PWM based DC electronic load

[krish2487]

Hello..

I need a DC electronic load for a specific load test jig and i wanted to adapt/adopt daves electronic load for use..
I did breadboarded the concept successfully but i noticed that the FET and the power resistors were getting too hot for my comfort.

It then struck me would the concept work with a comparator instead of an opamp..

i.e use a PWM signal to drive the FET.

I just wanted to sound whether this is a "workable" idea or am i just talking out of my hat.

I can see several advantages over using the mosfet in linear mode.

-> The fets run cooler
-> the Fets can be a garden variety type
(otherwise the DC D-S current needs to be high for the FET)
-> Thermal considerations like heatsink size and ventilation.

I shall be testing out this concept out later in the day.
(It is morning here in India and i am at work... :-) )
i just wanted to sound the concept and see if there is something that i being obviously blind to..

Thanks...

[BravoV]

Your idea is flaw and it wont work.

As long the fet or the resistor is taking the burden in dissipating the power, they will eventually heat up, no matter how you use it.

[krish2487]

Agreed...

It is precisely why i posted it online..
:-D

But it is definitely doable...

In amplifiers parlance
I would like to make a class D equivalent of a class A amplifier.
:-P

In retrospect the comparator was not probably the way to go about... but it should , be possible to come up with something that drives the FET in a PWM mode rather linear mode....

Regarding the FET and burden resistor dissipating the power.

Thats how class D / switchmode functions isnt it.. either in amplifiers, or UPS systems , motor drives, etc..
the FETS definitely run cooler than if they were run in linear mode..

agreed. i seem to be talking rubbish but it must be possible to come up with a switching electronic load.

[amspire]

The thing is, the heat has to be dissipated somewhere either in MOSFETs or resistors. If you have 100W of power to consume, something will be heating up with that 100W.

In many cases using  MOSFETs to dissipate heat can be cheaper then power resistors, so it may be cheaper to have a bank of MOSFETs on a big heatsink compared to PWM MOSFETs switching fixed power resistors along with the big capacitors and inductors that you will need to smooth the current.

How much does a IRFP460 cost? How much does a 150W - 200W power resistor cost? Both need to be bolted to a heatsink anyway.

MOSFETs have the big advantage over resistors in that they are variable devices, so they can dissipate anything from microamps to many amps easily.

Some of the CPU coolers are pretty amazing at dissipating heat, so multiple MOSFETs mounted on a series of good CPU coolers can make a pretty powerful electronic load.

Richard.

[BravoV]

Nope, still won't work, read here -> http://en.wikipedia.org/wiki/Conservation_of_energy

Apparently you've been misunderstanding the switching in the psu, I know what are you thinking with the correlation on power supply linear vs switching.

Linear psu generates more heat, switching type is more efficient generates much less heat, hence, if you somehow turn the mechanism inside the electronic load into switching method, then less heat. Again, this is just not right.

That "box" of electronic load, when we pumped the energy into it and if the energy is not going somewhere, that box will suck transform the energy into other energy form which in this case if you're using the fet or the resistor, this device only capable of transforming the electricity energy into heat.

No matter what magic you've put inside that "box", as long its receiving the power and only capable of dissipating the energy into heat, so it will heat up. You just can't reduce the heat or worst make it gone. Unless you managed to make that box able to transform the energy to other form rather than heat, like a light, then that box probably will have some light bulbs.

Again, as long the energy dissipating device is only fet or resistor, they only can transform into heat only, so if it consumed 100 watt of electricity energy, then it will release it as 100 watt of heat.

[krish2487]

Thanks for the information..
My understanding was indeed flawed..
(to a certain extent)

@armspire
So the resistors will run hot to the same temperature irrespective of the type of switching..
But fets are a different issue.


