High-power multimodular matrix converters (MMMCs) comprising multiple threephase to single-phase matrix converter modules have emerged as a viable topology candidate for medium-voltage adjustable speed drives. As a combination of direct power conversion and cascaded multilevel structure, the MMMCs inherit features such as elimination of dc capacitors, four quadrant operation capability, employment of low-voltage devices only, and superior output waveform quality under a limited device switching frequency. Due to their particular topological structure, modulation scheme design for the MMMCs is not straightforward and complicated. The presented work is mainly focused on development of suitable modulation schemes for the MMMCs. Several viable schemes as well as their corresponding switching patterns are proposed and verified by both simulation and experimental results.

In order for the MMMCs to produce sinusoidal waveforms at both input and output ac terminals, a direct transfer matrix based modulation scheme is presented. It is revealed that a suitable modulation strategy for the MMMCs should aim at fabricating the total input current on the primary side of the isolation transformer. For topologies with more than two modules in cascade on each output phase, switching period displacement is necessary among modules to generate multilevel output waveforms.

An indirect space vector based modulation scheme for the MMMCs is developed. With a few presumptions satisfied and viewed from a certain perspective, the MMMCs can still be modeled indirectly and be divided into fictitious rectifier and inverter stages. Therefore, space vector modulation methods can be independently applied to both stages for duty ratio calculation, before the results are converted and combined for determining per-phase output pulses. A new output switching pattern providing improved harmonic performance is also proposed.

A novel modulation scheme based on diode rectifier emulation and phase-shifted sinusoidal pulse-width modulation is proposed. The method sacrifices input power factor adjustment, but enables the use of an indirect module construction leading to significantly reduced device count and complexity. Strategy for reducing additional switchings caused by input voltage ripples is also implemented and explained.

In addition to simulation verifications, all the proposed schemes are further tested experimentally on a low-voltage prototype built in the lab. Details about the prototype implementation are introduced.