too long battery wires will kill controllers

Joined
Dec 20, 2008
Messages
116
Location
near Nijmegen, Netherlands
Battery wire inductance is the main cause of problem, not wire resistance. The wire-inductance causes spikes on the power input. Long motor wires however is not a problem. If you have to use long battery wires, install extra caps on the brushless controller. Reasons, manufacturer statements, rules of thumb, explanation, methods, manuals, suggestions for installing:
too long battery wires will kill ESC over time: precautions, solutions & workarounds - RCG
Contents
  • Solution I & II, rules of thumb
  • Problem
  • Capacitor type & polarity (orientation!)
  • How & where (not) to add extra capacitors
  • DIY capacitor pack pictures
  • Capacitor & pack suppliers
  • Expert/manufacturer opinions, they all say the same ... & their rules of thumb.
  • Explanation, water hammer/knock analogy, theory, references, measurements
All RC controllers with block or trapezoid commutation need extra caps, not only Castle. Also Schulze, Yge, Jeti, Shock, Markus, Scorpion etc. and the Chinese OEM manufacturers. That leaves us one controller that doesn't need extra caps, SLS controllers .
You guessed it, that controller uses sinussoidal commutation, therefore less/no harmonic frequencies. They tested the SLS-60-100 for one hour with a 70m (seventy meter) battery cable at 1500Watt load. One of the developers, Rolf Zimmermann, is user RogerZ at RCGroups. Some people in Japan are working on a RC sinussoidal controller too.

More discussion, also about cooling the input caps:
viewtopic.php?f=28&t=7504&start=15#p120502]
 
ron, thanks for highlighting the issue!

If I might add, the inductance is a function of loop area which is itself a function of cable length. I mention this because, while it's a good idea (and fairly intuitive) to minimize the distance between the controller and battery pack, it's important to control the lead separation as well, by either taping or zip-tieing the + and - run together. Particular attention should be paid to the cables around the battery itself - because due to the size of the battery, the loop the current travels is inherently large, and wherever possible, the leads should be run parallel and tied together.
What the inducance does is resist the sharp on/off demand for current that the controller requires. So while a controller is intended to feed sharp bursts of current to the motor, the current coming from the battery (fed through inductive cable runs) will actually be slightly rounded off, and 'sluggish'.
The large capacitors (bulk decoupling caps) on the controller are placed there (as close as possible to the FETs) to make up for the large inductance of the cable runs. In fact, if it weren't for the inductance, those caps wouldn't actually be necessary. While the sluggish current is making its way from the battery to the FETs, the capacitor is providing the 'sharp' current waveform. It follows that the more inductance there is in the line, the more current will be needed from the capacitors. As the current demand from the capacitors increases, more power will be dissipated in the capacitors due to the ESR (equivalent series resistance) in the cap. This is why the cap needs to have as low an ESR as possible. Too much heat can blow a cap (frying the controller), or at the very least reduce the useful life of the caps.
It's been discussed that in order to reduce the spark or arc when the battery pack is first plugged into the controller due to the inrush current needed to initially charge those decoupling caps, that a ferrite ring can be added to the power source leads. Though this will reduce the arc by reducing the inrush current, this increases the inductance in the line and increases the demand on those capacitors (reducing the life span, or increasing the risk of a fried controller). One option around this is a clip-on ferrite ring, that can be removed before the motor actually starts to turn. Also two other options are a soft-start circuit or a properly sized (not too big, not too small) bulk decoupling strategy.
The nice thing about sinusoidal commutation is that, rather than sharp on/off switching of conventional controllers (which have sharp current demands) the sinusoidal waveform is nice and smooth - so there isn't any significant high frequency (read sharp) current demand and inductance doesn't really become as much of a concern.
 
My battery cables are 4 feet long, :shock: 10-ga wire. Seems to work fine. But not for long?

So... now what do i do :?
 
Every bike will have some amount of inductance, no matter what you try and do. So the objective isn't to eliminate the inductance, but simply to minimize it. If your setup needs a 4 foot long length, then that is ok, but it will reduce the operating life of the caps. That impact might not be a really significant amount if the caps are good low ESR types. Otherwise, reduce the length if you can, or at the very least make sure the to lines (+ and -) always run parallel to eachother and don't separate forming a large loop.
 
