Calculating controller power input capacitance

[HighHopes]

there is also a wider picture to consider. as you know, arbitrarily selecting 160kHz switching frequency so that cap is good will leave you with a melting mosfet or hydrogen cooled heatsink to survive. on the good side, if you take the time to find the balance you will find that for all motor drives the results are always reasonable.

what i mean is.. you have 10uF phase current. so what is a good switching frequency? 20kHz? 160kHz? this is a question of control bandwidth for smooth action of the motor. so.. my guess is 100kHz would be in the range of acceptable. i say this because i have 50uh phase at 20kHz which is good, so your 10uH is 5x smaller (faster) so the control bandwidth has to be 5x higher, 100kHz. so i would sayy somewhere between 60kHz & 100kHz would be good for your system (you might calculate to get more accurate). assuming worse case 100kHz.. heck even best case 60khz.. what mosfet can switch at that frequency? how much phase current can you have while staying under thermal limits. when you switch that fast, what deadtime do you need, how will you manage that (100ns might be too much), with higher switching frequency & tighter design limits it becomes far more important to have a good design and absolutely nail perfect the layout & geometry .. are you up fort hat challenge? willing to put in the effort to learn so you design/build correct and minimize lab bench trouble shooting.

so you have to get "in the ball park" with these numbers before you start to calculate the capacitor. think about it.

you don't need to get to my Arms of cap, you need to get to YOUR Arms of cap. that might mean you need 5 parallel caps where i needed just one. the most parallel polypropelete caps i have used is 10, that was for 600V 600A system switching at blazing speeds. that's $500 just in caps + another $150 in snubber.

ps. for the math, use my attached PDF as a guide as it is more accurate. given in previous post in this thread.

 

[Lebowski]

what are the symptoms if you don't have enough cap ?

 

[Arlo1]

.

ps. for the math, use my attached PDF as a guide as it is more accurate. given in previous post in this thread.

That's what I used.

Is this all I need to look for is total ARMS??? Is this per H bridge or total on the inverter?


And why I'm here is I am wondering what will happen when you don't have enough caps... If my problem will occur.

 

[HighHopes]

if by "not enough cap" you mean, uF value is less than optimal calculation would recommend.. then the symptom is, under normal operation, excessive VOLTAGE ripple on the bus, i.e. greater than 5%. you can actually tolerate quite a bit of ripple at low speeds, but at higher RPM it starts to f#ck with your control because each time mosfet turns ON in MUST apply the bus voltage to the phase cable because motor bEMF is so high.. but if you have big ripple.. maybe you do not apply in that moment bus voltage.. perhaps it is only Vbatter - 20% .. or Vbattery + 20%. so what happens? that single pulse produces less, or more, torque than anticipated.. not intself a problem becaues the motor is like a big averaging device .. but your current feedbaack control .. it DOES see this and starts to Hunt. because it is expecting a certain performance but does not get and may go unstable because of it. that is normal operation symptom.

some lessor obvious impact of too small uF value.
other symptom, perhaps you have minimal DC bus uF value so during regenerative breaking if you have a lot of energy coming back and uF is small you can not store energy locally and battery can not take quick enough.. so voltage go up until destruction.

other symptom, you can not accomplish field weakening in smooth uniterrupted manor because your inverter struggles to source *imaginary* current, Id.

other symptom, your have a shoot-through event and you expect your desaturation detection circuit to kick in and save your mosfets from destruction, but unfortunately the DC link uF value is insufficient to hold the bus at high voltage for 10us time and so the voltage droops and affects your desat circuit so your desat circuit fails to actually protect mosfet and then destruction occurs.

Other question, of insufficient Arms rating, that is something else. this means your capacitor will heat up which takes away its life. if it is 5000Hour rated cap, it becomes less. 1/2 life for every 5deg C rise.

for voltage rating, it is not too big a concern for poly cap.. but electrolytic, it is well advised to have at least 50% margin. 100V battery pack = 200V DC link capacitor rating. Vcap/Vbattery >= 2. For any ratio <2 it will affect how quickly the electrolite dries over time.. so just another example of loss of expected endurance/life.

unrelated to this discussion, but interesting tid-bit of info.. you can not use electrolytic capacitors at all anywhere in aerospace (unless equipment is in a pressurized bay) because low atmosphsere pressure at high altitude means implosion of cap.

