How much of an issue is battery loop inductance?

[John in CR]

I always understood that if given an option, short battery leads and long phase wires is better than long battery leads and short phase wires. I question this conventional wisdom with moderate and high power rigs, since phase current is so much higher than battery current.

I've also heard that battery loop inductance can be a problem for controllers. I'm working on battery packs right now that will be installed within the bike frame, so they will be long and relatively thin, and installed in multiple areas of the bike (primarily the top tube and downtube) instead of a centralized block like with typical ebikes. For long rides a range extender pack would go on a back rack, opening up the potential for quite large loops. My plan is multiple long full voltage strings with cell level parallel structure connected between these separate packs. I realize that I shouldn't have my battery leads or packs form a loop around the perimeter of the large triangle.

My question is can I organize all this wiring in a manner that minimizes the loop inductance? I realize that I shouldn't have my battery leads or packs form a loop around the perimeter of the large triangle, but should I make a concerted effort to run the +/- leads as close together as practical to minimize the loop contained within each pack's flow? Is it significant enough to worry about? Are extra capacitors a sufficient solution? Will these long parallel wires for balance tap purposes play much of a factor if the current through them is low?

My main bike has been running with a conglomeration of separate packs in parallel, but I haven't installed extra caps on the battery side. Running at 8-10kw could the battery loop issue be part of what kills my controllers? What size and type of caps would be appropriate?

 

[Drunkskunk]

I read something on this maybe a decade and a half ago. I remember little about it but it effected the way Telco central offices place the batteries for backup power and load capacity balancing. the outcome of the issue was better placement of the wire and use of heavy guage wire for telco aplications. A quick google search brings up what may be the article I remember: http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=4794474

 

[ZOMGVTEK]

The issue is that longer wires and loops on the battery side of things add inductance where it is not needed or desired. This can lead to voltage spikes which kill MOSFET's.

Inductance on the motor side may limit performance, but it also makes the motor an easier load for the controller. It's designed to be driving an inductive load, but not designed to have an inductive power source. Running things at the limit really makes this show.

Never loop power wires around anything, and make them as short as possible, within reason. Ultra Low ESR and Low Inductance capacitors right on the MOSFET's power rail can really absorb some of the impact of the voltage spikes. If you have the space in the controller, something like this would be a good idea...

http://www.ebay.com/itm/250890586653?ssPageName=STRK:MEWAX:IT&_trksid=p3984.m1423.l2649

These are normally about $20 ea, so thats a great price.

 

[Njay]

John, do you remember how inductance slows down the rate of current rise? Inductance on the battery wiring will delay power arrival at the MOSFETs. And that's one of the input capacitor's function, to (locally) provide that power until the wiring inductance charges (when it's time to open the MOSFETs).

Do you know remember how inductance doesn't like to have current taken away from it and fights back by raising voltage? When the MOSFETs go off, the current "wants" to stop but wiring inductance doesn't want to. Input capacitors once again save the day, by absorbing the voltage spikes of the angry inductance.

Do you know what's the purpose of inductance raising the voltage? Breaking current flow means that you raise the resistance; if the inductance wants to keep the current, what can it do? Raise the voltage, keeping it at a level enough for the current to flow until all inductance's stored energy (in the magnetic field) is released.

 

[John in CR]

John, do you remember how inductance slows down the rate of current rise? Inductance on the battery wiring will delay power arrival at the MOSFETs. And that's one of the input capacitor's function, to (locally) provide that power until the wiring inductance charges (when it's time to open the MOSFETs).

Do you know remember how inductance doesn't like to have current taken away from it and fights back by raising voltage? When the MOSFETs go off, the current "wants" to stop but wiring inductance doesn't want to. Input capacitors once again save the day, by absorbing the voltage spikes of the angry inductance.

Do you know what's the purpose of inductance raising the voltage? Breaking current flow means that you raise the resistance; if the inductance wants to keep the current, what can it do? Raise the voltage, keeping it at a level enough for the current to flow until all inductance's stored energy (in the magnetic field) is released.

Yes, I think I have a handle on the why's. I don't have a handle on the magnitude of the issue though. I wasn't thinking to be sure that the battery wiring doesn't cross creating extra "turns", so I end up with just the single turn coil shaped which in my case will be shaped like a peace sign, plus the inductance of the wire itself. I still have question about how hard I should try to reduce the area contained within that single odd shaped loop.

