The excerpt below is from the book Pulsewidth Modulated DC-to-DC Power Conversion: Circuits, Dynamics, and Control Designs by Byungcho Choi.

It looks a bit long but my main question is why the implementation of current sensing network (CSN) by sensing the switch current is simpler than that of sensing inductor current?

Current mode control can be implemented in many different forms. The most popular among them is the peak current mode control, illustrated in Fig. 10.6. In this control scheme, the switch current is utilized in place of the inductor current. The switch current, which corresponds to the on-time inductor current, is sensed through CSN and blended with the compensation ramp. The peak value of the switch current, or equivalently the peak value of the inductor current, is used to determine the instant to turn off the switch; thus, the control scheme is called the peak current mode control. It should be noted that the switch current sensing is functionally identical to the inductor current sensing because the peak value of the inductor current is employed as the criterion to turn off the switch. There are several advantages in sensing the switch current rather than the inductor current. The first is the simplicity in CSN. An implementation of CSN for the peak current mode control is illustrated in a later example. As the second advantage, the sensed switch current can be used for the over-current protection for semiconductor switches. Due to these advantages, the peak current mode control is widely adapted to modern PWM dc-to-dc converters.

enter image description here


The switch current can be sensed by a small value current sense resistor between the source and ground. Since it's ground referenced it's easy to use in the control system. Also the temperature coefficient of the resistor is small, in contrast to using the RDSon of the FET or the ESR of the inductor. In integrated solutions a FET current mirror can provide low-loss sensing.

Inductor current sensing requires a resistor in series with the inductor, or a network around the inductor to cancel the reactance and provide an estimate of the DC current. Either way, it is not ground referenced and one end is switching rapidly. The circuitry to sense and translate the current signal requires a differential amp and possibly a filter to reject switching spikes.

If you use a network to cancel the reactance to get the inductor current, the temperature coefficient of copper can be a problem, and sometimes requires compensation by a thermistor.

[With inductor current sensing we are looking for the voltage across the ESR of the inductor. If we didn't use an external network to cancel the reactance we would have V=Ldi/dt+iESR. What we want is just i*ESR. This technique is fully explained HERE ]

Bu sensing the true inductor current however, you can use average current mode control. This avoids the subharmonic oscillation issue. It does make cycle-by-cycle current limit harder to implement though.

  • \$\begingroup\$ Thanks. I've got some good ideas now. Please clarify some points below: 1. How does ground reference related to temperature coefficient of the resistor? 2. With inductor current sensing why do we need "cancel the reactance and provide an estimate of the DC current"? I am confused because we want to send the inductor current not average inductor current. \$\endgroup\$ – anhnha Aug 9 '17 at 7:31
  • \$\begingroup\$ Sorry, poor wording in my answer, I've clarified and added more info about inductor current sensing. \$\endgroup\$ – John D Aug 9 '17 at 15:22

I think what the author is talking about is using the MOSFET Rds as a current sense resistor. This is better than adding a current sense resistor. The Rds of the MOSFET is inherent (you are stuck with it) so you have to accept the losses it causes. But if you add a current sense resistor, you will incur even more losses.

The Rds of the MOSFET follows the simple rule of Ohm's law: V = I * R. If you know R and measure V you can calculate I easily. In practice, for control, you may not need to know R precisely. The peak current control discussed by this article will adjust the peak current point up or down as needed until the output voltage is correct. So the output regulation will still be good, even if Rds varies a bit from unit to unit, or with temperature. This is just the nature of feedback control.

Some controllers also have cycle-by-cycle current limiting. The article alluded to this, calling it over-current protection. This is not part of the control, per-se, it is just a feature that can help avoid inductor saturation and limit short-circuit current. This feature DOES depend on knowing Rds accurately. Basically, this is an absolute upper limit on peak current. If this current is reached, the controller will instantly turn off the high-side switch, regardless of whether the output voltage is in regulation or not. So if Rds varies a bit from unit to unit, then the absolute upper limit will also vary a bit, and the over-current protection will kick in at slightly different levels from unit to unit.

Directly sensing instantaneous inductor current by measuring voltage across the inductor is not really practical. For the peak current control method, the instantaneous current must be sensed when the high side switch is closed and the inductor current is ramping upward. However, in this condition, the inductor voltage is just Vin - Vds - Vout. Vds is small, and Vin and Vout are pretty much fixed. So sensing a nearly fixed voltage will not help you figure out the current. John D mentioned that it is possible to use a network to cancel the reactance of the inductor. When you do this, the inductor resistance becomes your sense resistor. I don't know much about that, but it seems like such a network would be a filter, and would thus affect the signal bandwidth making instantaneous current sensing difficult or impossible. It would work for average current sensing, though.

So the only practical way to sense instantaneous inductor current is indirectly by using a current sense resistor in series with the inductor. As previously noted, this adds losses, so it is better to avoid doing it when possible.

In a boost converter there is generally an N-channel MOSFET with source grounded. This makes for easy low-side current sensing. A buck converter will usually have P or N MOSFET on the high side which can be used for high-side current sensing. While it is more difficult to design the high-side sense circuitry, it is still very doable. Many IC's incorporate this feature.

  • \$\begingroup\$ It's certainly possible to use the RDSon of the FET for current sensing, but it's more common to use a current sense resistor E.g. ti.com/lit/ds/symlink/lm5021-q1.pdf The tempco and part-to-part variaton of the RDSon can be problematic. It is possible to use the inductor for current sense, with the proper network to cancel the reactance. Sometimes also with a thermistor for temperature compensation. E.g ti.com/lit/ds/symlink/lm27402.pdf Of course integrated FET solutions have other possibilities for current sense like mirroring. \$\endgroup\$ – John D Aug 8 '17 at 16:10
  • \$\begingroup\$ I'd like to ask some questions. 1. "Directly sensing instantaneous inductor current by measuring voltage across the inductor is not really practical for obvious reasons." --- What is the reasons here? I couldn't get it. 2. " In practice, for control, you may not need to know R precisely. The control loop will still work because it adjusts the peak current target until the output is correct. Current limiting can only be as precise as Rds, however." --- Can you explain more detail about this point? \$\endgroup\$ – anhnha Aug 9 '17 at 7:35
  • \$\begingroup\$ @anhnha, I edited my answer. Hope it answers your questions. \$\endgroup\$ – mkeith Aug 9 '17 at 8:25

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.