Accuracy in LTspice Modeling

Sometimes we need to consider the human factor when using circuit design simulation tools. In this article, our hollow-state expert details an error that occurred in a past article, which is now being corrected.

- Richard Honeycutt

While I was in undergraduate school in the late 1960s, Hewlett Packard introduced its first minicomputer, which, if memory serves, had 8kB of RAM (twice that of the IBM 1620 the college used to teach programming), and a paper tape punch/reader for programming and output. The physics department bought one, and I programmed software to calculate the resonant modes of a rectangular room as part of my senior thesis project. A few years later, when I began reading scientific papers reporting the results of computer modeling, it did not yet dawn on me that what I had done on that project was in fact computer modeling. Consequently, I was hesitant at first to trust the conclusions drawn from computer modeling of physical systems.

At the time, I was blissfully unaware of the development of the Electrical Circuit Analysis Program (ECAP) by IBM in the 1960s, or the Simulation Program with Integrated Circuit Emphasis (SPICE) in 1973. The descendants of SPICE written for personal computers (PSPICE, LTSPICE, and several others) came along in the early to mid1980s. It was then that I became familiar with these circuit design tools as part of my duties as a college electronics professor. I found that while care was necessary in building circuit models to run on these simulation programs, they produced excellent results.

Some common error sources included:

• Poor component models, especially for transformers and vacuum tubes

• Parasitics such as stray capacitance, which were especially problematic in high-frequency analysis.

• Human error: undoubtedly the most prodigious error source.

Software companies provided component models for ideal resistors, capacitors, inductors, transistors, solid-state diodes, and later, ICs.