 

[Electroglide]

This topic has come up before, and earlier under a different name i had tried to argue the case of short as possible and heavy as possible twisted battery cables. My reasoning was sound but my logic was flawed at the time. I ran some tests then to show the effect of increasing the inductance in the battery line, and the effect on the ripple voltage seen at the decoupling electrolytic capacitors. So I think I can make a better explanation now then previously. Ok, so longer battery cables have 2 parameters of concern at the frequencies that we are talking about. Inductance and resistance. The inductance will be dependent on the effective loop area that you create when connecting your battery source and return wires to the controller. Twisting the cable minimizes that effective area because of the smaller loops that are created verses one long narrow loop that would be formed by a 2 conductor wire. There are 2 components to the inductance. The self inductance that is a function of the current flowing in a single conductor, and the mutual inductance that occurs between 2 conductors running in parallel in close proximity carrying current. For our case, you get the 2xLs+2xLm which gives you the total inductance for the loop. Moving on, the inductance in line along with the resistance of the wire create an impedance drop between your battery and your controller. It also is a energy storage component too. So keeping it simple, you can look at your load on the battery as a battery and battery line impedance in series with your controller, so using the voltage divider rule, you will have a lower average voltage at the controller than your battery open terminal voltage. ( z=L(s)+R), Vcontroller = Zcontroller/(Zbattery cable + Zcontroller + Zbattery) x Vbattery. The bad part comes in the form of the energy storage component of the inductance. Since current cannot change instantly because of the inductance, when your controller mosfets are in their conduction cycle..ie connecting the battery to motor phases, the current has to be drawn out of the capacitors in your controller until the inductance is overcome in your battery cables etc. This causes a dip in voltage at the capacitors that is reversed when the controller mosfets go into a freewheel state and the motor is no longer connected to the battery. Since the current cannot change instantly because of the inductance, the battery current flowing at the time of the end of the conduction state now flows into the capacitors in the controller, and they are charged up to a peak value above the mean voltage which is derived from the voltage divider mentioned above. This peak value will depend on the energy stored in the inductance (E=(1/2)*L*i^2) and the amount of capacitance you have on board the controller. What you can take from this is that having too little capacitance and or capacitors with too little ripple current rating will cause them to either blow up or generate lots of battery line noise, ground bounces, etc, when the controller is pwm-ing. For the most part, if you have too little capacitance, you most certainly have too little ripple current rating so a blow up of the capacitors is usually the end result. Ways to fix this problem is to reduce the battery line impedance where I think that the inductance plays a bigger role than the resistance. Since the effects on the capacitors are greatly influenced by frequency, you tend to not see as many problems when you operate your controller at WOT at top speed, since the mosfets are then switching at the commutation frequency and not the pwm frequency. Hence why these failures may not be as common as they should be based on what I have seen in Chinese controllers for the quantity of capacitors. Anyways, that is another 2 cents worth of opinion :lol:

 

[Njay]

Some very good 2 cents !

 

[John in CR]

Ok, so I should definitely beef up my capacitors. Do I have to replace the existing capacitors, or is just adding better ones fine?

 

[Electroglide]

Figuring out the required amount of capacitance is always a fun thing in controller design. Especially for the case where you are designing the controller, you don't know how the end user will connect it up. Sevcon use to have rules of thumb for determining how much ripple current you needed to stuff. They did a bunch of tests to derive those rules of thumb. I think that they used a ratio of 3.1 for DC motors and 4 for AC motors. So the way it works is that if your motor rms current is 100amps then you want to stuff at least 25Amp RMS ripple current capability. This may be higher if you think that BLDC is more like a DC motor because only one H-Bridge is in operation at one time. I tend to think that may be closer because the AC ratio was probably for the AC induction situation. In your case adding more capacitors with high ripple current ratings at the controller would give you the highest protection. Adding it at the battery doesn't really help the controller much. Panasonic make decent capacitors for 50VDC, 100VDC and 160VDC. Think that the families are FM, FC, and EB. Just a side note. If you stuff for ripple current rating based on the ratios listed above, the capacitance falls out the equation because you always have more capacitance than you need. Of course if you are almost always at WOT at high speed, the extra capacitors are not going to be doing too much, though they will smooth out some of the voltage ripple seen on the battery cables.

