![](http://datasheet.mmic.net.cn/Maxim-Integrated-Products/DS34T102GN-_datasheet_97090/DS34T102GN-_65.png)
____________________________________________________ DS34T101, DS34T102, DS34T104, DS34T108
65 of 366
10.6.3 Clock Recovery
The TDM-over-Packet block’s innovative clock recovery process is divided into two successive phases. In the
acquisition phase, rapid frequency lock is attained. In the tracking phase, frequency lock is sustained and phase is
also tracked. During the tracking phase, jitter is attenuated to comply with the relevant telecom standards even for
packet-switched networks with relatively large packet delay variation. Packet loss immunity is also significantly
improved.
During the acquisition phase, a direct estimation of the frequency discrepancy between the far-end and near-end
service clocks continuously drives an internal frequency synthesis device through a band-limited control loop. As a
result, frequency acquisition is achieved rapidly (typically less than 10 seconds). The clock recovery capture range
is
90 ppm around the nominal service clock for any supported clock rate.
Once the frequency-monitoring unit has detected a steady frequency lock, the system switches to its tracking
phase. During the tracking phase the fill level of the received-packet jitter buffer drives the internal frequency
synthesizer through a similar band-limited control loop.
While in the tracking phase, two tasks are performed. First, the far-end service clock frequency is slowly and
accurately tracked, while compelling the regenerated near-end service clock to have jitter and wander levels that
conform to ITU-T G.823/G.824 requirements, even for networks that introduce high packet delay variation and
packet loss. This performance can be attained due to a very efficient jitter attenuation mechanism, combined with a
high resolution internal digital PLL (=0.4 ppb). Second, the received-packet jitter buffer is maintained at its fill
level, regardless of the initial frequency discrepancy between the clocks. As a result, the latency added by the
mechanism is minimized, while immunity against overflow/underflow events (caused by extreme packet delay
variation events) is substantially enhanced.
The TDM-over-Packet block supports two clock recovery modes: common clock (differential) mode and adaptive
mode.
The common clock mode is used for applications where the TDMoP gateways at both ends of the PSN path have
access to the same high-quality reference clock. This mode makes use of RTP differential mode time-stamps and
therefore the RTP header must be present in TDMoP packets when this mode is used. The common reference
clock is provided to the chip on the
CLK_CMN input pin. The device is configured for common clock mode when
The adaptive clock mode is based solely on packet inter-arrival time and therefore can be used for applications
where a common reference clock is not available to both TDMoP gateways. This mode does not make use of time-
stamps and therefore the RTP header is not needed in the TDMoP packets when this mode is used. The device is
configured
for
adaptive
clock
mode
when
in
1
and
In adaptive mode, for low-speed interfaces (up to 4.6 MHz), an on-chip digital PLL, clocked by a 38.88MHz clock
derived from the
CLK_HIGH pin, synthesizes the recovered clock frequency. The frequency stability characteristics
of the
CLK_HIGH signal depend on the wander requirements of the recovered TDM clock. For applications where
the recovered TDM clock must comply with G.823/G.824 requirements for traffic interfaces, typically a TCXO can
be use as the source for the
CLK_HIGH signal. For applications where the recovered clock must comply with
G.823/G.824 requirements for synchronization interfaces, the
CLK_HIGH signal typically must come from an
OCXO.
In addition to performing clock recovery for up to eight low-speed (typically E1/T1) signals, the device can also be
configured in a high-speed mode in which it supports one E3, T3 or STS-1 signal in and out of port 1. In high-speed
mode, the on-chip digital PLL synthesizes the recovered clock frequency divided by 10 (for STS-1) or 12 (for E3 or
T3). This clock is available on the
TDM1_ACLK output pin and can be multiplied by an external PLL to get the
recovered clock of the high-speed signal (see section
15.3). High-speed mode is enabled when High_speed=1 in
2
For applications where the chip is used only for clock recovery purposes (i.e. data is not forwarded through the
chip) the external SDRAM is not needed.