Reflectivities and Doppler velocities from an airborne 95 GHz (3 mm) radar, as well as coincident aircraft data, are used to describe the detailed (~25 m) vertical structure of a cold front penetrating into an optically-clear convective boundary-layer (CBL). It has long been postulated that on the meso-g scale the cold air associated with a cold front may at times assume the vertical structure of a density (or gravity) current. Atmospheric density currents have received a great deal of attention, and they have been invoked in the interpretation of thunderstorm-generated gust fronts, sea and land breezes, katabatic winds spreading over level terrain, and also cold fronts. While the fine structure of density currents is well-known from laboratory experiments, the existence of such structure in the atmosphere has only been inferred, because of the lack of observations at sufficiently small scale. The observations on which this paper is based were collected as part of the International Water Vapor Project (IHOP), during the afternoon of 24 May 2002 in the Texas Panhandle. Frontogenesis along a cold front heading south across the Central Plains was apparent during the daytime as the cold air remained covered by a vast stratus cloud deck, while the warm side was mostly cloud-free. The aircraft made 8 transects across the front, at various levels between 2400 and 160 m above ground level, over a two hour period, before deep convection broke out at least 10 km ahead of the front. The Wyoming Cloud Radar (WCR) operated in two modes. One is the vertical-plane dual-Doppler mode, involving one antenna in the nadir, and a second antenna directed some 34 degrees forward of nadir. The second one is the up-looking mode, used on low-level flight legs. The cold front was associated with some stratus clouds, and the deep CBL ahead of the cold front was capped by some stratocumulus clouds, but broad evidence confirms that WCR echoes were largely due to insects. In order to assess to accuracy of the WCR radial velocity field, corrected for aircraft motion, as a proxy for air motion, we first examined some 10 hours of flight legs through the undisturbed CBL, far away from cold fronts or other radar fine-lines, with the WCR operating in profiling mode (one nadir antenna, one zenith antenna). This assessment shows that there is a systematic bias in the WCR vertical velocities, when compared to the gust-probe vertical velocities (see the companion paper by Geerts and Miao in this Conference). This bias, about 0.8 m/s on average, is larger in stronger updrafts, suggesting that insects actively oppose the updraft in which they are embedded. Transects of 2D vectors, corrected for insect motion, show the fine detail of the airflow across the cold front. The velocity and echo structures confirm the appearance of the cold front as a density current, with a well-defined, but highly variable head, rear-to-front flow within, and strong mixing at the shear zone above and behind the head through a series of large-amplitude Kelvin-Helmholz billows. An attempt will be made at estimating the shear vorticity and the Richardson number across the frontal surface. The cross sections, plus various kinematic and thermodynamic indicators, will be used to assess the analogy with density currents as simulated in the laboratory and by high-resolution models.