With advances in energy harvesting techniques, it is now feasible to build sustainable sensor networks (SSN) to support long-term applications. Unlike battery-powered sensor networks, the objective of sustainable sensor networks is to effectively utilize a continuous stream of ambient energy. Instead of pushing the limits of energy conservation, we are aiming at energy-synchronized designs 1 to keep energy supplies and demands in balance. Specifically, this work presents the Energy Synchronized Communication (ESC) as a transparent middleware between the network layer and data link layer that controls the amount and timing of RF activity at receiving nodes. In this work, we first derive a delay model for cross-traffic at individual nodes, which reveals an interesting stair effect in low-duty-cycle networks. This effect allows us to design a localized energy synchronization control with script O sign(1) time complexity that shuffles or adjusts the working schedule of a node to optimize crosstraffic delays in the presence of changing duty-cycle budgets. Under different rates of energy fluctuations, shuffle-based and adjustment-based methods have different influences on logical connectivity and cross-traffic delay, due to the inconsistent views of working schedules among neighboring nodes before schedule updates. We study the tradeoff between them and propose methods to update working schedules efficiently. To evaluate our work, ESC is implemented on MicaZ nodes with two state-of-theart routing protocols. Both test-bed experiment and large scale simulation results show significant performance improvements over randomized synchronization controls.