The diffuser is the gradually-expanding passage following the test section (e.g. Fig. 4), in which the flow speed decreases and the pressure rises. The recovery of pressure from kinetic energy reduces the power needed to drive the tunnel: in the case of open-circuit tunnels the diffuser also reduces drafts in the laboratory. The pressure rise is less than that given by Bernoulli's equation, because of losses due to skin friction and resulting growth of boundary-layer displacement thickness.
A "wide-angle" diffuser fitted with screens may precede the settling chamber (Fig. 2): it is intended to produce a rapid expansion in area, without any worthwhile pressure recovery (the dynamic pressure ahead of the settling chamber is so small compared to that in the test section that pressure gains or losses are almost negligible). Although the flow may separate from the walls of the rapid expansion, the extent of separation is limited by the screens, which smooth out velocity variations at the expense of a static pressure drop from one side of the screen to the other. Design rules for wide-angle diffusers are discussed by Mehta and Bradshaw "Design rules for small low speed wind tunnels, Aero. Journal (Royal Aeronautical Society)73, 443 (1979) (the remarks about wide-angle diffusers apply to any size of tunnel).
The usual design rule for subsonic diffusers is that the total included angle of a portion (frustrum) of a circular cone with the same length and area ratio as the diffuser should not exceed 5 deg. This is well below the angle for maximum pressure recovery, which is nearer 10 deg., but at angles of more than about 5 deg. the boundary layer is close enough to separation for the flow to be unsteady. The 5 deg. rule will fail if the test section is unusually long, so that the boundary-layer thickness at entry to the diffuser is unusually large. (We expect the behavior of boundary layers in adverse pressure gradient to depend on some dimensionless parameter such as
(
/UCL)dUCL/dx, which must be small enough to avoid separation. Here and UCL are strictly the local boundary layer thickness and centerline velocity - related by Bernoulli's equation to the pressure gradient, which is what really matters - but the evolution of is determined by its value at the beginning of the diffuser.)
The ideal diffuser shape is probably a gradually decreasing rate of expansion but this is difficult to build, except as a series of straight-walled sections. In general-purpose wind tunnels which may be used for tests of models big enough to disturb the flow in the diffuser, it is safest to keep to a conservative angle from the start.
It is unusual to have an area ratio of more than three at the fan: to an (optimistic) inviscid one-dimensional approximation, this recovers 95 percent of the kinetic energy in the flow in the form of a static-pressure rise. A well-designed fan restores a more uniform velocity profile, and further expansion can begin downstream of it, to match the area to that chosen for the settling chamber rather than to recover a further small amount of kinetic energy. If a further expansion ratio of more than three or four is required, the final stage usually takes place at a rapid expansion with screens (see above).
Supersonic tunnels, in which a diverging diffuser after the test section would produce a further increase in Mach number, are equipped with a second throat at the end of the test section (Fig. 6): the first throat is the one upstream of the test section through which the flow accelerates through the speed of sound. In the converging section leading to the second throat the flow is decelerated to slightly above sonic speed (obeying the one-dimensional inviscid compressible flow equations to a first approximation); in the diverging section downstream of the throat the Mach number rises again, until a shock wave or waves produce a reduction to subsonic speed. It may be shown that a shock wave in the converging portion of the second throat would be unstable, and in practice the second-throat Mach number is chosen large enough for the breakdown shock system to be located well downstream of the throat, to ensure stability under all operating conditions.
When the tunnel is started up, the second throat must be rather larger than the first throat in order for the latter to choke first, so that supersonic flow can be established in the test section. The inviscid-flow equations are not accurate during startup because a strong shock wave passes down the test section, reducing the speed below that of sound and usually causing massive boundary-layer separation. Therefore the second throat has to be significantly larger than the first, and to achieve the best possible supersonic diffusion in tunnels for high supersonic speeds the second throat is closed in after starting (reducing the local Mach number), by adjusting the shape of the walls. This is cost-effective only in large tunnels or if drive power is restricted.
A conventional subsonic diffuser is usually installed after the second throat, even in the case of an open-circuit blowdown or suckdown tunnel. The boundary layers on the tunnel walls will be driven close to separation by interaction with the shock wave(s) in the second throat, and may even separate if the Mach number entering the shock wave is more than about 1.3. Therefore the the subsonic diffuser (or its first leg, preceding the first corner, in the case of a closed-circuit tunnel) should have a smaller angle than the 5 deg. recommended for a low-speed tunnel, or be preceded by a constant-area section to allow the boundary layers to recover.