@bravoV
Shouldnt atleast the fets run cooler if you were to switch in the kilohertz range???
As far as the Fets are concerned the "load" in the drain and source is immaterial.
(correct me if i am again going wrong)

It is nevertheless a current flow from drain through source..
It is still a switching device and instead of applying a DC to the gate to partially switch on the gate, i am switching the fet between cut-off and saturation to average the current to the  desired level.

The resistors no doubt will get got by the same temperature difference as in linear op-amp mode.
But i have a hunch the fets might run cooler.
OK maybe using a direct buffered PWM output for the fet gate drive..

I dont mean to be obstinate but i just need to check it out for myself practically..
(Lets admit it, thats why we all are here arent we... :-P)

[BravoV]

Krish, its fine, everybody must learn something someday.

The basic flaw of your thinking is just these :

- Power supply
A device that transforms electricity from into different level/potential either lower or higher from the original power source.
Conversion method can be using linear method (not efficient and generates lots of heat) or switching type (efficient with less heat generated) in transforming the electricity.

- Electronic dummy load
A device that loads/sucks  electricity from a power source and the electricity consumed is not mean to be used again or to be wasted as most cases, the purpose is just to verify if the power source capable of sourcing up to certain level of energy.
No matter how you put the "magic method" in this box, in your case by switching the fet on and off even in gazillion hertz,   still its dissipating the same energy and again will generate the same heat level, you just can NOT reduce the heat generated.

See the basic fundamental different ?

[amspire]

@armspire
So the resistors will run hot to the same temperature irrespective of the type of switching..
But fets are a different issue.

If a MOSFET with a RDS of 10 milliohms switches a 10 ohm resistor on and off, then the resistor will get hot and the MOSFET will stay cool. But why would you want to do this? The MOSFET is cheap. The resistor is expensive. Why are you reducing the power in the cheap part so you can dissipate it in the expensive part?  If you use PWM, you need other parts as well - you need big electrolytic capacitors to average the load voltage, and probably inductors to smooth the current.

Isn't it cheaper, easier and better to use the MOSFETS to dissipate the heat in a linear electronic load circuit? A linear load using MOSFETs to dissipate the power is a superior solution to any switching PWM load circuit. The performance will be better, no big capacitors and inductors to cause a host of secondary problems, the response will be faster.

As other have explained, switching does not magically make power vanish, and to me the best place for the power to go is to the MOSFETs. I think you are imagining that somehow using PWM makes the power disappear somehow. It doesn't.

If things are getting to hot, then it is just that your heatsinks are nowhere good enough for the job, and whatever you do, you will have to get much better heatsinks, like the PC CPU coolers I mentioned. The good ones are considerably better then a 1kG slab of aluminum heatsink extrusion.

No circuit exists that will allow you to use a 10W capacity heatsink in a 500W electronic load. If you need to get rid of, say, 500W, you will need heatsinks/fan combinations that can get rid 500W without getting too hot. If the heatsink is too hot to touch, it is probably not suitable and you will need a much better one.

Richard.

[BravoV]

As other have explained, switching does not magically make power vanish, and to me the best place for the power to go is to the MOSFETs. I think you are imagining that somehow using PWM makes the power disappear somehow. It doesn't.

I believe the OP is confused and can't differentiate the basic concept between the "power supply" (transforming electricity) vs "electronic dummy load" (consuming electricity), I'm guessing thats why he has trouble getting it.

[krish2487]

Yep....

Burnt my hands..
Literally..

I come back to post my findings...
It worked with a PWM signal..
but the performance was the same..

The resistors did get hot...
but the fets got so hot i burnt my hands when i tried to see if there was any temperature increase...

BTW its a small heatsink for TO220

heres the link to it..
http://in.element14.com/aavid-thermalloy/581002b02500g/heat-sink-to-220-218-17-4-c-w/dp/1213467

I was loading a 3.3V PSU with about 400 mA.
about 1.4 watts.

I get the posts about trading dissipation between the resistors and FETS now.

Thank you for your help bravoV and armspire..

i ll need to get a really BIG heatsink now...
This was only a test run..
I need to test the same by about a couple of amperes now...

