Real World Mosfets
On the last page I mentioned that mosfets are like electronic switches. In many ways they are, but it is actually not that simple. Mosfets have many imperfections that can cause problems for H-Bridge designs. Each mosfet also has maximum current and voltage parameters, which must not be exceeded in your designs. It is a good idea to read through the mosfet datasheets to ensure that it is the right mosfet for your design. Nothing is more frustrating than freeing trapped smoke because you chose the wrong mosfet.
Mosfets have 3 regions of operation that we will concern ourselves with for H-Bridge design:
The first is cut-off mode, when the mosfet does not conduct.
The second is the saturation mode, when the mosfet is switched on and is conducting with a very low drain to source resistance.
The third region is in between cut-off and saturation, and is known as the triode region. This is the region where the mosfet begins to conduct, but has a high channel resistance (discussed below).
There are 3 key voltages that will determine what mode the mosfet is operating in. These are:
- VGS, the voltage from the gate to the source, taken with the gate as positive and source as negative
- VDS, the voltage from the drain to the source, taken with the drain as positive and the source as negative
- Vt, the mosfet’s threshold voltage. This value is a device parameter, and is roughly given in the device’s data sheets. (For example, an IRF540 N-Mos has a Vt of 2 to 4 volts, while an IRF9540 P-Mos has a Vt of (-2) to (-4) volts)
For an N-Mos device:
- Cutoff when VGS < Vt
- Triode when VGS > Vt, and VDS < (VGS – Vt)
- Saturation when VGS > Vt, and VDS > (VGS – Vt)
For an P-Mos device:
- Cutoff when VGS > Vt
- Triode when VGS < Vt, and VDS > (VGS – Vt)
- Saturation when VGS < Vt, and VDS < (VGS – Vt)
Ideally, for an H-Bridge we want to spend as little time in the triode region as possible, and maximize the time the mosfets are in the cutoff (off) and the saturation (on) regions. This will cause the mosfet to be the most efficient, since we avoid the losses caused by the higher drain to source resistance in the triode region.
Don’t get the idea, though, that the triode region is useless. For other types of circuits, such as analog amplifiers, the triode region can be quite useful. However, for an H-Bridge we treat it as a bad thing, and try to avoid it as much as possible.
Every mosfet has a drain to source resistance, or channel resistance. We will mainly concern ourselves with the saturation mode channel resistance, or RDS-ON. The higher the current rating of the mosfet, and the smaller the package, the smaller this channel resistance will be.
Why does the channel resistance matter? Every resistance has a voltage drop when current is passing through it, and this voltage drop will increase proportionally with current.
For example, if a motor draws one amp of current, and the high and low side transistors each have an RDS-ON of 0.5 ohms, Ohm’s law will give us a voltage drop of:
V = I * R = (1 Amp) * (2 * 0.5 Ohm) = 1 V.
So, if you are using a 7.2 volt battery pack, the motors would only be getting 6.2 volts. The higher the current, the lower the voltage across the motor terminals will be, and you will not be achieving maximum performance.
Surface mount mosfets also have smaller RDS-ON values, if you are willing to work with them. However surface mount components also need more care for heat dissipation than a component in a TO-220 package.
When the mosfet is in the triode region, the channel resistance is much higher, which is not desirable for H-Bridge design, and is why we try to avoid the triode region.
One very important thing to note is that all mosfets have a rated VGS-MAX. The value is usually between 15 and 20 volts, usually with an average of 18 volts. If you exceed this voltage limit, you will destroy your mosfets. It is not a problem when your power supply is below the rated VGS-MAX. However, if you are running with a voltage higher than the rated VGS-MAX, you need to limit the VGS voltage difference. The easiest way to do this is to use Zener diodes to limit the voltage.
MOSFET Reaction Times
If you look through a Mosfet’s data sheet, you will notice delay times for the mosfet to turn on and off. These delay times can be as long as 50 nano-seconds. Usually it takes loner for a mosfet to turn off than to turn on.
This is another reason why you should put a delay in-between switching mosfets on and off in your circuit. For example, if your h-bridge has a ‘forward’ signal, to go backward first turn all the mosfets off, wait a couple milliseconds, then switch on the appropriate combination of mosfets to reverse your motor.