I heard that twisting the wires together lowers inductance more than running them parallel, is that true? Is this something I should think of doing with my ECrazyman controller?
 
All goods points about the inductance issue!

Just one little thing though, smithinparis:

smithinparis said:
[...]The nice thing about sinusoidal commutation is that, rather than sharp on/off switching of conventional controllers (which have sharp current demands) the sinusoidal waveform is nice and smooth - so there isn't any significant high frequency (read sharp) current demand and inductance doesn't really become as much of a concern.
ALL switch mode controllers have sharp on/off pulses on the input. The difference between the sine variants such as the SinusLeistungsSteller (what fun to pronounce!) and regular old trapezoidal commutation controllers is more in the output than the input. Outputing a more appropriate waveform to what the motor wants (sine-like) results in the controller's peak input current spikes being of smaller amplitude and more uniform, but certainly not eliminated. It's still using pulse modulation after all...

BTW, good collection of info and links, Ron Van!
Pat
 
Since 2008, I added more info, and changed the layout.
too long battery wires will kill ESC over time: precautions, solutions & workarounds - RCG
Contents
  • Solution I & II, rules of thumb
  • Problem
  • Capacitor type & polarity (orientation!)
  • How & where (not) to add extra capacitors
  • DIY capacitor pack pictures
  • Capacitor & pack suppliers
  • Expert/manufacturer opinions, they all say the same ... & their rules of thumb.
  • Explanation, water hammer/knock analogy, theory, references, measurements
 
Surely, our controllers already have big capacitors directly across the battery wires. What's the problem. I can see a problem with R/C ESCs, but is it something for us to worry about?
 
d8veh said:
Surely, our controllers already have big capacitors directly across the battery wires. What's the problem. I can see a problem with R/C ESCs, but is it something for us to worry about?

i run about 1m of wire and in the last 10 years only killed 3 controllers
1 was running 48v on a 24v controller
2 got run over when the bracket holding it on snapped
3 got rattled so bad that the resistors started falling of the board lol
 
This, as with most rules of thumb, is a complex question. Inductance in the battery wiring, as well as the battery current and the FET switching speed as well as the battery voltage drives the voltage transients on the FETs. The capacitors only partially compensate for this, and there is a tendency to save money by using caps that don't have good properties. Ultimately this puts a stress on the controller. Controllers are also frequently designed or assembled with inadequate margins on the components, another way to save money.

Keeping the battery leads short and tightly paired maintains the safety margins.

Adding length to the motor wiring (which is already very inductive) is a much less stressful approach.
 
Very interesting discussion although mostly over my head. I have installed four Bafang BBS02 crank drives, one on a friends tadpole, one on my own tadpole, one on my own delta trike, and one on my sister's cruiser bike. All are used for pleasure riding with top motorized speed limited to 20MPH.

All the recumbent trikes have the motor/potted controller units near or at the front and the batteries at the rear, so five or six feet away. The bike battery is only a couple of feet away.

The only controller failures have occurred on the bike, one after a year and the other one after just a week. In both cases the potted controller was replaced since there is no real access to the components (one under warrantee), and no additional problems for over two years.

So my questions:

Should I twist the long runs together or just zip tie them together?

Is it worth my effort to remove a few inches of wire that was left for convenience when removing and installing batteries?

Will more harm result from sustained 20MPH motorized cruising as opposed to 10MPH pleasure riding with light pedal assist?

Thanks,
 
craigsj said:
For those that need to sound the alarm over power wire inductance in the lengths used here, I'd suggest:

1) characterizing the inductance, measuring and reporting the value, of such a wire over a typical length
2) demonstrate the claimed reduction in that inductance with tight pairing of wires (that'll be a good one), report methodology
3) show the threatening voltage transients at the FETs using a scope
4) prove the value of the recommended 220 uF of extra ESC capacitance per 4 inches of power wire

It's fine to talk about issues in an absolute sense, and many of these comments are true in an absolute sense, but it's also important to understand the magnitude of these effects and there seems to be a gross lack of appreciation for orders of magnitude here. Real data is called for. PWM frequencies, remember, are in the 10KHz range, that is slow.

P.S. For those convinced of the importance of tightly paired power wires, you should read up on transmission lines, linear inductance, linear capacitance, characteristic impedance, and why such things are important. Then hopefully you'll see why, at 10 KHz, this is a fool's errand. It's analogous to saying skin effect is a crucial consideration as some do.


