How arousal thresholds vary with different sounds is a critical issue for emergency awakenings, especially as sleepers are dying in fires despite having a working smoke alarm. Previous research shows that the current high-pitched (3000+ Hz) smoke alarm signal is significantly less effective than an alternative signal, the 520 Hz square wave, in all populations tested. However, as the number of sounds tested has been small further research is needed. Here we measured auditory arousal thresholds (AATs) across signals with a range of characteristics to determine the most effective waking signal. Thirty-nine young adults participated over three nights. In Part A, nine signals were presented in stage 4 sleep with ascending decibel levels. Signals were short beeps in the low- to mid-frequency range with different spectral complexities: square waves, pure tones, whoops and white noise. Part B manipulated temporal patterns, inserting silences of 0, 10 and 21 s after each 12 s of beeps. It was found that the low-frequency (400 and 520 Hz) square waves yielded significantly lower AATs than the alternatives. A trend was found across the three temporal manipulations, with a 10 s intervening silence showing some advantage. These findings support earlier research indicating that the best sound for awakening from deep sleep is a low-frequency square wave. It is argued that the signal with the lowest response threshold when awake may be the same as the most arousing signal when asleep, especially where the sleeper processes the signal as meaningful.