100WV 50A max 500W Active Load

[CamLight]

Nice Rick!
I'll dive into the schematics soon (still recovering after the fire...sheesh) but wanted to address your MOSFET questions.

You won't be able to use a single TO-220 case MOSFET for dissipating 250W.
Even assuming an infinite theoretically perfect heatsink, you're limited to about 87W.

Here are the numbers:
FDA59N30 junction-to-sink thermal resistance (theta-js) = 1.27 degrees-C/W
Max ambient = 30C.
Max allowable temp rise = 110C (for 140C max junction temp, for reliability).
Allowable power = 110/1.27 = 86.6W before you reach max junction temperature of 140C

If you factor in the thermal resistance for a REALLY good heat sink, let's say 0.15 degrees-C/W, you get a theta-ja = 1.27 + 0.15 = 1.42 and a max power limit of...
110 / 1.42 = 77.5W before you reach a junction temperature of 140C

Even of we made the design pretty unreliable, with a max allowed junction temperature of 175C, the max power level is...
140 / 1.42 = 98.6W before you reach 175C

You'll need more MOSFETs to lower the thermal resistance.
Let's assume four MOSFETs are being used and calculate the required heat sink thermal resistance needed to keep the junction temperature below 140C at a power level of 250W...
Theta-js for four MOSFETs = 1.27 / 4 = 0.318 degrees-C/W
Mac ambient = 30C
Allowable junction temp rise = 110C
Theta-ja = max junction-to-ambient thermal resistance

250 = 110 / theta-ja
theta-ja = 0.44

This means that the maximum allowable combined thermal resistance for the MOSFETs and heat sink = 0.44 degrees-C/W. Since the four MOSFETs theta-j-s = 0.318, then the heat sink's theta-sa cannot exceed 0.44 - 0.318 = 0.122 degrees-C/w. That's a darn good heat sink, but doable!

I'd start with four FDA59N30 MOSFETs on each heat sink and go from there. You can use a thermocouple on the rear metal plate (thru the heat sink) of one of the hottest MOSFET to see how well the heat sink is working. Be sure to wait until the heat sink has reached thermal equilibrium first, several minutes. This will tell you if you need to add additional heat sinks, a more efficient heat sink, or perhaps higher fan speed or different fan mounting.

Power limiting...a tough one.
There are analog multiplier chips out there that can take two voltages (in our case, pack voltage and shunt voltage) and multiply them to give you a power level representation, but you often need a lot of circuitry to put it all together (getting right voltages for multiplying, etc.). It gets complicated very, very quickly. I recommend using a microprocessor to read a stepped down pack-voltage and the shunt voltage to calculate the power level. Even a small 8-pin PIC like the 12F683 will work here. It has four 10-bit ADCs, a comparator, and can off the available +5V.

Using the temperature measurements mentioned above, you'll be able to sample the power level several hundred times a second and keep the power level low enough to prevent overheating. Lots of really good temp measurements will allow you to have a much higher starting power level, when the heat sink is cool. This is what I did for my CC-400 load. It handles up to 500W for 15 seconds and then 400W continuously after that. If those limits are exceeded, I turn off the MOSFETs and blink a red LED.

But, any power limiter will require a method of setting it for the different heat sink, fan, and MOSFET combinations other users will have. You could set up a block of header pins with a jumper to tell the PIC the max power level it should allow (perhaps in 25W increments?). But, this would require the user to measure their MOSFET temperature for the setup they have and that's very, very problematic. I don't think that many people are going to drill a hole in their heat sink and get a thermocouple to measure the temp.

I was thinking that in order for the design to be the most useful to others, it will need to be tested with a commercially available heat sink that others can buy. That way no temp measurements need to be taken and they will know the power level allowed by that design. If someone feels they need more power, they can build more units or find a better heat sink.

 

[rkosiorek]

how much will it change if they are TO3PN? max dissipation for the case i mean.

rick

 

[CamLight]

Rick, do you have a particular device in mind? The junction-to-sink thermal resistance is the critical item here.

[Edit] I forgot to answer part of your earlier questions. The SOA curve for the power limiting circuit can be developed from the MOSFET data sheet's curves. Both the current/voltage curve from the SOA graph (DC plot line) and the temperature-derived power curve need to be taken into account.

 

[rkosiorek]

john,

CPU manufacturers specify the number of watts the heatsink must be able to radiate. most of the new socket-T-P4 heatsinks are good for 100 to 130W and the older socket-N-Prescott CPU needs one capable of 190 to 210W. the old Prescott CPU's were comonly refered to as the PressHot or SpaceHeater. Max core temperature of the CPU for most desktop CPUs is 70C or 45C over ambient. so in that case we are looking at less than 0.5C/W. possibly as low as 0.3C/W.

but you are correct. some kind of standard heatsink should be chosen. i was thinking of the chinese clones of the stock Intel socket-T heatsinks. they are inexpensive and readily available.

