MOSFETs in parallel

[tpaairman]

One more question for tonight. Using the example schematic below, if I wanted to add more MOSFETs in parallel, what do I need to do?

When looking this up, I've seen mention of each MOSFET having it's own stopper resistor on their gates, which I can understand, but I'm not sure what value to use. Also, I saw one mention of having a small capacitor on each N Channel MOSFET between the gate and source, but none on the P Channel ones.

One other question is, where the feedback line goes to the inverting input of the op-amp, does there need to be a resistor in that line (in series between the speaker output and the op-amp) and would I also need a resistor from the inverting input to ground?

[J M Fahey]

Which schematic?

[tpaairman]

The one that's there now that I thought I had attached, but somehow missed.....

[J M Fahey]

Ok, now I see it.

That´s a school example, not a pactical one, but anyway:

1) to evenly share current, first you would have to match Mosfets, or use all same manufacturer and manufacturing batch.

2) besides that, and to force even current sharing, add "ballast" source emitters, typical general purpose value is n0.33 ohm, 5W each.

3) stopper resistors are needed when parallelling MosFets, specially Vertical ones (the cheaper IRFxxx type) , to avoid a dangerous self destroying situation where parallel Mosfets cross couple to each other and oscillate to death at very high frequencies, don´t be surpriosed at 50 to 200MHz range.
Stopper resistor very close to Gate leg, I solder them directly to it, period, combined to highish gate capacitance makes a low pass filter which stops the freakish oscillation.

4) what value? ... you might calcuate it knowing gate capacitance, it´s in the datasheet in a somewhat obscure way, "gate charge Coulombs" or similar, they don´t want to write "one" capacitance value because it varies with voltage, is not linear, etc. , but since these are Guitar Amps and not NASA Projects, I usually assume 1000 to 2000 pF and call it a day.

I have seen gate stoppers as low as 100 or 200 ohms ... but since I don´t care about ultra high frequencies I often use 1k or 2k2 and amps sound very good ... and I have never ever had that destructive oscillation.

5) as of NFB resistors, yes, you need a resistor from speaker out to negative input, another from there to ground, to define the gain of your amplifier.

Read about Op Amps, including the explanation of how NFB works.

Your example still *is* an Op Amp, just it´s current buffered by Mosfets which can handle , say, 5 or 10 Amperes to drive speakers, instead of typical feeble 5mA or so out of a standard Op Amp.

[tpaairman]

Thank You for the info. This is one of those things that when you look it up, you get a lot of "Oh, just throw a resistor here or that part there", but when you actually go to do it, there's much more to it than that. I've run into the same thing with tubes.

[tpaairman]

A side question on the op-amp. Over the years, I actually never worked with op-amps a whole lot, but I was always on the understanding that they had a positive and negative supply pin along with a ground. But I'm seeing lots of recommendations for audio purposes like the TL072 or NE5532, which show a + and - Vcc, but no ground. So the only thing I can think of is that the power supply negative goes to the Vcc-, but if that's the case, then would I need to have a voltage divider on the non-inverting input for the audio, or is that done internally?

[trobbins]

Adding parallel fets will add more capacitance for the opamp to drive. A slew rate limit will be hit at some frequency and voltage level.

[Dave H]

A side question on the op-amp. Over the years, I actually never worked with op-amps a whole lot, but I was always on the understanding that they had a positive and negative supply pin along with a ground. But I'm seeing lots of recommendations for audio purposes like the TL072 or NE5532, which show a + and - Vcc, but no ground. So the only thing I can think of is that the power supply negative goes to the Vcc-, but if that's the case, then would I need to have a voltage divider on the non-inverting input for the audio, or is that done internally?

There's no ground pin on op-amps. They normally operate from a dual +/- supply the centre tap of which is 'ground' (common to input and output) as below. There's no need for a voltage divider on the non inverting input. The circuit below has a voltage gain of 2. The circuit in your OP is unity gain.

[R.G.]

To do a good job of advising you, we would need to know a lot more about what you're trying to do. The advice you've been given is about as good as it can be, given how little is known about what you're trying do do.

The most obvious uses would be an audio semi-power amplifier, either a speaker driver or a 600 ohm line driver. Given how little a line driver is needed, the guess would be a speaker driver.

