Enhanced Multi-Rate DSL Data Pump Chip Set — SK70725A/SK70721
Datasheet 45
circuits based on its own transmitted signal. The Master moves between states based on its MAT.
When the Master timer has passed SMT2, it enters the SIGDET sub-state where it remains until
detection of an S0 signal from the Slave.
If Master and Slave are connected and the Slave is in the Inactive state, the Slave detects the S0
signal from the Master, and starts its MAT. The Slave enters the Wait sub-state, and begins training
its AGC. The Slave does not transmit any signal until MAT exceeds SMT2. At that time, the Slave
transmits an S0 signal and enters the EC sub-state where the Echo canceler and the Digital AGC
are trained.
The Master detects the S0 signal from the Slave, and resets the MAT to SMT2+1. This re-
synchronization process assures that Master and Slave state machines will be synchronized for the
remainder of the activation process. In most activation, attempts where the Slave is connected to
the line and is in the Inactive state at the beginning of the activation attempt, the Slave will begin to
transmit S0 just as the Master gets to the SIGDET state and the change in the Master MAT will be
minimal. If the Slave is reset or connected to the line sometime after the activation sequence has
begun at the Master, the Master will remain in the SIGDET state until S0 is received or the MAT
expires. If S0 is received after the Master has been in SIGDET for some time, the change in the
setting of the MAT at the Master may be significant. This change in the Master reference timer,
which is referred to as TDELTA in Figure 20 and Figure 21, allows the substates of the Master and
Slave Data Pumps to be synchronized for the remainder of the activation sequence.
The Master and Slave Data Pumps continue through the activation process as shown in Figure 20
and Figure 21. Both Data Pumps complete training of all the receiver components which can be
trained with a two-level (S0) signal relying on system timers SMT2, SMT3, SMT4, and SMT5 to
maintain synchronization between the Master and Slave.
After the SMT5 timers have expired, progression through the remainder of the activation states is
data driven, that is, it relies on receipt of a particular signal from the other end to move to the next
sub-state. The Slave enters the 4LVLDET sub-state and waits for receipt of a four level (S1) signal
from the Master. On receipt of this signal the Slave completes training of the DFE and 4 level slicer
then begins to transmit an S1 signal. The Master detects the S1 signal from the Slave and then
completes training of the DFE and 4 level slicer.
The EDSP also allows the designer to specify the length of time the Data Pump spends in each of
the activating substates. The default times have been optimized for best performance at either 784
kbps or 1168 kbps in the framed modes (modes 7 and 6). Reliable operation has been demonstrated
using the default values at all data rates between 272 kbps and 1168 kbps in unframed modes using
the default timing. Overall activation time can be reduced for certain applications by reducing the
activation times of the individual states. Table 9 gives the information required to select timer
values for each of the activating substates. After settings for each of the individual sub-state timers
has been optimized, the Master Activation Timer may be optimized to minimize the time required
between sequential activation attempts.
4.4.3 Transition to the Active State
The actions following detection of the S1 signal vary depending on the operating mode of the
system. The following sections describe the operation in both modes.
Unframed Signal (Modes 0,1,2,4,5)
Both Master and Slave move directly to the Active state on completion of the 4LVLDET training
functions. The devices remain in the Active state even after MAT expires. Received data are
available, and the ACTIVE indicator is asserted Active state.