#HardwareStack as a 12kW three phase active front end...

This afternoon has been an exciting (and hot) one in the lab at Converter Technology. Jon has now achieved 12.5kW from his HardwareStack based active front end (AFE). If you've not come across an AFE before, essentially it is a three phase power factor corrector which operates directly on the three phase supply to generate a high voltage DC link. This high voltage DC link is then used to drive another downstream converter or maybe an inverter for a motor. From a utility perspective, the benefits of an AFE are the same as a single phase power factor correction (PFC) system in that excellent power factor can be achieved.

If you are more used to the world of single phase PFC then the three phase active front end is rather different:-

• No bridge rectifier is used on the front end of the AFE power stage and in terms of power flow, it is natively bi-directional
• The DQ control scheme uses a phase locked loop (PLL) and generates sinusoidal current demand signals which have a controlled phase relationship to the voltage waveforms
• Space vector modulation is used to generate the required switching patterns on the six actively controlled power devices (SiC in our case)

The DQ control scheme uses a PLL implemented in the digital controller on our HardwareStack control board. The output of the PLL is a value from zero to 2*pi representing how far through the mains cycle we are at any point in time. Whilst this PLL value is an internal control variable, we are able to view it directly on a our 'scope by feeding it out via on of the three precision DAC's we have implemented on our control board. The plot below shows the PLL value (Green), the sector information for our space vector modulator (Blue) and the Vd Value (Red) used in the control scheme. These parameters are normally buried in the software of the control algorithm but we find it very useful to be able to display them as waveforms on a scope in real time for debug purposes using the advanced features of our HardwareStack platform.

In terms of performance, the power meter indicates an input power of 12.5kW and the output load is measured as 12.2kW. This gives an effective efficiency of 97.6%. The phase current (Red, 10A/div) and output voltage (Yellow, 200V/div) are shown below.

We are now working to improve the current waveform as there is some clear distortion present, especially around the zero cross point. Presently the power stage runs with relatively high dead-time in the switching legs and it is likely that this is creating non-linear behaviour around zero crossing of the current.

We will post more on this converter as we progress in the lab. It is really great to see it up and running. More importantly, this is another great example of how #HardwareStack can be used to rapidly prototype new converters in the early stages of a project in order to reduce risk. If you would like to learn about other converters we have implemented on the system, see the links below to our earlier posts.

98% Efficient 14kW LLC Converter using HardwareStack

99% Efficient High Voltage Buck Converter using HardwareStack

Thanks for reading!

19th January 2017 update

Follow on information - we now have some more data looking at the impact of deadtime on current harmonic distortion. The first two pictures below show the simulation results running with minimal deadtime and the second with 400ns (i.e. 4% of the 10us switching period)

You can see in these simulated results that the deadtime is a significant cause of distortion. Jon also captured the current waveform from the real hardware with 400ns deadtime and we get:-

The results are fairly similar, just a bit smoother in real life as we don't model the impact of switching node capacitance.

Interestingly, this deadtime effect is exactly the same problem faced by class-D audio amplifiers which was a research area interest of mine back in the 90's. I wrote a paper modelling the harmonic distortion effects so if you are interested, you can download this below.

Class-D Distortion and Dead Time Paper