
Airflow
C4 Chip
Ceramic Substrate
60
°
C
CBGA Package Ball
55
°
C
Printed-circuit Board
50
°
C
50
°
C
Airflow
60
°
C
55
°
C
Figure 2
. CBGA and Printed-circuit Board Contour Temperatures for Typical Computer Systems.
RESULTS
Printed-Circuit Temperatures
For the results presented in this paper, modeling and
experimental measurements have shown differences
between the solder joint and the PCB temperature. Figure
2. shows a temperature contour of a CBGA package under
computer-system operational conditions (e.g. air velocity is
1 m/s) showing the temperature difference is less than 5
°
C,
between the measured field data of the PCB and the
maximum solder-joint temperature. Again, the key
objective was to identify maximum printed-circuit boards
(PCBs) temperatures while the computer systems were in
typical use. Printed-circuit boards (PCB) were not uniform;
therefore, only maximum temperature regions of the board
were measured.
For uniprocessor computer systems, the maximum board
temperatures, were at or near the central-processing unit
(CPU). The data reported here is considered to be the
maximum board temperatures, one might expect in typical
office use, under stressed operational system loading
conditions. The maximum board temperature rise
measured was 28
°
C (Figure 3), and the mean for this
limited data set was 17
°
C (Figure 4). A frequency
histogram of this data is shown in Figure 4. These findings
concur with IPC-SM-785.
In most cases, the maximum temperature was at the
CPU. However, in several computer systems, the board in
the CPU region was actually at a lower temperature, than
ancillary components. In addition the following results are
noted:
Printed-circuit boards of operational computer
systems showed large temperaure variations.
That is, it is not too difficult to find regions that
were only slightly above ambient temperatures.
For the data reported here, there was a wide
range in CPU power dissipations (1 to 30
watts); however, this had little effect on the
board temperatures. That is, the high-power
CPUs typically used enhanced thermal
management schemes (larger heat sinks and/or
active heat sinks, e.g. fan sinks) to remove heat
off the component "case". Thus, the board
temperature at or near the CPUs did not exhibit
a significantly larger temperature rise, than
other systems.
Two systems used active muffin fans, attached
to a pin-fin heat sink.
For one tower workstation, the secondary cache
memory was operating at a higher temperature
than the CPU.
All CPU had active or passive heatsinks.
For one desktop workstation system, the region
of the CPU and the region of the graphics card
(PGA without heat sink), experienced
temperature rises of 11
°
C and 23
°
C;
respectively. Note, the board on the graphics
controller card, was 12
°
C higher than the CPU
region.
For one network server, the memory-
management unit (assumed) adjacent to the
CPU (with a fan-heatsink) was running 2
°
C
higher. Therefore, both data points are plotted.