Background and requirements
Modular control solutions for modular power converters
In response to the growing modularization of the power stage, some modularization of the control stage is required too! In practice, building large and/or complex digital control systems by simply assembling standardized control devices is uneasy. There several challenges to that:
Main challenges related to the modular design of power electronic control systems.
CLOSING FAST CONTROL LOOPS ANYWHERE ON THE NETWORK
One aspect of distributed converter control is particularly important so that the modularized controllers can be a true alternative to monolithic controllers. It is that the communication is sufficiently fast for supporting control loops that are closed across the network. In other words, inputs and outputs from anywhere on the network should not suffer from extra delays comparatively to a monolithic controller.
Fast local control loops
Local measurement inputs and modulator outputs can be used for local control loops.
Fast global control loops
Remote I/Os can also be involved in closed control loops, essentially with the same control bandwidth as local loops.
Using the most effective networking topology
Intuitively, when most delays occur inside the devices, it becomes obvious that the network itself isn’t fully utilized. By introducing the concept of topology efficiency (ET), this can be quantified as the percentage of time during which the communication links are used. Then, studying ET as a function of the network topology, it can be concluded that radial networks – i.e. tree-shaped – offer superior efficiency.
Intuitively, this can be relatively easily understood (as long as data congestion does not occur). Indeed, within a tree-shaped network, the longest path is proportional to the number of levels (i.e forks along the path) rather than the number of nodes. The closer are the slaves from the master, the shorter the delay.
1-tree (1 slave per node)
1-tree (1 slave per node)
2-tree (2 slaves per node)
The typical data traffic between distributed converter controllers is constituted by numerous small packets, which are exchanged at a relatively high frequency (up to >100kHz). In such a case, the overall network is best used when its nodes are interconnected by forming a pyramidal network.
Thanks to our unique tree-shaped control network, we can take the most of the improved DNL / DPT ratio, in other words, of the newest FPGA technology.
Topology is indifferent for very large payloads, but 2-tree is more than twice better for 25-40 Bytes!
Conventional high-speed serial communication
- 1-tree network topology (e.g. ring or string)
- Poorly exploited DNL / DPT ratio
Packets are processed one after the other. Therefore, the total delay corresponds to N+1 hops, which is close to the number of nodes N.
Low-latency communication with RealSync
- 2-tree network topology
- Takes advantage of high DNL / DPT
Packets that converge towards the master are aggregated. Besides, the total delay is proportional to the number of levels (rather than nodes).