Gate capacitance in mosfets is also an issue. If you are driving your h-bridges with a PWM signal, a higher gate capacitance will increase the time it takes for the mosfet to switch on and off. Ideally, you want to have the mosfet switching on and off as quickly as possible, in order to spend as little time as possible in the triode region. Slow switching will result in much higher channel resistances, which in turn will result in more power dissipation across the mosfet, causing the circuit to become less efficient, and the mosfets to heat up.
Also, you must make sure the gates of your mosfets see as little resistance as possible. For example, a high resistance to ground will cause the gate capacitance of an N-Mos to discharge much slower than if the resistance was low. This results in the mosfet spending more time in the triode region.
Mosfet drivers are one solution. I have successfully used the TC4427. It is very easy to use, and is placed in the circuit right before the mosfet gates. Other solutions are available, though I have not used any others as this one has worked fine for me. Note that the maximum supply voltage for the TC4427 is 18v. Make sure you DO NOT put any pull up or pull down resistors between the TC4427 and the mosfet’s gate, as this will severely reduce it’s effectiveness.
Also, choosing the correct PWM frequency is essential. If the frequency is too low, the motors will become too inefficient. If the frequency is too high, the h-bridge will become too inefficient for the reasons outlined above.
I have created a simple PSpice simulation, using the student version of OrCad PSpice, in order to demonstrate some of the problems of real PWM driven mosfets.
The circuit is a mosfet H-Bridge, using the IRF9540 as the high side P-Mos, and the IRF540 as the low side N-Mos. I attained spice models of these two components from International Rectifier’s web page.
The spice circuit I made can be downloaded here. (This file requires the mosfet models mentioned in the above paragraph. Click on the part numbers to download them) The circuit is set up to have the high side P-Mos connected directly to the 18v volt supply, while the low side N-Mos is connected to a 1kHz, 50% duty cycle square wave with an 18 volt amplitude. I put a resistance on the input signal in order to demonstrate the effect of not only this resistance, but the current limiting it causes and to exaggerate the distorted waveform caused by the gate capacitance.
In figure 4.1, the faint dotted square wave is the PWM input (V(7) in the circuit), and the solid line is the output, or voltage across the load ( V(2)-V(5) in the circuit). The gate capacitance is causing a brief delay while the mosfet is turning on, and a much larger delay while turning the mosfet off. When the mosfet is turning on and turning off, it is in the triode region, which is very inefficient and causes high power losses in the mosfets. It is important to minimize the time the mosfet is in the triode region.
It is hard to see in figure 4.1, but the output voltage maximum is slightly below the 18 volt supply voltage. This is caused by the channel resistance, RDS-ON. In this simulation, the load is drawing approximately 1.8 amperes. the IRF540 and IRF9540 have RDS-ON values of 0.077 ohms and 0.2 ohms, respectively. (Maximum current ratings are 27 amperes and -19 amperes)
Using Ohm’s law:
V = I*R = (1.8A) * (0.077 + 0.2) = 0.4986
This voltage drop of approximately half a volt may not seem like much, but it will increase proportionally with current draw. When larger motors are used, with higher currents at load, this voltage drop will be much greater.
For comparison, figure 4.2 shows the output when both the high side and low side MOSFETs are driven by the PWM signal. It is easy to see that when the high side is left fixed and only the low side has a PWM signal applied to it (as in figure 4.1), the output is much more desirable.
As a final comparison, Figure 4.3 shows the output when the high side mosfet is fixed to 5 volts, and the low side mosfet is driven with an 5 volt amplitude square wave (1kHz, 50% duty cycle). The rail voltages are still 18v. This demonstrates why it is a good idea to drive the mosfet gates with the same voltages as the H-bridge circuit is driven by, not a smaller logic voltage. (However keep in mind the VGS-MAX). The mosfet will switch on and off much slower when you drive the gates with your logic voltage (usually 5v). If you drive them with your motor voltage, you will have much better performance.
Spice simulations, however, are not perfect and do not always react like a real world circuit. I created this one simply to demonstrate some of the problems associated with h-bridges, and as such Spice can be a powerful tool.