ESCs and extended power wires: Should we add capacitors?

econfly

Member
This question came up in an ESC product review I wrote. Say you want to extend the power wires on your ESCs, maybe to mount them out on the arms under the motors. The general advice out there is to avoid this, extend motor wires if needed, but keep ESC power wires as short as possible. If absolutely necessary, extension of ESC power wires could require additional capacitors. Is this correct?

My thoughts follow. Comments and corrections are welcome.

The source of the problem is the very rapid switching that takes place in an ESC. Typically ESCs turn motors on/off anywhere from 8,000 to 16,000+ times per second. Within these brief on/off windows, the ESC adjusts the on time vs. the off time to control motor speed (see the linked review above for details). Skipping over the technical details behind the switching, it’s relatively accurate to say that an ESC’s pulsing demand for power results in voltage fluctuations on the input wires and the fluctuations are more severe the longer the length of power input wires to the ESC.

Wire length is an issue because the problem is driven by inductance on the power wires. The longer the run, the greater the inductance. The inductance resists changes to current flow and, in the presence of very rapid changes in ESC power demand, that ends up creating voltage fluctuations. By the nature of the problem, the peak voltage observed on the input lines at the ESC can be significantly higher than the battery source voltage. These voltage fluctuations can damage the ESC, with the primary point of initial failure being the capacitors themselves.

Conditional on an optimal build (more on this later), the broader layout of the multirotor (multiple ESCs, multiple power paths, etc.) does not change much about the nature of the problem. The fluctuations in voltage are most acute at the load — the ESC itself — and therefore any inherent capacitance that might be present in the build cannot directly address the issue. The only solutions are to limit the inductance and/or add capacitance at the ESC.

The best way to limit inductance on the ESC power connection is to limit the wire length, and, for any given length of connection, keep the wire pair tight (twisted or tight in parallel work about equally well). The worst outcome is a needlessly long power connection to the ESC with positive and ground wires separated.

Having limited inductance, one can add capacitors at the ESC. But whether this is really necessary is a complicated question. The answer, as best I can tell, is it depends.

There are many ways to examine this issue, and part of the problem is not directly observable. We can’t measure the extent to which the capacitor(s) on the ESC are internally fluctuating voltage. All we can observe is the voltage at the capacitor legs — i.e., the voltage at the ESC load point. As it turns out, this is does not appear to be a significant limitation.

To examine the question I have been testing the ESC I reviewed previously with various combinations of lead length to a battery and with various capacitors in place. The table below summarizes my results.

esc_length_zps52ce4a36.jpg


The values are voltage fluctuations as compared to the ESC’s stock capacitor setup (2 x 120μF Panasonics). The other capacitors in the table are what I had on hand and are all Rubycon ZLH series (a good choice for this application). For testing I removed the stock capacitors, so each item in the table reflects only the capacitors listed. The first line in the table reflects no capacitors at all (not recommended!). I tested with a battery, as opposed to a power supply, to best reflect real world experience. Voltage fluctuations were measured as AC peak-to-peak and AC RMS — both focused on deviations in voltage, ignoring the DC component. These measures presented highly correlated results, and the table is based on the more stable RMS measure.

First, note that the stock capacitors are not completely removing voltage fluctuations. We can improve things by adding capacitance, with improvement available all the way up to a huge 1200μF 63v capacitor (the largest I had).

Second, we can see that extending the power wires (either 4.5” or 9” — reasonable amounts for various arm lengths) results in higher voltage fluctuations, but we can mitigate those fluctuations by adding capacitors.

Third, compare the 6s results with no added length to the 4s results with 4.5” added length. The voltage fluctuations are nearly identical. Now one might say this is apples vs. oranges, and that would be correct. Yet, this is a major point of comparison and often ignored. True, adding length to power wires increases voltage fluctuations and puts more stress on the capacitors. But ESCs are made to operate under a variety of conditions. In particular, just about all ESCs operate over a significant voltage range and with tolerance of varying absolute deviations in voltage, including high frequency fluctuations, over that operating range. While the comparison is not perfect, there is some sense in saying that extending power wires at a given operating load is somewhat comparable to increasing the load itself if we focus solely on absolute voltage fluctuations at the ESC.

