Not All DC/DC Regulators Are Created Equal
This presentation covers the pros and cons between non-synchronous and synchronous switching regulators and the pros and cons between LDOs and the family of DC/DC power modules. It will also explore and provide application examples of the common system needs such as solution size, generated noise, efficiency, output capacitor bank selection, and how to select among the available power solutions. |
Resources
Hello. My name is Frank De Stasi. And today we will be looking at the different DC-to-DC regulator topologies, and also comparing them with some LDOs.
The agenda for today. We'll be looking at the difference between synchronous versus non-synchronous buck converters in terms of their size and cost, and ease of use and efficiency. And then we'll be looking at the difference between a switcher versus an LDO, and taking a look at their basic operation, the pros and cons, and also some example comparisons. And finally we'll take a look at some of the design resources that are available at TI.
So the first slide we show here is a typical non-synchronous buck regulator. And on the left side of the slide, we see the basic topology, where we have a diode connected from the switch node to ground, which is indicative of a non-synchronous type of topology. On the right-hand side, we see the load current. And the top figure shows the inductor current at heavy loads. And you can see that both the peak and the valley are above ground in that case.
The bottom of the figure, we see the case for very light loads. And the value of the inductor current actually reaches 0 and stays at 0 until the MOSFET transistor turns on. So this is indicative of the waveforms that you'll see in a non-synchronous type of converter. For a synchronous buck converter, that diode is replaced with an nMOS switch, as you see here on the left side of the diagram.
Again, on the right-hand side, we see the inductor current. And at heavy load, there's really no difference between the synchronous or the non-synchronous converter. Both the peak and the valley of the inductor current are above ground.
At very light loads, however, and if the regulator can't operate in what we call forced PWM mode, we see that the inductor current can actually go below ground and become negative. And of course, that's not possible with a diode, because the diode can only conduct in one direction. However, with a MOSFET turned on, the controller can allow the current to flow in both directions.
Another mode, a possible mode with a synchronous buck, is what we call diode emulation mode. And in this case, you can see at the bottom of the slide the inductor current reaches 0, and then it stops. And in this case, the controller is actually sensing when the inductor current reaches 0 and shuts off the bottom-side synchronous MOSFET. In this way, the controller emulates a diode, but with less losses.
So let's take a look at the comparison between solution size and cost for the two different typologies. In the non-synchronous topology, you obviously have to have your controller IC and an external diode. Normally it will be a Schottky diode. So of course, the total solution size is going to be a little bit larger than if you had just the IC by itself. And the cost, of course, is the cost of the IC plus the diode.
In the synchronous solution, we see here on the right-hand side both of the MOSFETs are integrated into the IC package. So of course, the solution size is going to be a little bit smaller. At the same time, since we have more components inside the IC, and they require more control of logic, the IC tends to be a little bit more expensive, because the controller is a little bit more complex. And you have a little bit more power dissipation in a smaller area, so you have to have a much better package.
Let's take a look at the layout differences between the non-synchronous buck and the synchronous buck. The most critical path in any buck regulator is what you see here in the slide that we've labeled the critical path. And essentially that's the current loop from the input capacitor through both of the MOSFET switches or the diode, if you're talking about a non-synchronous solution. So keeping that critical path area as small as possible not only improves the electrical performance, but it also improves the EMI performance.
So again, we'll take a quick look here. In the non-synchronous solution, it tends to be a little bit more difficult to optimize the layout, because you have an external diode that needs to be physically very close to the input capacitors in order to get their grounds close together again or to keep that critical loop area as small as possible.
With a synchronous regulator, on the other hand, typically if you have a good pinout in your IC, then your input and ground connections will be very close together. So the input capacitor can be very close to the IC. That's a little bit easier, usually, to keep that critical loop area as small as possible.
Let's take a look at light load efficiency. Starting with the non-synchronous version, we can see, as we saw before, at very light loads, the inductor current drops to 0. And it can't reverse back through the diode. So there can be no negative current. So in this case, the losses on the high-side MOSFET are relatively low, because the RMS currents are also relatively low.
Unfortunately, because of the relatively large diode drop, the conduction losses in the diode can be relatively large. The switching frequency is typically constant with this type of controller, so the overall losses are sort of a moderate and actually not too bad.
