mosfet capacitance - stupid question?

I'm displaying my ignorance here, I'm sure, but where better to erase it..

If one of the major problems with mosfets is their input capacitance, isn't a smaller capacitor in series with the input capacitance, in order to reduce the overall capacitance, an option? Is this stupid, commonplace, or impossible?
 
A series cap would not fix things because a large portion of the signal would be dropped across said cap. Say that the Cgs value is 4.5 nf (it's non-linear as well, but we're keeping it simple), and we add a 0.5 nf series capacitance. The overall equivalent is 0.45 nf, one tenth of Cgs. But nine tenths of the input signal is lost by being dropped across the 0.5 nf cap. Hence the FET receives an input that is 0.1 times the original signal (-20 dB).

Does this help?

Claude
 
The gate input capacitance can play havoc with a VAS stage as it loads the VAS output.
What people generally do is put a resistor in series (22R to 470R) with the gate to slightly decouple it from the VAS.

IRFP240 is quite a good MOSFET because its input capacitance is relatively low.
Watch out for high power MOSFETs like the IRFP250 which is two MOSFETs on the same die which gives twice as much capacitance as IRFP240.
 

homemodder

Disabled Account
2006-05-12 5:36 am
The gate input capacitance can play havoc with a VAS stage as it loads the VAS output.
What people generally do is put a resistor in series (22R to 470R) with the gate to slightly decouple it from the VAS.

IRFP240 is quite a good MOSFET because its input capacitance is relatively low.
Watch out for high power MOSFETs like the IRFP250 which is two MOSFETs on the same die which gives twice as much capacitance as IRFP240.


Nigel, I think what people should do is to always use a driver before the output mosfets.
 
Mosfets inherently form a Colpitts oscillator and/or essentially a Hartley oscillator when you take into account the lead inductance. Gate stopper resistors simply form a low pass filter in order to snub the resonance of the oscillator that exists in every mosfet. In addition a gate Zobel filter placed on the gate pin as close to the package as possible can help dampen the oscillation even better in turn allowing a smaller gate stopper, increasing the usable bandwidth of the mosfet. The real problem, as Eva stated, is the non-linear overall capacitance, which to a significant degree is dependent on Vds. You aren't just driving Cgs, but also Cds and Cgd. A quick look at a datasheet will show this to you.


:2c:
 
In this paper parasitic oscialltion and the cure using Zener diodes and ferrite beads are discussed:

http://www.microsemi.com/micnotes/APT0402.pdf

It seems it only refers to mosfets used for switching and I'm not sure whether the results are applicable to audio circuits at all. Adding a ferrite in series with the gate adds another nonlinear thing. On the other hand side the effects occur at frequencies well above the audio band...

What do you think? Is a ferrite bead in series with the gate a good idea or not? What about Zener diodes from gate to source?
 
Ferrite beads are good for high bandwidth applications, say where you're trying to get 5 nanosecond edges on your 2MHz switching power supply.

If you aren't making 5ns edges, don't bother using them.

In fact, gate resistors on the order of 100-1000 ohms are plenty for audio applications. In the linear range, gate voltage changes little, and only a few miliamps are necessary to change the gate at the amplifier's slew rate.

Zener diodes are for protection and have nothing to do with the switching speed of a MOSFET. If your circuit could apply excessive Vgs to the transistors, a zener might be wise. If the circuit is designed such that this can never happen, an additional zener is superfluous.

Tim
 
Zeners are also used to prevent spurious turn-on issues and gate overvoltage beause of dV/dT on the drain in switching circuits which hapen if a complementary gate driver passes through a non-conducting (high-Z) region while the voltage on the drain is going higher - if it's doing so quickly enough, the Cdg and Cgs may form a capacitive divider which places a voltage on the gate in excess of the treshold voltage, which will result in the MOSFET bucking the switch-off, resulting in severely increased switching loss. At extremes, it may even result in overvoltage at the gate, either positive or negative (depending on wether the MOSFET is being turned on or off).
However, this is exceedingly difficult to accomplish in linear operation at audio frequencies so zeners typically only have the effect of limiting current by limiting available Vgs. The problem with this is that it's quite unpredictable what the limit current will be if there is no source resistor. For IRF type MOSFETs the limit current also increases with temperature in practical circumstances as the Vgs tempco is negative up to currents near the maximum rated - and limiting is normally set lower.
 
Last edited:
MOSFET have internal gate resistance, usually in the 1 ohm to 50 ohm range. Newer parts tend to have much lower resistance for faster switching while older parts are usually high resistance.

A resonant method is required for measuring this resistance.

Ferrite beads are mostly useful for switching.

Mosfet transconductance varies a lot depending on drain current. This results in signal dependent open loop gain and crossover frequency, and thus lots of non-linearity. I would never use MOSFET in a linear audio amplifier without local error correction to avoid inserting the ugly MOSFET transfer function into the global feedback loop.
 
Mosfets inherently form a Colpitts oscillator and/or essentially a Hartley oscillator when you take into account the lead inductance. Gate stopper resistors simply form a low pass filter in order to snub the resonance of the oscillator that exists in every mosfet. In addition a gate Zobel filter placed on the gate pin as close to the package as possible can help dampen the oscillation even better in turn allowing a smaller gate stopper, increasing the usable bandwidth of the mosfet. The real problem, as Eva stated, is the non-linear overall capacitance, which to a significant degree is dependent on Vds. You aren't just driving Cgs, but also Cds and Cgd. A quick look at a datasheet will show this to you.


:2c:

Well-stated, CBS240. I believe these observations and fixes were first mentioned in the context of MOSFET audio amplifiers in my paper on a MOSFET amplifier with error correction published in the JAES in 1984. More detail on this can be found in that paper, which is available at Cordell Audio: Home Page.

Cheers,
Bob