| |
 | |  | | | | | AC stability analysis for closed-loop systems
An interview with Momchil Milev
High Performance Analog EDA Group, Texas Instruments, Inc. | |
| |  | |  |
Why is AC stability analysis important for linear circuits?
Use of feedback in amplifiers, biasing circuits etc. (linear circuits) is essential and common since very early days of electronic design. It is primary means of maintaining circuit's functionality over a wide range of operating conditions or input signal's frequency range. Systems with feedback, though, are inherently unstable, unless feedback's characteristics are maintained properly. AC stability analysis is important analysis for linear circuits because it shows how stable a circuit is under given operating conditions and how close its state is to a very undesirable state of generating self-induced oscillations which prevents the circuit to function normally.
What are the design challenges in doing this analysis for closed-loop systems?
Major design challenge is to maintain circuit's operation across wide range of changing conditions - temperature, supply voltage, load conditions etc. In order to meet these goals modern amplifiers and other linear analog circuits employ not just a single main-signal feedback loop, but also many local feedback loops which help the circuitry to adjust to particular changes in the circuit condition and, in this way, maintaining its designed functionality. Similarly, major challenge is to verify circuit's stable range of operation in terms of avoiding negative effects of feedback for different circuit conditions across a wide frequency range.
What are the advantage of using this new method over the traditional well-known methods that investigate Bode plot diagrams, phase and magnitude plots etc.?
Traditional methods for small-signal stability analysis of linear analog circuits are techniques of investigating the Bode plot diagram, phase and magnitude plots of the open loop circuit response or using means of pole-zero symbolic analysis to look for conditions that may cause circuit's oscillations (ringing). In this, it is important to underline "look for conditions" - as these techniques do not give a direct insight of which part of the circuit is really responsible for the circuit's undesirable behavior at a given condition. These methods look at the circuit as a "black box" and all they do is analyze circuit's behavior that is observed "on the outside". Years of experience and good theoretical understanding of circuit's operation are required to identify the source of the problem and compensate for it properly. This is what the proposed technique is about to change.
The technique developed by Rod Burt, which is used in this tool, gives us the insight of how circuit's feedback loops, both local and main-signal would respond to a condition that can cause small-signal oscillations. In this AC-domain analysis we evaluate every circuit's node sensitivity for oscillation by computing a number we call 'stability index' or 'stability factor'. Along with information on every node's natural frequency (of oscillations) we draw a complete and clear picture of where the oscillations may occur. This not only identifies where a problem may occur and how prone the circuit is to oscillation but also gives a better understanding of where circuit's feedback loops are - intentional or unintentional.
Traditional loop gain analysis, such as .stb in Spectre, is widely applied to analyze single or main loop effects. The main limitation of this method is that a single loop has to be defined and typically a path identified in the loop where a series element is inserted. Most practical feedback circuits have multiple loops making traditional loop gain analysis very difficult to apply. Our method determines the complex poles in the system or what happens when the loop gain becomes unity. Although this is what determines stability and phase margin, it does not give any open loop information. The main advantage to traditional loop gain analysis is the insight into the open loop response and because of this it may still be used for simple or main loop analysis in addition to our method.
In implementing this methodology, how were you able to leverage the Cadence Virtuoso platform?
Cadence DFII environment was our best choice for a platform for implementation of this technique and building this tool quickly and effectively. Cadence's Virtuoso platform along with the well integrated Virtuoso Analog Design Environment capabilities to use SKILL programming language and OCEAN API's to control and run simulation jobs on Virtuoso Spectre SPICE simulator offered an easy and straightforward way to create the tool's graphical interface (GUI), generate Virtuoso Spectre's input files, read and process the results by generating plots and results annotated directly on the design schematic. We had an initial version of the tool showing the full capability of the proposed method for stability analysis in less than a week. The tool has evolved since then in many different and significant ways due to the ease of use of DFII's API's, well integrated Virtuoso platform system, and always expanding programming capabilities that Cadence continues to offer for its tools. Without being able to leverage HNL customization under OSS, OCEAN scripting, integrated ADE calculator's processing capabilities, simulation control and Virtuoso Spectre simulator that are readily available under Cadence's DFII and Virtuoso platform, the job of programming such a tool as a standalone tool would have been nearly impossible to justify and accomplish in today's fast-paced EDA environment.
Read the paper, "A Tool and Methodology for AC-Stability Analysis of Continuous-Time Closed-Loop Systems"
| 
About the author
Momchil Milev has been with the High Performance Analog EDA Group, of Texas Instruments Inc (former Burr-Brown Corp) since 1998. He has been developing and delivering Process Design Kits, as well as implementing tools and methodologies for electronic design automation. Since May 2004 he is with the High Speed business unit, High Performance Analog (HPA) division of the company. Interests outside Electronic Design Automation include research in Artificial Neural Networks applications and programming of handheld and embedded application devices.
Rod Burt has been with Texas Instruments Inc (former Burr-Brown Corp) since 1982. As an Analog IC Design Engineer he has designed and led many successful projects helping to build Burr-Brown's high-precision linear products portfolio. |
| |
This content has not yet been rated by other users |
|
|
|
|
|
|
|