Just an idea.
Mounting the transistor on a quite thick and large copper plate without insulation for a best thermal contact.
Then mount the copper plate on the aluminum heat sink with insulation.
With a right choice of thickness and surface of the copper part, I think we can get an overall lower thermal resistance.
What do you think ?
May be this has been done or may be not worth the trouble.
Mounting the transistor on a quite thick and large copper plate without insulation for a best thermal contact.
Then mount the copper plate on the aluminum heat sink with insulation.
With a right choice of thickness and surface of the copper part, I think we can get an overall lower thermal resistance.
What do you think ?
May be this has been done or may be not worth the trouble.
So sorry but all that does is move the thermal resistance away from the active device. Like an electrical circuit, moving a resistor does not increase the current flow!
Sanken MT-200 packages work on this principle, heat spread to a bigger area of TIM to reduce thermal resistance to the heat-sink.
Indeed it moves the resistance, however the moved resistance is lower because of a larger surface. Doubling the contact area is about halving the thermal resistance.
Last edited:
Is not a bad idea , but using a large insulator will cost you more , on top of the cost of the copper plate.
Better solution is to use best insulator you could find lots of air flow and maybe active cooling.
Better solution is to use best insulator you could find lots of air flow and maybe active cooling.
Already done for you.
The insides of the transistor is the chip and a block of Copper. You nail the copper to your Aluminum.
The thermal resistance of the insulator is measurable but not "usually" a major bottleneck, until you got to way too much power per device on way enormous thick finny sinks. The #1 thermal resistance is usually fin to air. You have to spread heat to a LOT of air. With odd-lot extrusions you may have another bottleneck between device base and actual finning.
Unless you favor exotic devices, for less than the price of a big slab of copper you can run 2 or 4 devices, 1/2 or 1/4 the Rt in the mica, and generally more effective heat spreading.
Beautiful.
I see you have a copper bar for several transistors, so they cannot be mounted without insulation onto the copper. Do I miss something ?
An extreme design would be using silver instead of copper. Silver has a better thermal conductivity. 418 versus 390
I see you have a copper bar for several transistors, so they cannot be mounted without insulation onto the copper. Do I miss something ?
An extreme design would be using silver instead of copper. Silver has a better thermal conductivity. 418 versus 390
Last edited:
The technique is known, and used in some cases: for example when a chassis or case made of sheet steel is used as a heatsink, adding a spreader multiplies the effective area for little cost.
Also, when a very low Rth is paramount: the thermal resistance of insulators is rather poor, and vastly increasing their area can practically eliminate it.
Note that the technique might not be immediately visible: the copper plates used for that purpose are rarely bare, they are protected with a layer of tin, nickel or cadmium.
The power MOS on the right of these pictures is equipped with such a plate, although the main purpose is somewhat different here: it is an electrostatic guard.
Some of the first power IC's had a two stage spreader, a pure silver slab on top of a more conventional copper one
Also, when a very low Rth is paramount: the thermal resistance of insulators is rather poor, and vastly increasing their area can practically eliminate it.
Note that the technique might not be immediately visible: the copper plates used for that purpose are rarely bare, they are protected with a layer of tin, nickel or cadmium.
The power MOS on the right of these pictures is equipped with such a plate, although the main purpose is somewhat different here: it is an electrostatic guard.
Some of the first power IC's had a two stage spreader, a pure silver slab on top of a more conventional copper one
The F5 takes the audio output from the MOSFET drains so the drains are all conviently electrical tied together and no dielectric is used between the transistor case and the copper. There is a thermal interface product between the copper and the aluminum bar. All the screws are electrically isolated using nylon bushings through the copper.
Note, bare copper and bare aluminum do not play well together when they are bolted together. Always use an interface product or some sort of plating on the metals.
Excellent.
What do you recommend as thermal interface product(s). What is the thermal resistance per square inch ? ( or other per surface unit ).
Tin, Nickel , Cadmium have poor thermal conductivity compared to Copper, but used as plating is no issue about thermal resistance, l presume ? thanks of its thickness, presumably less than 10 micron meter ( 10 / 1,000,000 meter ).
Last edited:
This method it is very good for the Vbe multiplier. This one will react much quickly and the temperature for all the transistors will be the same.
Unfortunately, what matters is junction temperatures.
Junction temp is not case temperature, that is the question.
404 Not Found
I used above thermal interface product mostly because it was available in a 12" X 18" piece.
I used above thermal interface product mostly because it was available in a 12" X 18" piece.
I agree, it just moves the problem. Here is another example of using copper bar, but with insulators
- Status
- Not open for further replies.