0
\$\begingroup\$

I'm designing a current mode controlled buck converter IC which can work in either fixed frequency PWM in normal load, and pulse skipping mode in light load. I have a flag to indicate whether the load current is small enough to be a light low case or not by using a current controlled voltage source in series with the load to compare the load current with a reference threshold for light load current.

Now I'm thinking about how my controller can detect if the load is light enough because it is not possible to measure current directly from the load in an IC as in simulation. Is there a practical way to determine the load condition? Is comparing the compensator output to a threshold a good idea for a current mode control buck? In voltage mode, I think it is okay as this voltage also set the duty cycle directly through comparing with the PWM ramp comparator, but for a current mode, the duty cycle is indirectly controlled by this.

*Update: My converter is synchronous, and I used my zero crossing detector as a light load indicator as well. I counted if the zero crossing happens more than 4 continuous cycles then light load flag is triggered. For normal load detection to go back from light load mode to normal PWM, I compare the output voltage with a reference about 98% target output, so if light load mode cannot handle the load then the drop will trigger the comparator and the converter will jump back to PWM mode. I am not sure if this is the best scheme for toggling between modes, but for now it can do the job.

\$\endgroup\$
4
  • \$\begingroup\$ If you put the MOSFET and diode inside the IC then you can measure the inductor current. Zero-crossings of the inductor current might be an indication of no- or light-load. It gets easier if you are designing a synchronous buck as the diode is replaced by a low-side-driven MOSFET and its Rds-on can be used as a shunt for measuring the inductor current. \$\endgroup\$ Nov 2, 2021 at 13:25
  • \$\begingroup\$ You don't need to measure current, if your circuit is designed reasonably correctly it will traverse into and out of pulse skipping without batting so much as an eyelid. \$\endgroup\$
    – Andy aka
    Nov 2, 2021 at 13:39
  • \$\begingroup\$ If you think about it, what is the problem caused by too light load? The problem is if the load is too light then your smallest possible PWM signal still pushes too much current into the output, making the output voltage rise too high. And at that point your only option is to "dither" the PWM steps between 0 and 1. Actually you should be dithering them all the time, you probably just don't realize you are. \$\endgroup\$
    – user253751
    Nov 2, 2021 at 15:53
  • \$\begingroup\$ Thank guys, it is alright now. My converter is synchronous, and I used my zero crossing detector as a light load indicator as well. I counted if the zero crossing happens more than 4 continuous cycles then light load flag is triggered. For normal load detection to go back from light load mode to normal PWM, I compare the output voltage with a reference about 98% target output, so if light load mode cannot handle the load then the drop will trigger the comparator and the converter will jump back to PWM mode. \$\endgroup\$
    – aquoc
    Nov 10, 2021 at 8:53

1 Answer 1

1
\$\begingroup\$

You may consider what (almost) everyone else prefers, that is include high side differential amplifier for an external sense R.

This must also consider the DCR of the L chosen and \$I_{rms}^2R\$.
This R may be placed before or after the switch depending on optimal phase margin for damping low and high loads.

Practical recommendation

I would consider R=1mohm/uH for choosing this as a starting point, unless it affects your efficiency or power dissipation in R above 50mV. This ratio is 1/(63%)T which is also << switching rate.

The reason current sensing is important in CMC Bucks is this is akin to proportional feedback in a PID loop. 1st order feedback is always best for any servo and 2nd order derivative and slope compensation are to improve on margins and sub-harmonic instabilities with flyback types, but not without RC//RCD snubbers.

With \$f_{-3dB}= \dfrac{1}{2\pi ~T_{\text{63%}=\frac{L}{R}} }\$ =\$f_{-3dB}=\dfrac{0.35}{T_{\text{10~90%}}}\$ , it may interest you know remember considering phase-lead compensation to match this breakpoint in the Bode Plot for the damped L current sensor to improve on phase margin.

Free Ridley Software if using 32 bit Excel 2007+ old Guide New and guide

\$\endgroup\$
2
  • \$\begingroup\$ Thank you for the suggestion and good info. Actually I'm already using a high side differential amp for peak current feedback, and I used the Rds ON of the switch itself as a sensing resistor. I'm curious about the position of your high side sensing resistor, so if the damping is not good then the resistor should be placed after the switch to damp the ringing at the switching node, right? Speaking of slope compensation and sub-harmonic instability, is it fine to remove the slope comp in case the converter work in DCM or skip mode? \$\endgroup\$
    – aquoc
    Nov 10, 2021 at 10:07
  • \$\begingroup\$ That depends on your specs for overshoot and step unload. Slope compensation adds to predictive rather than reactive (too late) compensation. \$\endgroup\$ Nov 10, 2021 at 11:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.