Err, I was more referring to usual heatsinking, not watercooling* or using fancy liquid nitrogen.
For the Jfet I wait for the gold plated leads version
What's the advantage of a normally-off jfet in comparison to a mosfet? From the datasheet that looks very similar to a vertical mosfet, no transconductance given though.
Have fun, Hannes
*at a thermal resistance of about 0.6C/W I doubt water cooling is an option; maybe with water close to 4°C.
That kind of exotic cooling is not necessary at all. Just rough figures here, if you use a legal heat sink (with a max. end temperature of 60 deg C), and mount your MOSFET directly on the anodized surface, with only a good heat grease, you can drop 150W easily off one TO247 package.
And yes, a 20 um adodized surface isolates fine for like 150-200V.
And yes, a 20 um adodized surface isolates fine for like 150-200V.
Hi Lars,
bear me, I have no idea how that should work?
The TO-247 package has a thermal resistance of 0.6 C/W, so that alone amounts to a temperature rise of the silicon to 90 C above whatever the heatsink is and at 150W.
I just checked an IRF-datasheet, it gets even worse: max 0.8 C/W junction to case and another 0.2 C/W case to sink thermal resistance (some grease).
And that's already without considering the heatsink smaller than infinitely large
I guess the key lies in the maximum temperature of silicon, I thought it to be at about 120°C, but I guess - too lazy to look - it's higher. If 150°C is the max, you could maybe get away with that, but reliability is certainly not your goal.
Have fun, Hannes
bear me, I have no idea how that should work?
The TO-247 package has a thermal resistance of 0.6 C/W, so that alone amounts to a temperature rise of the silicon to 90 C above whatever the heatsink is and at 150W.
I just checked an IRF-datasheet, it gets even worse: max 0.8 C/W junction to case and another 0.2 C/W case to sink thermal resistance (some grease).
And that's already without considering the heatsink smaller than infinitely large
I guess the key lies in the maximum temperature of silicon, I thought it to be at about 120°C, but I guess - too lazy to look - it's higher. If 150°C is the max, you could maybe get away with that, but reliability is certainly not your goal.
Have fun, Hannes
Hannes: For a product, you are right, you wouldn't push to the limit like this. But we try to establish where the limit is, and then step back for reliability.
Take a look at IRFP260: http://www.datasheetcatalog.org/datasheet/irf/irfp260.pdf
Total die to case : 0.7 K/W, and max die temp. of 175 deg C.
So according to the datasheet: 175 - 60 = 115 K / 0,7 = 164 W at 60 deg. heat sink temperature. That's the absolute max. limit for reliable operation. But at this point there is no room for mistakes.
If you also want long lifespan, you should of course not push to 164W, but i recon half (~80W) wouldn't be a problem, even for reliability and long lifespan. I did some experiments with IRFP150 some years ago, pushing it right to the limit, with 175 deg die temperature, and after several weeks of operation, all test devices still worked fine.
I even did some experiments with some bipolar power transistors, heating them to 220 deg die temp, and they worked fine for weeks as well. Of course just an experiment.
In my amplifiers i design for less than 25W dissapation in a TO247 package, so lots of overhead there.
To estimate the thermal dissapation in a (correctly working) Class A/B amplifier, you should look at it's overall efficiency at max power. Depending on the idle current, and rail dropout, you get something like 60%. This means 40% of total power is heat, at 70% output power this dissapation is about 20% higher. So for 100W, you have 67W * 1.2 = 80 W worst case power dissapation in all. If you have 2 output transistors, this means each dissapate 40W heat onto the heat sink.
Edit:
This TO247 device only has 0.44 W/K die to case, so it will dump 204W. (Limit)
http://ixdev.ixys.com/DataSheet/IXKH70N60C5.pdf
Take a look at IRFP260: http://www.datasheetcatalog.org/datasheet/irf/irfp260.pdf
Total die to case : 0.7 K/W, and max die temp. of 175 deg C.
So according to the datasheet: 175 - 60 = 115 K / 0,7 = 164 W at 60 deg. heat sink temperature. That's the absolute max. limit for reliable operation. But at this point there is no room for mistakes.
If you also want long lifespan, you should of course not push to 164W, but i recon half (~80W) wouldn't be a problem, even for reliability and long lifespan. I did some experiments with IRFP150 some years ago, pushing it right to the limit, with 175 deg die temperature, and after several weeks of operation, all test devices still worked fine.
I even did some experiments with some bipolar power transistors, heating them to 220 deg die temp, and they worked fine for weeks as well. Of course just an experiment.
In my amplifiers i design for less than 25W dissapation in a TO247 package, so lots of overhead there.
To estimate the thermal dissapation in a (correctly working) Class A/B amplifier, you should look at it's overall efficiency at max power. Depending on the idle current, and rail dropout, you get something like 60%. This means 40% of total power is heat, at 70% output power this dissapation is about 20% higher. So for 100W, you have 67W * 1.2 = 80 W worst case power dissapation in all. If you have 2 output transistors, this means each dissapate 40W heat onto the heat sink.
Edit:
This TO247 device only has 0.44 W/K die to case, so it will dump 204W. (Limit)
http://ixdev.ixys.com/DataSheet/IXKH70N60C5.pdf
I once did a burn-in project that operated open die (TAB) devices submerged in a 65 degree C bath. The bath fluid had a boiling point of 170 C. The devices would be powered up to force boiling, regulating the temp. I suppose the details are process dependant but, at 170 we were expecting to accelerate the life of the devices to approximately 17 times normall. So, we were doing 48 hours of burn in to simulate a month of operation
Roughly extrapolating from that, a system expected to last 8-9 years may only last 6 months
Just another hopefully useful tid-bit
Roughly extrapolating from that, a system expected to last 8-9 years may only last 6 months
Just another hopefully useful tid-bit
Magura said:Hey, I just found out....I'm nobody
Magura
Not only you, I was doing it for 2 day's, just prior to final test and shipping...
Maybe we can start a new DiyAudio catagory, or, the Nobody thread
On a more serious note, my statement was regarding die, not case or heatsink. We had an onboard(in the die) diode correlated to temp so we new we were relatively accurate. We could even see the nucleate boiling causing noise in our temp measureing channel. Bare die remember. No die bond, no case. The die is suspended from the 360 .001-2" bumped copper leads, nothing else. The more intense guys were burning about 23 Watts! I think it was about .35"sq.
Regarding the Silicon Carbide JFETS, I do seem to remember the desire to run some of the diode SiC stuff pretty hot though... But, I would still stick to something less than 105 at the die or so. At least until I learned more. I hope the learning curve is quick though. $60 ea?
I also remember reading the Seimens Sic stuff. I think theirs was a cascoded JFET wasn't it?
Hmmmm, miore voltage, more current. But, maybe that was another thread???
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.