A Graphical Approach To MOSFET Evaluation
For the PDF version of this article, click here.
The demands of modern power supplies for increasingly higher efficiencies at higher frequencies have propelled MOSFET technology in the last few years to heights of performance that were hard to anticipate only a few years ago. The proliferation of these devices in almost all power applications requires a means of comparison to allow design engineers to choose the optimum device as easily and accurately as possible.
The product of a MOSFET's onresistance, R_{DS(ON)}, and total gate charge, Q_{G}, has been used as the figure of merit (FOM) for power MOSFETs for years. The lower the value, the better the MOSFET. In general, this value represents how good one fabrication process is compared to another. But, does this value indicate whether a given MOSFET will perform better than another in a particular application? The use of this FOM implies that two MOSFETs with the same value would perform the same in a given application. But, is that true regardless of the individual value of R_{DS(ON)} and Q_{G} that constitute this FOM?
To answer this question, we will show that we need a different tool to pinpoint the subtle differences. This tool is the FOM that results from multiplying the onresistance (referred to in the equations as R_{DS(ON)} for the highside MOSFET and R_{DS(ON)sr} for the lowside MOSFET), and the gatetodrain charge, Q_{GD}. We will show that this FOM works best to predict the performance of the highside control MOSFET, QHS (Fig. 1), in a synchronous buck converter, because its performance and losses are governed to a large degree by its dynamic losses. Meanwhile, Q_{G} × R_{DS(ON)} continues to predict the performance of the lowside synchronous rectifier, QLS, because the losses in the lowside device are mostly governed by ohmic losses. We also will demonstrate a simple method to evaluate a given MOSFET for a specific application. Furthermore, comparisons between several MOSFETs can be made to choose the most suitable device from a large selection of components.
For simplicity, consider only the common synchronous buck converter of Fig. 1 for the analysis. Assume that QLS is initially turned off and QHS is turned on. This applies the input voltage on one end of the inductor, causing the inductor current to rise. When QHS is turned off, the current will continue flowing through the inductor but now it flows through the diode D1. After a dead time on the order of a few tens of nanoseconds — dictated by the PWM controller — QLS turns on. This allows all the inductor current to flow through QLS rather than D1, since the voltage drop across its R_{DS(ON)} is lower than the diode voltage drop. Assuming that the current through the inductor does not reach zero (the continuous conduction mode), the voltage across QLS will simply be R_{DS(ON)} × I_{LOAD} during the full off period of QHS.
At the end of the off period of QHS, QLS will turn off, allowing the inductor current to flow in the diode D1. After the dead time, QHS will turn on and the cycle continues. The average voltage at the output will depend on the average ontime of QHS if the inductor current is continuous.
MOSFET Loss
Losses will be calculated based on the simple firstorder silicon losses without any consideration of the package's parasitic effects. This is done to simplify the equations and make them easy to implement in a spreadsheet while at the same time getting reasonable results.
Let's consider the two sources of losses in any switching MOSFET. The first is from conduction losses or ohmic losses, and the second is from dynamic losses. In conduction loss, power dissipates according to I^{2} × R × duty cycle, while the dynamic or switching losses are those that result when the MOSFET turns on or off, because there will be a finite voltage (a value between zero and the input voltage) between the MOSFET's drain and source while current is flowing through the transistor. The losses for the highside MOSFET may be calculated as follows:
Where t_{R} and t_{F} are the rise and fall times, f_{S} is the switching frequency, ΔPWM is the duty cycle, Q_{G} is the gate charge, V_{IN} is the input voltage, R_{DS(ON)} is the highside MOSFET onresistance, V_{DRIVE} is the gatedrive voltage and I_{LOAD} is the load current.
To calculate t_{R} and t_{F}, we need to make the following assumptions:

