Driving Mosfet On His Analog Region, Using mosfet as variable resistor

[Blue_Key]

Hello there,

I'm quite new here and it seems to be a great community.


I'm far to be knowledgeable when it gets to analog circuit so I'd like to post my design in order to get your advice whether this is doable or any improvement.

My goal is to have a power (250W) variable resistor. Note that it needs to be driven on that state only few tens of milli sec.

I know the MOSFET is not linear on that region and I count on handling that with the micro controller. I know that JFET can be used for this but the power here is too high.

I control it with a micro controller running a 3.3V using a PWM signal. I have a driving stage to fit the voltage range around the analog region of the MOSFET.

I use the drive stage to offset the voltage on the analog region of the mosfet in order to improve the definition of my PWM. PWM is filtered by an RC at the gate of the mosfet. A supply of 15V and few zener are there to drive the most between 2.7V to 6.2V

This circuit work on ISIS but simulation is not always showing all errors.

Thanks for your tips.

[Sch3mat1c]

1. Should the load, in fact, be resistive, or is a constant current source sufficient?

2. How much ripple or noise can you tolerate? Should it be absolutely steady or can it be chopping on and off without a problem?

I don't know if you intend to use IRFP4568, or if it's just representative. It looks like a very poor choice for this type of application. At a glance, it's much too large for linear purposes: the channel is too large (expensive, high capacitance, slow, hard to control). Looking in more depth, we see:
- Vgs(th) is 3V minimum, so a 2.7V minimum drive would work, but only as long as the device remains at exactly 25C. Vgs(th) has a strong negative tempco. Referring to the datasheet,
http://www.irf.com/product-info/datasheets...irfp4568pbf.pdf
Fig.3 shows *typical* transfer curves. The 25C curve should correspond with the typical Vgs(th) data, which is not given, but one can assume it's somewhere around 4V. If one extrapolates the curve down to 1mA, this is about right. This suggests that the curves should at least be shifted 1.0V to the left, assuming that variation in Vgs(th) is independent of other effects (like Rds(on) or gfs). The 175C "min" curve, then, crosses 2.7V at around 0.6A or so, rough guess. This guess could be off by a factor of two, easily.
- Although your stated goal is a 250W resistor, and this transistor is "rated" at 517W, one must keep in mind, maximum power ratings are completely unrepresentative of actual performance. A TO-247 package will dissipate 100W with a water-cooled heatsink, 50W with an air-cooled heatsink. That's it. If you must get 500W out of the thing, you have to reproduce the conditions which International Rectifier used to measure it: the entire device, leads and all, is immersed in a bath of boiling freon at 25C.
- Referring to Fig.8, the SOA delivers the final blow. MOSFETs are commonly touted as being free of second breakdown, the symptom which has long plagued bipolar transistors in higher voltage, linear applications. Unfortunately, modern devices, optimized for switching, are just as sensitive, if not moreso. The "DC" curve, up around 100A and under 5V, demonstrates the continuous power rating of 500W. At lower currents (out of saturation), current actually does not increase at all until the power is down to about 150W, and even then, it increases slowly. Down at 10A, it's about 7V (70W); 1A, 15V (15W!), etc. Even under pulsed conditions, this does not improve much: the 1ms curve shows 10A at 30V (300W) and such. So even if you want to pulse it very quickly (which will be challenging with R13-C4 as shown), you'll have trouble meeting your desired dissipation. Note that junction temperature increases during a pulse, so Vgs(th) is dropping all the while, making control difficult if not impossible.

Why microcontroller instead of op-amp?

Tim

[Blue_Key]

Hello Sch3mat1c,

Thanks for your comment.

Maybe I need to add some precision. Actually the maximum current going through the mosfet will be 10A and reduce as the voltage increase.

I've considered the Vth drift over temperature, that's why the control voltage range is wider than a usual 25°C control.

Also the mosfet will be driven 10-50ms every once in a while (like every 5 minutes), so it should be able to stand it.

I could not tolerate noise on this system and has to be steady during few ms.

The micro controller control the mosfet state from open circuit to C-C in few ms.

Your last point is interesting. In my application it seems to stay in the safe area, but still quite close. Do you have any suggestion about it ? (other components, other design ?)

thanks

[Sch3mat1c]

Nah, not nearly fast enough. You'll have better luck with something like this,

To dissipate 250W you'll need to use 5 x TO-247 devices and a big heatsink. This circuit will control the transistors within microseconds, keeping everything much more stable. Note: unless you have transistors that are exactly matched, and that stay matched in use, you must have one op-amp per transistor.

If your load is 10A and 25V, up to I don't know what, you'll need at least 2A rated transistors; anything rated up to, say, 10A would be plenty. Unfortunately, these are hard to find in a high-power package (I see transistors in TO-247 starting at 100V and 25A and going up from there..). BJTs might be more suitable; darlingtons are acceptable at this voltage drop. TIP142 might be good, and quite a bit cheaper than IRFP4568.

The crude-but-effective method is a "power DAC", using resistors with values in powers of two, switched by MOSFET or relay. The glitches may not be suitable for your purposes, but the reliability is unbeatable. Glitches, as in any DAC, can be minimized using suitable codes, for instance gray coding or thermometer codes.

Tim

[Blue_Key]

Good input thanks.

Although I can't have a R in serie with the mosfet on my circuit as the mini Rs has to be very low. That's also a reason of this mos, having a rs of 5mohm.

I've modified the circuit to have a control via a feed-back resistor. (just basic working principle)

But as there is feed-back, I'm a bit worry about oscillation on the system.


A power DAC will probably be more expensive than this, as voltage has to withstand 150V and 10A, mosfet gets expensive

This post has been edited by Blue_Key on January 28, 2013 01:50 pm

[Sch3mat1c]

150V 10A at the same time, or only the maximum of each? That would've been nice to know at the start.

The circuit shown is a voltage shunt regulator, plus R3.

If your maximum current is 10A, and maximum power 250W, then voltage does not drop below 25V at full load, and at voltages where series resistances matter (under 1V, perhaps), current will be so low that those resistances won't drop any additional voltage. If it must remain resistive down to very low voltages, darlington transistors won't work so well, but BJTs still aren't ruled out.

What source does this load work against, what application?

Tim

[Blue_Key]

Thanks for your reply.

Only one at a time.

R3 is out of the circuit, it simulate a big serial resistor.

Actually it's only one at a time.

My serial total resistance has to be below 0.1ohm, that's a need.