, there are 12 FETs per phase. I think we can agree on that. If in fact these ESCs are set up in an H bridge config, 6 of those would go from a phase to battery voltage, the other 6 would go from a phase to ground. This is so the ESC can reverse the polarity for reverse. At any one time, only two phases are "on"; one phase is connected to +V through its 6 "battery" FETs, and the other phase is connected to ground through its 6 "ground" FETs. So, current flows though 12 FETs in all, but through two sets of 6 FETs effectively in series. Wouldn't this mean that only 1/3 of the FETs are on at any one time?

@Serum: Yeah, I remember some discussion about this a while back, but couldn't find it using the search. It's either probably buried in an unrelated thread or my poor search phrase selection. :smile: I have no idea about the frequency for starting vs running, so I can't comment on that.

Programming/firmware DOES make a big difference in how an ESC operates. A while back, I posted a thread comparing how a MM works vs a Quark.

I guess I'm not quite following what you mean by the "saw-tooth" part of the signal. Do you mean the current/voltage spikes from driving an inductive load?

Hopefully GriffinRU pipes up here at some point...

On the 1/3, 2/3rd's total fet count thing. You're absolutely correct Brian. For some reason, I completely forgot about reverse. Sorry, Brainfart! LOL

Well, Artur says if the voltage coming off the motor from hard enough braking is high enough, and high enough to jump some 'diode' in the FET, then it charges the battery, I don't know if all of the energy at that point does, or if just the 'excess' does?

But I think it is worth noting, when I had my truck geared for roughly 50 mph... in my eagle tree data from that run I could see (if I was not mistaken) a blip of voltage spike on the battery (something like 0.25 volts) after a long speed straight, and after the voltage had immediately rose back up to no load... and it wasn't from applying throttle either. Maybe it was nothing, maybe it was a regenerative action? I don't know.

I was running 7s2p A123, 1512/3D (1700kv) and geared 18/51 with badlands (ballooning at that speed bad)... and this was with my 'super Quark'.

Something else I've been thinking about...

A motor of a given Kv will turn at X RPM for Y voltage across its terminals. The same motor would generate approximately Y voltage across its terminals if its stator is turned at X RPM by an outside force.

A battery will only draw current and store charge if a voltage potential greater than its own is connected across its terminals. Regenerative braking from a drive motor would require one or more of the following to recharge the battery under braking:

1. Accelerating the vehicle beyond the top speed at which the motor's RPM would be at its no load RPM given its Kv rating and pack voltage. ie. drive down a very very steep hill, maybe.

2. Have the ESC step up the output voltage of the motor under braking.

3. A drive system that mechanically regears the motor under a braking condition so that it spins faster.

I don't know how full scale regen braking works to recharge the battery, but I would guess (2) is the most practical...

I'd like to make an offer, I have a real nice dual channel oscilliscope, nice multimeter, bench vice, optical tachometer ,and compressed air... I also have a MM5700kv waiting on me to finish the rest of the car. I would be willing to do a few experements to help figure the braking/regen thing out. Give me a setup and what we are looking for and I'll find some time to get some numbers... (take into account the esc is down for mods)

I'm not sure that shunting the motor alone causes enough drag to stop the car like it does. I don't mean to be a sceptic or to challenge you gentelmen who are "in the know", but I like to have a full understanding of anything for which I claim to have knowledge.

@Aragon: I see what you are saying. However, a coil can deliver more voltage than what was put into it if the magnetic field is allowed to collapse faster than it was built up. Take any coil, apply a 9v battery on it, and then remove the battery. Depending on the coil rating (in milli-henries), a voltage MUCH MUCH higher can be generated. This is how the cow fences I've seen operate. Applied to R/C; depending on how fast you brake, the field can collapse faster than it was built up and can generate higher voltages than the battery.

@DrKnow65: I did an experiment a while back to test braking force of a motor with shorted phases. At low rpm (turning by hand), there IS resistance, but not seemingly enough to stop a vehicle. But spin that sucker as fast as we do (~30k rpm) and the braking force is high indeed! At 0 rpm, I would think that the ESC actually locks (or "pulse locks") the phases to keep the vehicle from rolling from a standstill since there is no braking force from a motor not spinning at all or very slowly.

Using an oscilloscope to measure 3 phase motors can be a bit tricky since there is no "common" point. Too bad wye motors don't have the common connection point accessible to the outside so you could use that as the common point. With delta motors, there is no common point at all other than what the ESC determines as common at any particular time. The best you could do is use CH1 on one phase and CH2 on another phase and use the delta function on the scope. This procedure is used for floating ground circuits or to measure signal across a specific device. Otherwise, using the scope's actual ground lead could short stuff out depending on the circuit design. Probably not a big deal for R/C use since the vehicle ground is totally isolated from the mains ground. Also, it might be hard to get a trigger since pulsewidth and even frequency may vary with throttle and load. Internal triggering uses the 60Hz wall frequency so that won't work. External trigger is what must be used for something like this. That said, you could still get a decent measurement on one phase using a dual-trace scope and just assume the other phases will be similar, just shifted in phase 120*.

Do you have more info on this? Or a better explanation? :)

Aragon, like a coil circuit in a car to fire the spark plugs, 12volts changed into 40,000volts.

