Bus capacitor placement for 5kW inverter

[jonathan_nadar]

Hi, I am currently developing a PMSM inverter rated for 5kW continuous power running off a 48V(nominal) Li-ion battery pack. I am using nine 330uF electrolytic capacitors across the bus along with few film capacitors and MLCCs. I want to understand the implications of using topology A instead of topology B attached to this post. Using topology A makes my PCB routing straightforward and I can keep the loop formed by the half bridge and the capacitor the smallest. I am using FOC, so current sensing is quite crucial to me. I am doing non isolated, differential current sensing across the shunt using TSV912.

Thanks in advance.

[Siwastaja]

With A, you cannot sense instantaneous current, only average current. With A, you can also just use one current sense resistor + amp instead of sensing 2 phases.

Average current control is workable and some inverters do that, but I don't see the point when it's not much more expensive to sense the two instantaneous phase currents, i.e., topology B.

I don't see large difference in layout. Just use shunt resistor as small as possible. I have used 1206 SMD parts up to some ~500W power level and 2512 rated for 3W up from there. But maybe the 5kW/48V needs a tad bigger, maybe not. It's about how much accuracy you need; but I wouldn't be scared of using some well-specified x100 shunt amplifiers just to get the resistor dissipation down, and package small. +/-10% current sense accuracy would be more than fine for basic torque control, after all.

[Robotec]

I would go with B as well if Resistor disspation is a problem you can always put severall in parallel.

[Siwastaja]

I would go with B as well if Resistor disspation is a problem you can always put severall in parallel.

Yeah, OTOH paralleling makes accurate Kelvin sensing more difficult. There are ways to separately sense the resistors and average the small voltages before amplification, but this takes space and increases cost, too.

So it's often best to use a single suitable resistor, even if that means you go to very small burden voltages and have to amplify more, sacrificing some accuracy, because alternative of parallel/series connection would also sacrifice accuracy just in different ways, plus have more power loss.

[jonathan_nadar]

To add on to the information, I have place sense resistors for two phases, and a x10 gain. I'll share an image of the layout tomorrow.