I found out there is a dead time in my steering controller after several runs. The transfer function for it in the $s$-domain is:

$$ H(s)=\frac{270 \cdot e^{-0.03s}}{0.015 \cdot s + 1} $$

To overcome this dead time, should I introduce a compensator? My control theory background is weak so I am unaware of a proper way to manage the dead time.

  • 1
    $\begingroup$ Use latex style to write equatiions. Transfer function of what? There are comensators like Smith predictor, but I would rather explore where the dead time came from and modify the system. $\endgroup$ Feb 25, 2019 at 7:41
  • $\begingroup$ Is it dead-time or deadband position from mechanical hysteresis and stiction. ? Can you verify? Deadband can be overcome by reducing actuator slack by design. Also transfer function looks wrong with e^ $\endgroup$ Feb 25, 2019 at 8:05
  • $\begingroup$ Other answers suggest DITHERING as a cure for deadband. $\endgroup$
    – analogsystemsrf
    Feb 25, 2019 at 12:35
  • $\begingroup$ @SunnyskyguyEE75 The exponential $e^{-\tau s}$ just means that there is a time delay of $\tau$ seconds before the output reacts to the input. $\endgroup$
    – Koen Tiels
    Feb 25, 2019 at 20:02
  • $\begingroup$ If the deadband excursion time delay is τ = 30ms then f-3dB ~0.3 / 30ms ~ 10Hz and a Gain of 270 or GBW~ 3kHz is unity gain instability frequency so a high order phase lead-lag compensation before 1kHz might work better to inject dither. noise by lead compensation ( HPF gain limited) R1C1//R2C2 ... althoough something tells me GBW ~ 1kHz So start with specs for Response time to step and ramp inputs then overshoot and small step response, reduced deadband, then solve. $\endgroup$ Feb 25, 2019 at 20:31


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.