Test Result Analysis:



Figures & Graphs:

Figure 1. Channel Throughput when only one node is transmiting
Figure 2. Channel Throughput when two nodes are transmitting simultaneously
Figure 3. Compare the channel throughput of 1-node Tx and 2=node Tx
Figure 4. Channel throughput when 1-node encounter bad-link-quality
Figure 5. Channel throughput when 1-node encounter bad-link-quality, but MAC scheduling enabled
Figure 6. Channel throughput Comparison: Master Node
Figure 7. Channel throughput Comparision: Slave Node 1 ( with good link quality)
Figure 8. Channel throughput Comparision: Slave Node 2 ( with bad link quality)
Figure 9. Channel throughput Comparision: Slave Node 2 (  bad link quality vs. good limk quality of the other node link)


Throughput Analysis:

802.11b provides 11Mbps in physical link channle capacity. But, It's only in a sense of time-instant.
However,  In practice, it cannnot achieve such a rate in a general sense.
Many factors compromise the data rate:
1. Channel Co-ordination Functions.
    the rules to be used to co-ordiante channel access wastes a size of channel capacity, such as slot times, contention wondows.
2. MAC header overhead
    Each packet has an overhead of  34 bytes
3. RTS/CTS
 

From Figure 1,
When packet  rate is more than 680 packet/sec, the Rx bitrate is keep in a stable value as near4Mbps
This is the case when only one node is acccesing channel, thus its real meaning is that
"The (CPU) cannot put more data in that link" . Two possible reason:
1. CPU( Hardware device)'s cabability is stressed!
2. the physical channel is already saturated with bits?
The real reason is 2. Actually, the link could achieve ~4.3Mbps.

From Figure 2,
When link is in good condition, two nodes divides the channel capacity, each has an upper limit of Tx rate as ~2.1Mbps.
In the experiment,  Tx node2 is a 486 CPU, whie Tx node is a PII machine. The different CPU power maybe accounts for
the slight difference of the channel throughput when packet rate is more than 340 packet/s.

As we adding all the overhaeds of the layer header:
UDP header : 8 bytes
IP header : 20 bytes
MAC header: 34 bytes
The corrected total  "datarate" over this wireless link is 4.3 *(750+20+8+34)/750 = 4.66 Mbps

From Figure 3:
 



Data loss when link is bad:
 " Data loss" has tow major forms when BER is low.
1. MAC retransmission failure. There is a upper limits for re-transmission, such as 16 -reTx. If all those attempts fails, the packet is lost. (we use a UDP protocol, so higher layer such as layer 4 will not re-transmit it)
2. MAC transimisson successful (MAC checksum is correct), but IP or UDP protocol rejects it , because the checksum is wrong. the statistical application only count thouse packets after UDP layer,So it is rejected!
Normally, the latter situation does not have a big chance to occur, but it might be present in my experiment.????