Because T4 and T1 are sampled at the same clock, there is no arbitrary phase between T4 and T1. Therefore, T4?T1 is accurate in time; the same principle applies to T3?T2. Therefore, this measurement is immune to any phase-walking between the two devices resulting from the asynchronous nature of this system. Similar to the previous embodiment, this measurement is limited by the resolution of the sampling clock period of T1/T2/T3/T4. In order to improve this accuracy, a frequency response measurement can be performed on both devices. Device 820 measures the channel response using the packet from device 810 and device 810 measures the channel response using the packet from device 820. Because the two devices are not synchronous, there is an uncertainty in the phase between the two clocks, annotated as Toffset here. This phase offset of the clock manifests itself as an extra phase of the channel response measurement on each side, but it can be eliminated by multiplying the channel responses from the two sides. Assuming the channel response is the same as before, then the measurement from device 820 will be H820(f)=H(f)e?j 2πfToffset
The measurement from the device #1 will be H810(f)=H(f)e+j 2πfToffset
The combined channel response is therefore H810(f)H820(f)=H(f)2+(ΣAke?j 2πf ΔTk)2