Header and Body 3


The core component of MegaMIMo is a lightweight distributed phase alignment algorithm that allows N single-antenna access points (APs) to behave as if they were a single giant MIMO AP with N packets concurrently to N clients. The algorithm works as follows: one of the APs acts as the lead, and all other APs adjust their transmitted signals to emulate a scenario where they share the same oscillator as the lead AP. Every joint transmission (i.e. a transmission of concurrent streams) starts by the lead AP transmitting a short synchronization header. The other APs listen to the synchronization header, and use it to discover the difference between their own oscillator phase and the oscillator phase of the lead AP. They can then align the phase of their own transmission with the lead AP by multiplying by the appropriate angle.


This synchronization cannot persist over time because different APs have different frequency offsets, which causes their signals to rotate at different speeds over time. To address this issue, the APs also use the synchronization header to estimate the frequency offset, predict changes in phase caused by this offset, and compensate for them during phase synchronization. MegaMIMO inserts resynchronization symbols periodically in a packet to allow APs to track the phase difference with the lead AP. Since the frequency of resynchronization symbols only depends on measurement error, and does not change with the increased number of APs, the overhead of such symbols is a small constant that becomes negligible as the number of APs increases.