Comparative study of low-pass filter and phase-locked loop type speed filters for sensorless control of AC drives

Research output: Contribution to journalJournal articleResearchpeer-review


High quality speed information is one of the key issues in machine sensorless drives, which often requires proper filtering of the estimated speed. This paper comparatively studies typical low-pass filters (LPF) and phase-locked loop (PLL) type filters with respect to ramp speed reference tracking and steady-state performances, as well as the achievement of adaptive cutoff frequency control. An improved LPF-based filter structure with no ramping and steady-state errors caused by filter parameter quantization effects is proposed, which is suitable for applying LPF for sensorless drives of AC machines, especially when fixed-point digital signal processor is selected e.g. in mass production. Furthermore, the potential of adopting PLL for speed filtering is explored. It is demonstrated that PLL type filters can well maintain the advantages offered by the improved LPF. Moreover, it is found that the PLL type filters exhibit almost linear relationship between the cutoff frequency of the PLL filter and its proportional-integral (PI) gains, which can ease the realization of speed filters with adaptive cutoff frequency for improving the speed transient performance. The proposed filters are verified experimentally. The PLL type filter with adaptive cutoff frequency can provide satisfactory performances under various operating conditions and is therefore recommended.
Original languageEnglish
JournalCES Transactions on Electrical Machines and Systems
Issue number2
Pages (from-to)207-215
Number of pages9
Publication statusPublished - Jul 2017


  • Adaptive cutoff frequency
  • Low-pass filter
  • Machine sensorless drive,
  • Phase-locked loop
  • Speed filter
  • Static error


Dive into the research topics of 'Comparative study of low-pass filter and phase-locked loop type speed filters for sensorless control of AC drives'. Together they form a unique fingerprint.

Cite this