Finally, note how much extra capacitance it takes to offset added wire length. For a 4.5” length you need to roughly double the capacitance to offset the added voltage fluctuations. For an added 9” wire length the required increase in capacitance is roughly quadruple the stock amount.

So what to do? My view, based on these results, is that just working well within the rating of your ESCs is a viable approach. If you are running 4s power and want to extend the ESC wire length just get ESCs that can handle 6s. That’s not a perfect solution, but it’s defensible and arguably sufficient in reality.

The problem is most acute if you are extending line length for ESCs operating at their upper limit. In that situation, adding capacitors could be necessary. However, keep in mind that the relevant issue is total run from the battery to the ESC, and ESC makers know this. No decent ESC is going to fail based on a battery mount that is three or four inches longer on one build than on another. Keep the build sensible and mount your battery centrally and close to the power distribution point. Eliminate extra wire, long harness leads, etc. If you do that, extending the ESC power wires a few inches is probably not going to be a problem.
 

Bartman

Welcome to MultiRotorForums.com!!
Very nicely done Rob! If anyone's wondering, Rob's been accumulating a shop full of electronic measuring and sensing equipment. He began by testing ESC's for our Jeti ESC product review and this is a continuation of the expertise he's developing.

I'm surprised by the results actually. There appears to be more of an effect at the ESC's than I would have guessed but your comment that upsizing the ESC's will deliver a reasonable approach to managing the problem.

Great job Rob and thanks for the follow through on this!
 

gregster

New Member
Hi Rob


This is a really interesting topic and your tests have shown improvements.


I am intrigued as to what effect the capacitor's internal impedance / ESR has on these fluctuations, I do understand that in high frequency switch mode power supplies it has a direct effect on the output ripple of the PSU.


Another test that may be worth looking at is performing the same test but with say a 560uF 35v capacitor as it has much lower impedance than the 63V.


The Rubycon 560uf 63v in a 12.5 x 30 case size has an impedance of 35mR ( or 43mR in the 16x20 case) which one did you use for your test


Would be interesting to repeat the test using say a Rubycon 560uf 35V which has a lower impedance of 20mR approximately half of the 63V.
 

econfly

Member
Hi Rob


This is a really interesting topic and your tests have shown improvements.


I am intrigued as to what effect the capacitor's internal impedance / ESR has on these fluctuations, I do understand that in high frequency switch mode power supplies it has a direct effect on the output ripple of the PSU.


Another test that may be worth looking at is performing the same test but with say a 560uF 35v capacitor as it has much lower impedance than the 63V.


The Rubycon 560uf 63v in a 12.5 x 30 case size has an impedance of 35mR ( or 43mR in the 16x20 case) which one did you use for your test


Would be interesting to repeat the test using say a Rubycon 560uf 35V which has a lower impedance of 20mR approximately half of the 63V.

Good points. I just used capacitors I had on hand. Surely the ESR matters, and these Rubycons I have are all relatively low internal ESR. For the 560uF 63v I have the 12x30 case size. One way to go with this (assuming the interest is out there) is to just order up a consistent series of caps and re-run the tests with an attempt to identify the relevance of small ESR changes on the results. I'm also interested in any feedback on how to measure the problem. One thought I had up-front was that this would be near impossible to quantify if all of the action were internal to the caps (i.e., we couldn't see ripple or spikes because the caps were fully correcting these issues, yet increased noise was stressing the caps to a greater degree). I was a little surprised that this was so clearly not the case -- at least for the ESC I'm using for the tests.
 


R_Lefebvre

Arducopter Developer
Excellent work! Thanks for doing this.