 

[HighHopes]

Is this per H bridge or total on the inverter?

it is for actually a single phase leg.. but .. i view it as conservative method for total Inverter instead. because in 3-phase system each leg has its own switching and DC link ripple associated ... but when combined into the common DC Link cap, then that DC link cap views all three phase legs with same voltage ripple but phase shifted in time. so from the DC link capacitor's point of view, actually the switching frequency is higher and the voltage ripple is smaller..

so you are clever to point out this in the math. you can then change the formula to suit, or what i do is just leave it as is and consider then this math to produce somewhat conservative results... like a safety margin.

for design with extreme hard to meet specification, perhaps you might go into the math more to reach the exact best solution imaginable, but this is not really necessary for DIY one-off. what we want here is something that works (functional + reliable) and we can be well happy with that.

 

[Teh Stork]

Phase ripple current and bus capacitor ripple current is not the same thing. Source inductance is what matters (the battery inductance), not the output inductance (motor phase inductance). In big battery packs (especially big ~100Ah cells) battery inductance can grow quite big if the layout is poor or the voltage is high. Ebikes stand out from the rest of the EV crowd in this regard.

Phase inductance and PWM freq have a direct effect on phase current ripple.

Phase inductance and bus-voltage have a direct effect on control requirements (control loop etc).

Imagine a good controller with only a decoupling issue due to too little capacitance on the bus. Simply upping the PWM frequency would be a solution. (At the same time you would decrease phase current ripple).

And the control loop and PWM frequency are not dependent. I have one TI instaspin-FOC based inverter running a 20kHz control loop with 60kHz PWM.

Also, MOSFETS switch much faster and better than IGBTs. Select a MOSFET with a "soft" body diode and you are good to go.

As for every good paper on bus capacitor sizing that is good, there is a bad one - so I could be wrong.

rHiiite (or something) should stop by this thread.

 

[HighHopes]

this thread is about sizing the DC link capacitor so i'll just stay on this topic.

Source inductance is what matters (the battery inductance), not the output inductance (motor phase inductance).

you'll notice in my attached PDF (previous post) this is mentioned as second last bullet. it is not usually the driving factor in determining DC link capacitor size in my experience.

if the source impedance was the driving factor.. how could that be? i can think of two examples where battery source impedance might become the dominant factor in sizing DC link capacitor, but i would argue that in both examples there is a better solution than to simply install a monster huge amount of cap.

scenario 1:
when the motor is asked to accelerate, the phase current goes up a lot to manage the torque, as phase current goes up the voltage droops and it is the job of the battery to supply the current requested. but what happens if the battery can not supply the current at the rate required (because internal source impedance is high).. well, then the DC link capacitor will begin to supply it. but this is not normal for cap to do, so if this is the case in your system.. then yes you would need a bigger DC link cap in order to allow it to assist. or you can buy better batteries because clearly they are the wrong type for your application, or you can accelerate to desired speed over longer period of time.

scenario 2:
on other hand, during heavy decleration and assume you have regenerative breaking.. the inertia of the motor and its stored kinetic energy will be transformerd by the motor (while it is acting as a generator) to supply current to your battery. but if your battery has high source impedance then the current will not be absorbed at a rate fast enough and so the bus voltage increases. so again one solution would be to have large DC link cap for purpose of storing this regen energy... or, you can buy better batteries or reduce the amount of regen energy allowable (software adjustable) which basically means decelerate slower and/or use mechanical breaks.

i'm happy with my method, it is proven with over a decade of successful application. but as my old professor says about design "there are many ways to skin a cat".

 

[Njay]

My friend h0tr0d sent me this paper some time ago, and he reminded me of it when he saw your method HH:

http://www.ecicaps.com/pdf/whitepapers/IEMDC_2009_11310_Final_Rev_4.pdf

 

[Teh Stork]

My friend h0tr0d sent me this paper some time ago, and he reminded me of it when he saw your method HH:

http://www.ecicaps.com/pdf/whitepapers/IEMDC_2009_11310_Final_Rev_4.pdf

This document:Selecting Film Bus Link Capacitors
For High Performance Inverter Applications
Why won't it link :S - Selecting Film Bus Link Capacitors
For High Performance Inverter Applications: from Epicaps

That article is fatally flawed and very, very wrong. Their entire analysis is based on the assumption that somehow the input current is equal to the output current, which is not at all correct. There are several other errors - or at least unstated assumptions - but that initial assumption makes the entire article worth exactly nothing. I could write quite extensively and explain exactly why that's wrong and how you actually need to do it, but I have a feeling that would be a waste of my time. I would suggest you start with some very basic reading on these topics - there are lots of good app notes and such available online, and anything related to buck converters is applicable to BLDC controllers as well (3-phase buck, basically). You're going to find yourself very far over your head very quickly (if not already) unless you get a better grasp of the basic theory at work here.

