I researched a bit on MOSFETs, and to judge by the safe operation areas, they're only good for switching applications.
So, does that mean that all the MOSFET amplifiers are class D? Or are some devices that can work with partial conduction?
Also, look at the safe operation area in the attach. It's from the NTY100N10, listed as "300w max power dissipation". Now, judging by the SOA curve, there isn't any combination of Vds and Id that can possibly result in 300w dissipation.
I assume this is peak power in switching operation. Is this correct?

 

I have seen a couple of class A headphone amps that run FET's in the linear region. Mostly, though, it's all on and off switching.

The transitions from one state to the other involve a transition through the ohmic region, which causes the device to heat. That is where the dissipation limit comes into play. Drivers try to turn them on and off as quickly as possible, but they do get hot.

International Rectifier has a number of papers available on their site that discuss this issue in some depth.

 

I have seen a couple of class A headphone amps that run FET's in the linear region. Mostly, though, it's all on and off switching.

The transitions from one state to the other involve a transition through the ohmic region, which causes the device to heat. That is where the dissipation limit comes into play. Drivers try to turn them on and off as quickly as possible, but they do get hot.

International Rectifier has a number of papers available on their site that discuss this issue in some depth.

I found some power MOSFETs (like the HUF75344A3) on the Fairchild site that have less drastic SOA curves, maybe they can be used for DC (the curve is attached).
I'm going to the IR site, to read that documents. Thank you.

 

MOSFETS are used as both switches and in linear operation, just like bjts. However, they have the negative that their turn-on voltages (Vgs) is not a very easily controlled parameter in manufacturing, so biasing them is a little more difficult. Also, their equivalent gain (called transconductance, gm) is not as high as a bjt's beta (hfe). I have used one in a discrete opamp as the output AB stage. Worked great. One of the best things is that the drain to source looks resistive when it is on. So you don't have the 0.3V saturation voltage like a bjt. That's how you create a rail-to-rail output.

As far as the maximum power read Note 1. It is a 10us 2% duty cycle dissipation limit. On the SOA graph, this is the level of the top-right 10us line (3 Amps, 100Volts = 300W)

 

MOSFETS are used as both switches and in linear operation, just like bjts. However, they have the negative that their turn-on voltages (Vgs) is not a very easily controlled parameter in manufacturing, so biasing them is a little more difficult. Also, their equivalent gain (called transconductance, gm) is not as high as a bjt's beta (hfe). I have used one in a discrete opamp as the output AB stage. Worked great. One of the best things is that the drain to source looks resistive when it is on. So you don't have the 0.3V saturation voltage like a bjt. That's how you create a rail-to-rail output.

As far as the maximum power read Note 1. It is a 10us 2% duty cycle dissipation limit. On the SOA graph, this is the level of the top-right 10us line (3 Amps, 100Volts = 300W)

Thank you, that gives me hope of using one for the power supply. I'm making a regulated power supply, and need to load the least possible an OPAMP, and a MOSFET would be very good for that.
What specific device did you use?

 

I think it was a 2n7000, but I was using it as a signal output, not power transfer.

So if I get what you are doing, you want to put an op-amp output into a circuit to drive a heavy load at the same voltage as the op-amp output? This is probably going to require a feedback loop.

To figure out which part you need, you have to determine what the maximum Vds * Id current that will happen in your application. And then make sure that all situations work in the SOA DC curve for that FET as well.

Does this make sense?

 

Something else to consider, almost all analog applications can be duplicated digitally. Regulators can be done with switching, and Class D amps are also a switching, high efficiency audio amp.

 

I think it was a 2n7000, but I was using it as a signal output, not power transfer.

So if I get what you are doing, you want to put an op-amp output into a circuit to drive a heavy load at the same voltage as the op-amp output? This is probably going to require a feedback loop.

To figure out which part you need, you have to determine what the maximum Vds * Id current that will happen in your application. And then make sure that all situations work in the SOA DC curve for that FET as well.

Does this make sense?

Yes, a lot of sense.
My idea is to use a LM317 to set the voltage reference, then feed an opamp with that reference, and drive the mosfet with the opamp. Then i intend to add current limiting, using a resistor to sense the current, and pull down the base of the mosfet (i attached the basic idea).

