Hi.
I would like to start a debate about the good old "if-the-heatsink-burns-you-its-to-hot-rule".
I'm doing a commercial design with the IRFP260N mosfet. The thermal resistance from chip to heatsink is pretty low and without any isolators ("live" heatsink) the heat-transfer to the heatsink is very low.
My calc shows that the silicium is around 110 degrees celsius when the heatsink is 80 deg! That means that I've got a burning hot heatsink but the chip is safe, - or what do you think.
On a side-note, check out this heatsink http://www.aavidthermalloy.com/products/maxclip/onepiece.shtml (scroll down to the 78240 "Big Chief".)
It's extremely efficient when fan-cooled. I'm dissipating 40W in 60mm of this in an ~ 2m/s airstream!
Any comments?
Regards TroelsM
I would like to start a debate about the good old "if-the-heatsink-burns-you-its-to-hot-rule".
I'm doing a commercial design with the IRFP260N mosfet. The thermal resistance from chip to heatsink is pretty low and without any isolators ("live" heatsink) the heat-transfer to the heatsink is very low.
My calc shows that the silicium is around 110 degrees celsius when the heatsink is 80 deg! That means that I've got a burning hot heatsink but the chip is safe, - or what do you think.
On a side-note, check out this heatsink http://www.aavidthermalloy.com/products/maxclip/onepiece.shtml (scroll down to the 78240 "Big Chief".)
It's extremely efficient when fan-cooled. I'm dissipating 40W in 60mm of this in an ~ 2m/s airstream!
Any comments?
Regards TroelsM
One more thing:
Any idea on how to add heat sink compound when using the "big Chief" profile?
A lot of the compound will simply be smeared off when the mosfet is inserted in the gab.
One more more thing... : If someone wants to see the calculations I can upload them later.
If these heatsinks are as good as the date suggest they should be the **** for low-profile high-power amps...
Regards TroelsM
Any idea on how to add heat sink compound when using the "big Chief" profile?
A lot of the compound will simply be smeared off when the mosfet is inserted in the gab.
One more more thing... : If someone wants to see the calculations I can upload them later.
If these heatsinks are as good as the date suggest they should be the **** for low-profile high-power amps...
Regards TroelsM
Try here:
http://www.bergquistcompany.com/thermal_materials.cfm
http://www.chomerics.com/products/documents/thermselguide2.PDF
and very interesting here:
http://www.kerafol.de/jml/
http://www.bergquistcompany.com/thermal_materials.cfm
http://www.chomerics.com/products/documents/thermselguide2.PDF
and very interesting here:
http://www.kerafol.de/jml/
TroelsM said:
Comments:
There is nothing to debate with regards to the safety of an 80C heatsink. You'll burn flesh with it and you'll cook your semiconductors (and probably destroy the speaker connected to the amp that is heatsinked at 80C).
It is simple physics- run semiconductors hot and they will die early. You can't change that by using thermal grease or funky heatsink profiles. Too hot is too hot.
I don't know about your customers, but when I buy a commercial product, I expect not to get burned by it and I expect it to work for a long time. Your design fails on both counts.
Skimping on a heatsink is a really bad way to economize.
I_F
Re: Re: Heatsink calculations
I guess you don't work for Samsung -- my LCD TV power supply failed in 1 year and it took 6 months to get a new one from Korea -- failure was heat related.
My car's electronic lock system instrument cluster, GPS system -- MB engineers say that the CPU/MCU is prone to failure due to thermal inadequacies.
I can recall that period in the distant past when controls in cars went from purely mechanical to hydraulically assisted -- door locks, speed controls, etc. -- the inadequacies of the technology and limited "learning curve" experience led to a flood of failures -- this when the Feds didn't enforce any recalls -- you were just stuck with it. Same thing is happening with MCU assisted devices and switching power supplies.
OK, off the soapbox -- a bigger heatsink is almost always a better solution -- but a little fan action will have an even more profound effect than square inches of surface area -- but put a fan into any device and you have to concern yourself with mechanical failure -- perhaps more important if you are running a server farm, or in a hospital ICU or operating room.
One thing which drives me nuts is folks trying to squeeze a 40 or 50 watt Class AB amplifier into a tea-tin. I guess it's a "guy-thing" in which case the product folder has only been delved in to the extent that there's a cookbook schematic.
