![](http://datasheet.mmic.net.cn/310000/AD9957_datasheet_16240198/AD9957_41.png)
AD9957
which three AD9957s are synchronized with one device operating
as a master timing unit and the others as slave units.
Rev. 0 | Page 41 of 60
The master device must have its SYNC_IN pins included as part
of the synchronization distribution and delay equalization mecha-
nism for it to be synchronized with the slave units.
The synchronization mechanism begins with the clock distribu-
tion and delay equalization block, which ensures that all devices
receive an edge aligned REFCLK signal. However, even though
the REFCLK signal is edge aligned among all devices, this alone
does not guarantee that the clock state of each internal clock
generator is coordinated with the others. This is the role of the
synchronization and delay equalization block. This block accepts
the SYNC_OUT signal generated by the master device and
redistributes it to the SYNC_IN input of the slave units (as well
as feeding it back to the master). The goal of the redistributed
SYNC_OUT signal from the master device is to deliver an edge
aligned SYNC_IN signal to all of the sync receivers.
Assuming that all devices share the same REFCLK edge (due to
the clock distribution and delay equalization block) and that all
devices share the same SYNC_IN edge (due to the synchroniza-
tion and delay equalization block), then all devices should be
generating an internal sync pulse in unison (assuming they all
have the same sync receiver delay value). With the further
stipulation that all devices have the same sync state preset value,
then the synchronized sync pulses cause all of the devices to
assume the same predefined clock state simultaneously. That is,
the internal clocks of all devices are fully synchronized.
The synchronization mechanism depends on the reliable
generation of a sync pulse by the edge detection block in the
sync receiver. Generation of a valid sync pulse, however,
requires proper sampling of the rising edge of the delayed
sync-in signal with the rising edge of the local SYSCLK. If the
edge timing of these signals fails to meet the setup or hold time
requirements of the internal latches in the edge detection
circuitry, then the proper generation of the sync pulse is in
jeopardy. The setup-and-hold validation block (see Figure 58)
gives the user a means to validate that proper edge timing exists
between the two signals. The setup-and-hold validation block
can be disabled via the sync timing validation disable bit in Control
Function Register 2.
The validation block makes use of a user-specified time window
(programmable in increments of ~150 ps via the 4-bit sync
validation delay word in the multichip sync register). The setup
validation and hold validation circuits use latches identical to
those in the rising edge detector and strobe generator. The
programmable time window is used to skew the timing between
the rising edges of the local SYSCLK signal and the rising edges
of the delayed sync-in signal. If either the hold or setup valida-
tion circuits fail to detect a valid edge sample, the condition is
indicated externally via the SYNC_SMP_ERR pin (active high).
The user must choose a sync validation delay value that is a
reasonable fraction of the SYSCLK period. For example, if the
SYSCLK frequency is 1 GHz (1 ns period), then a reasonable
value is 1 or 2 (150 ps or 300 ps). Choosing too large a value can
cause the SYNC_SMP_ERR pin to generate false error signals.
Choosing too small a value may cause instability.
SYNC
PULSE
SYSCLK
DELAY
DELAY
C
4
SYNC VALIDATION
DELAY
4
4
SYNC_SMP_ERR
SYNC RECEIVER
12
SYNC TIMING VALIDATION DISABLE
SETUP
VALIDATION
HOLD
VALIDATION
D Q
SETUP AND HOLD VALIDATION
TO
CLOCK
GENERATION
LOGIC
FROM
SYNC
RECEIVER
DELAY
LOGIC
D Q
D Q
RISING EDGE
DETECTOR
AND STROBE
GENERATOR
0
Figure 58. Sync Timing Validation Block