Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
802.16-2005 gives a complicated algorithm
to map subcarriers to subchannels (Section 8.4.6). (1) For the downlink, it is true that making
subcarriers for each subchannel (most likely used by one mobile station) scattered
among the available band (e.g., 10 MHz) will obtain a kind of frequency
diversity gain, as long as all the base stations sharing the same band are
accurately synchronized (in frequency domain). However, has anyone proved how
much gain can be achieved by using this complicated algorithm in Section 8.4.6
over a much simpler algorithm such as evenly interleaving subcarriers from
different subchannels over the entire band? My experience tells me that the
gain would be minimal, while the cost of implementation would easily out-weight
the small gain (extra memory, extra gates or MIPS needed and longer development
cycle). (2) For the uplink, the complicated
algorithm in Section 8.4.6 would almost certainly cause performance degradation
for high-speed-moving vehicle, such as a car on highway at 130 km/hour, a high-speed
train at 150 to 300 km/hour, as the Doppler frequency from different mobile station
would almost certainly cause unsynchronized frequency offset at the base
station receiver, resulting in significant inter-carrier-interference. Even
worse, if any one mobile has some frequency synchronization problem (this could
happen occasionally), the failed mobile could cause big trouble to all the
mobiles sharing the same band. I would suggest that the subcarriers for each subchannel
in the uplink should be adjacent to each other to avoid this Doppler frequency
shift/failed mobile problem. To still utilize the frequency diversity, a mobile
sending a large amount of data could use more than one subchannel, and the
interleaving depth should be increased to utilize this diversity. If the mobile
is sending a small amount of data, it is not necessary for him/her to occupy
more than one subchannel as he/she is not using much bandwidth anyway. (3) I am surprised by the fact that the
classical way of using time-domain interleaving (e.g., span 20ms or 40ms) to
combat fading is not used in Section 8.4. I doubt at 5MHz bandwidth and 130-300
km/hour moving speed, this standard would work without a classical time-domain
interleaving. If the goal of this standard is for slowly moving mobiles, then
this interleaving should not matter. Dr. Will Liew Andrew Corp. 434-386-5212
|