
Hardware Design Guide
31
IXP28XX Network Processor
Power Ratings and Requirements
2.5.1
di/dt Droop Analysis Results
An exhaustive droop analysis was performed using an Intel system validation platform. The goal of
analysis was to correlate the measured power-on di/dt with the theoretical design values. A detailed
model of the system board, power delivery circuit and package were developed and simulated
using the theoretical di/dt transient events to predict the voltage droop measured at the pins of the
device. Several correlation analysis were performed to ensure the integrity of the models. Once the
models were validated the simulation results were compared with measured results from the
system validation platform. The final step of the analysis was to perform a sensitivity analysis by
varying the di/dt stimulus to match the simulated results to the measured results.
The outcome of the analysis is the transient curve shown
Figure 6
. The following items summarize
the findings of the correlation study:
•
The First Droop magnitude is 34 mV (tr1 = 6 cycles, tr2 = 9 cycles, and tr3 = 1 cycle), which
occurs at 8 ns after transient is applied.
— The magnitude and time of the first droop depends on the combination of tr1 and tr2
transient events.
•
The Second Droop magnitude is 78.4 mV, which occurs at 0.67 µs after transient.
— The second droop depends on when the Microengines come out of reset, specifically how
much time this occurs after PLL locks and power starts decreasing.
•
The Third Droop magnitude is 67.6 mV, which occurs between 10.1 – 13.1 µs after transient.
— Assuming Voltage Regulator Bandwidth = 10 KHz.
— Cbulk is
only
50% of its total capacitance.
— I_final = 10.703 A (10% more than 9.73 A measured value).
— The number of clock cycles for tr3 does not impact the third droop.
— The amount of final current load and Cbulk will impact the amplitude of the third droop.
•
The first, second, and third droops are not affected by the duration of tr3.
2.5.2
Recommendations From Droop Analysis
There are two major recommendations from the correlation study:
•
Increase the system board cavity (under the processor) capacitors to address the second droop
amplitude. To reduce the magnitude of the second droop in the 0.67us timeframe, sufficient
decoupling capacitors should be placed under the processor; refer to
Figure 2
.
•
Increase bulk capacitors to address the third droop amplitude. As shown in
Figure 6
the current
will decrease after the PLL locks and the MicroEngines are taken out of reset, therefore
sufficient bulk capacitance is required to ride through the ~1uS high current transient event.
Downloaded from
Elcodis.com
electronic components distributor
Содержание IXP28 Series
Страница 10: ...IXP28XX Network Processor Revision History 10 Downloaded from Elcodis com electronic components distributor ...
Страница 126: ...126 Hardware Design Guide IXP28XX Network Processor PCI Downloaded from Elcodis com electronic components distributor ...
Страница 154: ...Index 154 Downloaded from Elcodis com electronic components distributor ...