Moving over to the synchronous topology, if we assume we have a forced PWM mode, in this case, as we saw before, the inductor current can be allowed to go negative, depending on your controller. In this case, in the high-side FET, you have much higher losses than before, because the RMS currents are higher, basically because you're sort of wasting a lot of power in those negative inductor currents.
And again, the low-side conduction losses are also large because of those negative currents. Typically, however, the switching frequency will be constant in this mode of operation, so the overall losses will be rather large.
Finally, if we look at the same synchronous converter, however, employing a diode emulation mode, as we mentioned before, the inductor current is not allowed to go negative. And when it reaches 0, it simply stops and stays there. So just like the non-synchronous case, we have very low losses in the high-side switch. Now we have low losses in the low-side synchronous switch as well, because the voltage drop due to the RDS on is much smaller than the diode.
Also in this case, it is possible for the controller to reduce the frequency at very light loads, and also reduce the losses as well. So in this case, the overall losses are very low. And you have very good efficiency at light load using a synchronous controller that utilizes diode emulation mode.
Let's take a look at the heavy load efficiency. Turns out that the heavy load efficiency is probably not that greatly affected, whether it's a synchronous or a non-synchronous. If we take a look at the non-synchronous side, we see that, again, the heavy load current inductor waveform, both the peak and valley of the inductor current are above ground.
Now, the low-side conduction losses, again, are relatively large due to the large diode voltage drop. The high-side FET switching losses, however, are relatively small, because the Schottky diode doesn't require any reverse recovery charge in order to turn the diode off.
If we look at the synchronous side, whether it's enforced PWM or whether it's diode emulation mode, the low-side conduction losses are actually very small, because the RDS on voltage drop is much smaller than the Schottky diode drop, giving it less loss. However, the high-side switching FET losses are much larger in the synchronous because the parasitic body diode of the low-side FET needs to be turned off. And this diode tends to have-- it's a parasitic diode. And it tends to have quite a bit of reverse recovery loss to that. So I think the losses are probably about a wash in this particular case.
So looking at a summary, we have the non-synchronous case. The light load efficiency is usually very good. The heavy load conduction losses are moderate compared to synchronous. The heavy load switching losses are low because of the Schottky diode. However, the efficiency at very low output voltages is rather poor. And this is because at the low output voltages, the voltage drop of the Schottky diode is a larger percentage of the output voltage. And that gives you a little bit poorer efficiency.
The synchronous converter, operating in forced PWM mode, has very poor light load efficiency because of the large negative currents. The heavy load conduction losses, however, are very low because of the synchronous FET. The heavy load switching losses are moderate because of the reverse recovery loss in the parasitic low-side diode. However, the efficiency at low output voltages is actually quite good because there's the voltage drop across the low-side switch is very small compared to the diode drop.
And finally, looking at the synchronous converter operating in diode emulation mode, this mode has the best light load efficiency and very low heavy load conduction losses. Still has some moderate switching losses on the high side. And it has also a very, very good efficiency at low output voltages.
Let's take a look at the low transients, specifically what we call the unloading transients. Here we have an example where the input voltage is 8 volts, and your output voltage is 5 volts, and your load current is stepping from 0 to 3 amps, and then back down to 0 again. So we'll be concentrating on the side where the load current drops from 3 amps to 0.
For a non-synchronous converter the output voltage overshoot is quite large. And it takes quite a long time for the output voltage to return to its nominal value. And this is simply because with no load current, there isn't any current in the inductor to discharge the output capacitor.
Moving over to the synchronous converter that uses forced PWM mode, we see that the unloading transient behavior is quite a bit better. This is because in forced PWM mode, the controller allows a negative inductor current. And that current can discharge the output capacitor very quickly and give you a very small low-transient response time.
Finally, if we look at the synchronous converter operating in diode emulation mode, it's pretty much exactly the same as the non-synchronous case, because when the inductor current reaches 0, the synchronous MOSFET is turned off. And there's nothing to discharge the output capacitor very quickly.
Let's take a look at current limit and short circuit protection. For a non-synchronous converter, pretty much the only type of current limit available is the peak-type current limit. And that's because there's only one MOSFET connected between VN and the switch. And that's used to measure the inductor current and to employ a current limit.
So overload protection for this particular topology is done on a cycle-by-cycle peak current basis. And that's used to limit the load current in case that the load is beyond the maximum capability of the device.
