North Korea detonated its first known nuclear device, a moderate-sized event in the northeastern corner of the Korean Peninsula, on 9 October 2006. A second one, several times more powerful, was set off nearby on 25 May 2009. Both were recorded at high signal-to-noise ratio (SNR) by a modern broadband (0.03–30 Hz) network of seismographs deployed since 2004 along the Sino–Korean border, and by station MDJ due north of ground zero. Spanning a wide range of station azimuth (259°–11°), the near-regional (1.3°–3.3°) paths are all purely continental and away from continental margins, making the resulting data uniquely suited for assessing the capabilities of an out-country network to verify the compliance by North Korea with a future comprehensive nuclear test ban treaty. The mb(Lg) magnitudes for the 2006 and 2009 nuclear explosions were determined anew, giving 4.32±0.13 and 4.86±0.13, respectively. We show that the MS–mb(Lg)method correctly differentiates these nuclear explosions from natural earthquakes, although the differentiation was not achieved in studies based on published mb(Lg) values. An analysis of the Pg:Lg ratio, derived from recordings of the two nuclear tests and of four regional earthquakes selected for their comparable magnitudes (4.2≤mb≤4.8), reveals that the ratios associated with the explosion and earthquake populations showed surprisingly little overlap for a broad frequency range of 3 to 11 Hz. In principle, MS–mb can also be used to correctly identify the nuclear tests as explosions, but the mb measurement for such moderate-sized events is not as robust as the mb(Lg) measurement, rendering this traditionally favored method less practical.