Wideband, high-power microwave pulses are expected to have important applications in the future. One of these applications is ultra-wideband radar. The wide bandwidth should generate increased information for targetcharacterization and identification. The high power should result in increased target detection range for conventional targets and targets with reduced signatures. A way to generate wideband, high-power microwave pulses with relatively conventional technology is to tail erode high-power pulses by passage through a low-pressure air cell. In this process, the tails ofshort(3 — 10 ns), high-amplitude (<1 MV/rn) pulses are removed. This erosion shortens thepulses and generates transmitted pulses with broadenedbandwidths. Thepressure mustbe matched to several incident pulse characteristics to create enough electron density to cause strong tail erosion. The important pulse characteristics are amplitude, frequency, pulse length, and pulse shape. Tail erosion of microwave pulses in the earth's atmosphere has previously been examined with one-dimensional, finite difference computer calculations.13 Experiments on tail erosion in a rectangular waveguide have verified two-dimensional (2D), finite difference computer calculations.4'5 We have shown experimentally that tail erosion from air breakdown broadens the 3 dB bandwidths of 2.8608 GHz incident pulses in a rectangular waveguide at 3.5 ton. The incident pulse amplitude varied from 0.67 —1.16 MV/rn. The pulse bandwidth increased from 0. 147 GHz by 0.0097 — 0.039 GHz or 0.34 — 1.4% relative. The incident bandwidth was 5. 12% relative to the incident carrierfrequency. This experimentalbroadening was simulated with a2D, electromagnetic, electron fluid computercode foravalanche ionization. The simulation predicted bandwidth broadening by 0.029 —0.13 GHz or 1.0 —4.4% relative for a peak initial electron density of 10 electrons/cm3. Although the measured and calculated transmitted electric field envelopes were in close agreement, the calculated bandwidths exceeded those measured by 13 —47%. Because the detectors were not fast enough to resolve individual cycles and therefore determine thelocalfrequency across thepulses, wepresently conclude that the simulation gives betterestimates of reality than do the measurements. The computer code gives predictions of the bandwidth broadening of 3.5 ns incident pulses at 3.5 torr for a Gaussian spatial background electron distribution with a 3 cm fullwidth at halfmaximum (FWHM)in the axialdirection. The peak valueofthe electron density distribution and its transverse FWHM are variable. Incident amplitudes of 1 —18 MV/rn and peak electron densities of 10 — 1011 electrons/cm3 are considered. The transmittedbandwidth varies from 0.352— 3.21 GHz or 12.3 —112% relative. The transmitted spectralcenter frequency varies from 2.87 —4.72 GHz. The transmitted amplitude at 54.6 cm from the input to thelow-pressure section varies from 0.263 — 12.7 MV/rn on the waveguide center line.