I_Forgot said:
I guess you don't work for Samsung -- my LCD TV power supply failed in 1 year and it took 6 months to get a new one from Korea -- failure was heat related.
My car's electronic lock system instrument cluster, GPS system -- MB engineers say that the CPU/MCU is prone to failure due to thermal inadequacies.
I can recall that period in the distant past when controls in cars went from purely mechanical to hydraulically assisted -- door locks, speed controls, etc. -- the inadequacies of the technology and limited "learning curve" experience led to a flood of failures -- this when the Feds didn't enforce any recalls -- you were just stuck with it. Same thing is happening with MCU assisted devices and switching power supplies.
OK, off the soapbox -- a bigger heatsink is almost always a better solution -- but a little fan action will have an even more profound effect than square inches of surface area -- but put a fan into any device and you have to concern yourself with mechanical failure -- perhaps more important if you are running a server farm, or in a hospital ICU or operating room.
One thing which drives me nuts is folks trying to squeeze a 40 or 50 watt Class AB amplifier into a tea-tin. I guess it's a "guy-thing" in which case the product folder has only been delved in to the extent that there's a cookbook schematic.
Hi I_forgot
I see your point with the 80 deg surface, but this heatsink is internal and cannot be touched.
My post was meant to start a debate in regards to the "old rule" vs the calculations.
What matters to the silicon is the junction temperature. My calculations (and measurements) indicate that the die is about 110deg and that should not be a problem.
The thing is that with a low thermal resistance from junction to case and no isopad ( live heatsink) there not much difference between junction and heatsink.
Therefore the heatsink could be 80 while the junction is a healthy 110 deg...
I have attached a sim of an IRF530 and a IRFP260N disapating the same W.
The first example is an IRF530 that's isolated from the heatsink. The Die-temp is 137 deg and thats high!
The last example is a IRFP260N that's not isolated from the heatsink. The die is only 103 deg, and that's not dangerous in any way.
The difference is that the IRFP260N is way better at transporting heat to the sink than the IRF530. Of course its not a "fair" comparison when the IFR530 is isolated. - I only did that to illustrate my point.
Hope this makes sense.
If there´s anything wrong with my model please say so.
Regards TroelsM
I see your point with the 80 deg surface, but this heatsink is internal and cannot be touched.
My post was meant to start a debate in regards to the "old rule" vs the calculations.
What matters to the silicon is the junction temperature. My calculations (and measurements) indicate that the die is about 110deg and that should not be a problem.
The thing is that with a low thermal resistance from junction to case and no isopad ( live heatsink) there not much difference between junction and heatsink.
Therefore the heatsink could be 80 while the junction is a healthy 110 deg...
I have attached a sim of an IRF530 and a IRFP260N disapating the same W.
The first example is an IRF530 that's isolated from the heatsink. The Die-temp is 137 deg and thats high!
The last example is a IRFP260N that's not isolated from the heatsink. The die is only 103 deg, and that's not dangerous in any way.
The difference is that the IRFP260N is way better at transporting heat to the sink than the IRF530. Of course its not a "fair" comparison when the IFR530 is isolated. - I only did that to illustrate my point.
Hope this makes sense.
If there´s anything wrong with my model please say so.
Regards TroelsM
Attachments
Calculations and sims are a good tool for estimating what to do. Often they simply do not reflect the real world operating environment. Customers have a really good oportunity to destroy all your work by doing the unplanned to your creation.
Have you actually bought the pieces and mounted them together and put it into operation? Put a thermometer on it and measure. Now you will know, not guess.
Have you actually bought the pieces and mounted them together and put it into operation? Put a thermometer on it and measure. Now you will know, not guess.
The die temp you calculate is the temperature at the die/package interface, not the junction temperature. There is a thermal gradient across the die and the peak temp located at very small locations along the drain or source fingers will be much higher. Small variations in finger length (width) due to imperfect lithography will cause hot spots to develop. When testing these things in the lab, semiconductor design or process engineers will use an IR camera and microscope to view the temperature variation across the die. The hot spots will be considerably hotter than the die/package interface- look up the thermal conductivity of silicon- it is not very high.
