Gain-optimal discretization and constrained MIACE tuning for rational loops with pure delay
This is a contunation of a research of mine, started in 2004, on generalised discretizaiton methods and applications to optimal control. Some results here for the optimal method were obtained as early as 2011 and are cited. The integration rule in a digital controller changes how much feedback gain a delayed loop can tolerate. For a pure delay equal to one sampling period, the optimal integratorputs weight 3/4 on the current error and 1/4 on the preceding error. Its dimensionless gainthreshold is 4. For every finer integer subdivision of the same delay, backward Euler givesthe largest threshold. We derive the exact nominal integration weight when the implementedweight can deviate by different amounts in either direction. The optimum changes characterwhen the total width of the uncertainty interval exceeds 1/2. For a cascade of any numberof positive first-order factors with one delay sample and independent integration weights in[1/2, 1], the unique optimum follows from a single scalar calculation. For general rationaldata, an exact, terminating projection algorithm finds the supremum over all stable gainintervals, including intervals separated from zero by instability. Explicit solutions for firstand second-order factors and for a sampled physical lag show how the answer depends on themodel. For the delayed integrator, we also give the operating gain that minimizes the Moduleof Integral Absolute Complex Error (MIACE) under monotone tracking and a prescribedgain reserve.
Authors
- L Tomov (ORCID: https://orcid.org/0000-0003-1902-6473)
Institutions
- New Bulgarian University (BG)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24414694
- Primary Topic
- Advanced Control Systems Design
- Type
- preprint