Electronic design relies heavily on simulation tools that offer greater insight into circuit and system behavior than might be obtained from direct hardware measurement. I do not meant to imply that physical measurement could be discarded; rather it represents a final confirmation test phase for any development project.
Simulation software tools include
|Small Signal AC Analysis - suitable for linear passive and active filters, or linear amplifiers operating at a predetermined DC bias setup.|
|Large Signal Analysis - suitable for "extremely" non linear behavior including magnetic hysterisis, switch mode power supplies etc|
|"Medium Signal" - Weakly Non Linear Analysis - suitable for "tame" non linear devices including mixers, amplifier distortion and oscillator phase noise|
|System Simulation - used for modulation format analysis, eye diagrams, Bit Error Rate performance, network availability and generic combinations of building blocks with arbitrary functional descriptions|
Early small signal analysis software includes TOUCHSTONE and later Eesoff. These tools are relatively "simple" and are based on a matrix solution formed by interconnections between components with a "linear" description. This matrix is solved at each defined frequency step and parameters such as impedance, loss or amplification and stability are then interpreted from the solution.
Large signal analysis began with SPICE and then PSPICE and other simulation tools based on a time domain step by step procedure. These produce an "oscilloscope" output which can then be passed on to a Fast Fourier Transform (FFT) algorithm to show harmonic content or intermodulation products as would be seen on a spectrum analyzer.
This direct time domain analysis was slow and computationally intensive. Other approaches were sought with relaxed processing overhead that could be applied to systems or circuits with relatively "benign" non linearity. Notable is the "harmonic balance" algorithm used in Libra and then Agilent ADS. These later tools also contain the earlier small signal analysis capability.
So far we see "circuit level simulation" but most systems are constructed from many interconnected circuits. It would be inappropriate to analysis one whole "chunk" in one go as internal insights would be obscured, and the time of simulation (and interpretation of results) would be excessive. Therefore "system level simulation" is required. In this approach, "building blocks" with "parameters" are connected, and many potential circuit topologies could fit inside each.
System level simulation precedes circuit level simulation, so it is logistically impossible to simulate (say) a complete radio without having designed all its circuits without simulation first!
Although large companies have adequate financial resources to afford "high end" simulation tools, the amateur scientist or electronic enthusiast may not. Simulation tools such as COMDISCO or ADS exceed $1,000,000 at an entry level. Few individuals have this amount of cash on hand. However, simulation software technology has matured to such an extent, and many of the "kernel" algorithms are in the public domain. Many companies now offer cost effective simulation tools along the following categories
|Low cost (< $1,000) simulation tools aimed at small companies or individuals|
|Low cost tools with a cut down "student version" that has defined constraints such as size, features or time limit|
|Free "student version" software|
|Free give away software|
The first three categories are recommended (by me) but anything given for free probably has little functionality but may be usable.
MATHCAD is a popular mathematic package used by scientists and engineers and can be used to analysis systems in an open ended fashion. This requires that the user can convert their "system" into equations and then run a program to simulate the predicted behavior. In another web chapter I show MATHCAD simulating a Neural Network - see Artificial Intelligence and NN Structure and Consensus Training.
MATLAB is "similar" but more advanced, however it is more "script based" whereas MATLAB works like a word document (you can cut and paste pictures, sound files and videos into your simulation program for fun) (Note LabView and HpVee offer something similar well worth looking into)
However many people are uncomfortable with mathematics and a direct circuit analysis package would be preferred. I recommend MicroCap for this.
MicroCap provides small signal frequency domain ac analysis and large signal analysis based on fixed time domain steps. A FFT is included. The student version is free and is unlimited in time. However their are limits on the maximum circuit size and also its model library is limited.
MicroCap is SPICE based and fairly easy to use. It can be applied to small circuits of transistors and OpAmps and Comparators and provides frequency sweeps, oscilloscope displays and spectrum analysis.
With a little imagination, it can be used well beyond its anticipated application, i.e. intended for lower frequencies such as audio. For the amateur experimenter, it can provide adequate predictions for RF behavior up to the GHz range!
These web chapters demonstrate MicroCap applied to extremes of its capability. It need not be used to this extent, however in comparison to expensive hardware test equipment, free analysis software is far better than nothing at all!
The NPN Bipolar Model web chapter shows how to model a transistor and how to make your own models, or repair inaccuracies in those included in the program. This models current gain hFE and Transconductance gm versus frequency at different bias currents, HFE versus bias current and collector voltage, and harmonic distortion products with an AC input signal applied for different bias currents.
The VCO (will) show a 3.7 MHz VCO with output power, harmonic content and spectral phase noise.
The RF PA will show a 10 Watt Peak Envelop Power (PEP) amplifier operating at 3.7 MHz with a two tone stimulus applied and showing Intermodulation sideband products.
Return to System Analysis
or Ian Scotts Technology Pages
© Ian R Scott 2007 - 2008