The question of how to turn on a MOSFET might sound trivial, since ease of switching is a major advantage of field-effect transistors. Since MOSFETs are voltage-driven, many users assume that they will turn on when a voltage, equal to or greater than the threshold, is applied to the gate.
However, the question of how to turn on a MOSFET or, at a more basic level, what is the minimum voltage that should be applied to the gate, needs reappraisal now that more and more converters are being controlled digitally. While digital control offers flexibility and functionality, the DSPs, FPGAs and other programmable devices with which it is implemented are designed to operate with low supply voltages. It is necessary to boost the final PWM signal to the level required by the MOSFET gate.
This is where things can go wrong: many digital designers look at the gate threshold voltage and assume that, just like a logic function, the MOSFET will change state as soon as the threshold is crossed. This assumption, unfortunately, is wrong.
In fact, the gate-source threshold voltage value is not even intended for use by system designers. It is the gate voltage at which the drain current crosses the threshold of 250μA. It is also measured under conditions that do not occur in real-world applications. The truth is that the threshold voltage is a MOSFET designer’s parameter. It defines the point at which the device is at the threshold of turning on. In other words, it is an indication of the beginning of the process, and is nowhere near the end of it.
Certainly, the gate voltage should be held below the threshold in the off state to minimise leakage current. But for the purposes of turning on the MOSFET, system designers can, and should, ignore the threshold value entirely.
So what information should the system designer turn to? A MOSFET datasheet will have a curve which shows the MOSFET turning on with increasing gate voltage: the transfer characteristics. This is illustrated for Vishay’s SiR826ADP MOSFET in Figure 1.
The transfer characteristics, however, are most useful as a measure of current variation with respect to temperature and applied gate voltage. More important, the curve showing data with the MOSFET fully on is called the output characteristics curve, as shown in Figure 2. Here, the MOSFET’s forward drop is measured as a function of current for different values of the gate-source voltage. System designers may refer to this curve when they wish to ensure that the gate voltage is sufficient.
As Figure 2 shows, for each gate voltage at which an on-resistance value is guaranteed, there is a range in which the drain-source voltage drop maintains strict linearity with current, beginning from zero. For lower values of gate voltage, as the current is increased the curve loses its linearity, goes through a knee, and flattens out.
A closer view of the output characteristics for gate voltages between 2.5V and 3.6V is shown in Figure 3. MOSFET users usually think of this as the linear mode. However, device designers refer to the grey area as the current saturation region: for the given gate voltage, the current that can be produced has reached its saturation limit.
Any increase in applied drain-source voltage will be sustained with only a slight increase in the current, whereas even a slight change in current can lead to a relatively large increase in the drain-source voltage. For higher gate voltages, when the MOSFET has been fully turned on, any operating point will be located in the area shaded in green to the left, marked as the resistive (or ohmic) region.
When confronted with the output characteristics, designers tend to demand to know the on-resistance at their particular operating conditions. Typically it will be at a combination of the gate-source voltage and the drain-source current when the curve has strayed from the straight and narrow into the grey area.
In fact, the real key to turning on the MOSFET is provided by the gate-charge curve shown in Figure 4.
While this curve is routinely provided in every MOSFET’s datasheet, its implications are not always understood by designers. In addition, recent developments in MOSFET technology, such as trench and shielded gates and charge-compensating superjunction structures, demand a fresh appraisal of this information.
To start with, the term ‘gate charge’ itself is somewhat misleading. The linearised and segmented curve does not look like the charging voltage of any capacitor, no matter how non-linear its value. In reality the gate-charge curve represents a superposition of two capacitors which are not in parallel, have different values, and carry different voltages. In the literature, the effective capacitance, Ciss, as seen from the gate terminal is defined as the sum of the gate-source capacitance and the gate-drain capacitance.
While this is a convenient entity to measure and specify in the datasheet, it is worth noting that gate charge is not a physical capacitance. It would be a misconception to imagine that the MOSFET is turned on by simply applying a voltage to ‘the gate capacitance Ciss’. Before turn-on, the gate-source capacitance is uncharged, but the gated-rain capacitance has a negative voltage/charge which needs to be removed. Both capacitors are non-linear; their values can vary widely with respect to applied voltage. The switching characteristics, therefore, are dependent more on their stored charges rather than the capacitance value at any given voltage.
Since the two component capacitances that make up gate capacitance are physically different and are charged to different voltages, the turn-on process also has two stages. The exact sequence is different for inductive and resistive loads; in most applications, however, the load is heavily inductive and can be described using the circuit model shown in Figure 5.
The timing diagram is shown in Figure 6:
T0 – T1: gate-source capacitance is charged from zero to the threshold voltage. There is no change in the drain-source voltage or current.
T1 – T2: current begins to rise in the device as the gate voltage rises from its threshold value to the plateau voltage. Drain-source current rises from 0A to the full load current, but there is no change in drain-source voltage. The charge associated with it is the integral of the gate-source voltage from 0V to the plateau voltage, and is specified in datasheets as ‘Qgs’.
T2 – T3: the flat region between T2 and T3 is also known as the Miller plateau. Before turn-on, the gate-drain capacitance is charged to the supply voltage and holds it until the current has peaked at T2. Between T2 and T3, the negative charge is converted to the positive charge corresponding to the plateau voltage. This is also seen as a fall of the drain voltage from the input voltage to near zero. The charge associated with this is approximately the integral of the gate-drain capacitance from zero to the input voltage, and is specified in datasheets as ‘Qgd’.
T3 – T4: as the gate voltage rises from the plateau voltage to the gate-source voltage, there is very little change in the drain-source voltage or current. The effective on-resistance, however, reduces marginally with the rising gate voltage. At some voltage above the plateau voltage, MOSFET manufacturers feel confident enough to guarantee an upper limit to the effective on-resistance.
In the real world, then, turning on a MOSFET is not an event but a process. It is not a question of applying a voltage as an input at the gate which will toggle the output from high to low on-resistance. It is the two charges, Qgs and Qgd, injected into the device through the gate pin, which do the job.
The gate voltage will rise above the threshold and plateau values in the process, but that is a by-product of the turn-on process.
In addition, the speed with which a modern power MOSFET turns on or off is not a simple function of Qgs or Qgd. A detailed study of both the gate-charge curve and capacitance characteristics is necessary to compare switching speeds, especially for superjunction MOSFETs.