Hi. I have an LT1084 12V/5A linear regulator dropping about 3V and passing 2.2A that is heatsinked nicely to one of these:
http://media.digikey.com/photos/Aavid Thermalloy Photos/531102B02500.jpg
1.75" tall and rated at 10W/50degC rise
I have measured up to 98degC at the heatsink... dangerously close it would seem to the 105C rating of the TO-220 package. Should I be worried or looking at a bigger heatsink?
http://media.digikey.com/photos/Aavid Thermalloy Photos/531102B02500.jpg
1.75" tall and rated at 10W/50degC rise
I have measured up to 98degC at the heatsink... dangerously close it would seem to the 105C rating of the TO-220 package. Should I be worried or looking at a bigger heatsink?
On the Aavid Thermalloy web site 531102B02500G is listed as 10.4 C/W, so you should expect a 73C rise at 7 Watts dissipation. With a room temperature of 25C, you should measure 98C at the heatsink so your results are consistent. Based on the thermal considerations section of the LT1084 datasheet, that puts the junction of the power transistor in the IC at about 117C which really is too hot. It is probably a few degrees higher than that since the thermal resistance of the case to heatsink interface is not zero. I would suggest a bigger heatsink.
Thanks, guys.
That picture was of a slightly smaller heatsink, the one I am using is rated at 10C/W but hey that's pretty much the same. Bigger heatsink it will have to be, and I should have done the simple calcs first. I just thought that the ratings were "worst case" and I did not expect quite so much heat. Now I have to redesign the pcb to make room for a bigger footprint, I have limited vertical space.
That picture was of a slightly smaller heatsink, the one I am using is rated at 10C/W but hey that's pretty much the same. Bigger heatsink it will have to be, and I should have done the simple calcs first. I just thought that the ratings were "worst case" and I did not expect quite so much heat. Now I have to redesign the pcb to make room for a bigger footprint, I have limited vertical space.
It isn't perfectly ok to run anywhere near 150C if you want reliability. Rule of thumb is that failure rate doubles for each 10C increase in temperature. You would have to request a full reliability report for the LT1084 from Linear Tech to see what the projected life time is at 150C. From there back the temperature down as needed to get the component lifetime needed for the application. Conservatively I try to keep junction temps below 100C at the maximum room temperature expected.
As I say a bit of heat is OK.
All you have to do is look at some of the cheap amps on the market with high watts and small heatsinks. They seem to get away with it.
If you dont mind paying a small fortune for a heatsink and fans then you can run at 30C all day.
As the LT1084 will operate down to 1V in/out differential, you could try something like a 0.68 Ohm 10W resistor before the regulator. At 2.2 Amps you will be taking 7W of dissipation away from the regulator and a resistor may be easy to hide somewhere out of the way.
My old mentor (God Bless him) used to design with quick "rules of thumb" and his designs always worked superbly. He taught me that to find a heat sink, a quick way was to take the maximum device temperature (150deg C here), subtract the ambient (he always used 50 deg C for super reliability) then divide the answer by the dissipation. Here we have 150-50 = 100 divided by 7W = 14.3 deg C/W.
Subtract the makers Jc-case figure (2.3) plus an additional 1.5 for the insulating pad so the heat sink must be 14.3 - 2.3+1.5) = 10.5.
As your heat sink is lower than this (10deg C/W) then it should be fine.
Not the purists way of working it out I suppose but it has passed the test of time on many projects and is easy to memorise.
Les
Should also mention that working out the heat sink is just part of the story. You should arrange for the vanes to be vertical to get a good chimney effect for best air movement. The fitting of a fan, even at slow speed, greatly assists the convection of heat away from the sink. The dear old 2N3055 was very popular because its maximum junction temperature was 200 deg C which made it possible to use a smaller heat sink than the general 150 deg C devices.
My old mentor (God Bless him) used to design with quick "rules of thumb" and his designs always worked superbly. He taught me that to find a heat sink, a quick way was to take the maximum device temperature (150deg C here), subtract the ambient (he always used 50 deg C for super reliability) then divide the answer by the dissipation. Here we have 150-50 = 100 divided by 7W = 14.3 deg C/W.
Subtract the makers Jc-case figure (2.3) plus an additional 1.5 for the insulating pad so the heat sink must be 14.3 - 2.3+1.5) = 10.5.
As your heat sink is lower than this (10deg C/W) then it should be fine.
Not the purists way of working it out I suppose but it has passed the test of time on many projects and is easy to memorise.
Les
Should also mention that working out the heat sink is just part of the story. You should arrange for the vanes to be vertical to get a good chimney effect for best air movement. The fitting of a fan, even at slow speed, greatly assists the convection of heat away from the sink. The dear old 2N3055 was very popular because its maximum junction temperature was 200 deg C which made it possible to use a smaller heat sink than the general 150 deg C devices.
Last edited:
Sorry, my PC Calculator screwed me up, not the first time, dissipation in the resistor is half as quoted but still may be worth while. I had a similar problem on a DC heater supply circuit. The heat sink was pretty hot, especially being in a case with just barely adequate ventilation so I added a resistor. Its effect was very noticeable and of course a wire wound resistor will sit sweating quite happily for as long as you want it too 
.
Les
.Les
Getting away with something doesn't mean it's okay -
http://www.designnews.com/blog/Made..._a_Meltdown.php?rssid=20254&text=Emerson+ES50
Its a whole different ball game in high volume production.
Saving a penny on a million items is £10,000 a shot.
So a 20p heatsink saves £200,000 !!!!!!!!!!!
Really, it's quite amazing that consumer electronics last as long as they do. ~100C internal temps cooking transistors, ICs, and capacitors most notoriously. A 10khr lifetime isn't anything to scoff at, and is fairly typical among the tube guys.
Not that it's any kind of favorable goal to shoot for a fairly short design life. Best engineering practices as always. Plus, those tube guys have sockets...
Tim
Not that it's any kind of favorable goal to shoot for a fairly short design life. Best engineering practices as always. Plus, those tube guys have sockets...
Tim
No one has ever put too much heatsink on a device temperature-wise though it can get expensive. My personal target is to keep semiconductors the temperature of a hot shower or lower. Keep in mind the internal temperature of the part is often far higher than the heatsink temp.
G²
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.