C H A P T E R 4 F E A T U R E S
V100 Versatile Multiplexer Technical Manual Version 2.2
Page 158 of 231
In Raw mode, async data is transferred transparently. The compression mode intrinsically uses error-
correction, since the compression tables cannot work reliably otherwise.
4.10.1
Error-correction
The error-correction function is based on 100mS time periods. Data received by the tributary port is
buffered into a 100mS superframe irrespective of channel rate, which is then split into a number of
smaller sub-packets according to Dynamic Bandwidth Allocation (DBA) requirements. For example, if the
channel is configured with a baud rate of 115,200bps then there will be approximately 800 characters per
superframe buffer, which is then sent as a number of small HDLC frames depending on the internal DBA
rate, or the size of the logical connection on the aggregate which is allocated to the channel. The
underlying rule here is that no subframe or packet on the aggregate should occupy more than 20mS on
the aggregate.
The error-correction has a window size of 128, so it can correct up to 128 100mS buffers of continuous
data, representing roughly 128Kbytes of data at 115,200bps. For surfing/browsing this would improve,
since the data is much more sporadic.
In practice, only 75% of this buffer is used for safety reasons, before flow control is asserted at the port.
This means that at worst case, the error-correction can survive up to 9 seconds without receiving a good
frame.
The principle used for error-correction uses a NACK-only protocol. This means that a negative
acknowledgement packet is sent back whenever a good frame is received out of sequence, since at least
one frame in the middle must be missing. This NACK asks for the retransmission of frames starting with
the first missing one.
Errors experienced on the link affect the whole 100mS superframe, which has to be retransmitted. When
the error rate approaches 1 in 10e4 to 1 in 10e5, the error-correction will start to fail since every
superframe will be received in error. The technique therefore starts to reach a useful limit at error rates of
around 1 in 10e5.
When the channel input buffer reaches 75%, as caused by a poor link, insufficient bandwidth or loss of
carrier, the CTS signal on the port is dropped by the multiplexer. This signal must be connected to the
DTE for the error-correction to work. Loss of carrier is then handled by asserting flow control to stop
the input data. Realistically, this means that a channel should continue without data loss through carrier
losses of up to at least 5 seconds PROVIDED it is configured so that it does not lose any data. NOTE:
“The Data Channel Flag” parameter on the SYSTEM menu must be set to “Follows Alarms” for
the error-correction to work properly. If the connection is lost for more than 30 seconds then the
channel dumps all data, deletes the compression and restarts.
