Dummy Load II

[jocanon]

Building it for the long haul is a good idea. As much as I like you guys and like building dummy loads I don't really want to be doing it every 6 months .

 

[()blivion]

*THIS* looks like a good transistor, higher voltage too. Good for your higher voltage mode.

Here are the specs that matter...

Drain-Source Breakdown Voltage:.... 150 V........ (3X target)
Continuous Drain Current:.............. 50 A......... (10X target)
Maximum Operating Temperature:... +175 C...... (nothing should ever get this hot)
Package / Case:.......................... TO-3P....... (Fairly big)
Power Dissipation:........................ 250 W....... (realistic for the packaging it's in)

The TO-247 package is about the same size/power. That would mean *THIS* one.

Specs...

Drain-Source Breakdown Voltage:..... 100 V.......... (2X target)
Continuous Drain Current:............... 56 A............ (More than 10X target)
Maximum Operating Temperature:.... +175 C......... (super hot sauce)
Package / Case:........................... TO-247........ (same size. :-/)
Power Dissipation:......................... 200 W......... (seems close, a tad low if anything :shrug


To compare, here is a picture of the three different transistors next to each other, so you can get a feel for how much better/bigger we are talking here. The order is TO-220, TO-3P, TO-247.

 

[jocanon]

Those are big but they seem workable enough to me. What does everyone else think? Ron, could these easily replace the FETs in the schematic with a little tweaking?


I am thinking to get more pipe space I am going to just build the water cooled copper pipe in a flat square so I can solder FETs and the sense resistors all around each side if the square with the logic circuitry and PCB in the center. I'll just build a plywood box to encase the whole thing. I don't really care what it looks like, I just want it to work.

 

[Mr RB]

RthJC (thermal resistance junction-case) for the FET is 0.5 °C/W. RthCS (case-sink) isn't stated but it's not zero and is likely about the same 0.5 °C/W. So, at 125 watts, the junction is probably running 125° C above the 25° of the water pipe or close to 150°

Yep, that's why I would never run a TO-220 pack over about 50W dissipation. But we've gone round in circles now? I thought people had already decided it was ok to try to get about 1700W dissipation out of 8 FETs?

As one idea, you could submerge the entire FET assembly, running tap water is a pretty good insulator at 24v DC.

 

[ronv]

The plastic case is probably a good reflection of the junction temperature while the tab is the case temperature.

Place probes at nodes as close as practical to the die of interest.
This measurement location need not be in the primary thermal
path. For instance, the temperature at the top of the package or
at an exposed tab may be very close to the die temperature even
though the measurement location is not in the primary
thermal path.

I thought people had already decided it was ok to try to get about 1700W dissipation out of 8 FETs?

No, the initial design was for 1200 watts with 10 FETs. - 120 watts each.
So here we are talking 15 FETs at 60 amps - 1440 watts - 96 watts each.
So basically we are expecting 100 watts from a 300 watt part.

I think during testing Jeremy ran it up to the 70C pipe temperature trip point.

I don't know what the thermal resistance is case to pipe Probably not much since they are soldered to the pipe.

https://www.electro-tech-online.com/custompdfs/2013/02/CD00003180-81350.pdf

 

[jocanon]

Ron, so are you saying you think those are fine?

 

[ronv]

Yep. I see nothing in the specs that we are close to.

 

[()blivion]

I think the reason we are all divided on the issue of what is safe power and what isn't is because the dang overkill heat sinking is throwing every thing off. It's hard to imagine that the chip could be very hot when it is mere millimeters away from ice cold water. So, the real heart of the issue all comes down to...

"Given an infinitely good heatsink with infinitely good coupling to the tab, at what wattage will the thermal impedance of the chip start to create a thermal gradient that exceeds T[SUB]j[/SUB] max?"

In theory, this should have been given to us on a silver platter as the parts rated wattage, as that's the exactly the main variable we are talking about in the above. But it could also be expressed as the SOA, because SOA should include the wattage and anything else. AND, it can also be expressed as the package max power, because the thermal transfer area is going to be about the same size for all TO-220 FETs no matter what the exact part is.

What we would really need to know to answer the question once and for all is... What is the thermal impedance from junction to tab? And what is the die area? I think the most important part would be the die thermal impedance. Seems to me that would be the highest number. And it is probably approximately the thermal impedance of silicon X the thickness.

 

[jocanon]

Just throwing it out there, but does it make any difference that we are using the MOSFET linearly versus switching?

 

[()blivion]

Using the MOSFET linearly makes all the difference. It'a apples and oranges.

If we were to PWM into a load, then the only part specs that we would really have to consider would be max volts and amps. This is because when doing PWM, the MOSFET is typically only ever fully OFF, or fully ON. When fully OFF, it has such a high resistance that no power is able to move through it, so no power is being lost. When fully ON, it has a very low resistance, so even though lots of power may be going through it, that power is being dissipated in other parts.

By driving the FETs in the linear however, we are intentionally making the parts resistance such that the total power going through the system is almost all lost directly in the FET. This of course makes all the heat lost in them, which is why we need them to be heatsinked extremely well. Otherwise it would be some other part that was getting hot and needing cooling.



