Enabling long range (>10m) wireless communication in non-line-of sight (NLOS) scenarios would dramatically expand the application space and usability of mm-scale wireless sensor nodes. The major technical challenges posed by a mm-scale form-factor are poor antenna efficiency and the small instantaneous current limit (~10s of μA) of thin-film batteries. We address these challenges in several ways: 1) We found that a magnetic dipole antenna achieves better efficiency at an electrically-small size than an electric dipole, when the antennas are resonated with off-chip lumped components. In addition, the high impedance of electrically-small electric dipoles (~4kΩ compared to 10Ω for the magnetic antenna) requires an impractically large off-chip inductor to resonate. 2) By simultaneously considering the magnetic dipole efficiency, frequency-dependent path-loss, and wall penetration loss, we found that a 915MHz carrier frequency is optimal for a 3×3×3mm ³ sensor node in NLOS asymmetric communication with a gateway. This is despite the resulting low antenna efficiency (0.21%) which typically drives mm-scale radios to operate at ≫1GHz frequency [1]. 3) In transmit (TX) mode, instead of using a PA and PLL, we utilize a cross-coupled driver to resonate the magnetic antenna at 915MHz with a quality factor (Q) of 110 in order to reduce overall power consumption. 4) In receive (RX) mode, we propose an approach of reusing the cross-coupled driver in a non-oscillating mode to raise the Q of the resonant tank to 300, resulting in 49dB voltage gain at 43μW, thereby replacing a power-hungry LNA and bulky off-chip filter. 5) A sparse pulse-position modulation (PPM) combined with a sensor-initiation communication protocol [2] shifts the power-hungry calibration of frequency offset to the gateway, enabling crystal-free radio design. The complete radio, including the transceiver IC, a 3D antenna, off-chip capacitors, a processor, a power management unit (PMU) and memory, is integrated within a 3×3×3mm ³ sensor node, demonstrating stand-alone bi-directional 20m NLOS wireless communication with variable data rates of 30b/s to 30.3kb/s for TX and 7.8kb/s to 62.5kb/s for RX. The transmitter generates -26.9 dBm equivalent isotropically radiated power (EIRP) consuming 2mW power and the receiver has a sensitivity of -93dBm consuming 1.85mW.