But there's another situation you need to look at. I have read that the problem with long lead lengths isn't the voltage fluctuation per-se, but actually the ripple currents in the capacitors due to the ripple current. ie: While they are busy balancing the voltage, charge is flowing in and out of them. Current. In both directions. The current flowing through them leads to heating of the caps due to ESR. What I've read is that the real problem is heating of the caps, which will eventually fail because of it.

This is a problem important to me, as I have an ESC running the tail rotor on a SRH, and the lead length is very long. At least 16". Maybe 24". I've bought a capacitor bank to help but haven't added it yet. I'll probably be OK though as the ESC I'm using is rated at 80A and 50V, but I'm only running it on 34V, and typically only 5-10A.
 

econfly

Member
Please define "impedance" for myself and others, but mostly for me. :)

Short answer: resistance to current flow that changes in response to frequency.

Excellent work! Thanks for doing this.

But there's another situation you need to look at. I have read that the problem with long lead lengths isn't the voltage fluctuation per-se, but actually the ripple currents in the capacitors due to the ripple current. ie: While they are busy balancing the voltage, charge is flowing in and out of them. Current. In both directions. The current flowing through them leads to heating of the caps due to ESR. What I've read is that the real problem is heating of the caps, which will eventually fail because of it.

This is a problem important to me, as I have an ESC running the tail rotor on a SRH, and the lead length is very long. At least 16". Maybe 24". I've bought a capacitor bank to help but haven't added it yet. I'll probably be OK though as the ESC I'm using is rated at 80A and 50V, but I'm only running it on 34V, and typically only 5-10A.

The problem is measuring the current flow in/out of the capacitor. It's so much easier to measure voltage on the oscilloscope (and, assuming fixed circuit effective resistance we know that voltage and current are in fixed proportion). But ideally I agree -- what we want to know is the fluctuation of current in and out of the capacitor.

What do you think of the fact that we have voltage fluctuations even with the caps in place? I'm taking this to mean that one of two things is true. Either (a) there just isn't enough capacitance there to absorb the ripple/spikes, and/or (b) the frequency is high enough that the caps can't react fast enough to do the job.

We know at least some of this is due to (b) -- otherwise what's the point of chasing the lowest ESR caps we can get? But this raises another question: If we are at the limit of the capacitor to change in the time available (either the limit of its total capacitance or the limit of its ability to react), what more material harm can be achieved by increasing ripple/spikes? I'm sure the answer is "some harm", but how can we quantify it? Put another way, if we can't shove any more current in or out of the capacitor in, say, 1/16,000th of a second, then can we really hurt it much by increasing the level of ripple in the circuit?

I know this is over-simplification, but I think the answer to this whole issue has to begin with the observation that we have plenty of ripple in place even in an idealized case of direct ESC to battery connection.

Finally, we know for sure that the caps will fail. It's just a matter of time. That's part of what makes all of this so complex and superficially simple. Yes, adding caps can't hurt. Yes, adding length to the power leads can't help. But how much? And, all things considered, should we care? All very interesting.

I like your answer as much as any: Just run an ESC that can handle the job and quite a bit more. Than we can probably safely assume the rig will crash or get re-built in one way or another before the capacitors are likely to fail...
 
Last edited by a moderator:


econfly

Member
reading that made my face pucker :(

I'll just quote wikipedia:

It is necessary to introduce the concept of impedance in AC circuits because there are two additional impeding mechanisms to be taken into account besides the normal resistance of DC circuits: the induction of voltages in conductors self-induced by the magnetic fields of currents (inductance), and the electrostatic storage of charge induced by voltages between conductors (capacitance). The impedance caused by these two effects is collectively referred to as reactance and forms the imaginary part of complex impedance whereas resistance forms the real part.

Yeah, how about that? By the way, this is the sort of stuff you run into on the HAM radio exams (I just took and passed the three of them a few weeks ago).