Short circuit protection, on the other hand, requires an additional circuitry, usually in the form of a short circuit frequency fold-back type of topology. So when the output voltage is shorted and it becomes very near 0 volts, with this primitive topology, we normally would have to reduce the switching frequency in order to protect the converter from that kind of a hard short.
In the synchronous case, we have two MOSFET switches always working in conjunction. So we can sense the current in both of those switches, and thereby we can use both a peak current limit and a valley current limit in order to limit our maximum load current.
So for overload protection, we'll typically select a valley-type current limit in order to limit the maximum output current. And we'll use the peak current limit just to protect the high-side switch in case, for some reason, the inductor current becomes much higher than we planned.
Short circuit protection is also inherent if we use a valley current limit, because essentially the controller will wait for the valley of the inductor current to drop below a certain level before the high-side switch is turned on. So short circuit protection is actually-- we get that for free when we use the valley current limit.
So just as a quick summary here, for the non-synchronous case, in terms of size and ease of use, usually a little bit larger than a synchronous case. The light load efficiency is very good. The heavy load efficiency, well, you have a little bit more conduction loss, but you also have a little less switching loss. The efficiency at low output voltages is poor, as we said.
The low transients on the unloading side is poor, as we said. And short circuit protection requires some kind of a frequency fold-back to be built into the controller, which is just extra control circuitry that can typically make the die larger or maybe make the Q-Kernel little bit larger.
On the synchronous side, typically it's a smaller-size package or a smaller-size solution. And it's a little bit easier to use. You don't have to place the diode or design for the diode. The light load efficiency can be very good when the controller uses diode emulation mode. At heavy loads, heavy lower efficiency, well, the conduction losses are a little bit less. But on the other hand, you have a little bit more switching losses.
The efficiency at low output voltage is very good. The low-transient overshoot can be very good if your controller uses a forced PWM mode. And short circuit protection is automatically implemented when we use valley current mode control.
So let's take a look at an example of wide V in synchronous buck converters. On this slide, we see the LM43603 and LM46002 devices. These two devices are the latest in our family of Simple Switcher products available from TI. They are fully synchronous converters that employ diode emulation mode. They have an input voltage range of 3 and 1/2 to 36 volts, or up to 60 volts for the LM46002. Output currents range from 1/2 amp, 1 amp, 2 amp, and 3 amps. We have a very efficient light load mode, where we use diode emulation mode in conjunction with a PFM mode. And this gives us an efficiency of about 75% at 1 milliamp load for a 12-volt to 3.3-volt conversion.
And of course, these devices have full features that you come to expect from a modern converter-- soft start, tracking. You can synchronize them to an external clock. They have Power Good. And of course, they have the usual protection features such as current limit and thermal shutdown.
The pinout of these devices are also optimized to give good electrical performance, a good thermal performance, and good EMI performance. So if you look here at the right side of the package, you'll see that the input pins and the power ground pins are very close together. And this allows your input capacitors to be physically very close to the input and the ground of the converter. This in turn keeps the critical path loop very small that we talked about earlier.
Also, the analog ground pin and the feedback pin are also right next to each other. So you can place your feedback divider very close to the device itself, which helps to minimize noise. Again, on the left-hand side, we see the CBOOT pin very close to the switch pin. So your CBOOT capacitor can be placed very close together, which helps to minimize noise in EMI.
Also, this device has a thermal pad on the bottom, which connects to the ground plane, and also allows for a very large unbroken ground plane from the top of the layout to the bottom of the layout here, which gives good electrical and good thermal performance as well.
As an example, we have a couple of curves for EMI. On the left-hand side we see the LM43603 radiate EMI emissions. This converter is producing 3.3 volts out at 3 amps. And essentially it's well within the Class B limits just using our EVM board with a default BOM and no input filter.
On the right-hand side, we see the conducted EMI results. Essentially it's the same BOM and the same EVM. Here we're using a very small input filter. And as you can see, it's well below any of the limit lines.
Taking a look at the thermal performance for the LM44603 buck regulator, 25-degree ambient and 500 kilohertz, we can see here that even at its highest input voltage, 24 volts, and its maximum output current of 3 amps, that the IC gets only about 64 degrees C above the ambient. So this indicates that we have a pretty efficient regulator here in this case.
