Transistor vs MOSFET as Switch, which one's better ?

[electrosam]

Hello !

I have been using Transistor as a switch in many of the designs, as a convention.
Nothing fancy, just a BC547 and a base resistor and a controller to toggle.

Now, MOSFET does not need a Gate (or Base) resistor because it already has very high input impedance.
So is it a good choice than a transistor in case of Switch application ?

[Kremmen]

Sure, MOSFETs are good switches, but not for the reason you mention.
You see, while the gate of a FET is insulated from the conduction channel and no gate current will flow to the main channel, a FET gate has a not insignificant capacitance. At higher frequencies charging and discharging that capacitance can require considerable currents.
So, while FETs and IGBTs have largely replaced BJTs in switching applications it is for other reasons than just lack of base current.
A proper FET switch design does contain a gate resistor to limit the charging current spikes and eliminate or minimize ringing in the drain circuit. Heavily overdriving the gate usually results in oscillations in the MHz to GHz range subject to details of the circuit. You don't necessarily want that.

P.S. There is no "better". For small signals and low frequencies either will work. For higher power FETs tend to dominate and for the real high current applications, IGBT are used, combining some of the desirable features of both BJT and FET. (Also combining some of the less desirable features such as slow turn-off). So in a critical case, study the literature. There is no 1 right answer.

[olsenn]

First of all, a MOSFET is a transistor; the difference you are referring to is between a BJT and a MOSFET. Both can do the trick for you, but historically BJTs have been cheaper and more common. However, FETS can usually handle greater amounts of power. Unfortunately, there is no easy answer to your question; there are many properties to consider, such as switching time (max frequency), voltage, current draw, quiescent current, physical size, cost...

[AndyC_772]

[quote author=Kremmen link=topic=10663.msg144700#msg144700 date=1347202078A proper FET switch design does contain a gate resistor to limit the charging current spikes and eliminate or minimize ringing in the drain circuit. Heavily overdriving the gate usually results in oscillations in the MHz to GHz range subject to details of the circuit. You don't necessarily want that.[/quote]

On the other hand, a gate resistor can also result in the FET spending much longer in between the fully-on and fully-off states than it could do, resulting in a huge increase in switching losses. A couple of misplaced ohms here can be the difference between a switch that works well and gets a bit warm, vs one which overheats and lets the magic smoke out within seconds of being switched on.

I use power MOSFETs in dc/dc converters a lot; normally the problem is turning them off and on fast enough.

[Kremmen]

proper FET switch design does contain a gate resistor to limit the charging current spikes and eliminate or minimize ringing in the drain circuit. Heavily overdriving the gate usually results in oscillations in the MHz to GHz range subject to details of the circuit. You don't necessarily want that.

On the other hand, a gate resistor can also result in the FET spending much longer in between the fully-on and fully-off states than it could do, resulting in a huge increase in switching losses. A couple of misplaced ohms here can be the difference between a switch that works well and gets a bit warm, vs one which overheats and lets the magic smoke out within seconds of being switched on.

I use power MOSFETs in dc/dc converters a lot; normally the problem is turning them off and on fast enough.

Limiting the gate charge/discharge currents will slow down FET turn-on and turn-off, that much is true.
Whether this has an impact on performance and what kind of impact, is a topic for circuit analysis and design in each case, there is no single right answer.
So far we have been discussing the so called hard-switching case where the FET switches full drain voltages and currents simultaneously. I attached a diagram i have made some time ago, outlining the normal turn-on sequence of a hard switched FET. You can see that the full drain current has developed before any appreciable decrease in drain-source voltage Vds. During this time the device will dissipate almost full Vds x Id, emphasizing the need for rapid switching.

In practice there is always a "gate resistor" of some kind, if only the source impedance of the output feeding into the gate and parasitic impedances of the actual circuit. Often this impedance is low causing the gate to be heavily overdriven. Now a small/reasonable amount of overdrive is desirable to account for component parameter variations and to ensure full switching. Gross overdrive will result in oscillation of the drain circuit, sometimes severely.
The idea to optimally drive a FET gate is to provide the electric charges Qgs and Qgd in minimum time to ensure a quick turn-on of the main channel and vice versa. Once that is done, further driving the gate will result in overcharge. Since the channel is already fully on, no further increase in conduction is possible.

