IBM21P100BGB
IBM 133 PCI-X Bridge R1.1
Clocking and Reset
Page 90 of 131
ppb11_clock_reset.fm.03
July 9, 2001
case for fixed frequency applications with simple clock generators or oscillators. A third possibility may be to
use a “power good” indicator, again if the proper stability assurances can be made. Other ways to provide the
S_CLK_STABLE input signal may also be possible.
The S_CLK_STABLE input provides a further measure of control for cases where the secondary bus mode
and clock frequency could vary from reset to reset, as might occur in motherboard applications with pluggable
slots. Here, the clock generator will need to adapt to the changes along with the bridge. By holding the
S_CLK_STABLE signal initially low during reset, the bridge will not be controlling the S_PCIXCAP network
and the clock generation circuitry is free to do its own mode and frequency determination sequence. The
clock frequency may also be adjusted based on the number of populated slots, determined by the PRSNT
pins of the bus. Then, once the frequency of the S_CLK input is stable, the clock circuit can assert the
S_CLK_STABLE signal to allow the bridge to complete the reset sequence. In any event, though, the clock
generation circuitry must guarantee that the clock frequency it provides falls within the range that the bridge
will ultimately determine and broadcast on the initialization pattern. To do this, the clock generator may need
to be capable of driving proper values on the S_SEL100 and S_PCIXCAP inputs, in addition to controlling the
S_CLK_STABLE signal. A mismatch between the broadcast initialization pattern and the actual operating
mode and frequency of the bus is a violation of the architecture and will cause unpredictable results.
6.5 Driver Impedance Selection
On the IBM 133 PCI-X Bridge R1.1, the output drivers for the bussed PCI / PCI-X interface signals are
capable of two different output impedances: a 40 ohm output impedance for point-to-point applications, and a
20 ohm output impedance for multi-point configurations. The output impedance for the primary and
secondary interfaces is separately controlled, and the bridge selects a default impedance value at the
de-assertion of the bus reset on the basis of the bus mode and frequency initialization pattern which was
received (on the primary interface) or generated (on the secondary interface). The bridge makes the assump-
tion that if a bus is configured to be in PCI-X 133 mode, it will be lightly loaded and therefore have a higher
impedance. Hence, the drivers are put into point-to-point mode for this case. For all other PCI-X and all PCI
configurations, the bridge assumes that the bus is more heavily loaded and has a lower impedance, so the
drivers are set to multi-point mode.
There may be some applications, however, for which these assumptions are inaccurate. For example, one
might want to connect to a conventional PCI device in a point-to-point manner. For exception cases like this,
two control input signals are provided, P_DRVR_MODE for the primary interface and S_DRVR_MODE for
the secondary interface. When these inputs are pulled high, the bridge will change the output impedance of
the drivers on their respective interfaces to the opposite state than was assumed by default, as shown in
Table 10: Driver Impedance Selection
Primary Bus Mode
Default Driver Mode
(P_DRVR_MODE=0)
Driver Mode if
P_DRVR_MODE=1
Secondary Bus
Mode
Default Driver Mode
(S_DRVR_MODE=0)
Driver Mode if
S_DRVR_MODE=1
Conventional PCI
Multi-point (20
W)
Point-to-point (40
W)
Conventional PCI
Multi-point (20
W)
Point-to-point (40
W)
PCI-X 66
Multi-point (20
W)
Point-to-point (40
W)
PCI-X 66
Multi-point (20
W)
Point-to-point (40
W)
PCI-X 100
Multi-point (20
W)
Point-to-point (40
W)
PCI-X 100
Multi-point (20
W)
Point-to-point (40
W)
PCI-X 133
Point-to-point (40
W)
Multi-point (20
W)
PCI-X 133
Point-to-point (40
W)
Multi-point (20
W)