The proper understanding of gravel-bed river dynamics is a crucial issue for the effective protection against related natural hazards, design of hydraulic structures, and preservation of their high ecological value in mountain regions. However, despite more than one century of research in the field, most available models fail to accurately predict bedload transport rates in such alluvial rivers because of the complex relationships between the flow, channel morphology, and sediment transport. It is now recognized that spatio-temporal variability is an inherent property of bedload transport in gravel-bed rivers which results in its pulsating character even under steady flow conditions. This experimental study aims to better understand the physical mechanisms involved in sediment transport in gravel-bed channels characterized by alternate bars. More specifically, it is concerned with the origins of the pulsating nature of bedload transport under steady external conditions in relation to bed macro-forms. Experiments were conducted over long time periods, in the order of hundreds of hours, in order to investigate transport rate fluctuations over a wide spectrum of time scales and if any dynamic equilibrium state was reached. Three experiments were altogether performed, each characterized by a different sediment feed rate, in a 16-m long and 60-cm wide tilting flume using moderately-sorted gravel. The bedload transport rates were continuously recorded at the flume outlet during the runs using vertical impact plates. Additionally, the bed and water elevations were measured every ten minutes using ultrasonic probes and a laser-sheet imaging technique both mounted on an automated moving cart. The joint analysis of the topographical and bedload transport measurements demonstrated that sediment waves migrated in a step like motion from pool to pool inducing most recorded pulses. They were thus identified as the primary mode of sediment transport in the alternate bar system. Additionally, these migrating low-relief bedforms were found to cause occasional bar failures which generated particularly large pulses. At the largest fluctuation time scale (about 10 h), bedload pulses were associated with quasi-periodic variations in the global bed volume. This observation suggests that the sediment storage capacity of the bed, for a given bed configuration and external conditions, may govern and set an upper limit to the system fluctuations. The comparison between the experiments showed that the bed responded to the increase in sediment supply by increasing its average slope and/or evolving toward a more braided configuration. In addition, this adjustment of the bed transport capacity was found to be associated with a smoothing of the bedload transport pulsating regime resulting in shorter and more frequent pulses of lower magnitude. In conclusion, this study shed new light on bedload transport in gravel-bed rivers by documenting several of its aspects under controlled conditions. More specifically, it bears experimental evidence of the presence of sediment waves in alternate bar systems, and show how the dynamics of these two types of bedform drive sediment transport and control bedload macro-pulse characteristics in gravel-bed channels.