I'm final year student currently designing a buck converter with an analogue controller (LM2743). What my supervisor told me is to used a PI controller using bode plot. My design spec:
1. Vin=12V
2. Vout=5V
3. Pout=10Watt
4. Fsw=100KHZ
From there I obtained my inductor (48.6uH), output capacitor (15uF) an load resistor at (2.5ohm). From there my transfer function is:
From there I'm quite stuck on how to do it. Where I know the T.F for PI controller is Gc(s)=Kp+(ki/s). After that, lets say I've managed top design the PI controller. From datasheet LM2743, there are few components need to be connected to the main controller which is pin FB and EAO.
Its mentioned in the data sheet that the calculation can be made using the equatio given but didn't tell on how to get the Fz1, Fz2, Fp1 and Fp2 where these value that can decide the components.
-
\$\begingroup\$ I used to learned how to design PI controller using damping factor and freq using a P factor point, but my transfer function got imaginary number so thats why my SV asked to design it using frequency response. \$\endgroup\$– Taufiq MazdiApr 16, 2018 at 2:08
1 Answer
Although P feedback is stable and simple, PI feedback reduces gain error thus better load regulation but increases overshoot. While PID offers a better compromise. The RC filters in series are Lead-Lag compensation being a limited D frequency range.
The design must start with tolerance specs for each source of error: load limits, overshoot transient and time; load regulation error ;input sensitivity, and temperture rise worst case, for each range of expected variables.
It is a good practice to get used to doing things this way by learning how formal PSU's summary specs are done and follow these definitions. An ATX spec might be >10 pages and have a dozen parameters that you can summarize including environment specs.
The lead-lag filter is described with the phase-gain margin of the response in the App. Note.
Integrators reduce ripple but decrease stability.
Differentiators, increase stability and step response time but add ripple.
A Lead-Lag filter usually adds enough derivative gain just before unity gain to boost the phase margin. It may only be a 10dB boost in gain in the HF range just before unity gain.
The basic process of choosing these filters requires the entire loop response in order to apply asymptotic lines with the understanding that the breakpoint applies a 45 deg. phase shift and goes to 0 and 90 deg exceeding a +/- decade span. You can find a full step-by-step description elewhere, such as your Control Systems textbook.
Generally when you do this on a regular basis, you will rely on a tool to visualize the results in sumulation.
Although this one, I did for you, (Falstad) although no phase, it is interactive and you can edit any components with a Bode plot. It can be exported and do DSO time plots as well.
You can use Vspice or any others you have for Bode Analysis or Nyquist analysis.
You might also look at my choice of a more efficient TI regulator in WebBench.