MPC8548E PowerQUICC III Integrated Processor Hardware Specifications, Rev. 9
Freescale Semiconductor
79
PCI Express
17.5
Receiver Compliance Eye Diagrams
The RX eye diagram in Figure 49 is specified using the passive compliance/test measurement load (see
Figure 50) in place of any real PCI Express RX component.
Note: In general, the minimum receiver eye diagram measured with the compliance/test measurement load
(see Figure 50) is larger than the minimum receiver eye diagram measured over a range of systems at the
input receiver of any real PCI Express component. The degraded eye diagram at the input receiver is due
to traces internal to the package as well as silicon parasitic characteristics which cause the real PCI Express
component to vary in impedance from the compliance/test measurement load. The input receiver eye
diagram is implementation specific and is not specified. RX component designer must provide additional
margin to adequately compensate for the degraded minimum receiver eye diagram (shown in Figure 49)
expected at the input receiver based on some adequate combination of system simulations and the return
loss measured looking into the RX package and silicon. The RX eye diagram must be aligned in time using
the jitter median to locate the center of the eye diagram.
LTX-SKEW
Total Skew
—
20
ns
Skew across all lanes on a Link. This includes
variation in the length of SKP ordered set (for
example, COM and one to five symbols) at the RX
as well as any delay differences arising from the
interconnect itself.
Notes:
1. No test load is necessarily associated with this value.
2. Specified at the measurement point and measured over any 250 consecutive UIs. The test load in Figure 50 must be used
as the RX device when taking measurements (also see the receiver compliance eye diagram shown in Figure 49). If the
clocks to the RX and TX are not derived from the same reference clock, the TX UI recovered from 3500 consecutive UI must
be used as a reference for the eye diagram.
3. A TRX-EYE = 0.40 UI provides for a total sum of 0.60 UI deterministic and random jitter budget for the transmitter and
interconnect collected any 250 consecutive UIs. The TRX-EYE-MEDIAN-to-MAX-JITTER specification ensures a jitter distribution
in which the median and the maximum deviation from the median is less than half of the total. UI jitter budget collected over
any 250 consecutive TX UIs. Note that the median is not the same as the mean. The jitter median describes the point in time
where the number of jitter points on either side is approximately equal as opposed to the averaged time value. If the clocks
to the RX and TX are not derived from the same reference clock, the TX UI recovered from 3500 consecutive UI must be
used as the reference for the eye diagram.
4. The receiver input impedance shall result in a differential return loss greater than or equal to 15 dB with the D+ line biased
to 300 mV and the D– line biased to –{300 mV and a common mode return loss greater than or equal to 6 dB (no bias
required) over a frequency range of 50 MHz to 1.25 GHz. This input impedance requirement applies to all valid input levels.
The reference impedance for return loss measurements for is 50
to ground for both the D+ and D– line (that is, as
measured by a vector network analyzer with 50-
probes—see Figure 50). Note: that the series capacitors CTX is optional
for the return loss measurement.
5. Impedance during all LTSSM states. When transitioning from a fundamental reset to detect (the initial state of the LTSSM)
there is a 5 ms transition time before receiver termination values must be met on all unconfigured lanes of a port.
6. The RX DC common mode Impedance that exists when no power is present or fundamental reset is asserted. This helps
ensure that the receiver detect circuit does not falsely assume a receiver is powered on when it is not. This term must be
measured at 300 mV above the RX ground.
7. It is recommended that the recovered TX UI is calculated using all edges in the 3500 consecutive UI interval with a fit
algorithm using a minimization merit function. Least squares and median deviation fits have worked well with experimental
and simulated data.
Table 57. Differential Receiver (RX) Input Specifications (continued)
Symbol
Parameter
Min
Nom
Max
Unit
Comments