angle of attack from 7°-10° increases the pressure level in the 100-Hz frequency band by almost 30 db and that a change from 10°-12° decreases the level in the 1000-Hz band by about 20 db. It appears that a stall-warning system need not necessarily use the FPL spectrum in the entire audio-frequency range. One may generally obtain positive identification of a spectrum at a given air speed by observing the fluctuating-pressure signal in only two one-third octave bands. In the design of a warning system for stall at a wide range of speeds, the effects of flowspeed are also an important consideration. In scaling the FPL spectra, one takes the flow speed into account by referring the fluctuating pressure to the freestream dynamic pressure q = p Uoo/2 and referring the frequency to £/«>// (where / represents a characteristic length dimension). Figure 6 indicates how a spectrum may be expected to change with airspeed and also shows a number of possibilities as to how one might shift a level/frequency detection window of a stall-warning system to account for these speed -associated spectrum shifts. Thus, detection of incipient stall by means of microphones flush-mounted at properly selected locations on the upper surfaces of aircraft wings appears feasible. The greatest and most easily detectable changes in the fluctuating-pressure spectrum sensed by a microphone occur as the flow reattachment line passes over the microphone; therefore, microphones located where reattachment occurs at a critical angle of attack (at a given speed) are particularly well-suited for detecting when the aircraft reaches that critical angle. In designing a stall-detection system, one must also ensure that acoustic noise, such as results primarily from the propulsion system, does not mask the fluctuating-pressure signal used as the basis for stall detection. In many aircraft, the fluctuating-pressure spectrum at reattachment greatly exceeds that due to noise and thus the problem of masking is resolved; however, where this condition does not exist, additional signal processing (e.g., narrowband filtering, correlation, or averaging of signals from two sensors) may be used to extract the desired signal.