François Boige, Power Design Technologies, France
Efficiency vs cost is THE dilemma of any power conversion stage design. The competition for the best offer in terms of power converter efficiency and the final price, is the key for being competitive in the energy transition. On the other hand, the converter losses (Power x efficiency) are paid by the end user. Most of the time, the best converter is the one with the highest efficiency and most competitive cost. However, those two parameters are opposed, and reaching the right efficiency point can be really expensive for companies: jumping from 98% to 99% efficiency, could double the converter cost.
In power electronic conversion, the losses are mainly created by the power semiconductor devices in conduction and switching. Let’s see how reducing these losses impact the cost.
- The conduction losses can be reduced by using a larger semiconductor section but it increases the cost as a bigger die or module need to be used. Moreover, this solution impacts negatively the switching losses as the parasitic elements are increased.
The switching losses can be reduced by:
- Lowering the switching frequency, but it will increase the filtering needs as it needs more inductance and capacitance to keep the same ripple level. In that case, the inductors and capacitors are going to be more expensive.
- Using wide bandgap (WBG) semiconductor devices as their switching speed is higher but they are more expensive (5 to 10 times silicon devices).
- Using multi-level topologies as the voltage switched by the semiconductor is lower with a lower voltage rating. In term of cost, more semiconductor devices are used but with a lower voltage rating. As the devices cost is non-linear to the voltage rating, the semiconductor overall cost will depend on the applications. The filtering multilevel needs is usually more important in term of input filter (Flying Capacitance, NPC, T-type topologies) but with smaller output filter. The complexity cost is also to be considered.
- The overall semiconductor loss reduction decreases the cooling device size and cost needed to limit the semiconductor junction temperature.
The case of a 3-phase inverter with a sine filter has been studied with PowerForge in order to illustrate the previous assertions.
The inverter nominal operating point is at 200kW with a DC bus of 800V, 400V between phases, a frequency of 50Hz, cos(φ)=0.8 and Space vector modulation.
Si IGBT vs SiC MOSFET
The figure of merit between Si IGBT and SiC MOSFET in the efficiency vs cost plan allows to separate the plan in two :
- the solutions with efficiency below 98.3% equipped with Si IGBT device have the lowest cost
- the solutions above this efficiency equipped with SiC MOSFET have the lowest cost. To explain this result, SiC MOSFET is more expensive but increasing the switching frequency has a small impact on the loss, while reducing greatly the sine filter cost. At high switching frequency, the gain on filter will be reduced by an additional heatsink price and additional losses.
2 Level topologies vs Multilevel topologies
The result from this comparison are not intuitive at a first glance. Below 98.7% efficiency, 4L Flying capacitor with Si IGBT are less expensive. Indeed, while being more expensive in terms of semiconductor devices and capacitor, the AC filter is greatly reduced by the apparent output switching frequency. At low frequency, the Si IGBT is more performant in terms of efficiency and cost. In this application T-Type and NPC are not competitive because they need a large amount of capacitor in the DC bus to keep the imposed 1% DC bus voltage ripple between the middle point and the bus voltage, which greatly impacts the cost.
2 Level topologies vs Multilevel topologies vs Si IGBT vs SiC MOSFET
The combination of the topologies and the semiconductor device technology has been tested as shown in Fig. 3. In term of efficiency vs cost, Flying Capacitor with Si and 2 Levels with SiC are forming best solutions.
The solutions presented do not exclude the existence of a better solution. In particular, loosen the ripple constraints on NPC and T-Type middle point, evaluate the modulation strategy influence or other semiconductor device as well as refine the optimal switching frequencies for this converter. Finally, the best presented conclusion only have a meaning for this specific converter specifications and cannot be generalized.
PowerForge is meant to address power converter early phase design and find the BEST POWER CONVERTER TRADE-OFF.
The best challenge is to find the best Trade-off between COST and EFFICIENCY