Download Drives, data-acquisition systems and control electronics all require

Drives, data-acquisition systems and control electronics all require energy to do their job. This
energy comes from a source: a battery, an accumulator or a direct connection to the electrical
network. In some cases, the energy source is connected directly to the device. In others, the
available voltage must be adjusted to your needs.
Paul van Haren
Alex van den Heuvel
There is a wide range of power-supply products available on the market and, for a variety of devices,
there are also suitable ready-made components available. For cost reasons, if these ready-made
components are suitable for use with the application , we usually recommended using them. However,
sometimes an application has special requirements which an off-the-shelf power supply cannot
achieve. It is then necessary to develop an application-specific power supply.
Engineers place many different requirements on power supplies, which ultimately affect the final
design. These requirements fall into five categories: function, power, quality, environmental factors
and safety. The basic function of a power supply is to deliver power to a device in the form of voltage
and current. However, there may also be other functional requirements such as insulation between
primary and secondary voltages, diagnostics and monitoring, and control interfaces. Quality standards
describe the required accuracy and the amount of ripple and noise permitted on the supplied voltage
and current. Some devices, such as motors, can deliver short-term energy back to the power supply.
When a motor like this stops quickly, the kinetic energy is transformed into electrical energy and is
delivered back to the power supply. These types of devices must therefore not only be able to supply
energy, but must also be able to receive it. If a power supply is used to energise an actuator such as a
coil or a motor, the control speed of the supply has a direct impact on the characteristics of the final
control loop. And of course, a power supply must always operate stably, even during variable loading.
This is more easily said than done. Stability often poses limits on achieving other requirements such
as control speed.
The constraints on the power supply are largely determined by environmental factors. The operating
temperature and the potential for cooling will impact the design. In many cases, they give rise to high
efficiency requirements. Indeed, the greater the efficiency of a power supply, the less heat there is to
be dissipated and the easier it is to cool. In high-humidity environments, additional requirements arise
with respect to isolation. Additional measures for the housing or filling the power supply with insulating
materials are often possible solutions. Finally, it is important to have a good idea of the amount of
vibration and shock the power supply will be exposed to. The weight of the components also varies
with the capacity of the power supply. A high-capacity supply contains heavier components. These
then are at risk of vibrating loose or being damaged by shock or vibration. Such risks particularly play
a role in automotive applications.
Product safety also deserves significant attention. The scope and intended use translate into
standards that a product must meet. In some cases, the safety of the product must be demonstrated
by obtaining approval from an authorised agency. For small or medium production components, this is
often a very large hurdle. A great deal can be gained by making a number of design choices early,
during the architecture phase.
Business Case
Once the product requirements are known, it is time to carefully examine the possibility of purchasing
a suitable ready-made power supply. It may be necessary to reduce particular product requirements
slightly in order to fit within the range of a standard power supply. A precise understanding of why the
requirements are as they are helps to assess whether purchased products are suitable for the
application.
If it is not possible to fulfil all the requirements using a purchased product - for example, in the case of
special power supplies - it may still be possible to use purchased components for the major parts of
the power supply. In our experience, purchased components, modules and semi-finished products that
meet the requirements are an attractive alternative to developing the power supplies ourselves. It
allows the designers time and space to concentrate on those features that cannot be achieved with
purchased components.
The business case will ultimately determine the choice between make or buy. For small quantities, the
cost of developing a specific power supply often weighs heavily on the total price. As the quantity
increases, it may be that optimising the power supply for the application in-house results in a
significant product-cost savings.
Ripple
Many of the power supply designs developed at Technolution are developed similar to "three-stage
rockets". In the first stage, the externally supplied voltage - the mains power line or other energy
source - is transformed into a safe and manageable voltage. Any fluctuations from the source are
cleaned up to produce a stable operating voltage for the following stages. The second stage is the
switched-mode power supply which uses a highly efficient manner to adjust the operating voltage to
achieve the lowest-possible dissipation in the third step. The third stage is a fine tuning that aims to
suppress the ripple and noise from the second stage.
During the architecture phase, a number of choices can be made that are important for the
requirements of the various stages. How are the requirements divided across the various stages?
What will the output voltage of the first stage be? And if a power supply must produce several output
voltages, it is helpful if the first stage can be shared between the outputs?
For the first stage, purchased products are often sufficient. Because safety requirements are important
in this stage, especially in medical applications, we often decide to use purchased modules with the
proper certification. This makes product-level certification easier. The additional costs of a purchased
module usually outweigh the design and certification costs.
The second stage is a switched-mode power supply which achieves the main functional requirements
of the whole. The third stage is an analogue regulator to suppress the ripple from the second stage,
which is intrinsic to switched-mode power supplies. In this way, a power supply with extremely low
ripple and noise levels is created while achieving high output at the same time.
The photo shows a detail of an
automotive power supply that can
deliver approximately 270 watts in
total. The source voltage in this
application can fluctuate
significantly. When the engine is
started, the vehicle battery voltage
will drop substantially. During
normal operation, while the vehicle
battery is charging, the voltage is
higher. Extremely high demands are
also placed on isolating the power
supply from disturbances in the
input voltage. In addition to the
stability of the various stages, it is
therefore also very important to
control the input impedance of this power supply properly. Oscillations are lurking everywhere and
are detrimental to the quality of the output voltages. Existing measuring equipment proved to be
inadequate for this design and it was necessary to develop our own tools to measure both the input
impedance and the transfer from the control loop. Furthermore, extremely strict conditions were made
regarding shock and vibration resistance and the power supply had to handle high operating
temperatures in a very small volume. The familiar production method of using both small as well as
large and heavy components next to each other ensured a great deal of attention had to be paid to
the arrangement of the components. Large components require large amounts of heat during the
soldering process, which affects the soldering of small components. The design and production of
such power supplies is not easy.