 

[dnmun]

why not just set up the circuit equations and calculate the value of the inductance and whether it is significant?

i really cannot imagine it is significant when compared to the battery voltage or the resistance in the contacts of the connectors. jmho.

 

[John in CR]

why not just set up the circuit equations and calculate the value of the inductance and whether it is significant?

i really cannot imagine it is significant when compared to the battery voltage or the resistance in the contacts of the connectors. jmho.

I stopped using connectors and simply hardwire it. Since I disconnect so rarely, it's easier than installing connectors. The battery is 23s A123 and 82V off the charge.

Electroglide,
Thanks for all the info. BTW, I spend very little time at WOT, and only approach full duty once every week or two when I get on the highway. My bikes are too fast for full throttle except during acceleration for the vast majority of my errand runs, and generally keep it under 50mph except on the highway. Typical cruising is 35-40, except for my school taxi runs where I ride a bit slower and rarely pass.

 

[Electroglide]

If you are at WOT most of the time I bet that you could get away with adding a bit more ripple current capability in the capacitors. When Sevcon came up with those ratios I think they were trying to cover the case where forklifts would be going up an incline fully loaded, and with them doing this on a repetitive basis. Going up one hill at maximum current will heat up the caps but if you don't keep hitting big hills and you are at WOT and back emf limited on the current going into the motor the rest of the time, then they have some time to cool down . The question is do you think you have enough ripple current capability already, and one way to find out would be to throw a temperature sensor on the end cap of one of them and ride up a big hill at WOT where you know that you are battery current limited. If you see the cap quickly rise in temperature, then you probably don't have enough ripple current rating. If it is slow, like 0.5C or so a second then you are probably safe. If you are going to be doing a lot of hill climbing, I would be more conservative on the temperature rise. That is just a swag on my part.

Dnmum,
We could calculate that inductance for twisted cables in air. I haven't done that sort of stuff in a long time. Probably there are calculators out there for figuring out the inductance for a twisted pair already. The resistance of the path will dominate at near DC conditions, but I think that when you are talking about the switching pwm frequency, it plays a greater role and dominates when the cable length begins to get long. I found a calculator for the inductance of a straight wire, and got 1.3uH for a 1 meter length of 10AWG wire. So that would be 2.6uH for a source and return wire. This doesn't include the mutual inductance. Ok, so with a switching frequency of 20kHz, at 50% duty cycle, that would be a dt = 25usec. With a switch current of di = 100A, we can calculate the voltage drop across the line inductance without mutual inductance included. VL = L di/dt = 1.3uH * 100/25usec = 5.2V Now if you add the effects from the mutual inductance(I suspect a larger component), that drop will increase. Unless I screwed up something in my logic. The bigger that number gets the more your capacitors have to work. I have seen the effects of longer battery cables on scissor lift controllers (250A). At 50kHz switching, a designer had melted the plastic sleeves off his electrolytic capacitors because his ripple current had been too high. That was a bit of a stinker..literally :lol:

 

[rhitee05]

For steady-state, the inductance of the wires won't matter at all. Z=j*w*L, so we're talking about a few uH times maybe 20 kHz, so that will end up being tens of milliohms. Very small unless you're running huge gauge wires. Also, since it's reactive, there's no power loss so it's not heating the wires.

The issue is the voltage spikes that occur in the controller at switching instants, as Electroglide already explained. John, what you probably need are a few large electrolytic capacitors and several smaller ceramic caps. The electrolytics take the brunt of the ripple current, but the ceramics will do more to knock down the large, fast voltage spikes that threaten to pop your FETs. The temp measuring method he suggested is probably as good as any way to figure out whether you need more electrolytics. Unless you happen to have an oscilloscope handy and you can measure the voltage ripple directly.

You could also add something like an MOV across the FETs for some extra protection. They look like large disc capacitors and will absorb voltage spikes. They'll pop if they have to absorb too much, so best used in addition to an adequate number of caps.

