Error analysis and improvement of FILT

✍ Scribed by Toshio Hosono

• Publisher: John Wiley and Sons
• Year: 1998
• Tongue: English
• Weight: 158 KB
• Volume: 81
• Category: Article
• ISSN: 8756-663X
• DOI: 10.1002/(sici)1520-6432(199808)81:8<10::aid-ecjb2>3.0.co;2-5

✦ Synopsis

The main errors of FILT (fast inversion of Laplace transform, developed by T. Hosono) are approximation error, truncation error, and cancellation error, which are analyzed theoretically and numerically. The following are clarified: (1) the truncation error is easily controlled by increasing the number of terms used and by use of Euler transformation; (2) the approximation and cancellation errors compete with each other, so that the maximum number of significant digits is decreased to 70% of the mantissa digits; (3) proposed higher-order FILTs give better cancellation errors, and the second-order FILT gives a number of significant digits that is greater than 80% of the mantissa digits.