You are not logged in.
Hi,
We would like to understand more details about the backoff mechanism, and set up a experiment in which a TX always transmits with 0 backoff slot and a RX has a large initial backoff number. We generate large throughput and make the TX transmit packets to the RX back to back.
We expect to see that the RX will never count down its backoff number.
However, it still counts down the backoff number once about every 500 packets. We would like to know is it normal ?(Assuming there is no other interference)
If it is, what are the possibilities that cause the RX has a shorter DIFS and starting backoff faster than the TX? clock drift?
Thank you.
Offline
Is the traffic you are sending unicast (i.e. does it require an ACK)? If so, the failure to receive an ACK would cause a timeout at the transmitter. A timeout (T_SIFS+T_SLOT+T_PHY_RX_START_DLY) would cause your RX node to decrement its backoff counter. One way to rule this out is to send multicast traffic that requires no ACK reception.
Offline
It is unicast, but the RX does send ACK, so there will be no timeout at the transmitter. What we do is check the backoff number when the RX transmit ACK (BKack) and the backoff number when the RX receive next packet (BKnextdata), and see if (BKack-BKnextdata=0). In most of the time, it is 0, but sometimes (once every 500 packets) it is not 0.
Offline
The ACK itself could be lost at the original TX side due to a decoding failure. Even if your RX is sending an ACK 100% of the time, there could still be a timeout on the TX side.
At the RX side, are the decremented backoffs periodic? In other words, do they occur every 500 packets exactly? Or is it random and roughly once every 500 packets? If it is random, does this ratio depend on any other factors like SNR? If this is an over-the-air experiment, try moving your kits closer or further away to see if this ratio changes. If it does change, that means it probably is related to some kind of OTA loss of a packet.
Offline
I see. Thank you so much for answering. We will do more experiments and see the results.
Offline