Thank you sincerely for your efforts to help me understand thermal dynamics better..
:-)

[amspire]

The resistors did get hot...
but the fets got so hot i burnt my hands when i tried to see if there was any temperature increase...

BTW its a small heatsink for TO220

heres the link to it..
http://in.element14.com/aavid-thermalloy/581002b02500g/heat-sink-to-220-218-17-4-c-w/dp/1213467

That heatsink is 17 degC/W. if the room temperature is 40 degrees, and a heatsink can burn at 70 degrees, then that heatsink is good for up to about 2Watts of power.

A cooler from an old computer like ebay item 130685997762 (an AMD athlon/sempron cooler) should be good for at least 50 watts, and people are throwing stuff like this out.  Better modern coolers can be good for hundreds of watts.

Richard.

[nctnico]

There is an interesting advantage of using PWM in a dummy load. Most cheap Mosfets can't handle much DC current (see SOA graph). If the Mosfet is pulsed it can handle much more power. If you use multiple Mosfets and switch them in a round-robin fashion you should be able to push more current through them than is possible with pure DC. The disadvantage is that this setup requires a filter on the input which probably makes it harder for a control loop to work.

[ejeffrey]

Another way to get a much higher power rating is to dunk the heatsink, mosfet attached, into a bucket of distilled water or oil...

To echo what everyone else has said, PWM only helps you move the power dissipation from the MOSFET to the resistor, and requires some high current inductors and capacitors to boot.  This is not going to be a win if you are using standard power resistor as your load.  However, you could use something like an incandescent lamp or an electrical heating element as your load.  If you need hundreds or thousands of watts load and can tolerate a fair bit of residual ripple current from the PWM, this may be a win.  For 10s of watts or less, one or more MOSFETs is going to be the cheapest and easiest way to dissipate the power.

[nctnico]

Another way to get a much higher power rating is to dunk the heatsink, mosfet attached, into a bucket of distilled water or oil...

That won't defeat the SOA limits.

[amspire]

Another way to get a much higher power rating is to dunk the heatsink, mosfet attached, into a bucket of distilled water or oil...

That won't defeat the SOA limits.

Electronic loads are often used in constant resistance or constant current mode. In both these modes, there are no great surges, and there shouldn't be any SOA, as long as you have chosen devices and cooling that is right for the job. To use multiple parallel MOSFETs without rediculously big source current sense resistors, you can just have a dedicated opamp current regulator for each MOSFET, rather then trying to switch from m MOSFET to MOSFET in a round robin fashion.

Not sure I see any difficult SOA problems at all, unless you are talking about an environment with big inductive voltage spikes possible.

Richard.

[nctnico]

The problem is that at higher voltages the current handling of Mosfets and transistors gets really low. A transistor's die doesn't heat in an uniform way. Due to impurities you get hot spots. This is what the SOA graph is about. If the current is pulsed the hot spots have less time to form.

[ejeffrey]

Another way to get a much higher power rating is to dunk the heatsink, mosfet attached, into a bucket of distilled water or oil...

That won't defeat the SOA limits.

I should have said it is an easy way to reach (near) the rated power of the device.  Obviously a device with poor power handling can't be improved.

The problem is that at higher voltages the current handling of Mosfets and transistors gets really low. A transistor's die doesn't heat in an uniform way. Due to impurities you get hot spots. This is what the SOA graph is about. If the current is pulsed the hot spots have less time to form.

That is mostly a problem of bipolar transistors, not MOSFETs. MOSFETs (at least power MOSFETS) are mostly only limited by temperature, breakdown voltage, and fusing current.  Second breakdown is a problem of bipolar transistor because their transconductance increases with temperature.

The SOA graph in most MOSFET data sheets is for pulsed applications with low duty cycle, and is essentially a measurement of thermal inertia.  You cannot increase the total power dissipation of a MOSFET by pulsing it.