Heat kills. If you want long life, you have to reduce the operating temperature. If you are making a ground to air missile that will be in flight for 3 minutes before running out of fuel or destroying itself, you don't need long life and can afford to run the temperature higher than in a piece of consumer equipment that the customer expects to work properly for anywhere between 5 and 20 years.
I_F
Heat kills. If you want long life, you have to reduce the operating temperature. If you are making a ground to air missile that will be in flight for 3 minutes before running out of fuel or destroying itself, you don't need long life and can afford to run the temperature higher than in a piece of consumer equipment that the customer expects to work properly for anywhere between 5 and 20 years.
I_F
With all respect to your opinions I think your are missing my point. When the thermal resistance of different packages are so very different, the heatsink-temperature is a very poor measure of the actual temperature inside the transistor.
We can all agree that the electronics will be a lot better of when it is cold. BUT, - when you design power-electronics things will get hot. Sometimes very hot and sometimes that is OK!
My point is that we need to be able to calculate the die-temperature and that the heat-sink temperature is a very poor indicator of the silicium well-being.
I_Forgot: If there´s a factor that I'm missing in the model please tell me. I'm having a hard time accepting that my model is "just wrong". If there's is an unknown in the equation we need to find it. Otherwise the whole idea about calculating temperature is bad.
Regards TroelsM
We can all agree that the electronics will be a lot better of when it is cold. BUT, - when you design power-electronics things will get hot. Sometimes very hot and sometimes that is OK!
My point is that we need to be able to calculate the die-temperature and that the heat-sink temperature is a very poor indicator of the silicium well-being.
I_Forgot: If there´s a factor that I'm missing in the model please tell me. I'm having a hard time accepting that my model is "just wrong". If there's is an unknown in the equation we need to find it. Otherwise the whole idea about calculating temperature is bad.
Regards TroelsM
Hi,
I have read a few times that junction temps should not exceed 100degC.
If I were designing I would try to ensure that the device dissipation NEVER exceeded half the device rating at the elevated running temperature.
ie. a 175degC device running at 80degC case temperature must be derated down to 63.3% of it's 25degC rating.
half of 63% is 31.7%. So a 125W device is by my reckoning used at a maximum dissipation of 39.6W call it 40W.
With this philosophy you will never get a Tj to Ts differential of 30Cdeg. I usually aim for no more than 10Cdeg Tj to Ts at normal dissipation.
I have read a few times that junction temps should not exceed 100degC.
If I were designing I would try to ensure that the device dissipation NEVER exceeded half the device rating at the elevated running temperature.
ie. a 175degC device running at 80degC case temperature must be derated down to 63.3% of it's 25degC rating.
half of 63% is 31.7%. So a 125W device is by my reckoning used at a maximum dissipation of 39.6W call it 40W.
With this philosophy you will never get a Tj to Ts differential of 30Cdeg. I usually aim for no more than 10Cdeg Tj to Ts at normal dissipation.
Many things that I have built for the US military require that junctions temperatures be maintained below 110C. I apply this as well to industrial designs. Many would consider 125 C very conservative and "high-quality"for a consumer device. 150 C is bad ju ju... 175 C is just garbage.
The heat sink temperaure is immaterial unless it is exposed. I often solder TO220 & TO247's directly to heatsinks in order to buy down the interface resistance. Not only are they HOT but they are LIVE and carrying a few hundred amps to boot... works awesome.
Now, there is a beauty thing about FETs and you can find similar traits with BJT and IGBTS. The on resistance, Rdson, of a FET follows a nice mathematical formula... don't know what it is exactly... doesn't matter. IR and others provide a chart of normalized Rdson vs. temp. Measure Rdson at room temp... you have to do this very quickly... think storage scope. Perform the same measurement at your peak design power... do this after heat sink temp has reached equilibrium at peak design ambient ambient. The ratio of Rdson can then be used to determine die temperature... right at the business part of the die that I_Forgot mentions.
Is this easy? No. Is is hard... not always. Get clever about it. Unsurpassed for accuracy in my book.
OH... the heatsink, "Big Chief", using aluminum for a spring... yuk (WTF were they thinking?). Check out Aavid's "Max-clip" hardware and supporting extrsusion. Much better stuff.
The heat sink temperaure is immaterial unless it is exposed. I often solder TO220 & TO247's directly to heatsinks in order to buy down the interface resistance. Not only are they HOT but they are LIVE and carrying a few hundred amps to boot... works awesome.