Even mm matter when you get to real power.

Wires are like a swinging whip with respect to swinging a magnetic field around them, you can pick however much burden you want for your caps and how much you want to derate the DC bus voltage to protect the caps.
 
craigsj said:
Great, then producing quantifiable evidence should be easy. You'd think in the 10 years this thread has existed someone has done a measurement. I feel people are better served with data, not FUD.

While I agree with the sentiment of your post, I think the tone could use a little work. Especially with L4P, it's been a while since i've been active, but he had contributed an absolute ton to the knowledge pool for ebikes when I was active. You're just as much able to do these measurements as others.
 
Arlo's threads on building a controller contain a lot of good information on this, both in terms of actual data (with scope traces) and in terms of advice from experts who design and build this gear professionally. There are also many other excellent controller design threads. The spike that even a few feet of wire makes when significant currents are interrupted is significant, and the energy content in these spikes rises quickly as the wire is lengthened. The effective series inductance and resistance of the commonly used (cheap) capacitors in controllers are not very good, so they are only partially effective at protecting against these spikes. The resulting ripple heats these capacitors, shortening their life, and as they age their ability to protect is reduced, eventually leading to controller failure. The FETs have very limited capability for absorbing these spikes. When that is exceeded they fail dramatically. We've seen plenty of failed FETs for several reasons, and this is one of them.

But the average DIY ebike maker is not too interested in the engineering details. This thread presents "rule of thumb" information. Good practices to follow.

The bottom line is to make the battery side cables short and paired. All cabling should be minimal in length, but if additional length is needed, put it on the motor side, and again, the cables should be kept together. These are simple things to do that reduce the transients the FETs see, which is good for controller health. They become more important as the current and/or voltage levels are increased.

The low power crowd can probably ignore these guidelines and not suffer much. But why not follow good practices, even at low power, since it is easy to do?
 
Admittedly, these issues are most concerning for RC motors and controllers. A long time ago, I had an ebike with an ESC-style controller and a RC motor (an outrunner, likely with lots of inductance). The power cable from the battery to the ESC was only about a foot long.

It seemed to run fine without any mods, but when I added two low-ESR capacitors (the most frequently recommended ones), the voltage ripple went down from 4V to 1V.

I never measured the voltage spikes, but...I was using 36V on an ESC that was capable of running 50V max, so I had plenty of headroom. There are many stories of RC drives using 12S, and at 4.2V per cell, the battery pack was at 50.4V with an ESC that had 50V capacitors. There is apparently a safety margin there, but 4V of voltage ripple would place the spikes at 54-ish volts?

Some of them blew up, and some of them didn't. A Castle Creations HV-160 ESC is more than $200.

Jeremy Harris doesn't post much anymore, but...he wrote a detailed and thoughtful thread on the differences between RC ESC's and common ebike controllers.
 
craigsj said:
tostino said:
While I agree with the sentiment of your post, I think the tone could use a little work. Especially with L4P, it's been a while since i've been active, but he had contributed an absolute ton to the knowledge pool for ebikes when I was active. You're just as much able to do these measurements as others.
The burden of proof is not on me. What can be asserted without evidence can be dismissed without evidence. I don't need the measurements.

Don't agree on tone either. I responded to a comment that contained no useful information. No poster is immune to challenge, he no doubt could provide input but chose condescension. In my brief history here, it's the second time I've experienced this from him.

There are assertions made here that are outright false. I could name some but I'd rather "experts" that support these sacred cows defend them. You can't have dialog with closed minds.

If you want to contribute or be taken seriously, you may have to provide some real data. Otherwise readers will rightly continue to believe the known experts (who have already demonstrated expertise in this forum and elsewhere) and correctly disregard the unsubstantiated comments from unproven posters. The experts are under no burden to prove their comments in every thread.

If anyone has any doubt regarding transients destroying FETs, a more extreme example is discussed in the DIY battery spot welding threads, where a couple of feet of wire has sufficient inductance to destroy a half dozen or more parallel FETs when turning off currents of a few hundred amps even at low voltage, if not properly managed. We never know how much margin our controllers have designed into them, so adding to the transient voltage burden is always a concern. Also keep in mind that the controller FETs are actually switching motor phase current, which (at low speed) is many times larger than the battery current, so they are subject to a lot of stress, and they do it a lot more frequently than a battery welder, in addition to operating at much higher voltages and temperatures.
 