 

[dnmun]

doesn't anyone have access to SPICE or other canned circuit programs?

try using the real numbers. 2000uF of capacitance at the end of two individual 20cm of 10G cable, .3mm separation between the two if they are a twisted pair, and use an estimate of about 1R for total resistance and see what the ripple will be at 20kHz?

ripple at the base of the capacitor since that is what the mosfets will see.

 

[Njay]

http://endless-sphere.com/forums/viewtopic.php?f=30&t=31804

 

[Electroglide]

Ok, so we want to talk about inductance that the mosfets see? There is a macro effect from inductance coming form the battery lines connected and there is a micro effect caused by the circuit inductance within the controller. The voltage over shoot that you see on your mosfets when hard switching off a bank of mosfets comes primarily from the local inductance of the traces between the mosfet upper and lower switches, the freewheel path on the supply rail and the freewheel path on the ground rail, and the path to your capacitor bank, and the lead inductance of the devices. The faster that you try to turn off those devices with current flowing through them, the larger that spike gets until you reach the avalanche voltage of the mosfets for that bank. This overshoot voltage rides on top of the bus voltage of the capacitors plus a diode drop, give or take a few hundred millivolts. The absolute value is affected by the line inductance to a certain extent depending on how much capacitance you have in your controller ie the voltage ripple seen on the caps. For a bldc motor, I swag that the hard switch occurs at the minimum of the voltage ripple on the capacitors. The dt for a switch transition can be in the range of 5 to 3 usec for our chinese controllers to 30 or so nanoseconds for good dc-dc converters. A few hundred Nano heneries of inductance can result in 5 to 30 volts overshoot easily if you have high phase currents and a fast turn off time. For this sort of modeling, you don't need a device level pspice to see the effects. There is simulation software called PSIM which has an unlimited time for use eval that is limited to a fix number of devices and nodes. Like the student version of pspice, it is useful for learning about fundamental items like this. It is more geared toward power electronic as well, and allows relatively painless experimentation with different concepts. My previous company used it to model a fuel cell DC-AC controller which made it easier to go to a real implementation with a good degree of confidence. Like Rhitee mentioned a way to minimize the overshoots is to provide low esr capacitors close to the switching banks across the bus. In my designs, I used 200V 1U0 Ceramic capacitors pretty much electrically on top of the half bridges. This had the advantage of reducing the peak value of the overshoots and allowing me to switch off the mosfets quicker. Since you seldom gain something for nothing, this is also the case for using ceramics on the bus. What happens is that the frequency content of the overshoot voltage shifts down, which can cause problems with conductive emissions if you care about such things :lol:

 

[dnmun]

i had wondered why they put the ceramic capacitor across the S/D busses. makes a lot more sense there with this observation.

originally i was thinking of how the inductance of the other wire in the twisted pair, or parallel inductor, would somehow counteract the induced voltage but if this is charge delivered to the controller on the main wire, is there still current flowing in the opposite direction on that other conductor, to complete the circuit so it would follow kirchoff's rule? now i am confused.

 

[rhitee05]

Please forget anything anyone ever tried to tell you about calculating the inductance of a single wire. Bunk.

Current flows in loops. Inductance only has meaning in loops. You simply can't calculate the inductance of a single wire, and any formula which claims to do so is relying on assumptions about the other half of the loop.

 

[John in CR]

You guys are over my head, so I'm playing catch up here.

None of my bike have compact battery packs, and I really want the freedom to spread them out even more. It sounds like I definitely need extra caps, some electolytics and some rail caps like BigMoose mentioned here that squash voltage spikes. http://endless-sphere.com/forums/viewtopic.php?f=2&t=32853&p=480054#p479857 . What I need isn't going to fit in any of my controllers, so can I just mount them immediately in front of the controller, or would that effectively be like leads being way too long?

Also, I'd just shoot for overkill, so please toss out some values I need for up to $100 worth or slightly more.

 

[ZapPat]

You guys are over my head, so I'm playing catch up here.

None of my bike have compact battery packs, and I really want the freedom to spread them out even more. It sounds like I definitely need extra caps, some electolytics and some rail caps like BigMoose mentioned here that squash voltage spikes. http://endless-sphere.com/forums/viewtopic.php?f=2&t=32853&p=480054#p479857 . What I need isn't going to fit in any of my controllers, so can I just mount them immediately in front of the controller, or would that effectively be like leads being way too long?