That film capacitance calculation paper is shit. You can visit my own thread early on when I started to work with 3-phase controllers. In that topic you will find a novice electrical engineer high on application notes and white papers, with some of the basics all wrong. Thats me, in case you didn't get it - two years ago.

i'm happy with my method, it is proven with over a decade of successful application. but as my old professor says about design "there are many ways to skin a cat".

Show me the paper where calculated voltage ripple and measured voltage ripple show the same. That Ecicaps document certainly does not describe it, I've tried.

 

[h0tr0d]

My friend h0tr0d sent me this paper some time ago, and he reminded me of it when he saw your method HH:

http://www.ecicaps.com/pdf/whitepapers/IEMDC_2009_11310_Final_Rev_4.pdf

If I read correctly, the formulas are the same as HighHopes wrote.

Ripple Current:
delta_I = d*(1-d)*Vbus/(f*L)
delta_I = 0.5*(1-0.5)*72V/(100kHz*2.64uH)

That film capacitance calculation paper is shit.

Please don't express yourself like that.
If your objective is just to share knowledge, phrase construction like that just "distracts" the reader from your objective, you been either right or wrong, it doesn't matter...

 

[h0tr0d]

what are the symptoms if you don't have enough cap ?

I've experienced that in my ebike (chinese) controller, 48V, 30A (battery)
I had 2 shunts and added a third for more current (42A measured afterwards) and, with 3x470uF caps, the performance off-the-line was the same as before, kidda disappointing... Only mid-range accel was improved.
Then I replaced 2x470uF for 2x2200uF and then I really felt the intended increase in performance from the extra shunt.

Regarding ESL, I think it's much more important to use a bunch caps in parallel then selecting an ubber special ultra low ESL cap (10 caps in parallel have 1/10 of the ESL of 1).

My 2 cents.

Edit: Found this - http://www.eetimes.com/document.asp?doc_id=1273212
And this - http://www.ti.com/lit/an/slta055/slta055.pdf

 

[PowerPedro]

Interresting topic!! I wonder if anyone can share real measurements of ripple current?

 

[Njay]

The closest I've been from that was seeing the transient voltage drop between the input point and the caps legs. If I knew the loop's inductance I would know the current. It's a difficult measure to make, specially without influencing the layout.

So that currently leaves me to seeing it in simulation, which leaves me worried about excessive ripple current in 2 fronts:

1) Small values of capacitance can resonate badly with the inductance to the power source (just as I have mentioned in the 1st post on this topic). If reality agrees with the sims, I think that kind of excludes using small values on at least any motorcycle or bigger size electric vehicle.

2) Ripple currents travel between the high value caps and the smaller "snubber" caps increasing RMS ripple current on the caps by possibly substantial amounts, unless there's some relevant ESR in one of the cap "levels". This one was totally unexpected, but still it is only simulation, not seen in reality.

Most part of the difficulty in clarifying this lies in the lack of real ESR/ESL data from real caps and doing the measurements.

 

[HighHopes]

resonance and balance of energy are two key issues when sticking an inverter into a wider system, you are right to be thoughtful on the matter. i've worked with experts on this subject and major capacitor suppliers directly with custom design stuff and have gone through all this before. many inverters all power levels between 1kW & 100kW and always with the "small" capacitors & snubbers. all evaulated in EMI chambers the size of a 3 car garage, two stories tall. what i've found is that unless you have big $ and expert assistance it is virtually impossible to simulate ahead of time what the real resonant frequency(s) would be and if that frequency would be excited by something in the system. in my experience, its a real issue but is VERY RARE to be a complication when designing in this way, poly caps, snubber & laminated bus bars. the DC supply cables need to go first to the DC link caps (poly caps) and the snubbers (also special type) must be physically across each phase leg which means you need 3. in most cases, unless big budget project, its just easier to design in normal way and hope for no problem and in rare case that you have a problem make some quick last minute changes to solve and move on.

but its a really interesting subject and worth learning about. perhaps in the future our tools and prediction methods will improve, but for now, its good to be aware and to have some tricks up your sleeve ready for the unlikely possibility you have this problem but otherwise don't overly concern yourself.

$0.02

 

[Njay]

Really interesting indeed. Regarding the DC link caps, the thing is, it seems easier to just stick a big capacitance there, as the power source output impedance will always be happy even when laying out a loop of 2 meters of cable with a "big" loop area. And, of course, if going to "big" capacitance, economics at our level says it has to be eletrolytic.