Just think of this problem as additional resistance on the power wires caused by the ESC's pulsing current. V/I = R. So if R is bigger (more effective resistance) then any change in I (current) must be accompanied by a bigger change in V (voltage) than if R were smaller (dV = R * dI if R is fixed; bigger R, then bigger change in voltage for any given change in current). The wrinkle here is that the current changes are themselves creating higher effective resistance over the longer power wires. That's the problem. Longer wires in the presence of pulsing current means more effective resistance and hence bigger voltage changes. That puts a larger burden on the capacitors.

There probably is some nice clear water-through-a-pipe analogy for this, but I'm terrible at those. Maybe someone can offer one up.
 
Last edited by a moderator:

Bartman

Welcome to MultiRotorForums.com!!
the wikipedia explanation makes much more sense.

as for the water reference, i'd just say that the AC current being imagined as water would be very turbulent with the mass of the water having to reverse direction with a set/high frequency instead of flowing smoothly and consistently through the pipe in one direction. the turbulence would induce greater resistance that the mass of water would have to overcome to move through the pipe compared to the smooth flow of the DC current.

maybe?
 

gregster

New Member
Please define "impedance" for myself and others, but mostly for me. :)



Here is a bit from wiki

""Electrical impedance is the measure of the opposition that a circuit presents to a current when a voltage is applied""

In respect of a capacitor the impedance of it is known as ESR ( Equivalent Series Resistance) and is measured in ohms, a perfect capacitor would have no ESR and would not be effected by high frequency waveforms, however its there and each capacitor has it, the higher the value the worse the ESR has an effect on the voltage / current. In fast switching power supplies this is seen as ripple voltage on the output rail.

With regards to ESC's the results above look interesting as it would seem the ESR of a capacitor has a similar effect just like a fast switching power supply, so with this in mind a small capacitor with a very low ESR may have a similar result as a much bigger one.
 

R_Lefebvre

Arducopter Developer
The problem is measuring the current flow in/out of the capacitor. It's so much easier to measure voltage on the oscilloscope (and, assuming fixed circuit effective resistance we know that voltage and current are in fixed proportion). But ideally I agree -- what we want to know is the fluctuation of current in and out of the capacitor.

Well, according to the ESC manufacturers, what you should really be measuring is the temperature of the capacitors, as that is the failure mode.

I guess the temperature would be related to the voltage fluctuation, and frequency. That would tell you what the average current is? Then factor in ESR to get power dissipated as heat. Something like that. 'Damnit Jim, I'm a mechanical engineer, not electrical!'

What do you think of the fact that we have voltage fluctuations even with the caps in place? I'm taking this to mean that one of two things is true. Either (a) there just isn't enough capacitance there to absorb the ripple/spikes, and/or (b) the frequency is high enough that the caps can't react fast enough to do the job.

I think it's really both things. ESR comes into play. ESR prevents the caps from charging infinitely fast. It takes voltage differential to make them accept and release their charge.

We know at least some of this is due to (b) -- otherwise what's the point of chasing the lowest ESR caps we can get? But this raises another question: If we are at the limit of the capacitor to change in the time available (either the limit of its total capacitance or the limit of its ability to react), what more material harm can be achieved by increasing ripple/spikes? I'm sure the answer is "some harm", but how can we quantify it? Put another way, if we can't shove any more current in or out of the capacitor in, say, 1/16,000th of a second, then can we really hurt it much by increasing the level of ripple in the circuit?

As stated, the harm is the ripple current induced heating of the caps. But I would also think that higher voltage fluctuations would also eventually cause some device (caps or fets most likely) to exceed their voltage rating on the high side.

