VoxSolaris: The Voice of the Sun
  The VoxSolaris Variable Voltage Project

Project Origins

The usual method of converting DC to DC at another voltage is to use a switch mode converter of some description. Such devices can be designed to supply a fixed or variable voltage output, accept a fixed or variable input and if need be, handle several inputs and outputs at the same time. The flexibility of the switch mode converter is perfect for a very wide range of applications. The efficiency however falls short, and this can matter in high power applications. A good design will have an efficiency of the order of 95 percent. That is still a toll of 5 percent for the sake of having the correct voltage. In an electric car 5 percent of the range would be lost and that could make the difference between a straight drive home or a half-hour diversion to a charging station.

In an electric car with a DC motor, speed is ultimately controlled by setting the voltage. Accelerator position sets the required torque which in turn sets the required current. At any given time the required voltage will be that needed to overcome back emf (directly proportional to current speed), plus that needed to give the required current on the basis of the ohmic resistance of the curcuit as a whole. In a typical arrangement a fixed voltage battery feeds a switch mode converter which by means of rapid switching, charges short term energy stores (inductors and capacitors). Then the required voltage, calculated every 10 milliseconds or so from a reading from a sensor on the accelerator, is assembled, again by rapid switching, from the charged energy stores.

But in an electric car we already have charged energy stores in the form of battery cells so we can cut out the middleman and assemble the output directly. With this approach, half the switching is bypassed and in the half that remains, the hysteresis losses that bedevil the inductors and capacitors of the switch mode converter are essentially eliminated. For discharge leveling, cells are switched in on a rotating basis. While switched out, the cells can be individually tested. When the voltage of a cell falls below a safe threshold it is excluded from further dischage but cell-wise testing also facilitates superior battery management. Regular testing provides a picture of cell deterioration over time and cells can be excluded altogether if they fail to accept charge efficently. This contrasts sharply with batteries comprising a fixed number of cells connected in series where, depending on the type of cell used, one 'dead' cell can seriously undermine the performance of the battery as a whole. The output voltage is granular as it is essentially always a multiple of the cell voltage but for an electric car with a high voltage motor this granularity is inperceptable.

The Retro Experiment

By assembling the output voltage directly from battery cells rather than the very short term energy stores of the switch mode converter, much less switching takes place. Instead of switching at 20KHz or more, switching only takes place every 10 milliseconds in response to the accelerator sensor readings. And although the cells are rotated, they are generally left switched in for many 10 milisecond cycles at a time. Slower switching did not improve the effeciency of the voltage assembly process as the power transistors still exhibited the characteristic voltage drop and consumed much the same power. But in the slower switching, of one of our committee members with a self confessed mistrust of electronics and of in-car electronics in particular, saw an opportunity to add a manual overide to the otherwise completely computer controlled system. At his request we benchmarked the transistors against solenoid operated mechanical switches that were switched on by a pulse and switched off by a pulse, rather than relays that are energized throughout the on phase. To our surprise, the switches performed far better than we expected. They took too long to switch (about 1 millisecond) which looked horrible on the oscilloscope and detracted from power output. But in terms of electrical efficiency, the switches comfortably beat the transistors. Experiments involving combinations of mechanical switches and power transistors are continuing. The current format is to have one power transistor between the power pack and the motor. This can be switched on and off to use the motor inductance as a means of reducing granularity and as a means of protecting the switches from sparking. The approach works better with large format cells as there are two switches per cell. Assuming large format cells the unit is smaller, weighs less and costs less than a switch mode converter.


The above describes how translation losses between battery cells and the variable voltage load, in this case an electric motor, are reduced. But the method of matching voltage by switching in an appropriate number of cells can equally be applied to the charge cycle if the input is DC. This is of relevence when considering range extenders, regenerative braking and other possible inputs such as the incorporation of solar body panels. With regenerative braking, the output voltage is cut below the back emf and tracked lower as the car slows. We have assumed there are no circumstances in which the brake and accelerator would be applied simultaneously so when the brake is applied the accelerator is ignored and treated as though the foot is off. Solar body pannels are just set to charge cells that are rotated out from the task of providing current to the motor.

The more complicated input to handle is that from the IC engine range extender. For highest efficiency two conditions need to be met. Firstly the IC engine needs to run within the most fuel efficient part of its operating range as much as possible. Secondly, while there are circumstances where it is appropriate to have the range extender recharge the cells, the power is better directed to propulsion motor as the cells don't give back more than around 85 percent of what is put in.

While the origins of the project owed much to an interest in more efficient electric cars, the project was given the name 'Variable Voltage' to acknowledge a wider application base. It is a comprehensive system for matching multiple power sources that may or may not have stable voltages to multiple power sinks that may or may not have stable voltage requirements. As well as a means to manage power in a car, its other, somewhat simpler objective is to manage power in stationary micro-generation scenarios with or without a grid connection.

Field Trails

The project is controlled by software from HadronZoo which is written in C++ and runs on the Arduino processor. The first part of this is complete but is aimed at prototyping. Before we publish anything we will need to test out several variations on the design and conduct formal field trials on the more promising designs. We are hoping to have this done before the end of 2020. Pending legal advice we will at that point publish the circuit diagrams, configuration documentation and the software under a general public license.