What kind of power source you had designed your inverters with, HH? And at what kind of lengths are your inverters from the power source? Being able to choose, one can always choose to have longer wiring to the motor (where more inductance actually "helps") than to the power source. But in some systems like electric vehicles you may have a big loop area in the battery-controller due to physical constraints (on one side, tenth's of Ah batteries are big and is not easy to have a small wiring loop area, on another side it is not uncommon to see let's say car conversions where the battery pack is split between front and rear).

The resonance between the 2 levels of caps was a really surprise for me. It's also interesting because it's just the same problem all over again but at a different scale.

 

[HighHopes]

in the EMI chamber i had luxury to test all sorts of combinations of cable lengths & source type. i never tried battery though while in EMI chamber, which unfortunately is most intersting for you. fairly uncommon to have batteries in aerospace (see news article about boeing 787 "more electric aircraft" for examples of why this has been avoided for many years). most my inverters were powered from a generator rectified into DC. i have used batteries only once or twice.

maybe it makes a really big difference when supplied by a battery which i do not have much experience with.

if you put big electolytic caps just make sure you put enough in parallel to deal with the ripple current.. you must avoid cap heating

 

[Njay]

maybe it makes a really big difference when supplied by a battery which i do not have much experience with.

At 1st sight I would say the generator is worse. I guess concerning just the wiring connecting power source and controller it's the same being batteries or generator, but the generator has "issues" not present in batteries like the "load dump" and the sort I believe, which I believe you had to take care of too.

 

[h0tr0d]

is VERY RARE to be a complication when designing in this way, poly caps, snubber & laminated bus bars

NJay, move along, nothing to see here... :wink:
Let's burn some FET's, mate!!!

 

[Njay]

Too expensive to burn without knowing why

 

[h0tr0d]

I know the way you'll test your system, so no problem.

You will burn stuff, BUT you WILL know why, I'm bloody sure of that... :wink:

 

[okashira]

BLDC/Induction:
BLDC motors have magnets so no need to ask inverter to supply magnetic field (medium through which torque is transferred to rotor). But, for induction motor this current (Id) comes from inverter since the machine has no magnets. It is not possible for the battery to supply reactive energy so it all comes from the cap meaning the capacity has to be about 15% higher.

Can you clarify this more? How is the 15% number determined? I like equations. )

 

[mrass]

this means your capacitor will heat up which takes away its life. if it is 5000Hour rated cap, it becomes less. 1/2 life for every 5deg C rise.

Yes and Luckily not, a 5000 hour rated electrolytic cap is rated for 5000 hour of "useful life" at the rated temperature (typically 85 or 105dgr.C) and at the rated voltage.
Any deviation below that will expand its expected life by a huge amount. (think of servers running 24/7 at 40dgr.C for several years).

for voltage rating, it is not too big a concern for poly cap.. but electrolytic, it is well advised to have at least 50% margin. 100V battery pack = 200V DC link capacitor rating. Vcap/Vbattery >= 2. For any ratio <2 it will affect how quickly the electrolite dries over time.. so just another example of loss of expected endurance/life.

Also luckily not, surely a margin is needed, at least 10% worst case (so including spikes) should be enough, bigger margin longer life.

Earlier in this thread somebody stated that ESR goes down with temperature, this true and also the opposite is true, It could double when going from 20 dgr.C to -10 dgr.C!
Sure they would warm up from the increased loss but already messed up your circuit, my rule of thumb: get as much capacitors of the highest quality in as possible.

 

[HighHopes]

Can you clarify this more? How is the 15% number determined?

there may be an equation, but i don't know what it is off the top of my head. this number comes from looking at the relative magnitude in Iq (talking about FOC here). Id is the field producing, and i think this comes primarily from the capacitor (not the traction pack). what portion is Id to Iq? you can see this in matlab/simulink. might be you can do more research here and find for us a more suitable % when driving induction machines.

 

[okashira]

Can you clarify this more? How is the 15% number determined?

there may be an equation, but i don't know what it is off the top of my head. this number comes from looking at the relative magnitude in Iq (talking about FOC here). Id is the field producing, and i think this comes primarily from the capacitor (not the traction pack). what portion is Id to Iq? you can see this in matlab/simulink. might be you can do more research here and find for us a more suitable % when driving induction machines.

Yup I figured it. It's totally dependent on your Id

 

[madin88]

without reading the whole thread: i am right that the more µF and the lower the ESR, the better it is in gernally for the FET's?
at least thats what i catched in another controller discussion and from adaptto engineers.. or could there be a disadcantage of having such "stiff" supply line to the FET's?