We propose a cross-layer approach with tightly-coupled time synchronization for real-time support and predictable lifetime in battery-operated sensor networks. Our design spans a sensor hardware platform with hardware-based global time synchroniza-tion, a TDMA link layer protocol with collision-free multi-hop support and node scheduling algorithms for maximum concurrency and streaming. Our dual-radio sensor platform, FireFly, features an IEEE 802.15.4 transceiver and supports global time syn-chronization indoors by using an AM radio carrier-current method and an atomic clock receiver for outdoors. A TDMA-based link protocol, RT-Link, leverages the hardware for fixed and mobile nodes with a predictable node lifetime of over 2 years. It outperforms comparable sensor network link protocols such as B-MAC and S-MAC in terms of end-to-end latency and throughput and node lifetime across all duty cycle ratios. Operating over RT-Link is MAX, a scheduling framework which offers optimal transmission concurrency and bandwidth management for networks with regular structure. Through analysis and experiments we show that global time sync is a robust, economical and scalable alternative to in-band software-based techniques. To illus-trate the capabilities and flexibility of our platform, we describe our experiences with two-way voice streaming over multiple hops. We have deployed a 42-node network with sub-20µs synchronization accuracy in the NIOSH experimental coal mine for people-tracking and voice communication. 1.