Fortunately, the controllers need none of your understanding operate, and will continue to operate in a mode with much less cap stress on short battery side leads vs phase leads for those who are seeking performance.
 
Alan B said:
If anyone has any doubt regarding transients destroying FETs, a more extreme example is discussed in the DIY battery spot welding threads, where a couple of feet of wire has sufficient inductance to destroy a half dozen or more parallel FETs when turning off currents of a few hundred amps even at low voltage, if not properly managed. We never know how much margin our controllers have designed into them, so adding to the transient voltage burden is always a concern. Also keep in mind that the controller FETs are actually switching motor phase current, which (at low speed) is many times larger than the battery current, so they are subject to a lot of stress, and they do it a lot more frequently than a battery welder, in addition to operating at much higher voltages and temperatures.
You are mixing up things here.

The reason spotwelders have problems is that they are wired so that they are suddenly shutting off a current with a single ended switch. That's how you generate a large voltage with an inductor - put a lot of current through it and stop it suddenly.

In inverters, there are a series of complementary switches to V+ and return (usually 3.) This means that when you suddenly cut off current via one side of the switch, the flyback current has a place to go - the body diode of the other switch. Some inverters rely on this path, but better inverters actively drive one of the two switches to reduce diode losses (which can cause the FETs to heat up.) BTW this is why you start getting regen over base speed even if you don't have a regen controller - those FETs start acting like rectifiers, and your motor becomes a generator even if you are not driving the FETs.

So the two problems don't have much to do with each other.

That's not to say that there is no benefit to be gained from snubbers, of course. Reducing HF energy content reduces EMI and reduces false triggering/bad A/D readings from easily coupled HF noise.
 
billvon said:
Alan B said:
If anyone has any doubt regarding transients destroying FETs, a more extreme example is discussed in the DIY battery spot welding threads, where a couple of feet of wire has sufficient inductance to destroy a half dozen or more parallel FETs when turning off currents of a few hundred amps even at low voltage, if not properly managed. We never know how much margin our controllers have designed into them, so adding to the transient voltage burden is always a concern. Also keep in mind that the controller FETs are actually switching motor phase current, which (at low speed) is many times larger than the battery current, so they are subject to a lot of stress, and they do it a lot more frequently than a battery welder, in addition to operating at much higher voltages and temperatures.
You are mixing up things here.

The reason spotwelders have problems is that they are wired so that they are suddenly shutting off a current with a single ended switch. That's how you generate a large voltage with an inductor - put a lot of current through it and stop it suddenly.

In inverters, there are a series of complementary switches to V+ and return (usually 3.) This means that when you suddenly cut off current via one side of the switch, the flyback current has a place to go - the body diode of the other switch. Some inverters rely on this path, but better inverters actively drive one of the two switches to reduce diode losses (which can cause the FETs to heat up.) BTW this is why you start getting regen over base speed even if you don't have a regen controller - those FETs start acting like rectifiers, and your motor becomes a generator even if you are not driving the FETs.

So the two problems don't have much to do with each other.

That's not to say that there is no benefit to be gained from snubbers, of course. Reducing HF energy content reduces EMI and reduces false triggering/bad A/D readings from easily coupled HF noise.

The example serves to show there is considerable energy developed by changing current in the inductance of even short cabling. Circuit configuration is different, but the motor controller's body diodes do not absorb voltage transients and ripple from battery circuit inductance.

The body diodes of the FETs capture the flyback from the motor inductance (that's why we can add wire on the motor side with little concern), and handle it nicely. This is not the case for the energy in the battery side inductance, which adds to the battery voltage and increases the system voltage when the battery current is reduced. The FET body diodes provide no help here. This ripple is moderated by the controller capacitors, which are only partially effective due to their resistance and inductance which limits their effective capacitance. The resulting ripple current heats the capacitors (which are very poor heat dissipators) and threatens the FET voltage margins and adds noise to the system. The more cable inductance, the more effective capacitance is needed. Hence the rule of thumb in this thread, and the experience of folks who have damaged controllers in this way.
 
Back
Top