Cirrus Logic AN Integrating CobraNet into Audio Products - Manual (Page 9)

Thermal P RELIMINARY
7
CS18101 (CM-2) AppNote1 - rev 1.1 Jan, 2004
w w w . c i r r u s . c o m
The Practice
Theory is good, but it falls well short in practice. For starters, the results are only as good
as the data and the data is often missing or questionable. T
ja
, T
jc
and T
j
(max) are often
missing from the data sheets altogether. And when they are listed T
ja
and T
jc
often have a
wide error margin. To make matters worse, P
d
is sometimes no better than a guess based
on some ad-hoc test dreamed up by the semiconductor maker.
There are many different numbers that can be used for T
ja
. Good data sheets list several
types of thermal coefficients for different airflow rates. Mediocre data sheets will list only
one T
ja
, and it will probably be unqualified as to what conditions for which that coefficient
is valid.
Then there are the factors that are not accounted for. For example, the PCB itself will act
as a heat sink of sorts. This is especially true of multi-layer PCB's with many power and
ground planes and BGAs that have good thermal coupling between the die and the PCB.
Humidity will greatly effect the thermal conductivity of the air, as anyone living in Phoenix,
Arizona can attest. And the turbulence of the air flow can play a huge role in cooling off a
chip.
So what is an engineer to do? Going through the numbers will put you into the right
ballpark, but testing is required to be confident that it will work.
CM-1
The CM-1 is a good example of why thermal theory and practice are not the same thing.
Of the major components on the CM-1, the SRAM, Ethernet Controller, and Ethernet are
all rated at a maximum of 70°C ambient. Their data sheets list no more information on the
maximum junction temperature or thermal coefficients. Further, the ambient temperature
specification makes no mention of the ambient conditions (air flow, humidity, etc.). For
these parts we have to assume that 70°C max. ambient under any condition is valid.
On the other hand, the FPGA and DSP have some data, which we'll go over here.
According to the Motorola 56303 data sheets:
T
j
(max) = 100°C.
P
d
(typ) = 0.5 watts.
T
ja
= 58°C/Watt.
Therefore:
T
a
(max) = T
j
(max) - (T
ja
(max) * P
d
(typ)) = 100 - (58 * 0.5) = 71°C
From this we can conclude that 71°C is the maximum ambient temperature allowed with
this DSP. But keep in mind that P
d
is a rough guess and T
ja
is unqualified as to air flow
and humidity.
Similarly, according to the Xilinx documentation:
T
j
(max) = 100°C.
P
d
(typ) = 1.0 watt.