As an example of a passive building facade, the Trombe Wall was first developed some fifty years ago. However, this type of passive solar heating system, using buoyancy-driven convection, is rarely included in building designs. This is partly because their operation is poorly understood. This paper reports on an experimental investigation into buoyancy-driven convection in a test rig designed to simulate the operation of a passive solar collector. The test rig comprised a vertical open-ended channel, approximately 1m square, heated from one side. The channel depth could be varied from 20mm to 110mm, and heating inputs varied from 200W to 1000W. Temperatures and airflow rates were measured and recorded, to characterise both steady-state and transient performance. The principal findings are: • Time constants (for heating) ranged typically between 30 and 70 minutes. • Flow regimes were mainly laminar (Reynolds numbers varying from ~500 to ~4000, depending on heat input and channel depth) • The thermal efficiency (as a solar collector) and the overall heat transfer were functions of heat input, and were not dependent on the channel depth. • The mass flow rate through the channel increased both as the heat input increased and as the channel depth increased. The paper presents these findings and discusses their implications in more detail.