Cell Relay Archive

["wal" <Your.Name@research.kpn.com>]

dgriffin@gio.com.au wrote in article <872600538.26388@dejanews.com>...
> The Telcos claim that fiber is the only viable
> transport because of BER characteristics of
> microwave.  Of course the Telcos charge ten
> times as much for fiber as a private microwave
> system would cost.
> 
> I quote one Telco's statement:
> 
> "The impact on ATM cell loss is significantly
> worse on a microwave system.  A BER of  1x10(-8)
> on optical fiber typically results in a Cell
> Loss Rate of 5x10(-13).  The same BER for
> microwave link results in a CLR of 1x10(-7) for
> a 2 bit error burst improving to 1x10(-8)
> for a 40 bit error burst.  Microwave radio
> systems typically have 4 bit error bursts.  In
> principle CLR should be confined to traffic
> shaping, node congestion but not transmission
> path.  This is an important factor in choosing
> the optimum transmission medium for an ATM
> network." (note:  The "2 bit error, 40 bit error
> and 4 bit error" are exact quotes and not my
> typos.)
> 
> Can anyone help me sort out if the above is
> correct and what the heck it really means?  Any
> comments, suggestions, or help would be most
> appreciated.

It appears to me that the differences are due to assumptions on the number
of errored bits that appear in groups. If optical systems show a typical
behaviour of 1 bit error randomly dispersed, ATM's HEC can correct a single
bit error anywhere in the 40-bit header. Thus, no cell loss. The
probability of more than 1 bit errored in the 40-bit ATM header (under the
assumption of independent singl-bit errors) is several orders of magnitude
smaller. A graphs showing the relation between error probability and the
uncorrectable headers (lost cells) is found in ITU-T recommendation I.432,
Annex A. 

For other transmission systems where the errored bits typically occur in
groups, the HEC does not work. Your Telco has made the calculations for any
number of grouped bit errors between 2 bits (the smallest group) and 40
bits (a large group; coincidently the ATM cell header is 5 bytes). Assume
the size of a bit error cluster to be G [2..40]. Assume there are BER bit
errors, then there are BER/G groups of errors. Each group has a probability
of  roughly 5/53 to hit a cell header. Assume that a hit cell header
results in a lost cell.  Thus the number of cells lost can be estimated. To
convert BER and group size G into Cell Loss Ratio (CLR), would thus
approximate to CLR ~~ BER/G * 5/53 * 53*8 = 40*BER/G.

ATM has been used over (geostationary) satellite links and over microwave
links without prohibitive large cell losses, even under less than optimal
conditions regarding SNR. Given a reasonable estimate for the BER it's up
to you to decide whether you can live with the resulting CLR. If so, I see
no problem. 

Kind regards, Kees van der Wal.
KPN Research, Leidschendam, the Netherlands.