Hello. I'm Alan Martin. And welcome to the topic of Buck Regulator Architectures. This is an overview of buck regulator architectures discussing the choice between power switches and control loop methods. 

Let's begin with the power switch choices. There are synchronous and non-synchronous buck regulators. Those are shown in the two columns here. And then you can either have internal FETs inside a regulator, an integrated circuit-type regulator, or you may have an integrated circuit controller driving external MOSFETs. So those are the various choices you have in terms of power switch arrangements. 

Let's discuss the simplest, which is a single-phase non-synchronous buck converter. Schematically, this is a representation. There's an input switch shown in the upper left-hand corner. You've got an input supply. And on the right-hand side is the output load. The voltage across the output load is always lower than the input voltage in any buck converter. 

In the middle of the schematic is the commutating diode. That is what supports inductor current during the off time of the switch. The most important equation on here is the Vout is the duty factor times Vin, or re-expressed as the duty cycle is Vout over Vin. That's a very important equation. It's a good approximation. In reality, the duty cycle will be slightly wider to make up for losses in the stage. 

Here's some key waveforms off of that schematic. Notice that the input current is discontinuous. The inductor current, which is the third waveform, is always continuous. It does have inflection points, but there are no discontinuities in the inductor path. 

The bottom waveform is the switch waveform. That's the junction between the commutating diode, the input switch, and the inductor input side. This is a rectangular waveform. The duty cycle of the switch, again, is Vout over Vin plus a little bit more to make up for losses. 

Notice that the discontinuous input waveform in the top trace is what is seen across the input capacitor of the stage. The RMS value of that waveform is what sets the physical size of the input capacitor. The input capacitor is based on RMS ripple current rating, not on its capacitance value. 

Let's look at the difference between light-load operation and full-load operation. The upper waveforms show what is occurring in a synchronous buck regulator power stage topology. The blue trace shows inductor current. The minimum and the maximum of that inductor current waveform is what's called the ripple current of the inductor. The midline, or the average, through the blue waveform is the average output current from the supply. 

Now, as your load current decreases, that blue waveform moves down and crosses the zero axis. If, in fact, it's a synchronous regulator, such as a voltage mode regulator running in PFM mode-- excuse me, PWM mode, you will get negative inductor current. 

And many synchronous regulator control architectures look for the zero crossing or the negative inductor current and turn the bottom MOSFET off in order to provide diode or diode emulation mode, similar to the way non-synchronous buck regulators operate. That improves efficiency at light load. 

There's another anomaly that can occur in a synchronous buck. The two MOSFETs are arranged directly between Vin and ground. If both switches manage to turn on simultaneously, you've applied a direct short between the input supply and ground. This causes the MOSFETs to overheat. They will get slower, and the situation can become destructive. So it's important to not allow the MOSFETs to turn on simultaneously. 

The time in which both MOSFETs are turned off is referred to as dead time. There are several approaches to controlling dead time. The most common is a fixed dead time. That's essentially done by a one shot inside the control loop, and it makes sure that the bottom switch is off previous to the top switch turning on and tying the switch node to Vin. 

There are other schemes known as adaptive dead time. These are more advanced controllers, and they allow a lot more flexibility in MOSFET selection. But it's necessary that the detection circuitry detect that both MOSFETs are off before turning one of them on, and this must occur in both the leading and the falling edge of the switch waveform. 

Now let's talk about the choices of control mode. The power stages can all be controlled by one of various choices of control mode. And each control mode has its advantage or disadvantage, and we will cover those in the following slides. 

Current mode control, there are several particular approaches-- peak current mode, valley current mode, average current mode. There's a number of them. All of them have their various advantages. That's a more advanced topic. And then there's a hysteretic mode control. Constant on-time type parts are a derivation of hysteretic mode. 

