It's time to SPICE up your life!

-February 19, 2015

We all have our favourite SPICE circuit simulators and I decided to compare the performance of Hyperlynx Analog, PSpice, and LTspice to predict the gain and phase margins of a buck converter. Poor stability of the feedback-control loop affects the output noise, the PSRR, the reverse transfer, the output impedance, and the step-load response of a voltage regulator, impacting overall system performance.

The purpose of this post is not to endorse one circuit simulator over another, nor to promote a specific type of modelling, e.g. state space or switching, or discuss the effectiveness of using the Bode plot to predict stability, but to share my experiences of comparing three popular SPICE engines to develop reliable power electronics for space applications. There are many other 'flavours' of SPICE simulators available such as TINA, IsSPICE, ICAPS and µCAP, as well as competing tools which I really enjoy using, like ADS by Keysight Technologies, which provides good support for high-frequency and RF modelling.

We all have requirements and constraints when selecting EDA toolscost, familiarity, existing library of models. It's what a project/company mandates. Personal preference, the ease and intuitiveness of using the software are also key decision factors. In my own case, quite often I find myself drawn to the simplicity of LTspice to quickly prove a concept without having to bother with initial conditions or convergence issues, where the use of the word 'simplicity' is not meant to detract from the capability of LTspice, but a compliment to Mike and his team (it's also free and has a very large and helpful user support group!). PSpice is a popular tool and a 'Lite' flavour can be downloaded free-of-charge with limits on the size and complexity of designs that can be modelled compared to its licensed version. Sometimes, but not so much today, I'll enter a text netlist and simulate thisit's very insightful when extracting layout parasitics to export a netlist to understand and quantify unwanted transmission and via effects.

The circuit to be simulated is a simple buck converter whose unregulated input ranges from +10 to +15V, required to supply a load with a regulated +5V rail at 10A. The switching frequency is 250 kHz and voltage-controlled feedback is used to provide a stable output as shown below:

Figure 1 Buck converter with voltage-controlled feedback.

Each of the three simulators has its unique quirks, some of which have kept me up almost all night on several occasions. It's not fun when your design simulates in one tool but the identical circuit fails to run on another because of very subtle differences between the software. Normally this is when you discover how good or bad the manuals are!

All of the simulators can be used to determine control-loop stability and allow parameter sweeps to model worst-case and end-of-life effects. I plan to write an Applications Note to show users of Hyperlynx Analog how to predict the gain and phase margin of switched-mode and linear regulators, which will discuss how to simulate DC-DC converters, where and how to access the feedback loop, the pros and cons of using a Bode plot to predict loop stability, and how to import third-party PSpice models. Some providers of space-grade voltage regulators provide a simulation model which you can 'hook-up' to a control loop within a simulator. If you would like to be notified when this tutorial is ready, please send me an email.

For each simulator, I ran transient analysis to verify the functionality of the buck converter as shown in EzWave below:

Figure 2 Predicted +5V output as the unregulated, line, voltage varies.

Some modifications have to be made to the original schematic to access the feedback loop, measure its gain and predict stability, and I will cover these in my next post.

I am a great believer of design re-use and both Hyperlynx Analog and PSpice allow the design drawn for circuit simulation to be re-used as the schematic for PCB floor-planning, layout, and product design. In the real world, a single stage of entry avoids unnecessary duplication and errors, helping space-electronic products get to market faster and be developed to cost and schedule. Both LTspice and Hyperlynx Analog support hierarchical simulation using symbols while PSpice prefers a flat design as it tends to repeat references throughout a hierarchy.

Hyperlynx Analog allows users to back-annotate and import SPICE or S-parameter trace data from the routing to account for its parasitic effects. Mentor's PCB flow also permits the prediction of EMI, thermal, and power-integrity effects, and if you are producing spacecraft avionics, these tools are essential to ensure your hardware is delivered right-first-time.

Unlike the other tools in this comparison, Hyperlynx Analog allows me to simulate low-frequency, DC-DC converters, as well as high-speed, mixed-signal circuits where I can import third-party S-parameter component models, or extract the S-parameters of a complete network to account for layout effects. For this comparison, I used the Eldo simulator within Hyperlynx Analog. The three tools permit analog circuits and logic to be co-simulated which is useful for digital control of voltage regulators and ADCs/DACs.

The quality of simulation models provided by suppliers of power microelectronics varies immensely and I encourage vendors of DC-DC converters to get in touch if they need help to produce models that are fit for the space industry, and compatible with the EDA tools used by manufacturers of satellite sub-systems and spacecraft.

A word of caution: simulation does not replace breadboarding prior to a production build! Modelling allows you to predict the behaviour of your regulator and debug flaws quickly and at low cost, permits what-if analyses without the fear of blowing anything or anyone up, lets you understand how a design reacts to faults such as a short-circuit, and simulates test measurements if you do not have access to the relevant equipment. A working SPICE model doesn't necessarily mean your prototype hardware will work first time: if you 'wire' your model as you would in a lab accounting for ohmic losses, leakages, and parasitics, the predicted performance should correlate with that measured from your Engineering Model. In the real world of product development, simulation MUST be supplemented by measurement of loop stability from your hardware to minimise risk. Component parasitics, temperature, and PCB layout all affect control-loop stability, and no experienced designer would rely solely on modelling to predict critical performance. Measurement should be performed on the prototype, qualification, and flight builds.

To help designers predict loop gain and phase margin, several vendors provide free modelling software that uses proprietary simulation models specific to their parts, e.g. Linear Technology's PowerCAD and Analog Devices' ADIsimPower. SIMPLIS, PSIM, and Power 4-5-6 are also very useful and quicker than SPICE-based modelling, and top-down tools such as Simulink use behavioural models to predict the stability of the feedback-control loop and I really like the speed of this approach.

Space-electronics companies invest a lot in EDA software and many do not understand the capability or potential of such tools to get their products to market right-first-time, delivered to cost and schedule. In many instances, businesses use simulation software from one supplier and schematic entry and PCB layout software from another, unnecessarily duplicating design entry, buying multiple licences, and having to learn two flows. Please don't make this expensive mistake!

Until next month, get your SPICEs out and happy simulating!

Also see:

Loading comments...

Write a Comment

To comment please Log In