The term baryon acoustic oscillations (BAO) refers to sound (acoustic) waves that traveled across the early universe between the times of inflation and recombination. The term is also freely used to indicate the imprint that these waves left on the universe. BAOs are of much interest, as the imprints still visible help confirm cosmological theories, and also represent a standard ruler, a phenomenon that can be viewed at cosmological distances with a characteristic length: relating the length to the angle of separation on the celestial sphere, offering a means of determining how far away the instance of the phenomenon is.
The waves arose because the end of inflation left the universe with some dense regions (the outgrowth of the initial fluctuations) as well as a plasma of protons, electrons, and photons that was out of equilibrium (not in hydrostatic equilibrium), the gravity of dense regions pulling material toward them, and the plasma's inertia and the springiness of its pressure caused them to rebound, a portion of the plasma in the shape of a sphere growing smaller, then larger. I presume this pulsing repeated with some damping, each time setting off an outward sound wave, i.e., acting like a musical instrument or loudspeaker. The initial wave, the strongest, had the time from inflation to recombination to travel. At recombination, the plasma changed phase, the charged particles joining to become neutral particles, distinctly changing the pressure and tension forces that were the basis for the outwardly-moving waves (the resulting sharp drop in opacity made the radiation-pressure-component shrink to insignificance) and the waves froze in place, leaving dense regions in the shape of spherical shells, which often had traveled far enough to overlap with others. These somewhat-dense regions show a very slightly higher temperature, contributing to the CMB anisotropies observable today, and this presence of more baryons triggered the growth of more galaxies at these locations, leading to patterns in the positions of galaxies, as well as in the effect of galaxies on the CMB. The view of recombination that we see via the CMB (the spherical shell-shape around us just distant enough that the CMB EMR is just reaching us now is termed the surface of last scattering) shows circles of higher temperature from the higher density anywhere that surface coincided with the freezing of an acoustic wave. Given the nature of the plasma and the time between inflation and recombination, the waves are computed to have traveled about 130 kpc (the sound horizon), forming spheres of that radius. Hubble expansion from z = 1100 has expanded that to about 150 Mpc now. There are far too many such circles overlapping to be directly discernible, but they should leave statistical signs in the CMB anisotropies (one of the reasons there is so much interest in studying it) and also in the placement of galaxies, discernible in surveys such as SDSS (In both cases, researchers have announced they have identified the signs). The term acoustic scale refers to this distance or to the angular acoustic scale, the angular distance corresponding to this distance. The high-density locations and resulting galaxies have moved since then, but that movement puts them on the order of 5-10 Mpc from where they would otherwise be. Analysis of current peculiar velocities evident from observation (and of the dynamics that must have taken place) can compensate somewhat for this.
Of course, the above description presumes that the density peak that initiated the wave was (roughly) on the surface of last scattering so the entire 130 kpc is transverse (from our point of view). BAOs from density peaks a bit in front or behind the surface of last scattering will produce "smaller circles in the sky". My best guess is the largest (130 kpc) circles have the most extra density right on surface of last scattering, for similar reasons to spherical structures such as supernova remnants and planetary nebulae appearing as circles rather than simple disks. But such smaller circles would contribute to a smoothing effect and be part of the reason why statistical analysis is required to discern the imprint.