Also, I'd just shoot for overkill, so please toss out some values I need for up to $100 worth or slightly more.

Forget those low capacity caps if you're going to mount them any distance from the MOSFETs, they wont do anything outside your controller. Just stick with large, low ESR electrolitics in this case, mounted as close to your controller as possible. Your controller's internal capacitors will do the rest of the job fine as usual.

 

[Electroglide]

Please forget anything anyone ever tried to tell you about calculating the inductance of a single wire. Bunk.

Current flows in loops. Inductance only has meaning in loops. You simply can't calculate the inductance of a single wire, and any formula which claims to do so is relying on assumptions about the other half of the loop.

Just pulled out the text book...should have done that first :lol: I was incorrect for the line inductance from the point of self and mutual...by strict definition mutual is between two separate circuits and self is between loops of the same circuit. So really we are only dealing with self inductance. I agree totally that the inductance of a single wire is a special case. I am guessing that the calculator that I referenced was probably for parallel wire transmission line and based on the formula that I see in my textbook which is similar, and it might still apply for our battery line connection. The formula gives an inductance/meter result...probably still valid for our case if our wires maintain a specific distance apart.

 

[rhitee05]

For the case of battery wires, the two-wire case for inductance is appropriate:

http://www.cvel.clemson.edu/emc/calculators/Inductance_Calculator/wire2.html

Forget those low capacity caps if you're going to mount them any distance from the MOSFETs, they wont do anything outside your controller. Just stick with large, low ESR electrolitics in this case, mounted as close to your controller as possible. Your controller's internal capacitors will do the rest of the job fine as usual.

+1

You could add electrolytics just outside the case if needed, but ceramics need to go right at the FETs to do any good. They're small so you could probably squeeze a handful in if you need to. The leads need to be kept as short as possible, too.

 

[Electroglide]

That calculator looks valid. My text uses an approximation for the inductance per meter because they assume for practical lines that the distance is much greater than the radius of the line ie "d>>a". So for a twisted pair, the distance would be at a minimum. Probably for vehicle battery cables, the "d" is no longer much greater than "a", so the approximation would no longer be valid and you would need a better approximation. Good find.

As for the caps, I agree with you guys that having them at the entry point into the controller would be sufficient, and good quality low ESR electrolytics are good enough. The film caps that Big Moose showed are intended for high voltage applications where the electrolytics characteristics are poor in terms of ESR. Putting ceramics or film caps inside would be good too. Ceramics are smaller, but films caps are safer. If you have a crack in a high C SMD ceramic, it can burn a hole through a PCB if it is connected to a high capacity bus.

 

[Electroglide]

Just a clarification on the ceramic capacitors. When they are machined assembled SMT on a PCB, and the board has little flex in them, they are quite reliable. When you hand solder them, and have large amounts of solder on the end terminations, they are less reliable and may see fractures from stress of the solder fillets which will lead to a failure. Thats why I would suggest small film capacitors instead of the ceramics for a retrofit.

 

[John in CR]

Until a new controller comes out, it's sounding more and more like I need to split the strands in the motor phases into 2 or 3 groups to run with 3 controllers. I want to get to 150A battery side reliably without paying close to $1k. I can get controllers that handle 50A with no issues for $30-50/ea. 3 run in parallel gets me to my 150A target. If I want 200A battery side, then I just split the winding strands into 4 groups instead. Does the inductance of the resulting parallel coils of the motor windings and imperfect timing between the controllers shoot this idea down?

I'm sick and tired of having to baby controllers just because my motors are low resistance and inductance. Why is this so difficult? Scooter controllers with el cheapo fets can handle 50A or 70A at 60V-74V nominal with 100V rated components, yet nothing short of controller overkill can reliably push past the 100A threshold. The really fun stuff is up there above 100A and at 100-120V true highway speeds are possible too, all even with a fat guy aboard and a motor inside the wheel which including brake, tire, motor, etc is a sub 30lb wheel assembly. What's the best way to get there without spending over $500 on a controller, since the motor and factory controller cost only $125 plus shipping and taxes to begin with?