FPGAs
After defining the architecture and requirements for the various stages, the next step is to develop the
detailed design of the stages.
Dimensioning components requires calculations and simulations. There are two general techniques
used for simulations: detailed simulations based on cycle-by-cycle techniques and more general
techniques using average models with state-space averaging. The techniques are complementary.
The cycle-by-cycle techniques are very well-suited to calculating the effects of parasitic component
properties, but they require long calculation times. Simulations based on an average model provide
insight into the control characteristics of the power supply. These simulations show whether the design
will provide a stable power supply over the entire range. With long supply lines, there is a risk that the
cables will begin to oscillate with the power supply. Simulations highlight these types of undesirable
situations early and allow the effectiveness of countermeasures to be assessed quickly. The
simulations also provide an overall picture of the effects of component variation caused by tolerances.
The quality of the simulation results depends on many factors. The component models have limited
accuracy and the manufacturer does not specify some properties. With simulations, the old adage is
true - garbage in, garbage out. It is important the designer always understands what he is doing and
checks that the results do not contradict actual experience and conceptual logic.
In addition to all the simulations, calculations are also required. The expected life of components is
determined from the specifications and the characteristics in the application. The load on the
components and the operating temperature have a major impact on life. Calculations can be used to
demonstrate that the design meets the reliability requirements.
During the detailed design, the architecture choices made are also verified and validated. Sometimes
improvements are possible by modifying or redistributing the requirements of the various stages. As a
result, the design is simpler and thus less expensive.
To control power levels, ready-made building blocks can be implemented or controllers can be
developed using FPGAs. Using field-programmable gate arrays provides design flexibility for more
advanced controls. For example, with the aid of an FPGA, it is possible to produce a four-quadrant
power supply with outputs that must supply both positive and negative voltages as well as deliver or
absorb power.
Despite using state-space averaging techniques to implement simplifications, there is significant
agreement between the design simulation and the measurements of an early prototype, as can be
seen here in a response to a step change in the output load.
Constraints
Technolution normally implements a "first-time-right" design methodology. We challenge our designers
to create a design that meets all the requirements right from the start. For power supplies, we make an
exception. In our experience, it is wiser to produce a functional prototype that allows for variation and
experimentation with alternative components. Proper access to conduct measurements is also
provided. We assemble critical components in the circuit that are as true-to-life, i.e. as compact, as
possible. This allows parasitic components included in the final design to also be evaluated in the
prototype. Less critical components are provided more space to allow freedom to experiment.
We extensively evaluate and optimise the prototype. First we test the individual stages separately.
This phase concentrates on the parasitic properties of the components. The aim is to gather enough
insight that the circuit is working properly and that the simulation results are consistent with the
prototype measurements. We then connect the stages together to create the final circuit. It is now
possible to conduct measurements of the control behaviour of the final power supply. The entire chain
is tested using this step-by-step process.
There are also a number of constraints under which the power supply must demonstrate good
behavioural characteristics. One such constraint is when switching the power supply on and off. When
switching off, the second and the third stages will attempt to deliver energy to the output for as long as
possible while their input voltage is decreasing. As a result, the power increases and changes the
loading of the components. The switching on and off behaviour is difficult, if not impossible, to
simulate, especially because proper models do not exist. It is therefore necessary to thoroughly test
these constraints within the scope of the application using the prototype.
The photo shows a section of a multi-channel highvoltage power supply. This board is specially designed to
power capacitive loads and can operate in four quadrants
- thus both supplying and absorbing power. As a special
feature, the regulator in the power supply can introduce a
high-frequency signal into the output voltage. The
resulting high-frequency current is measured and an
accurate complex impedance is determined using realtime cross-correlation techniques. Measurement of the
complex impedance forms the input for optimising the
control loop parameters. The control loop then self-
adjusts to the properties of the load. On the picture, the Perspex partitions required for the electrical
insulation of the high-voltage stages, as specified in the European standards, are visible. The
partitions prevent transfer between the high voltage components and sensitive digital control logic.
Close collaboration
When the prototype phase has been completed successfully and the prototype functions in
accordance with the requirements, it is time to design the final device. This design step involves
bringing together the blocks from the prototype and fitting them into the available space. Cooling
aspects are greatly emphasised during this step. Despite efforts to maximise efficiency, heat always
dissipates in components and that is precisely what we want to prevent. Heat removal requires space
and this is contrary to the desire for compact designs. Is there space for a heat sink on the
components? Should the heat be lead through the PCB to the back? Is the heat sink part of the power
supply housing and should special mechanical constraints be considered? Is cooling provided by air?
Is forced cooling or cooling water provided? All of these questions determine the thermal
characteristics of the design. Thermal simulations identify interactions between the various
components and allow verifications to be made as to whether the individual components remain within
specification. If necessary, the arrangement of the components must be further optimised in
combination with the cooling.
It goes without saying that designing and manufacturing power supplies requires a multidisciplinary
approach. The importance of close collaboration between designers and manufacturers is described in
the article "Together from idea to PCB" in Bits&Chips 5, 2012. Similarly, it is clear that cooperation
with engineers and parts suppliers right from the start will contribute to developing the best product.
June 15, 2012
Alex van den Heuvel is a senior consultant and technology manager for electronics at Technolution.
One of his specialist areas is architecture of complex high-power electronics such as those found in
power supplies and motion-control systems. Paul van Haren is a programme manager at Technolution
and focuses on defining new projects.