t_{R} ≈ t_{F}

For switching, consider only the gatetodrain charge component Q_{GD}, as the rest of the gate charge Q_{G} does not play any role in switching:
Empirically, the constant K_{RF} ≈ 4.6, and I_{DRIVE} is the gatedrive current, R_{G} is the gatepath resistance and V_{DRIVE} is the gatedrive voltage. Using the same treatment as above for the lowside MOSFET, we can derive the equations for the synchronous rectifier:
Where V_{D} is the voltage drop across the body diode or a Schottky diode in parallel. The draintosource capacitor loss will not be considered in this analysis because it represents a small percent of the total power loss compared to the dynamic and conduction losses.
A Device Evaluation Chart
Based on Eqs. 1 and 2, it is clear that R_{DS(ON)} and Q_{GD}, and not Q_{G}, play a major role in determining the power losses in the highside MOSFET in synchronous buck converters. While both R_{DS(ON)} and Q_{GD} determine how much conduction loss and switching loss is to be expected, the balance between the two losses is different in a given MOSFET used in the highside position compared to the synchronous rectifier position in a synchronous buck converter. With that in mind, a new tool is needed that allows us to readily evaluate whether a given MOSFET is well suited for operation in a given position (highside or synchronous rectifier) based on the power dissipation. This tool should also help in comparing two MOSFETs headtohead for a specific application using both Q_{GD} and R_{DS(ON)}.
In order to evaluate MOSFET power dissipation, we need to create a reference circuit representative of a commonly used buck converter. The following specifications are typical for a buck converter application:
Input voltage = 12 V
Output voltage = 1.5 V
Load current = 15 A
This represents a typical phase of a multiphase power supply for the new generation of microprocessors.
Using the power dissipation equations listed previously, in an Excel spreadsheet, one can create a set of constantpower dissipation lines using a MOSFET's R_{DS(ON)} and Q_{GD}. This set of lines represents possible combinations of Q_{GD} and R_{DS(ON)} that will produce the same power dissipation if used in the reference circuit. This should be repeated several times using power dissipation values from 0.05 W to 2.2 W. These values represent a realistic MOSFET power dissipation. Separate graphs are needed for the highside and the lowside MOSFETs.
A spreadsheetgenerated graph should represent R_{DS(ON)} on the Xaxis and Q_{GD} on the Yaxis. Several constantpower dissipation lines are drawn on the graph. Fig. 2 is the highside MOSFET graphs where the thin slanted lines are the constantpower dissipation lines calculated at a 300kHz switching frequency. The associated value of power dissipation is in black on the bottom and righthand side of the graph. Several constantFOM lines are also drawn on the graph; those are the thick blue hyperbolas with the associated FOM value on the top left side of the graph.
Fig. 2 is calculated at load current of 15 A and a switching frequency of 300 kHz at a duty cycle, ΔPWM, of 0.125. Fig. 3 depicts constantpower dissipation graphs for the control MOSFET at load current of 30 A and a switching frequency of 500 kHz. It is worthwhile noting that even a Q_{GD} × R_{DS(ON)} FOM of 10 is not capable of delivering power dissipation under 1 W, and in reality delivers a power dissipation of 1.4 W only at the optimum combination of Q_{GD} and R_{DS(ON)}.
In Fig. 3, a power dissipation of less than 1 W at 30 A and 500 kHz is to be found only in future developments in this fastmoving semiconductor industry. The optimum mix of R_{DS(ON)} and Q_{GD} for a given FOM may be obtained by plotting the constantpower dissipation graph and a set of FOM curves of interest. It can be easily observed that any constantpower dissipation line will intersect with the FOM hyperbola in two points, each with a different mix of R_{DS(ON)} and Q_{GD}. Ultimately, there is one constantpower dissipation line with a tangent to this FOM hyperbola that clearly from the graph has the lowest power dissipation and would have a single pair of optimum values for Q_{GD} and R_{DS(ON)}. Fig. 4 depicts this situation for FOMs of 10 and 15 and constantpower dissipation lines of 1.4 W and 1.7 W. It is clear that for an FOM of 15, one can never get a power dissipation of less than 1.7 W (i.e., 1.4 W is unattainable for an FOM of 15).
For synchronous rectifiers, we need a different chart because the loss equations are different from those of the highside MOSFET. Fig. 5 depicts data from this new chart as a set of constantpower dissipation lines calculated at 15 A and 300kHz switching frequency, as well as a set of FOM hyperbolas for the synchronous rectifier. Note that the constantpower dissipation lines are almost vertical, indicating that they are mostly dependant on R_{DS(ON)} and, to a much lesser degree, on Q_{GD} and dynamic losses.
MOSFET Evaluation
To compare two MOSFETs:

