| The LEM (Link Error Monitor) counts symbols in the receive data path that
shouldn't have been transmitted by its upstream station. It only counts
these invalid symbols when the PCM (connection managment state machine) is
in the Active, Next or Maintenance State. The reason that there are
invalid symbols which shouldn't be transmitted is that FDDI uses a 4B/5B
encoding scheme. Which means that for every 4 bits of real data you
want to transmit 5 actual bits are sent to represent those 4 bits. So
doing the quick binary math tells that if you start with (2)**4 = 16
and convert to (2)**5 = 32 that we now have 16 extra symbols. Actually,
FDDI uses 8 of these extra symbols (Quiet,Halt,Idle,J,K,R,S,T) to make Line
States. This leaves 8 garbage or unwanted symbols which NO FDDI STATION
should ever transmit. Basically, it is these symbols that get counted
as errors by the Link Error Monitor state machine when the Line State
Machine and the Physical Connection Management state machine are in the
appropriate states. Actually, it is a bit more complicated than this
and if you want to know more please contact me directly and I will send
you copies of pages out of the ELM chip specification.
As far as seeing 10-15/day this shouldn't hurt anything. The LEM
doesn't count every error that could occur and so there are other
checks in the MAC chip (a CRC check on the whole packet) and almost all
network software has additional data corruption checks. So any packet
that has an error will be retransmitted by these upper layer software
protocols. The LEM can't detect a valid data symbol getting changed to
another valid data symbol - but these other checks can. The LEM is only
meant to give a rough estimte of the quality of a link to the PCM to
let it know if it should bother to try and use the link to send data.
I'm not the best person to try and guess what may be causing these
errors. It could be the length of the link - which may cause a weak
optical signal at the receiver. Some dust in a optical connector could
also do this as well as a connector that is not pushed in all the way.
It could be a crystal oscillator that is slightly out of spec and the
station is just slightly too fast or slow. I'm sure there are many other
possible causes. This amount of errors shouldn't cause any problems or any
degradation in network performance. If the customer is worried then have
the link checked out with specialized test equipment.
I hope this helps,
Chuck Weaver
|