View attachment 2

the contact patch on these is large enough for a pair of TO-3PN transistors side by side. I was considering an aluminum bar with a couple of screws to clamp the transistors down.

the specs i have are for the Fairchild FDA59N30 in a TO-3PN case. the specs do not explicitly list the Theta-js. they do list the (theta-jc) = 0.25 C/W MAX and the (theta-cs) = 0.24C/W MIN. that works out to a (theta-js) of 0.25 + 0.24 = 0.49 C/W. or 0.25C/W for a pair of them.

for a max allowable temp rise of 110C that works out to 110/0.49 = 225W which would be considerably higher than the TO220 case. a pair mounted to a cheap heatsink would be 0.25 + 0.5 = 0.75C/W. so the max power would be 110/0.75 = 150W using an absolute bottom end heatsink designed to barely meet the minimum spec. i think that most of these heatsinks will perform much better than that.

being new at this "Teach" i think i understood your math example correctly. attached is a pdf of the Fairchild spec sheet. also i attached a pdf of the schematic.

what do you recommend for the temperature sensor? NTC thermistor? some kind of chip? LM35?

thanks,
rick

 

[CamLight]

Rick,
I agree and good idea on those heat sinks. I've seen several CPU coolers rated at 0.5C/W and a bit lower. That is for a very large contact patch though (size of the CPU). You'll have to derate the thermal resistance rating by a good 20% (at least). But, that can be compensated for by using a heat sink/fan combo that uses a common fan size that can be replaced with a higher CFM/higher static pressure model. The CPU fans are CFM-rated very low so they're quiet. We don't care about that.

The aluminum bar clamp can work but have a screw in between the FETs too. Otherwise, the bar will bow up in the middle and tilt the FETs. Even rising up 1 mil will significantly degrade the contact they have with the heat sink.

Aha, the TO-3P case version...oops. MUCH better than the TO-220 cased version FDAF59N30 I was mistakenly looking at.

Math looks good (twin FETs = 0.25C/W theta-js, assuming no insulator and good thermal compound), but IMHO too optimistic for the heat sink rating. Keep pushing the power expectations down, not up. It's the only way that this thing will stay reliable as there are so many other variables that will increase the FET temperature (incorrect grease application, incorrect mounting screw torque, thermal cycling loosening of the screws, higher ambient temperature, dust in the heat sink fins, fan slowing down as it ages, etc.).

Assuming a stock heat sink rated at 0.5C/W, I'd derate it as follows:
- Smaller patch size, raise sink to 0.6C/W.
- 110 / (0.25 + 0.6) = 130W for 140C junction temp.
- 145 / (0.25 + 0.6) = 171W for 175W junction temp, absolute max power (this unit WILL eventually burn out).

There might be some better thermally-rated sink, but be sure to derate it due to the small contact area of the FETs vs. CPU. Or, add a much more powerful fan and use the sink's original theta-sa (sink-to-ambient rating).

IMHO, don't have a thermal sensor for the unit. It depends VERY heavily on the skill of the user to mount it and that's critical. Perhaps drilling a hole in the heat sink between the FETs and epoxy the sensor in there but the thermal lag won't allow the sensor to protect the FETs unless the overload is small. Otherwise the FETs will burn out long before the heat sink gets too hot. Seen it happen wayyyyyyy too many times during my load development. The sensor doesn't work very well in a certain commercially available electronic-load/analyzer product either, IMHO. Ended up not using any thermal sensing at all in my loads, just the curves/settings in the PIC firmware.

If you did use a sensor, I'd recommend a thermistor in the sink itself. But, these have tolerances (10%?) and unless a calibration step is added to the load construction, you'll have to derate the FET/heat sink assembly because of this tolerance. The LM35, and related chips, are quite accurate but you can't get them close enough to the FETs to get a good reading. Very slow too as the heat will have to go from the junction, to the sink, through the LM35 case. Perhaps more reasons to go without a sensor?

[Edit] SAFETY WARNING: The MOSFETs will not have an insulator between them and the heat sink so the heat sink may very well be at the pack positive voltage level when the pack is connected. If the sinks are anodized (most likely) that MIGHT insulate the FETs, but you can't count on that. Be sure to include incredibly clear instructions and warnings about this....over and over and over again. Someone will eventually zap themselves or damage some equipment and you want their lawyer to immediately say that you're not worth suing because your warnings were so clear. I'd normally put a couple of smilies here but this kind of stuff will happen if...excuse me, when...someone goofs up.

 

[rkosiorek]

John,

realised that i missed a very important fact in one of your previous posts. FIRE? what Fire? what happened? everyone okay?

rick

 

[CamLight]

Hi Rick,
Everything is OK, no big damage, but one heck of a mess to clean up. Not from the fire, from the darn powder extinguishers. Who knew those little things held so much (nasty, corrosive) powder!!

Well...OK, I did. but I never suspected I would have to damp sponge every surface in the entire place ten times!

Worn cable near a power strip cable clamp caused overheating or a spark and lit up the back panel of my (former) entertainment center. That's about when I came out of the next room wondering why the smoke alarm was screaming at me. Two extinguishers later, the fire is out and I'm standing there dumbfounded as a HUGE cloud of powder makes it way through the apartment.

<deep sigh>

Almost done cleaning up and ready to get back to Doc's CC-400 (and others too)!

 

[rkosiorek]

another bit designed - a temp controlled FAN.



this way the fan can be set to turn on when the heatsink gets warm. say about body temperature. can't wait to try it out on the load when i get that a bit more done.

rick

 

[CamLight]

Nice rick!

I was thinking...
Would it be worth it to have an external ENABLE input line to the circuit so that the PIC could turn the fan on/off based on the power level of the discharge. If the stock heat sink/fan combo was being used, then you'd know what the FET temperatures would be for various power levels and the user wouldn't need to add most of the fan control circuitry, just the transistor/FET switch to turn the fan on/off.

If the user wanted to use a different heat sink, or PWM or linear-bias (a FET or transistor) control its speed, then the full fan-control circuit could be installed.

 

[jateureka]

Sorry to dig up an old thread, but has any one built/tested any of the constant current load circuits in this post? ANy photos or results?

 

[katou]

I'm pretty interested in this one too, so thanks jateureka.

Katou