A practical first approximation in designing the gate circuit would be to consider the required switching time and gate charges (yes - charges, NOT capacitances. The capacitances vary as a function of the drain-source voltage). So do the charges of course as functions of Vgs but those are usually available in the datasheets. Lets take an example, say the regular run of the mill IRF540N that we have learned to trust and love. Say you want to turn it on and off in 1 microsecond (let's be reasonable here, this is an older spec component). From the IR datasheet you can see that assuming our Id would be say 20A max, the device will be fully on when Vgs => 5V. Lets use a comfortable margin and say we drive the FET with a gate voltage of 8V. Further checking the datasheet we see that the total gate charge at Vgs=8V will be around 35-38 nC subject to Vds. Lets use 38 to be on the safe side and now we can right away calculate, that to transfer 38 nC of charge in 1 us, we need 38 mA of current. Next, again checking the gate charge vs Vgs diagram we see that the plateau t2 - t3 occurs at around 4.5V Vgs. This is the voltage available to drive the turn-on current. Now, simple division gives us R = 4.5V / 38mA = 118 ohms. This is the first approximation for the gate resistor.

Further complications may require that the charge and discharge currents be made unequal and then the next step is to add series diode-resistors in antiparallel to control the gate resistance in both directions separately. After that it gets more complicated and laziness prevents me from going there in this post.

So, while overdriving a FET gate to the max does result in the shortest switching time, you don't usually want the side effects that the high dv/dt and resulting displacement currents bring. In a straightforward hard-switching case the reasoning i outlined above should be applied and if not good enough, improved designs should be considered. A better controlled gate charging circuit may be used, or one can try to suppress the ringing resulting from overdriving. Or finally a more clever switching strategy could be attempted such as a ZVS scheme or synchronous/resonant switching, avoiding the hard switching issue altogether.

Meanwhile the attached IR app note clarifies my point in some more detail.

[electrosam]

Thanks all for comprehensive information !

[jackbob]

Quick simple  mosfets are commonly used as switchtes, in fact probably 99% of the time that's what they are doing in one form or another. Bjt's are mainly used with analog signals (between 0 and the peak voltage). Mosfets should ideally only be all the way on or off which is exactly what a switch does, if they are not all the way on or off they get really hot and will fail quickly. Bread boarding I never use gate resistors but in practice at least a small resistance is good to have. Always use a pull down resistor though when in doubt 4k7 or 10k are great values. A pull down ensures the gate voltage is 0 when you turn off the gate voltage. It goes between the gate and ground. Those are the basics which of course can get much more advanced especially as the frequency goes up.

[Rudolfo]

Hello,

just in case short and/or overheating protection is needed, a smart MOSFET switch, like the iC-DN/DP, could make sense as well. Specifically when interfacing the microcontroller to the industrial environment. Here are some interesting application ideas for that: http://www.ichaus.biz/mcu_interface , or the smart MOSFET switches: http://www.ichaus.biz/keyword/Output%20Stages .

[jackbob]

Your question was very vague, mosfet being used as a switch can be as simple as a on off light switch or used in countless switching applications such as pwm, this is where it gets more complex and rise and fall time as well as capacitance come into big play. Rudolfo, I really like that link to that pdf about monitoring mosfets. Yes in the industrial world there needs to be extra safety (no room for errors when you are driving large appliances with these).

[AndyC_772]

Quick simple  mosfets are commonly used as switchtes, in fact probably 99% of the time that's what they are doing in one form or another. Bjt's are mainly used with analog signals (between 0 and the peak voltage). Mosfets should ideally only be all the way on or off which is exactly what a switch does, if they are not all the way on or off they get really hot and will fail quickly.

It's just a question of power dissipation, just as it is with a BJT. A BJT will get just as hot as a MOSFET if it has the same current through it and the same voltage drop, and there's nothing about a MOSFET that makes them inherently more likely to overheat. In extreme cases (thermal runaway), the reverse can be true.

[jackbob]

That is true the power dissipation will be the same, but mosfets are more commonly used for switching applications and are optimized for that purpose (much lower drain source resistance). They have very fast rise and fall times making them respond quicker, wasting less power. At higher frequencies with more rise and fall times per second, mosfets perform better and waste less power. Besides their performance in these conditions, and their lower gate source resistance, the power dissipation is the same. But for many switching applications these conditions apply and make the mosfet ideal for the job.

[ptricks]

Asking that question is about like asking people which is better, AC or DC ? Both have their uses.

[AndyC_772]

I can feel a Harry Hill sketch coming on...