Audio power amplifiers get all the press, but it's the power supply that does the work. With the 12V supply for the output stage, and an opamp capable of pulling the gates above the power supply, you can only optimistically get 12V peak on the speaker output. That amounts to 9W into an 8 ohm speaker, and a peak current of about 1A into a resistive load, maybe 2A in a reactive speaker type load. That analysis holds no matter how many or what kind of MOSFETs you have there, given that they can supply the load current and withstand the voltage, while getting rid of maybe 5W of heat. If you want more power out, you either have to lower the load impedance or raise both the power supply voltage and current capability.

People (including myself) get sucked in by how simple an amplifier circuit is. The problem is that power amplifiers should rightfully be considered a power supply, with the amplifier being a little wart on the power supply that lets the power out in a controlled way.

If you changed to a 4 ohm speaker you might get 18W if everything else is PERFECT. That would double the current into the load (AND out of the power supply) while putting out the same voltage.

Paralleling MOSFETs at low voltages like under 30V total is either unnecessary or requires very good engineering indeed. You can get 50V/75A (!) single device MOSFETs at reasonable prices. Why parallel?

But if you really, really need to parallel, here's how. The reason for tinkering with devices for paralleling is that each MOSFET is not exactly like the others with the same part number. That's why data sheets give you a minimum and maximum on threshold voltage and gm. On top of that, the current you get per increment of increase of gate voltage changes with with changes in current and temperature. So the current gain of the device wanders around within the same device, and can be substantially different from device to device with the same type number.

The way you fix this is with feedback. Low value source resistors have a voltage that increases purely linearly with increasing current. If you put a source resistor in where the increase in source resistor voltage with current is bigger than the change in gate-source voltage over current and temperature. The size of the resistor depends on the rate of change of transconductance of the device over devices and temperature, and you have to consult the datasheet for the change in threshold voltage from device to device and the temperature coefficient of the transconductance. This takes some arithmetic. General guidelines are OK in most cases.

You will find references that say that no source resistors are needed, because the temperature coefficient of the gate-source voltage and transconductance is positive. That's true for lateral MOSFETs, but not the vertical MOSFETs you commonly find. So for the most common power types, you need source resistors to parallel up to several amps of current, at which point the hot ones naturally lower their gain so the cold ones can catch up.

Gate resistors are a necessity for nearly all MOSFETs used as linear amplifiers, and especially if they're used as followers and super especially if they're paralleled. The exact value is hard to compute, but having some kind of resistor right at the grid is critical, and very often anything between 100 ohms and 1K is "enough". The resistor is in series with the gate capacitance. Unfortunately the gate capacitance of a MOSFET is not constant, but varies with voltage and current levels, so it's really hard to pick the perfect one. In this we are lucky because a rule of thumb of 100 to 1K is generally good enough.

[tpaairman]

To do a good job of advising you, we would need to know a lot more about what you're trying to do. The advice you've been given is about as good as it can be, given how little is known about what you're trying do do.

The most obvious uses would be an audio semi-power amplifier, either a speaker driver or a 600 ohm line driver. Given how little a line driver is needed, the guess would be a speaker driver.

This would be a guitar amp that I'm building for my son for his bass guitar. He's a teenager so I'm not wanting to put the money into building a tube amp just yet.

The parallel MOSFETs was kind of a side idea that I had to maybe make something that would make the neighbors wet the bed, or at least give his mother a reason to cuss me out. And also partly, how hard would it be to do. But I think at this point I have to agree with you that ultimately it's really not worth the extra effort.

As one last question, if someone could check my math here. To calculate the power output, my understanding is (Vcc * .7) / speaker load. And maybe knock off a couple volts from the Vcc first. I saw that in an example, but I've tried looking that up and I can't seem to find it.

[Dave H]

As one last question, if someone could check my math here. To calculate the power output, my understanding is (Vcc * .7) / speaker load. And maybe knock off a couple volts from the Vcc first. I saw that in an example, but I've tried looking that up and I can't seem to find it.

That should be (Vcc * .7)^2 / speaker load, and knock off a couple of volts from Vcc first

[R.G.]

If you want the quick and dirty answer, go to ebay, buy one of the PCBs or kits to make a power amplifier out of an LM3886, and construct a bipolar power supply for it. Again, getting a power supply made and a neat sink made up for it vastly outweighs the considerations of the circuit itself. I recommend the LM3886 mostly because they work, and work very well. Get an Antek toroidal transformer of maybe 50 or 100VA and 20-0-20 V for it. This set up is a kind of sweet spot for an amplifier of 30-50W. The Antek toroid is about $20-30 depending on the choice, and is again a sweet spot for powering this chip. The PCBs and kits are cheap and work. It is very easy to waste a lot of time and part on amplifying the output of an opamp to drive a speaker.

Ask me how I know this...