Now, there is a beauty thing about FETs and you can find similar traits with BJT and IGBTS. The on resistance, Rdson, of a FET follows a nice mathematical formula... don't know what it is exactly... doesn't matter. IR and others provide a chart of normalized Rdson vs. temp. Measure Rdson at room temp... you have to do this very quickly... think storage scope. Perform the same measurement at your peak design power... do this after heat sink temp has reached equilibrium at peak design ambient ambient. The ratio of Rdson can then be used to determine die temperature... right at the business part of the die that I_Forgot mentions.
Is this easy? No. Is is hard... not always. Get clever about it. Unsurpassed for accuracy in my book.
OH... the heatsink, "Big Chief", using aluminum for a spring... yuk (WTF were they thinking?). Check out Aavid's "Max-clip" hardware and supporting extrsusion. Much better stuff.
It's a valid point that the temperature in the case is higher than the temperature outside the case. But I think that I've got that part covered.
The heatsink will only "see" the cold airflow from the outside and some radiated heat from near-by heatsinks. The heatsinks are "silver-finish" and I'm guessing that radiated heat from sink-to-sink won't matter much.
I'm using 35 Deg C as the max ambient. (very, very, very unlikely here in Denmark. We have approx 30 on the hottest summer-days).
I'm attaching a sim of my newest thermal model. 4 fets are dissipating 75W (modelled as one source) and the fets are attached to a heat-sink with a thermal resistance of 0.75w/deg. All thermal resistances have been made larger than real-life (hopefully).
This gives a heatsink-temp of 92 (ninety-two!) and a junction-temp of 115 deg. We can agree that touching the heat-sink will be very unpleasant, but I´m not designing a commercial product. This will only be used by professionals.
Another consideration is the solder-joint-temperature. If this is above 90 deg, the joint could get "unstable" in time. Any ideas?
Regards TroelsM
The heatsink will only "see" the cold airflow from the outside and some radiated heat from near-by heatsinks. The heatsinks are "silver-finish" and I'm guessing that radiated heat from sink-to-sink won't matter much.
I'm using 35 Deg C as the max ambient. (very, very, very unlikely here in Denmark. We have approx 30 on the hottest summer-days).
I'm attaching a sim of my newest thermal model. 4 fets are dissipating 75W (modelled as one source) and the fets are attached to a heat-sink with a thermal resistance of 0.75w/deg. All thermal resistances have been made larger than real-life (hopefully).
This gives a heatsink-temp of 92 (ninety-two!) and a junction-temp of 115 deg. We can agree that touching the heat-sink will be very unpleasant, but I´m not designing a commercial product. This will only be used by professionals.
Another consideration is the solder-joint-temperature. If this is above 90 deg, the joint could get "unstable" in time. Any ideas?
Regards TroelsM
Hi Poobah.
That's clever! I hadn't thought about using the on-resistance as a measure of the temperature. Nice!
About the "big chief" profile: I see your point about using aluminium as a spring, but the alu will "spring" more than a normal mounting and as far as I can see it works great in real life.
My only worry is that I'll be doing a design that takes about 800 fets in all, and I don't know how I'll get the "assembly-crew" to put the fet in the clip the right way....
Regards TroelsM
That's clever! I hadn't thought about using the on-resistance as a measure of the temperature. Nice!
About the "big chief" profile: I see your point about using aluminium as a spring, but the alu will "spring" more than a normal mounting and as far as I can see it works great in real life.
My only worry is that I'll be doing a design that takes about 800 fets in all, and I don't know how I'll get the "assembly-crew" to put the fet in the clip the right way....
Regards TroelsM
For BJT's you can use Vbe at fixed base (emittor? can't remember) currents.
You'll have to use your own judgement on the aluminum... I use some here... but the power levels are lower. I sorta don't trust it... alumiunum springs that is.
If you are in design though... the "max-clips" are worth looking at... they are super easy to use... and you can't do them wrong. High marks for idiot proofing.
You'll have to use your own judgement on the aluminum... I use some here... but the power levels are lower. I sorta don't trust it... alumiunum springs that is.
If you are in design though... the "max-clips" are worth looking at... they are super easy to use... and you can't do them wrong. High marks for idiot proofing.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.