Plot the values of R_{DS(ON)} and Q_{GD} on the graph.

Run a line parallel to the constantpower dissipation line through each of the points plotted.

Observe the value of the power dissipation on the topside. Obviously, the lower the power dissipation, the better the MOSFET.
To determine the best mix between the R_{DS(ON)} and Q_{GD}:

Determine the FOM of the process.

Plot a curve R_{DS(ON)} × Q_{GD} = FOM, which is a hyperbola.

On the graph, draw a line parallel to the constantpower line as a tangent to this hyperbola.

Read the associated value of R_{DS(ON)} and Q_{GD}.
As an example, the table includes three MOSFETs, M1, M2 and M3, and using the typical values in the data sheets, we get that M1 FOM is about 30% better than M3 and 42% better than M2 (Fig. 2). This might cause the user to expect at least that much improvement in performance. However, according to the graph, the differences between the devices from the power dissipation, and hence efficiency, point of view is minimal. These results are supported by efficiency tests that we performed on all of these MOSFETs. Please remember that in the graph I am using the typical values of Q_{GD} and R_{DS(ON)} and not the actual devices with their spreads in parameters, which should account for the difference. Still, remember that the difference in the power dissipation is about 0.1 W between M1, M2 and M3.
An implementation of this idea is quite possible using the complete set of switching equations together with the package and dieparasitics to get a more accurate comparison. Such a model will be so complex that the use of an advanced mathematical sheet such as Maple becomes mandatory. This ambitious approach is probably worthwhile, but it is not likely to be easy to fully understand and use.
Based on the results derived in this article, I propose the use of the more accurate FOM, which is the gatedrain charge times MOSFET onresistance (Q_{GD} × R_{DS(ON)}). As you can see, for a given constant “new” FOM, the power dissipation on the highside MOSFET will vary significantly. For example, with an FOM of 100, the power dissipation can be as low as 1.3 W or as high as 2.2 W, depending on the individual MOSFET value of R_{DS(ON)} and Q_{G} = FOM/ R_{DS(ON)}.
It is clear now that FOM is not a very precise tool to measure whether a given MOSFET with a given FOM will perform well or not in a given topology and a given position (e.g. in a synchronous buck converter, the highside MOSFET requires a different mix of Q_{GD} and R_{DS(ON)} compared to the synchronous rectifier). We actually need the individual value of R_{DS(ON)} and Q_{GD} for such a MOSFET to know how good this MOSFET will perform by plotting the point on the graph and reading the value of the power dissipation.
Comparison between two given MOSFETs is very easily achieved by plotting both of them on the graph and calculating the power dissipation. A secondary benefit is actually finding the best combination of Q_{GD} and R_{DS(ON)} for a given FOM by finding the tangent parallel to the constantpower lines of FOM curve and directly reading the combination of Q_{GD} and R_{DS(ON)}.
The use of a spreadsheet lends itself to “whatif” analysis by allowing the user to change the parameters in the reference design and generating a new graph. For example, one might examine the effects of changing the frequency on the performance of any given MOSFET. A separate spreadsheet and graph should be used for the lowside MOSFET (Fig. 5).
Table. Representative MOSFETs and their figures of merit.Part Number Total Q_{G} Q_{GD} R_{DS(ON)} R_{DS(ON)} × Q_{G} M1 13 4.7 11.2 146 M2 18 5 11.5 207 M3 19 4.5 10 190