Taking a look at the LM46002, a higher voltage unit with 2 amps, even at an input voltage of 48 volts, and its maximum output current of 2 amps, the maximum temperature rises only about 74 degrees C in this case.
Let's take a look at the differences between an LDO and a DC-to-DC converter. First of all, an LDO is a linear regulator, essentially. And mostly your regulators are termed LDO. And the term LDO stands for Low Dropout, even though some regulators may not have a very low-dropout voltage.
The definition of dropout voltage is the lowest input to output differential that the regulator requires to still remain in regulation. So this essentially means that if you have a very low-dropout voltage on your LDO, you can work more closely to your input and output specifications, which can be important sometimes in battery-operated applications. AS your battery voltage is dropping off and dying, it allows you to operate for a longer period of time.
LDOs are usually very simple. They usually only require just an input and an output capacitor. And in some cases, you don't even need an output capacitor. Unfortunately, of course, an LDO, being a linear regulator, the output voltage must always be less than the input voltage.
Let's take a quick look at LDO operation. And as I said, it's a linear closed-loop system. So if you take a look at the diagram here, we see that the output voltage is sampled and compared to a reference. And then the op-amp that you see here controls the gate of the power MOSFET transistor. And essentially the power MOSFET is being controlled as a variable resistor between the input and the output.
And as you can see, the full load current has to flow through that MOSFET. So the losses in that MOSFET, with the load current multiplied by the difference between V in and V out, produces quite a bit of power dissipation.
And as an example, let's suppose we have an input voltage of 5 volts and an output voltage of 1.8 volts, with a load current of 1 amp. In this case, the differential between the FET is 3.2 volts, with a 1 amp of current flowing through there. So the power dissipation of the FET is 3.2 watts.
And that's actually quite a bit of power. That's actually more than the load power, which is only 1.8 watts. So in this case, if you calculate it, you get an efficiency of 36%, which is rather low.
Another example. Suppose your input is 3.3 volts and your output is 3 volts, with a load current of 1 amp. In this case, the voltage drop across the FET is only 0.3 volts. So the power dissipation in the FET is only 0.3 watts. And if you calculate the efficiency in this case, it's up to 91%.
So this pretty much demonstrates the strength of the LDO. Essentially when your input and output voltages are very close together, you can get some high efficiency out of it, even in LDO.
The pros and cons of an LDO? Usually small size. They're usually available in smaller package sizes. Very low noise. There's obviously no EMI generated with an LDO. Usually it's simpler to use. The PC board layout is much simpler than a switching converter. And they can be less expensive than a DC-to-DC converter.
On the con side, the efficiency is obviously very low when the input voltage is greater than V out. That leads to higher board temperatures as well. And finally, the LDO, being a linear regulator, can only reduce the input voltage.
So let's take a look a another member of our Simple Switcher family. This one is the LMZ20501 Simple Switcher Nano Module. And essentially we have a regulator that is embedded into a PC board substrate with the inductor mounted on top. You have one complete package that gives you an entire regulator with the inductor already integrated and built in.
In this case, the package size is very small. It's 3 and 1/2 millimeters on each side and 1.75 millimeters tall. The maximum load current available on this device is 1 amp. And the input voltage range is from 2.7 volts to 5 and 1/2 volts.
Another member of this same family, the LMZ21700, has exactly the same package and exactly the same package size. In this case, we can get 650 milliamps out of the regulator for an input voltage range of 3 volts to 17 volts.
So let's quickly just look at a step-down regulator. Essentially the DC-to-DC converter can provide very high efficiency over a wide range of input, low voltages, and loads. And it does this by storing energy in the inductor, which obviously an LDO cannot do. The power MOSFETs, in the case of DC-to-DC regulator, are used as switches rather than as a resistor. And that also improves the efficiency greatly.
So again, we take a look at the diagram here. The output voltage is sampled through a voltage divider. And it's compared to a reference voltage in the op-amp that you see in the diagram here. And the output of the op-amp controls a modulator, which turns the two switches on and off. And as you can see from what we talked about previously, these devices that I'm using as an example here are also fully synchronous.
So the fact that we have an inductor to store energy, and the fact that the MOSFETs are used as switches rather than as a resistor, means that we can achieve very high efficiency in a DC-to-DC converter.
