Please do not adjust the level so that the first tones are "normal listening volume". Start with REALLY LOW volumes, because the last tone is really loud, even though you don't hear it as loud.
If you crank it up, the last 19 kHz tone might not only destroy your tweeters (if you use loudspeakers, something which I definitely recommend AGAINST), or it could damage your hearing.
The 19kHz sound is LOUD!! We just don't hear it, because the frequency is so high and our hearing sensitivity drops towards higher frequencies. Just because you don't hear it as loud, doesn't mean it can't cause damage!
Damage is related to playback intensity levels (dB) not perceived loudness (sone).
So, don't be fooled even if the last 19 kHz doesn't "sound like" it's loud. Believe me, it is VERY LOUD! Loud enough to damage your hearing, if you play it too loud.
So, be careful: don't crank it up and don't do many repeats in a row.
I'm not kidding.Additional tidbits
Now that I have hopefully scared you, here are some additional tidbits:
0) You should hear PURE simple tones. No warping, no "up and down sine-wave like sounds", no ambulance, absolutely no noise, no distortion. Just pure unadultered simple tones (like phone DMTF tones).
1) All sound cards that resample will produce artifacts with this sample. If you can't hear them, it could be your headphones.
2) This sample can also be a useful test for high frequency playback ability of your headphones. I have tested this with HFI-650, HD-600, AKG 271S and Grado SR-60. The best headphone to spot the artifacts from my Audigy 2 is Ultrasone HFI-650. AKG 271S masks the artifacts almost completely. So, test with high-frequency capable, not-rolled-off headphones (in that regard this is a test for headphones as well, if your soundcard resamples/makes artifacts).
3) Many people will be happy and say "Hey, I can hear 19kHz". Some will say, "it's almost more like a feeling than hearing". This latter is a sign you are already playing it at dangerous volume levels. Also, let me just remind that our hearing is not a brick wall filter. I could make some of you hear 21kHz, if I played it back at insane levels. However, that would just permanently destroy part of your hearing and possible even cause you a nose bleed, so it's not very useful
So, to recap: the sensitivity of our hearing diminishes towards higher frequencies at "normal / safe" playback intensity levels. If you can't initially hear it, it's normal. Many 20 year olds can't hear it at normal play back levels, they need to raise the volume (which can be dangerous).
4) When using upsampling/re-quantisation (like with Foobar) you should hear artifacts, although they are more than likely to be diminished from the artifacts you hear with a resampling soundcard with no additional upsampling. So, if you don't hear any artifacts with upsampling, you are missing something. There are still some artifacts, they can be just very hard to notice. But again, don't crank it up and don't play this in a loop many times. I'm not just saying it, I mean it.
5) This sample can also be used to test one part of the quality of the upsampling engine. For example, I've used the Foobar SSRC engine (good), Wavelab (horrible results) and Sound Forge (best I've found) upsampling engines (with different settings, but all upsampling to 96kHz). The differences vary quite a lot and some engines do it much better than others. You could also use this to test your cd-players upsampling, but PLEASE DO NOT use it to play back on your speakers. Especially ribbon tweeters are really not meant for insane playback levels and you can easily kill them. And boy, are they expensive.
6) On Creative cards (sans EMU) you can reduce the distortion by using relatively high quality upsampling (say SSRC in Foobar 2000) combined with Windows/Creative Mixer settings in Windows: set the volume to 75% max or even as low as 50%. Even if you compensate this lowering of software volume by pumping up the headphone output volume, you can get reduced clipping. So, this actually shows that all creative cards (sans EMU) have two problems: they have resampling related aliasing distortion (IMD) and they have also volume control related clipping distortion. You can minimize both by upsampling to 96kHz and by setting volume control to 50%/75% (try both) in Windows.
7) There is another sample, that is again recorded at full volume. It's also very good for testing your sound cards resampling algorithm and clipping problems:
10 second clipping distortion test sample at full amplitude (c. 170-200 Hz tones) - FLAC 528 KBhttp://www21.brinkster.com/roina/cd-check_test_10s.zip
(Note: That is free hosting space with limited download quota. If it gets clogged up, somebody else has to host the file)
This time there are not any super high frequency tones, but the same warnings still aply: don't crank it up, don't repeat listen to it.
With this sample, if you hear anything except steady dual tone (to most it sounds like one steady sound - not rising/falling), then you have problems. Many will hear horrible noise or modulated distortion that is played on top of the original signal, masking it.
8) There is another older thread about this same issue here are Head-fi, but with the search disabled...
That's all I can think of now... Play safe!