Also, there is an effect where, even if you are operating within the voltage limits of an electronic design, sharply rising voltage rates cause insulation to break down over time. I know this because I deal with industrial AC motors powered by variable speed drives, sort of like giant ESCs. I'm in Canada where the supply voltage is 600V. Typically in the US it's 480V, and most motors we have to buy are rated for that. A 600V AC drive could be used to drive a 480V motor, simply by chopping down the voltage by using a lower PWM duty cycle when creating the waveform. Similar to a BEC. So in theory, a 600V drive can output a 480V waveform. But the thing is it's not really 480V. It's got peaks (above) 600V. Some motors are designed for this. But some have insulation that will break down due to the rise-time of the voltage spikes being used for the 480V waveform.
 

econfly

Member
Measuring capacitor temp is a nice idea, but of course more difficult. The temp part is easy, but it's the waiting and test time that would make it a good sized task to take on. At that point, I think the way to go would be to test a full build over several flights, varying lead lengths and capacitance, and logging capacitor temp and anything else related to ESC stress that is feasible for each one. That would really be a step up in getting at the issues here. Also a big job.

In the mean time I do think observed voltage fluctuations stand as a decent proxy for what we really care about. And the main points (at least to my mind) are that:

  1. There is plenty of focus on adding length to ESC leads, but that needs to be considered in light of the entire build. Limiting wire runs in general, mounting batteries centrally, etc., are relevant as well. ESCs are built to perform in a range of situations. Any decent ESC maker must take this into account. So, it stands to reason that a very well considered build with minimal power paths elsewhere in the design can almost certainly allow some acceptable latitude in extending the ESC power connection.
  2. Similarly, ESCs are built to operate over a range of voltages and current demands. While likely not a perfect offset to the ills of adding lead length, choosing an ESC with specifications in excess of the build's requirements is a good idea if a longer power path is desired.
  3. If the prior two cases / approaches don't apply, and the build has long power paths to ESCs and the ESCs are operating at the limit of their specifications, then adding capacitors is a good idea and one that can go a long way to addressing any potential problems.
 

Motopreserve

Drone Enthusiast
Great stuff Econ. Very cool of you to take the time to test all this. Really appreciate your effort and time.

Do you have a link to a US retailer for the specific caps you recommend? I have been following the main thread over on the other forum (I believe it may have kicked off all this debate) and the links provided early in the thread are to caps that I think are not ideal (at least the person on the Digikey live chat said as much).

Where/which are you getting low ESR? Digikey has them - but it seems product pages don't necessarily list "low ESR" on products that may in fact fall under that category.

Thanks!
 

R_Lefebvre

Arducopter Developer
I've got the same question. ESR doesn't seem to be something that is commonly used for cap selection, so it's hard to shop for them based on that.
 

Motopreserve

Drone Enthusiast
I've got the same question. ESR doesn't seem to be something that is commonly used for cap selection, so it's hard to shop for them based on that.

I literally have 6 different caps bookmarked that I have found based on inconsistent links/posts. I'm honestly not sure if ANY of them would be appropriate :)
 

econfly

Member
The ESR (equivalent series resistance) is just a way to think of the internal resistance of a capacitor in the real world. An ideal (i.e., theoretical) capacitor has zero internal resistance. The ESR is the resister you can imagine as being in series with an ideal capacitor, where that combination of resistance and capacitance is the approximate description of the actual (non-ideal) capacitor.

When the capacitor is charging and discharging at low frequencies, the internal resistance isn't very important. There is plenty of time for the capacitor to charge/discharge, so a little extra resistance inside the capacitor slowing things down isn't an issue.

But at high frequency things change. Now the capacitor has to charge and discharge very rapidly. For our ESCs a complete charge/discharge cycle typically takes place in 1/8000th to 1/16,000th of a second or faster.

This brings up one more complication, which is that the capacitor's internal resistance is a function of the charge/discharge frequency (this is the impedance concept discussed earlier). So, we want capacitors with low impedance at high frequency. That translates to low ESR for our purposes.

I tested using Rubycon ZLH series aluminum electrolytic capacitors. These are used in many ESCs and are widely available.

