
RELEASED
PM73122 AAL1GATOR-32
DATASHEET
PMC-1981419
ISSUE 7
32 LINK CES/DBCES AAL1 SAR PROCESSOR
PROPRIETARY AND CONFIDENTIAL TO PMC-SIERRA, INC., AND FOR ITS CUSTOMERS’ INTERNAL USE
125
into underrun. In order to prevent the underrun, (47 - frames_per_cell) frames
are buffered before data is sent out. This buffering when combined with the
frames_per_cell + 1 that is buffered on the transmit side results in a total of 48
frames of buffering. The result is every DBCES connection has the same
amount of delay in starting up regardless of the number of active DS0’s. This
delay is (48 + CDVT) * 125 us. The calculation to determine how much extra
buffering is needed in DBCES mode is calculated by RALP every time the queue
exits underrun or exits from an all idle state. This buffer adjustment guarantees
that the delay for a given channel does not change regardless of how many other
channels are active or inactive. This is due to the fact the total buffering between
the transmit side and receive side is always equal to 48 frames. If there is only
one channel active, 48 frames will be buffered on the transmit side and 0 frames
(excluding CDVT) on the receive side. If 31 channels are active, then 3 frames
will be buffered on the transmit side and 45 frames will be buffered.
The overrun condition is declared when the data in the receive buffer exceeds
the maximum specified buffer. When a cell is received that causes the maximum
buffer depth to be exceeded, the OVERRUN sticky bit is set and the AAL1gator-
32 enters the forced underrun condition. The incoming cells for the queue are
dropped until underrun occurs. Each time a cell is received and dropped in the
forced underrun condition, the FORCED_UNDERRUN sticky bit is set. Once
underrun occurs, the overrun flag is cleared and the same algorithm used in
underrun is followed. Figure 53 on page 126 describes overrun detection, the
underrun and recovery process.
Overruns can also occur due to lost/misinserted cells when robust SN processing
is done if the buffer is set too small. This is because when the SN processing
detects a potentially lost cell event, the cell will be written into the buffer at the
correct position assuming that cells have been lost. When Robust SN
processing is enabled, the R_MAX_BUF should be equal to or greater than two
times CDVT or CDVT plus 9 times the number of frames required per cell to
allow for the processing performed on lost/misinserted cells.
Anytime an overflow occurs, the R_OVERRUNS counter is incremented. When
an overrun is detected an entry will be made into the RCVN_STAT_FIFO, unless
the corresponding enable bit is not set or this is not the first enabled sticky bit to
be set for this queue since the sticky bit register was last cleared.
Note:
Inserting cells can cause an overrun. The threshold is checked as each byte
is written into memory. If an overflow occurs in the middle of a cell, the
remainder of the cell will be dropped.