First, let's talk about voltage mode control and its advantages and disadvantages. Schematically, all of these approaches look nearly the same. Except for current mode there will be some current-sensing path. The advantage of voltage mode control is its stable modulation. It's not very sensitive to noise. You only have a single control path, and it works over a very wide range of duty cycles. And if you have a P-type top switch, it may even go to 100% duty factor. 

There are disadvantages, though. The loop gain  is proportional to Vin, and that's understandable. As you increase Vin, the gain of the loop increases. Typically, the output filter, the LC filter formed by the output capacitors and the power inductor, that is a two-pole, low-pass filter. And to properly compensate that, you need what's called Type III compensation. And this involves a number of small resistor and capacitor values that must be calculated for proper operation and need to be measured to verify proper operation once the design is done. 

A difficulty that you find is that the continuous inductor current mode and the discontinuous inductor current mode that we discussed previously results in a compensation challenge, that the compensation network you came up with may not be appropriate for both modes of operation, and therefore you must make some sort of a compromise. 

The other disadvantage is that since the loop gain is proportional to Vin, if you have a fast step change in input voltage, that will probably show up on the output waveform as some sort of undershoot or overshoot that is undesirable. In addition, because the switches and the inductor and the control loop are unaware of what the output current actually is, additional circuitry must be applied if you're going to guard against destructive events that occur in the event you overcurrent the supply or accidentally short the output. 

In contrast, current mode control has its set of advantages and disadvantages. The output LC filter, because of the inner current control loop, changes the output LC characteristic to that of a single-pole low pass. Therefore, it's easier to come up with an appropriate set of compensation, usually Type II compensation, and that may be more stable and easier to derive. 

This control method has very good line rejection as well. So step changes in the input voltage don't show up nearly as badly in the output waveform as a result of using this control technique. Inherent in this control technique is that you get cycle-by-cycle current limiting. So in the event of short circuit or overcurrent events, the power supply will behave properly and non-destructively. 

In the case of using it in multi-phase operation, which we'll discuss in a later module, current sharing can be implemented because you provide independent current information for each phase of a multi-phase design. 

One of the disadvantages is that they may be susceptible to noise. This is because the current feedback signal that enters the control loop is quite small in nature because you're trying to minimize the amount of power wasted in the current-sensing element. 

It's susceptible to minimum ON-time issues. So if you've got a peak current mode control loop, it may be inconvenient or very hard at minimum ON-time to properly sense the inductor current and end the switching cycle. This tends to happen at high input voltages and low output voltages where that event occurs. 

And as I alluded to, you need some sort of sense resistor mechanism to actually sense the current in each phase of a particular design. That can either be a discrete resistor. There are occasions where you can use the inductor resistance as part of the control loop. And there are other topologies that use MOSFET on-resistances instead of a discrete resistor. 

And then the third most common control method is that of hysteretic mode. This is unique in that rather than an error amplifier, an error comparator is used, and that is what provides the mechanism for setting the output voltage. A big advantage of hysteretic control is they have extremely fast transient response. Essentially, the loop is looking at the output voltage, and if it's too low it turns on the switch. And when the voltage reaches the necessary output level, it turns the switch back off. 

There's no phase compensation required in these. In other words, they're inherently stable, if you can use that, because it could also be stated that they're inherently unstable, and hence the name hysteretic mode. Disadvantages to these are essentially ripple regulator topologies. And any extra noise on the output may cause malfunction of the loop. 

Additionally, they do require some output ESR of the output capacitors in order to actually sense changes in output voltage. So output capacitor selection is key to the proper performance. They're very sensitive to layout. They're very sensitive to component selections. They don't scale well, and they don't migrate well from design to design. So you may get it working the first time. You try migrating the very same schematic into a different layout, and you run into problems. 

Also, there's large frequency variations in these. Some of them are essentially spread spectrum or chaotic. And some of them, such as minimal on-time, are pseudo-constant frequency. And there are approaches for mitigating issues there. 

So there you have it. Those are the various control methods that you may find in a synchronous or non-synchronous buck regulator. Thank you for joining us on this topic.