User's Manual Part 1

xxvi APCD-LM043-8.0 (DRAFT C)
:
Previously, the CCU chose, at the beginning of each polling cycle, which EUM (or random
access) to poll. If no EUM was ready to be polled, a random access poll was sent. Now, the
CCU both chooses which EUM (or random access) to poll and when to do so. So, if no EUM
(or random access) is ready to be polled immediately, it determines which will be ready
soonest and schedules the poll for that time.
In a heavily loaded system, the result is nearly identical (see minIPS change below) to the
previous system—there is always an EUM ready to be polled, so the spaces between poll
cycles are the minimum possible. No change in capacity or responsiveness occurs.
In a lightly loaded or idle system, the difference in spectral occupancy is profound, without a
noticeable change in system responsiveness. In an idle system, random access polls are
spaced 55 ms apart (the same spacing as in a heavily loaded system) and are less than 1 ms
long, so the channel is occupied less than two percent of the time. When only one or two users
are active, the polls occur at exactly the desired rate—every 29 ms for default best effort
users—leaving large gaps between cycles.
One parameter has changed meaning slightly. The minIPS parameter previously indicated the
minimum allowed spacing between two polls of the same EUM and was primarily used to set
an upper limit on the polling rate. The minIPS parameter now indicates the minimum allowed
average spacing between two polls of the same EUM. That is, the EUM will be polled no more
often, on average, than every minIPS. The average is over a short time span of about 10 polls.
This improvement more accurately provides the usually desired control—which is a transfer
rate cap—as the effects of transient traffic spikes are averaged out.
To further reduce the impact of a lightly loaded system on other users of the spectrum, gaps
between poll cycles are constrained to be at least 10 ms long. Polling cycles that would have
been closer together are re-arranged ahead and behind these gaps. Random access polls are
treated slightly differently than others to enhance this.
The ideal and max IPS violation counts in the load meter are slightly more sensitive.
NOTE: Low impact polling will be immediately noticeable in the rad
rssi display: the number of received packets will be as low as
16 per second in an idle system. Users used to seeing 900+ may
be concerned.
Long Reply Timeout
The CCU reply timeout has been increased by 200 us, which extends the timeout radius from
10 km to 40 km. Note that it does not improve the RF range.
Duplicate IP Detection and Protection
The CCU discards ARPs from the radio with a duplicate source IP address to its own, while
printing out a warning including the EUM ID of the station the packet came from. The routing
protocol statistic “Rx Radio Err – Duplicate IP address” counts these duplicates. Previously,
setting an EUM or user-PC IP address the same as the CCU would disrupt traffic as these
ARP requests and responses would corrupt ARP tables across the network.