Thermal Resistance (Semi-Packages)

[adyeng]

Hi all,
I am in the process of modeling the thermal resistance of a surface mount (SMD1) package as used by IR and Semelab and Sensitron.
I am basically running a spread sheet that accounts for the various layers within the package including the material thermal conductivity, spread angles and effective areas. The problem I have is that the results do not match the IRs datasheets. Is this likely to be due to IR rounding or "guestimating" figures?
Should the method I am using provide a reasonable answer?
Does anyone have any other methods that may be of use (beside measuring the devices).
Regards,
Adyeng

 

[IRstuff]

How far off are you and in what direction?

TTFN

 

[adyeng]

Hi IRStuff,

The Value given in the App Note for the SMD1 is given as 0.83.
The value I obtain for the same sized die is 0.26?
I am suprised as this is such a basic package (No vias etc) just the following:
Si Die = 250um
Eutectic = 10um or preform of 50um
Gold flash = 3.81um
Nickle = 8.89um
Copper Tungsten = 274.6um
Nickle = 8.89um
Gold flash = 3.81um

 

[IRstuff]

Do you still have the link for the app note?

TTFN

 

[adyeng]

The Datasheet can be obtained from

http://www.irf.com/technical-info/appnotes/an-1016.pdf

cheers

 

[IRstuff]

What are you assuming to power generating area of the die and how are you determining the spreading resistance through the die?

Additionally, you are probably assuming 100% effective heat flow through the area of the die and substrate. As an example, there may be derating factors to account for statistical possibility of voids in the die bond as well as voids in the bonding of the external pad.

TTFN

 

[IRstuff]

As an aside, since IR's number is higher, that's probably a good/conservative thing. Bad news would be if your number was 4 times worse than their's.

You should also read IR's AN-994. They indicate that their thermal resistances are measured, using packaged die clamped to an infinite heat sink. The temperature difference between the heat sink and the die surface is then used in the calculation of the thermal resistance. This means that in addition to whatever geometrical factors there might be, there's also a contribution from the "infinite" heat sink used in the measurement as well as the efficacy of the thermal path from the die pad to the heat sink, e.g., did they use thermal grease? or was it strictly CuW to Cu?

TTFN