So in this example, the same one we used a moment ago, we have an input voltage of 5 volts, an output voltage of 1.8 volts, and a load current of 1 amp. And if you go through the math that you see in the slide here, we discover that the losses in the top-side switch is only about 36 milliwatts. And the loss on the bottom side of the synchronous FET is only about 64 milliwatts. So the total loss in this approximate example is about 100 milliwatts under these conditions.
So if we calculate the efficiency, the approximate efficiency for this converter with about 100 milliwatts of loss, under these conditions, the efficiency is about 95%-- whereas we saw a moment ago under exactly the same conditions, the LDO only has an efficiency of 36%.
So the pros and cons between the switcher and the LDO. Very high efficiency over a large voltage range, which means it's easier to dissipate the heat, which in turn means your PC board temperature is lower. There's much more flexible conversions available with a DC-to-DC converter. In fact, a buck synchronous converter that we just showed a moment ago can be used as an inverter to produce a negative output voltage if you wish. You certainly can't do that with an LDO.
On the con side, it's a little bit more complex layout for a DC-to-DC converter. The PC board layout is much more critical for a buck converter than it is for an LDO. Of course, there is switching noise, which will generate EMI, which you don't have on an LDO. And typically a DC-to-DC converter can be a little bit more expensive in some cases than an LDO.
Let's take a look at an example between the two here. Let's say we have an input voltage range of 3.3 volts to 5 and 1/2 volts. Our output is 1.8 volts. And our load current is 1/2 an amp. In this particular example, we are going to say that our key issues in the application are power dissipation and size.
On the left, we have the LMZ21700 Nano Module, with its EVM of about 1,700 square millimeters. The total solution size is about 100 square millimeters. And we can see operating under these conditions from the thermal camera the maximum temperature is less than 31 degrees C.
On the right-hand side, we have the LP38693 LDO. And the PC board area is actually a little bit larger. The solution size is the same, but obviously the power dissipation is much larger. So the temperature of the die and the PC board is much hotter in this case.
So even though the LDO in this particular case has a larger PC board than the Nano Module, the Nano Module is still the winner because it's much more efficient and the PCB temperature is much less.
Second example. Let's say we have an input of 3.3 volts and an output of 1.8 volts, with a load current of 1 amp. In this case, the key issues in the application are the minimum input voltage required for proper regulation.
Here we compare the LMZ20501 Nano Module with the TPS72518 LDO. And both of these devices can meet the specifications. Both of them have about the same features and about the same accuracy. However, the Nano Module is rated for a minimum of input voltage of 2.7 volts, whereas the LDO can operate all the way down to 2.01 volts.
Now, essentially that's because the Nano Module has much more complex circuitry inside, because it's a switching regulator. It's not a simple LDO. So these circuits require a certain amount of headroom from the input voltage in order to function correctly. So regardless of the load or output voltage, the minimum input voltage for the LMZ20501 is 2.7 volts. And below that, the device will shut itself off. So in this case, the TPS72518 is the winner because it can operate all the way down to 2 volts.
The final example. V in equals 3.3 volts and V out equals 1.8 volts, with a load current of 1 amp. In this particular example, our key issues will be the application size or the solution size.
Here we compare the LMZ20501 Nano Module to the TPS7A37 LDO. And again, both these solutions meet the specifications. They have almost the same features, almost the same accuracy. However, the Nano Module comes in at about 49 square millimeters, whereas the LDO, with all of its output capacitors, comes in at about 65 square millimeters.
So we can see that all in all, the LMZ20501 requires about 25% less board space than the LDO in this case. So the Nano Module is the winner in this category.
So this is a quick summary. Use an LDO when your application requires low output voltage noise, no EMI, operation at very low input voltages, or if the input voltage is close to the output voltage. Use a Nano Module DC-to-DC converter when your application requires very high efficiency, low PC board temperature, small solution size, or you have very large variations in the input voltage.
Now let's take a look at some of the design tools and support that TI provides for our Simple Switcher products. So if you go to the simpleswitcher.com web page, you'll find quite a few tools and support training. We have WEBENCH, which I hope everyone has used. We provide reference designs, application notes, training videos, and of course, there's always the E2E form, where you can ask questions.
So in conclusion, I'd just like to say thank you. And for more information, please visit the simpleswitcher.com website. And thank you for listening.