Hardware design

Power Electronics Design

Kyros Engineering designs switching power electronics — DC-DC converters, inverters, motor drives, EV charging and battery systems — from topology selection and magnetics through gate drive, thermal design, control-loop compensation and EMC. The work is led by an engineer whose graduate training is in control systems and power conversion, which matters because most power designs that misbehave on the bench are failing in the loop rather than in the parts list.

Scope this with us
What we work inBuck / boost / flyback / forwardLLC & phase-shifted full bridgeThree-phase invertersGate drive & deadtime designMagnetics designControl-loop compensationDigital control (STM32, HRTIM)Current & voltage sensingThermal design & validationPower-factor correctionLTspice simulationEMC-aware layout

Power electronics punishes the gap between a simulation that converges and a board that survives. The topology is usually the easy decision. What breaks programs is everything downstream of it: the gate loop that rings, the transformer that saturates at the corner case nobody swept, the compensation that is stable at nominal load and oscillates at ten percent.

We are usually brought in at one of two moments — before the topology is committed, when the decision still costs nothing, or after a prototype starts doing something the simulation said it would not.

Where power designs actually come apart

The failure modes repeat. After enough bring-ups you check the same short list before believing anything else.

  • Gate-loop inductance — layout, not the driver, is usually what is causing the ringing and the lost shoot-through margin
  • Magnetics designed for the nominal operating point and never checked at startup, at fault, or at the thermal corner
  • Control-loop compensation tuned at one load and never characterised across the range, so stability is a coincidence
  • Thermal paths validated by a spreadsheet rather than by a thermocouple on a running board in the actual enclosure
  • Current sensing whose bandwidth and delay were never budgeted into the loop that depends on it
  • Startup, inrush and fault behaviour treated as edge cases instead of as requirements

What we design

Converters and drives where efficiency, thermal behaviour or control quality are real requirements rather than a datasheet number.

  • Isolated and non-isolated DC-DC — buck, boost, flyback, forward, LLC and phase-shifted full bridge
  • Inverters and motor drives, including multi-motor coordination and high-resolution PWM
  • EV charging power stages and the control that supervises them
  • Battery systems: charge control, cell monitoring, balancing, contactor and fault logic
  • Precision and low-noise rails for instrumentation, where the supply is part of the measurement chain
  • AC front ends with power-factor correction and realistic inrush and holdup behaviour

Control loops treated as engineering, not tuning

We model the plant, design the compensator against a stability margin you can defend, and then measure the loop on hardware rather than declaring victory because the output looks flat. Where the control is digital, the sampling delay, the ADC acquisition window and the PWM update boundary are budgeted into the model instead of being discovered later as phase margin that went missing. Protection gets the same treatment: overcurrent, overtemperature and fault response are designed with the loop, not bolted on after it.

You probably want this if…

A converter works at nominal load and misbehaves at the extremes of the range
The design passes on the bench and fails EMC pre-compliance
Efficiency or thermal numbers are short of target and nobody can say which loss dominates
You are choosing a topology and the decision drives the magnetics, the enclosure and the cost
Something is ringing, running hot, or failing intermittently and the simulation says it should not

Frequently asked

Do you design the power stage, the control, or both?

Both, and that is usually the point of hiring us. Splitting the power stage from its control across two vendors is how a design ends up with a compensator that is correct for a plant nobody re-measured after the magnetics changed. Where a client already has one half in hand, we take the other and re-verify the interface between them.

Can you help before we have committed to a topology?

That is the cheapest hour you will spend on the whole program. A two-week architecture diagnostic at $6,000 puts the topology trade — efficiency, thermal, cost, size, EMC risk and part availability — in writing, and you keep the document whether or not we do the build.

Do you take designs through EMC?

We design for it from the start and run pre-compliance before a test-house booking is made. Finding a 20 dB emissions problem in a chamber at a day rate is the expensive way to learn something a near-field probe and a spectrum analyser would have shown a month earlier.

What size of power design do you take on?

From sub-watt precision rails to multi-kilowatt drives and charging stages. The size matters less than whether the requirements are real. A 5 W design with a genuine noise budget is more engineering than a 5 kW design that only has to be roughly efficient.

Power design doing something the simulation didn't predict?

Send the schematic and the waveform. We will tell you what we think it is before you commit to a scope.