The vast majority of projects involving power MOSFETs are using them for switching. There are some very valuable uses for them in power audio amplification that use them in the linear, but these applications are the minority. The parts we have chosen were intended for switching application. Luckily, (as far as I know), the only difference between FETs for switching and FETs for amplification is that FETs for amplification are usually more liner over their range, where as switching FETs are usually faster and have a larger min and max resistance. In the end it is perfectly safe to use the switching type FETs for linear mode as we are not interested in accurately reproducing some analog signal without distortion. We are just trying to make a variable resistance that lets us balance the power going through it.

 

[ronv]

In theory, this should have been given to us on a silver platter as the parts rated wattage, as that's the exactly the main variable we are talking about in the above. But it could also be expressed as the SOA, because SOA should include the wattage and anything else. AND, it can also be expressed as the package max power, because the thermal transfer area is going to be about the same size for all TO-220 FETs no matter what the exact part is.

What we would really need to know to answer the question once and for all is... What is the thermal impedance from junction to tab? And what is the die area? I think the most important part would be the die thermal impedance. Seems to me that would be the highest number. And it is probably approximately the thermal impedance of silicon X the thickness.

I agree. This is how I see it.

We do have the value for the die temperature in RthJC (Thermal resistance junction to case) of 0.5 degree C per watt. So in our case the tab is 25C so the die at 5 amps at 25 volts across the FET (120 watts) should be at about 85C - 60 from heating and 25 from ambient. This matches up pretty well with Jeremy's temperature reading of the plastic of 88C. So then we can go down to Fig 2 in the datasheet and look at the SOA and at 25 volts it shows we can have about 12 amps thru the FET. In reality it won't have that much margin but well below where we will be running it - 4amps.
That is my read on it. The magic is that the case temperature is at 25 C. which nobody gets with an air cooled heatsink in a 25C room.

 

[jocanon]

Should we increase the temperature shut off point and attach the temp probe to the plastic case for a quicker response, or leave it on the copper and leave it lower? I am fine with either, but putting on the plastic case would require some re-work of the schematic.

 

[()blivion]

Well, in a perfect world, we would put one thermal sensor on each FET. We would want each sensor to only respond to that one FET, so we would not want to put it on the pipe where all the heat is shared. Reversing this logic, given just one heat sensor, what we would want would be to put the thermal sensor somewhere that it actually did get all the combined heat of the FETs.

Given just this logic, I would think the best solution would be to install the sensor on the pipe, toward the exhaust end, and have it set to go off at a quite low temp. Just a tad above the water resting temp.

However, there is the matter of how well the copper pipe conducts heat vs how well the moving water steals and redistributes it. This sets up all kinds of crazy thermal gradients at different points in the pipe, with all sorts of different temp curves. A few mm could possibly change the temp a lot, as could putting it at the end rather than in the center between two FETS, then again, maybe not. Finally, there is the possibility of the abnormal situation where you may accidentally forget to turn on the water, then we no longer see heat being redistributed from one point to another in the pipe thanks to the water. All these things make the matter surprisingly complicated. How fast is the thermal sensor going to heat up when on the pipe vs when on the FET back? Which way is going to be faster to respond? Which way gets a better reading of the actual junction temp?

With all this new logic, the best place would seem to be the hottest place you can find. Or the place that gets hot the quickest. Which, given the gathered data, seems to be the FET backs for both cases.

In the end, I think it all depends on what kind of overheat scenario we are trying to protect the system from. If we want to catch the event that the cooling system has failed, then the second option is probably best. If we are after catching an overload with a 100% functional cooling system, the first option is better.

Finally, we could always get two sensors and do both, hedging our bets to some degree.

 

[ronv]

Maybe a compromise. How about in the space between the mounting tab and the plastic body. Can you measure the temperature there? Then we can adjust the trip point.

 

[jocanon]

Do you mean the top of the FET like where the plastic meets the metal tab? Second question, what is the best way that you recommend I measure the temperature? I have been using an infrared thermometer but it is hard to pin poit that exact of a location, even though it has a red laser, the laser is not in the right spot.

 

[ronv]

Yes, that might be a good spot. My voltmeter has a temperature probe on it. Wish I had an infared. We may not be that far off with 50C at that point. We can always tweek it.

 

[jocanon]

Hm, I need a new voltmeter anyway. What one do you have that reads temperatures?

 

[()blivion]

I have been using an infrared thermometer but it is hard to pin point that exact of a location, even though it has a red laser, the laser is not in the right spot.

I hate those thermometers. Not only does it always seem like that same problem effects every one of them, but even worse, they are not very accurate to begin with. It goes without saying, they work by capturing the light that is emitted by an object because of how hot it is. That's great for quick tests, tests from a distance, or when physical contact is a practical problem. The crux with this method though, is that not everything gives off the same level of IR at the same temperature. So something that is really hot, but is not very emissive, may read several degrees colder than it actually is. This could possibly be why you are reading much hotter temps off the back of the FETs than off the pipe. Black bodies emit more light.

 

[ronv]

I think it is this one that I have.

**broken link removed**

 

[jocanon]

...This could possibly be why you are reading much hotter temps off the back of the FETs than off the pipe. Black bodies emit more light.

Yeah, not sure how accurate the temperature readings are, I think they are in the ball park. As far a the copper pipe, I am getting those readings from the temperature probe that' glued to the pipe right next to the FET. Also, when I touch the pipe or the metal tab of the FET I can tell by feel that it is very cool.