I see your point BrianG, and knowing that you have tested this makes me feel better- ideas are good, but not good like tested ideas :) I knew there had to be some charging of the coils during braking and it makes good sence that this would only happen when the rpm drops below where the inductive charge is no longer doing the work of slowing the motor.

You can do a Google search on inductor formulas, inductive kickback, back-EMF, etc to find all kinds of information about this phenomenon.

When you apply a voltage to a coil, a magnetic field will build up over time. The time it takes for the coil to magnetically saturate depends on the coil rating (henry value), core type (air vs iron), winding resistance, and voltage supplied. This built-up field is actually "storing" a charge, somewhat like a capacitor does. When you remove the battery from the coil, the field collapses instantly generating a voltage higher than the one applied because the collapse is quicker than the build-up. This voltage generated is opposite in polarity than the one applied because of the direction of the magnetic field motion cutting the coil windings. Incidentally, many times in circuits that energize a relay, you'll see a diode placed across the coil in reverse polarity, which shunts the back-EMF generated by the relay coil to prevent the V spike from damaging connected circuitry.

Ok cool, that makes sense, but how will one control the speed at which the magnetic field builds or collapses in a motor winding if the magnets are moving over it at a set speed?

Well, that's what the FETs do. If you totally short the windings, it will be full brake. If you short the windings in pulses, you can control the intensity.

From what I understand of coils, the amount of voltage produced by a coil is proportional to the rate of change of magnetic flux. The magnets moving over the motor windings cause magnetic flux around the windings which creates a voltage potential between their terminals, the magnitude of which is proportional to the speed of the magnets' motion over the windings. (ie. the speed at which the motor is spinning)

I don't understand what you suggest though - How will the FETs pulsing a short between the windings change the rate of magnetic flux induced by the magnets?

It will switch the coils on and off at a fast rate to regulate the amount of braking force. This will load the coils down when on and unload them when off. Ever notice in a 1:1 car how engine rpms drop when there is a large load on the alternator? Same thing here.

Yup, I understand how it would regulate braking force. What I don't understand is how it would increase the rate of magnetic flux generated by the magnets, and consequently increase the voltage potential across the motor windings enough for current to flow back into the battery pack...

Well, if the vehicle were moving fast enough (high motor rpms) and then you brake hard allowing the field to collapse very quickly, the induced back EMF could be greater than the battery voltage.

Think average voltage, if you have a longer Pulse width when switching the short for braking, then you have a higher average voltage than a shorter pulse width. Correct me if I mistaken what you are talking about?

Perhaps, but that possibility could only last for a split second then - the time taken for the ESC to stop supplying power to the windings. It would be ideal if as much of the car's kinetic energy transfered during braking as possible could be converted back into electrical energy that charged the pack...

Mmm, what I'm ultimately talking about is converting kinetic energy into electrical energy and storing it in the battery pack. An RC car in motion is in a state of stored kinetic energy, the magnitude of which is proportional to its momentum, and the source of which ultimately came from the electrical energy that the system took from the battery to accelerate the car to whatever velocity. If the object's motion is stopped, its kinetic energy is transferred elsewhere. The key is to transfer as much of that energy back into the battery pack.

But the problem as I see it is as I outlined in post #14. The voltage generated by the motor needs to be stepped up somehow in order to charge the battery...

I just did a little test just to see how it went. I ran my DPR in live mode and hooked it up to my MBX5T that is now equiped with a Schulze 18.97 and a 7XL motor. I gave it some throttle and let it go so it would freewheel and the graph stayed flat. I then gave it some throttle and hit the brakes and saw a little spike on the graph. Just to make sure I hooked the DPR to my BPP truck that has a Schulze 40.160 and Neu 1521/1Y. Gave it some throttle and let it freewheel and the graph stayed flat. Then I gave it some throttle and hit the brakes, same thing happened a little spike on the graph.

it was matthias schulze who told me they used the battery trick for braking.

That's great. Are you able to see the direction of current flow in the graph? The DPR records in both directions, so that spike could be current drawn from the battery or current drawn by the battery...

The spike was upwards so it actually went to the battery.

Excellent! Do you plan to test it in driving? Braking at varying levels downhill would be interesting to see.

How much does the voltage spike?

I think my brain just exploded..... please, continue- this is extremely interesting stuff.

Today I actually got everything back together. The rear diff on the BPP let go and I had an issue with my Aveox in the MBX5T. So I will get some data on Monday if the weather is good. I didn't notice what + volts the spikes were. I just noticed they were there. I'm going to look in my previous data to see if it is there and I'll "prt sc" on a close up.

So, you are seeing what I saw then, as far as a voltage spike to the battery? I saw, IIRC, about 0.25 - 0.5 volts spike, so even a little potential over the battery packs voltage will result in decent current levels going into the pack.

I love being corrected, that means I'm a little bit smarter now :)

I will have to agree. The spikes seems like 0.25-0.5. I also noticed a sound coming from the ESC (Schulze 18.97). I listened closely to figure out where it was coming from and it was from the capacitors. It sounded almost like when you take a picture and the capacitor inside the camera recharges after the flash, but it was more like a blow-off valve. I have to see if the same spikes appers with the MM. If so, I think setting a little drag brake might throw some charge back into the pack.

Brian great thread!
I suggest to check this pdf before we start with Q on smoke, brakes, fets, caps, regeneration etc.

Nice PDF there! I just skimmed over it a little and it seems to have some REALLY good info! Thanks for that link!