[BravoV]

Regarding mosfet in linear mode, I posted a thread discussing about it while ago here -> https://www.eevblog.com/forum/beginners/question-on-mosfet-in-linear-mode-does-'hot-spotting'-a-serious-problem/msg99612/#msg99612

According Achmed99 that was experiencing problem using plain mosfet, and he substituted with more expensive special made "linear" mosfet, though he hasn't update the progress yet.

PS : Don't mean to hijack this thread, if you some thought or experience, share it there, please.

[krish2487]

@nctnico

If you look at the original post that was one of my primary points.
DC current through the drain.
:-)

I did manage to pump the current thru a much higher level.
but i had to trade the heat in the resistors instead..

which kinda defeats the original purpose anyway..

It is possible as others have said to put a small inductor the drain-source leg to smoothen the current but then the magnetics too will need to rated for the load current.

I was actually thinking of stealing several ideas mentioned from several above posters..

1. aluminium housed power resistor 5% tolerance (25 or 50W).
2. run of the mill mosfets with individual buffers.
3. 25/12 Khz PWM with variable duty cycle to control current.
4. all housed on a massive heatsink with a 12V DC fan (120mm) to cool.

:-D

It wont have any significant improvement over a linear mode current load except the fets life, can be increased.
(not critical.. just a fancy requirement...)
and using an intelligent control for the PWM i can implement pulse/constant power loads which "might" unduly burden the FET.

PS: for those calculating the cost of the above project..
I work for a Power electronics industry manufacturing motor drives and UPS systems upto 45 KW 3 Phase.

So most the above mentioned stuff, i can get for nearly next to nothing...

:-D

[krish2487]

@achmed99
(any inspiration from jeff dunham??? :-D)

none taken.
It is by looking at others that we learn..
Nice work btw on your APFC.
All the best with it!!!

[nctnico]

That is mostly a problem of bipolar transistors, not MOSFETs. MOSFETs (at least power MOSFETS) are mostly only limited by temperature, breakdown voltage, and fusing current.  Second breakdown is a problem of bipolar transistor because their transconductance increases with temperature.

Actually a Mosfet has a negative tempco in the linear region (like a transistor). So a hot spot will cause thermal runaway. You'd need to run the numbers on the thermal dynamics to see whether you can dissipate more or less power in a Mosfet when pulsing it in its linear region.

[amspire]

What you are saying is that for making a linear 1 A active load, it is much better to use something like a IRF840 with a .85 ohm resistance or a IRF830 with 1.5 ohms resistance and 8A maximum current instead of something like an IRF540 with 20 milliohm resistance and 36A maximum current. Thanks for the application note. It makes sense.

[amspire]

I have seen lots of switching MOSFET circuits where there is a weak drive circuit for the gate - often a low current output, of just a pull down resistor to turn the MOSFET off.

In these cases, the Miller effect can turn the MOSFET back as the drain voltage rises as it is turned off. You end up with the MOSFET still conducting as the current rises all the way up to maximum. Even worse, if high speed standard diodes are used instead of schottky diodes, then there can be a considerable current spike as the MOSFET turns off due to the stored charge in the diode. It is easy not to know it is happening, other then the MOSFET is a little warmer then expected.

What AcHmed99 is saying is that a bad gate drive in a switching circuit can easily stress a spot on the MOSFET and ruining reliablility. I had always know a good drive was important for efficiency, but I hadn't worried about reliability as long as temperatures seemed OK.

Very valuable information.

Richard.

[NiHaoMike]

How much does a IRFP460 cost? How much does a 150W - 200W power resistor cost? Both need to be bolted to a heatsink anyway.

A cheap space heater can dissipate 1500W continuously and costs $15 or so. The heating element can be used as-is or rewired for lower voltage.

[nctnico]

The whole point of an electronic load is that is can be regulated for constant current, constant voltage, constant resistance or constant power. Most commercial loads also allow to create load transients. Just having a big resistor as a load doesn't help. You'll need many different resistors to get the same flexibility.