To those new to switching power
supplies, the inevitable question comes up: Look, your circuit
has one power FET, one diode and a few comparators with some
logic. Why can't you run a simulation and go straight to a production
board? The IC guys can do it, simulating 100,000 transistors
or more, so WHAT'S THE PROBLEM?
It's a good question. The answer
is complex, and some of the reasons are presented below. But
the reality is that there is no simulation program available
that is going to do this for you. It will also not be available
in the foreseeable future, regardless of your budget for computers,
software, and an engineer to run them.
The best and most accurate simulation
tool that you have is an analog computer that
you must build yourself - it's called a breadboard. If you only have one tool available, this is the
one you MUST use. You can and should use other simulation programs
such as
POWER 4-5-6 to help you design, understand, and characterize
the breadboard properly, but do not expect to skip the lab-testing
phase.
Limits of Simulation
tools for Power Supplies
It is not possible to simulate
a complete switching power supply in any reasonable amount of
time. You can't just enter a schematic and expect your circuit
analysis program to do anything useful with it. Simulation still
cannot replace empirical measurements of most of the waveforms
of switching power supplies-- it is used as a guide to proper
design, but you still need to build a test circuit.
Even with a powerful computer,
and the most advanced software, you will not be able to predict
every parameter of a switching power supply down to the last
detail. Here are some of the more difficult aspects of power
supplies that you will need to measure in the lab:
- EMI
- Device switching waveforms
- Transient peak stresses
- Cross regulation of multiple
outputs
- Control and loop characteristics
- Transformer performance
- Thermal profiles
- Start up and protection circuit
operation
Problem number one is that component
and PCB parasitics play a major role in determining some of
these parameters, and cannot be predicted completely before
the test circuit is built.
Problem number two is that most
simulation programs were not designed to handle the fast large-signal
transients of the power devices, and they may not always converge.
Simulation
as Part of the Design Cycle
Given the limitations, should
you even bother with simulation? The answer is Yes! But be realistic
about what you can and cannot do with the simulation tools available.
Here's the generally recommended approach:
- Get 90% of the electrical design
completed on the computer
- Choose the basic parameters
of power components and the control loop (Be careful when
scheduling: 90% of the design does not mean 90% of the job
is done - the remaining 10% of the job will take much longer
than most people plan for!)
- Predict the first-order and
some second-order characteristics of many waveforms to verify
proper operation with your selected values, and to ensure
specification compliance as much as possible
- Based on the predicted waveform,
select the details of your components, especially magnetics,
power switches, and capacitors
- Reiterate the simulation to
check values and power component dissipation
- Build your first test circuit;
Depending on your particular experience with the topology
and technology you are planning to use, this may be a breadboard,
or a PC board
- If you see any unexpected behavior,
go back to simulation tools to try and explain as much as
possible the causes and implications; Do not waste time trying
to get a perfect match from simulation to test circuit, just
make sure you understand what is going on and you can control
it in production
POWER 4-5-6 was designed to work with you with this kind of design
approach in mind.
The Simulation
Tools You Need
POWER 4-5-6 should be used to design and simulate the basic converter
and control. You can go straight from this to the breadboard.
The program was designed specifically for PWM power supplies
to help you rapidly design and simulate a converter as fast
as possible to get on with the work of breadboarding.
You can use a Spice-based package
to simulate more detailed parts of the circuit, and general
aspects that POWER 4-5-6 cannot handle.
Be aware that these packages will be over 100 times slower,
but they offer the advantage of being able to simulate any generalized
circuit. They do not provide any design capability. They also
frequently suffer from convergence problems when trying to simulate
the fast voltage and current transitions of PWM converters.
Recommended packages: Intusoft's
ICAP and MicroSim's PSpice. If you are working on a complex
power system such as a space station that may have serious system
interaction issues, you may want to get Analogy's Saber. This
is a general purpose simulator that is less prone to convergence
problems than Spice-based packages. Saber is designed
to handle large systems very effectively. However, it does require
a workstation, a skilled operator, and substantially more investment.
Different
Simulation Models
There are many different ways
to simulate a switching power supply, and you may need to work
with several different models to help your design process. Different
types of models are used to find a good balance between simulation
speed and accuracy.
- Small-signal
modeling is done with
linearized circuit models that help you design the control
system. This is built into POWER 4-5-6 which also takes the
results of these models, and extracts the important performance
parameters for you, saving a lot of time and energy. The small-signal
models can also be implemented effectively in any general-purpose
circuit simulator with excellent results.
Small-signal modeling on a general
circuit simulation package is fast, with no convergence problems,
but it requires the most analytical insight from the user.
Since the small-signal models
are linear, they can also be analyzed by hand with conventional
circuit analysis techniques, and are often used to teach the
theory of PWM converters
- Large-signal
modeling with averaged models
replaces the switching elements of a power supply with averaged
nonlinear circuits. This is very useful for assessing the
performance of large and complex power systems, and can be
implemented on Spice packages or Saber. The simulation is
fast, but details of switching waveforms and peak stresses
are lost, so it is of limited use to many circuit designers.
It will also frequently misinterpret control events such as
peak current mode control, and peak-current limiting.
- This kind of model will also
execute quickly on any circuit analysis package, but still
requires substantial insight on the part of the user.
- Large-signal
modeling with ideal switches.
The power switches and diodes are replaced with equivalent
resistances and voltage drops representing their on and off
states. This provides the most rapid simulation while retaining
the important details of switching waveforms, but not the
details of the switch transitions themselves. POWER
4-5-6 uses this technique, and supplements it with algorithms
that take advantage of the special features and characteristics
of PWM power
supplies to produce the fastest possible simulation.
This kind of simulation is very
useful for seeing start-up performance, step loads, transitions
from CCM to DCM, and other large-signal transients.
Step-Load Simulation
from Power 4-5-6
This simulation from Power 4-5-6
takes a fraction of a second to run. A good ideal switch simulator
will show peak currents and ripple voltages, and properly
simulate CCM and DCM operation. In this figure, a converter
is stepping from full load to light load, and the closed-loop
response of the voltage and current are shown. Switching spikes
are not shown in the simulation - these are almost impossible
to predict accurately, so there is no point in wasting time
trying to simulate them.
Spice and Saber ideal-switch
models can also be built, but they will run considerably slower
and can sometimes be prone to convergence problems. User input
is needed to define the switches well, but this is an effective
and useful simulation approach.
- Large-signal
modeling with detailed power devices. This is where Spice models can be very useful. Simulation
times become very long, so you will typically only simulate
a small portion of the power circuit for a few cycles. You
will sometimes use this type of simulation to understand the
complex waveforms you see on a power device.
This kind of simulation requires
the least experience from the user, but is the slowest of
all.
Remember: whichever method you
use, the simulation will only reflect the model that is put
into it. It is very time-consuming and usually impractical to
enter every parasitic component in order to exactly simulate
what you see on the bench.
© copyrightdley
Engineering, Inc. 2002