Pick one with well more than enough voltage rating for your system. For example, a 6s setup with fully charged batteries has system voltage of 6 x 4.2v = 25.2v. Voltage can spike a few volts above that given the problem we are trying to solve, so the typical capacitor chosen by ESC makers is 30-35v rated.

Assuming 6s power, here are the caps I would order (links to Digi-Key). These are fairly common sizes (particularly the 220μF and 470μF) and Digi-Key has plenty of stock of all of them right now.

35v, 220μF : 35ZLH220MEFC8X11.5
35v, 470μF : 35ZLH470MEFC10X16
35v, 560μF : 35ZLH560MEFC10X20
35v, 1000μF : 35ZLH1000MEFC12.5X20
 

Motopreserve

Drone Enthusiast
Econ,

once again sir, you rock! Not sure why it was so difficult for the other thread to hone in on some decent links - but you knocked it out of the park in one fell swoop. ;)
 

econfly

Member
Econ,

once again sir, you rock! Not sure why it was so difficult for the other thread to hone in on some decent links - but you knocked it out of the park in one fell swoop.
wink.png

Thanks! I really appreciate the positive feedback.

A few points that are probably obvious to most, but just in case not... If you do decide to install more/different capacitors (and as I note in comments earlier, the simple choice to lengthen ESC power wires is not necessarily indicative of a need to do so -- generally I think this alleged problem is over-blown for all but extreme cases), solder the caps as close as possible to the ESC load point (where the two incoming power wires terminate on the ESC). Cut the capacitor legs to as short as possible. Be sure to install the caps in parallel, and avoid a branching or common-load point installation if using more than one cap. By this, I mean each capacitor should be directly connected to the load in parallel (across the positive/ground incoming lines). Don't run one pair of wires and put caps on it one behind the other. The idea is to get each capacitor to operate as close to the load as possible and with minimal shared paths to that load. The choice of, say 2 x 220μF or a single, say, 470μF cap should be close to irrelevant if the caps are chosen optimally, and particularly when adding capacitor(s) to an existing ESC.

If I were going to the trouble to do this I would add a single cap to each ESC, right to the solder points used by the existing capacitor(s), with the legs trimmed as much as possible to get it to fit. To simplify things, go with the voltage of the existing caps (as noted above, this would be 30-35v for 6s, a few volts difference won't matter, but err on the high side). The capacitance of that single cap should be whatever gets the job done, but of course that is about impossible to know without test equipment and trial and error. For just about any conceivable sensible situation I think doubling the existing capacitance should get the job done. Go a little over that if you want (it's only a few dollars in cost anyway), but apart from extreme situations I would be truly surprised if that is necessary.
 

econfly

Member
A few more things as I think about what I will do in future builds:

First of all, I'm not going to worry about extending ESC power connections if that makes sense given the build. But if I do it, I will be certain to keep the extension to a minimum length and the two extension wires tight together.

Second, I will be very aware of any unnecessary wire in the power path. I've been lax about this in the past, with, e.g., a two-battery harness having several inches of extra length. This is the easiest and most overlooked way to improve the build and decrease stress on the ESCs, and there is little to no cost involved in getting this right.

Third, I will buy ESCs that are rated above my build requirements. If I'm building a 6s rig, I'll look for ESCs rated, say, 4s-8s. It's not just the capacitors that can fail, and for the cost and minor weight issues I see this as a generally good idea now -- not to say this is a necessity, of course, particularly if the ESC power path is not going to be extended.

Finally, and because I like to tinker around with things and enjoy the build process, I might add a capacitor to each ESC if I end up lengthening the power connection in a big way. But then again, I might not. I really do believe that the three approaches above will make this step unnecessary. But, because it can't hurt anything, involves just a little soldering, and the cost of capacitors is cheap, I can't see any problem with it if it adds a little peace of mind.
 
Last edited by a moderator:

Top