Something else to consider, almost all analog applications can be duplicated digitally. Regulators can be done with switching, and Class D amps are also a switching, high efficiency audio amp.

Yes, i know they are a lot more efficient, but this is going to be a bench supply, used for protoboarding, etc. I don't want switching noises in that kind of supply, and it won't be running a lot of time.

 

Norfindel,

I like it. Of course, you most likely need some sort of feedback for stability. But the architecture looks generally sound.

 

Norfindel,

I like it. Of course, you most likely need some sort of feedback for stability. But the architecture looks generally sound.

Thank you. Can i expect problems with Vout directly connected to the inverting input of the OPAMP? I fear i don't know too much about frequency response with OPAMPs, to say something

 

Probably, that's the feedback stuff you'll need to add. Likely, if you powered this up, it would oscillate. Now you need to compensate it. You can mathematically go through this, or "play with it".

To play with it, you need to understand exactly what the problem is. Basically, the op amp output will probably move too fast. Let's say that the opamp thinks the output is too low. It will raise its output. The fastest it can do this is its slew rate. But there is a lot of "stuff" between that and the feedback. By the time the output responds, the op amp has gone too far. So it will drop it, and so on, chasing its tail, AKA oscillating. You could do something like put a resistor between Vout and the inverting input of U1. Then feedback a capacitor from the output of U1 to the inverting input. This is a standard integrator. This will effectively slow down U1's slew rate. The loop will balance the integrator.

Now here's how you do it. Go too big first. It'll respond like crap, but at least the oscillation won't happen. Then put some pulses into the input or pulse the load. Drop the integrator RC until it responds better, but doesn't ring. Ringing means you are getting closer to the oscillation point. But it also means you have a faster response. It's all about tradeoffs.

BTW, if you want to learn the math of what's going on, bode plots are the way to go.

Oh and another thing. Put an output capacitor on this thing.
One more, you may have to add some more to handle when you are current limiting. The loop will unbalance causing havoc. But, I would worry about that later. Just know it is coming.

 

Seems like i better wear safety googles, just in case...
Thank you for your help, i'm going to take a look to see what devices can i found here, and start doing some tests.

 

i dont mean to sound down, but all this for a homemade bench power supply? all the time and $ spent building it you could just buy a power supply to be used for powering other custom circuits you are building.....

 

Any circuit where you learn something is worth it.
Norfindel,
I've been thinking about it, and I think having the current limit circuit inside the loop is bad. You will be breaking the loop on the current limit, which will cause a significant overshoot when the load goes away.

I think the current sense should go inside the loop, but it should pull down the voltage input ot the opamp. Basically, you are taking the inner loop and putting it on the outside.

That make sense?

 

i dont mean to sound down, but all this for a homemade bench power supply? all the time and $ spent building it you could just buy a power supply to be used for powering other custom circuits you are building.....

Yes, a good point, but i am upgrading a power supply i done some years ago in school, so i have the costly things, and the power supply is one of the circuits i'm building . Besides that, i'm not in a hurry, i can wait until this finally works, or at least, explodes a few times

Norfindel,
I've been thinking about it, and I think having the current limit circuit inside the loop is bad. You will be breaking the loop on the current limit, which will cause a significant overshoot when the load goes away.

I think the current sense should go inside the loop, but it should pull down the voltage input ot the opamp. Basically, you are taking the inner loop and putting it on the outside.

That make sense?

Yep, i was thinking about that, also. At first, i thinked about using a transistor to pull the opamp input to 0v, but the Vce sat would be a problem. So, what about using the first Opamp to do the work, amplifying the difference between the LM317 output and the output of the current limiter?
I simulated the the attached circuit in ltspice, and seems to work. Probably ltspice is a bit optimistic, however, but i will give it a try on the protoboard.
Here, Vref is the output of the LM317, U1 is the Opamp i just talked about, U2 would amplify the voltage drop thru Rlim (and thus, measure the output current), and U3 is measuring this voltage against a reference. R7 and R8 would really be a potentiometer, and probably the 1N750 would finally be a low power (and cheap) voltage regulator IC.
U3 "looks" like a comparator, but really it's non-inverting input depends on it's output.
If this don't oscillates, i'm going to get drunk
By the way, i found this while researching, nice stuff: http://www.edn.com/archives/1998/030298/05df_05.pdf