The task at hand is to build a 50kW (e.g. say 125VDC 400A out) DC-DC converter that charges a large lead-acid stack with said 400A and accepts a somewhat wide input voltage range. With that in mind, if we allow the input voltage range to be 50-260V, the converter topology will need to be zeta (or similar - buck-boost, SEPIC, Ćuk). If I instead ensure the input voltage is always higher than the output (e.g. 150-800), then the buck topology is likely the best choice (multi-phase buck, a vastly scaled up version of the 12V->Vcore part in modern PC motherboards). The zeta topology however gives me more flexibility in the specific application, so I'm wondering - could it also be scaled to such power levels?


Some guys are building an electric motor testing bench, and the part that I might get involved is the powerful DC-DC converter for energy recuperation. The idea is that a stack of power lead-acid cells (forklift-duty types) would power an ESC, which commands the Motor Under Test. Another, larger motor, would act as a mechanical load to the M.U.T. That second motor, operated as a generator, would recuperate the energy back into the battery stack through a DC-DC converter, which should have programmable target output power (so that we can modulate the torque that the M.U.T. has to provide in order to keep the set RPM).

More current to the battery leads to more mechanical torque, so the M.U.T. will draw even more current to keep up, so in the end the battery will still be drained, but only with the power lost as inefficiencies in the system - not the full 50 kW.


  • The generator will be a 3-phase permanent-magnet type
  • The M.U.T. will be up to 50kW, so realistically the DC-DC converter only needs to handle 40kW. But I'll aim for 50kW to have some margin.
  • The battery voltage could be as low as 24V, and as high as 120V. Of course, 24V batteries will only be used for testing low-power low-voltage motors, which are nowhere near 50 kW.
  • 400A will be the max amps we'd want to get into the battery, regardless of its voltage. So 50kW are only with the highest battery voltages.
  • 50kW will only need to be sustained for a few seconds, and the entire test run (with varying power levels) is less than 1 minute.
  • We'll only build one, or a handful, of these converters. Component cost is not a huge priority. Development cost is more relevant.
  • Other guys are handling the mechanical part (motors, bearings, torque measurement, shafts, ...)

Zeta topology option

I've read a Microchip app note (High-Power CC/CV Battery Charger Using an Inverse SEPIC (Zeta) Topology) which presents a 100W charger in the same general direction. The central part of the converter is this:

Zeta converter presented in the Microchip app note I was thinking that maybe, maybe, if I uprate the power components for voltage and current, and provide many more parallel copies of C1,C7,C8, Q1, Q2, the circuit can be scaled to tens of kW. Then it would give me more flexibility, as this way I don't need to worry about the corner cases that the other option has.

Buck topology option

Instead I can push the guys to select a motor with a very low Kv for the generator (or use a transformer after it), so that the input voltage to the DC-DC converter is always higher than the desired output. This either calls for a motor with ridiculously low Kv, or a 3-phase transformer, if I want to cover the worst possible case: battery voltage 120V, test at low RPM. The transformer options is likely the best, if also provided with bypass contactors to handle the low battery voltage case as well (to avoid ridiculously large input:output voltage ratios).

Other thoughts

  • Of course the option to lose the energy from the generator in a resistive grid is possible, but is viewed as inefficient and will cost more to provide the necessary cooling
  • probably a permanent-magnet generator is not the best idea. Come to think of it, using an alternator is maybe a fruitful direction, I'm not sure why the guys are fixated on PM motors, maybe I'll be able to persuade them otherwise.
  • \$\begingroup\$ I’m curious who defined the low battery voltage and high current that neglects the preferred higher impedance used by industry of e-cars batteryuniversity.com/learn/article/electric_vehicle_ev \$\endgroup\$ Dec 18, 2019 at 4:26
  • \$\begingroup\$ See here for P/I preferred slope engineeringtoolbox.com/… \$\endgroup\$ Dec 18, 2019 at 4:46
  • \$\begingroup\$ Yeah, beats me too... but keep in mind these are PM machines, and are targeted more at electric forklifts, karts, and such. That might be part of the answer. \$\endgroup\$
    – anrieff
    Dec 18, 2019 at 5:51
  • \$\begingroup\$ Will you rewire the batteries or only use part of the stack when testing a low voltage motor? Won't that result in imbalance between your batteries? \$\endgroup\$
    – bobflux
    Dec 18, 2019 at 14:01
  • \$\begingroup\$ How low do you have to go wrt rpm/frequency? If you want to measure motor stall torque you will have to use the generator as a motor to create opposing torque to the MUT... This is probably true at low rpm too \$\endgroup\$
    – bobflux
    Dec 18, 2019 at 14:18

3 Answers 3


Is the (insert name) DC-DC converter topology scalable to (insert impressive power level)?

Yes. OK, next question.

Oh, you meant reasonably scalable, didn't you?

The problem with the Ćuk, SEPIC & zeta converters are that they carry considerable power in the filter capacitors, and they need multiple inductors. So they may not scale as well as other options.

A 4-switch buck-boost should scale better than the above options, because you'll only need one inductor and one set of filter caps, at the expense of needing four switches.

I think if I were you I'd explore a multi-phase 4-switch (per bank) buck boost. I know that if I were the one doing this there'd be a microcontroller controlling it, but whether you do that or seek out some controller chips that'll do the job for you is a strategic decision that depends on the team you have available.

  • \$\begingroup\$ Yes, you guessed correctly. Reasonably scalable is it. And thank you for the 4-switch buck-boost, I wasn't familiar with it, seems to fit the bill quite well. \$\endgroup\$
    – anrieff
    Dec 19, 2019 at 21:24

Sure the Zeta does not scale too well to 50Kw. Buck or boost would scale better.

If you want to load up your DC motors and recycle energy then have you considered a wound field machine? This could be another forklift motor or maybe an alternator. With this approach the power that you need to control is a small fraction of 50 kW.

  • \$\begingroup\$ Yes, as stated in the question, I'll pursue the guys to use an alternator, if it's feasible to build one. \$\endgroup\$
    – anrieff
    Dec 19, 2019 at 21:25

Frame challenge : A 50kW test lasting 1 minute uses less than 1 kWh, or about 15c in electricity.

how many of these tests will you be doing?

Before choosing anything more complex than a resistive load, how many tests will it take to pay back the cost of a recuperator?

EDIT : for cooling, site the resistor bank outdoors in summer.

This assumes the lead-acid bank is still used for powering tests; if there is a long time between tests, it can be charged relatively slowly; 1kW would suffice if setup/teardown took an hour. (I'd be looking at LiFePO4 nowadays for its much better cycle life and its currently low and falling cost)

Alternatively given an AC output from the generator that is higher than the battery voltage, regeneration can be simple, just a 3-phase rectifier and a resistor (bank) to control (adjust) charge current.

If voltage boost is necessary, there are other ways to achieve it; either a transformer between generator and rectifier, or a manual gearbox between motor and generator.

  • \$\begingroup\$ Then you'd need to find a good AC/DC converter with 50kW capability and these are much more expensive than 50kW DC/DC converters (I believe due to the fact AC/DC is usually isolated). \$\endgroup\$
    – anrieff
    Jun 16, 2021 at 23:30
  • \$\begingroup\$ Ah, I thought from the question, you were powering the 50kW test from lead-acid batteries, which can be charged at much lower power (e.g. 1kW if it takes an hour to setup each test) A 1kW charger won't cost much. \$\endgroup\$
    – user16324
    Jun 17, 2021 at 11:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.