The following files are WAV files that contain weak signals in white noise.
The audio format is 16 bit mono at a sample rate of 44100 Hz.
The audio signal consists of a white noise background and a pulsed sine
wave signal. The sine wave has a frequency of 800 Hz and is pulsed at
2 Hz with 50% duty cycle. Each file has a duration of 10 s.
The white noise background has a level of 36 QU / sqrt(Hz), where QU
is one quantisation unit of the digitized signal. The amplitude of the sine
wave signal (during the periods where it is on) was adjusted to be equal
to the noise power within varoius bandwidths. The bandwidth in Hz is
indicated in the file name. For example, in the file bw1000.wav, the
sine wave has an RMS amplitude of 36 QU * sqrt(1000) = 1138.42 QU,
which corresponds to a maximum of 1609.97 QU. If this signal was filtered
with a noise bandwith of 1000 Hz, an SNR of 0 dB would result.

You can check how well your brain filters the signal out of the noise by
listening to the following WAV files:

bw50.wav
bw75.wav
bw100.wav
bw150.wav
bw200.wav
bw300.wav
bw500.wav
bw1000.wav
bw1500.wav
bw2000.wav
bw3000.wav
bw5000.wav

A nice tool that can be used to generate such data is R. You can obtain it
from www.r-project.org. Once you have R installed, you can use the
script weaksig.R to create WAV files like those above. After starting R,
just type

   source('weaksig.R')
   doit(1000)

to create a file like the bw1000.wav above.