Originally Posted by arnyk
OK, here's my first shot at an IM test.
There are two files in the zip file - both 2496, both containing the usual keys jangling sound, but at the end are 4 seconds of test tones.
(2) Now move on to listening to just the last 4 seconds of the files. The contents of the last 4 seconds of the file should sound like a -30 dB 4 KHz tone (not at all ear splitting), followed by a click, 1 second of silence, a click, 1 second of silence, a click, 1 second of silence, a click and end, whether you listen to the 2496 or the 1644 version. Please report what you hear for both files. Do not change your volume control between the keys jangling ABX listening and an the test tone listening. You do not need to ABX the test tone segments, just do sighted listening and report your subjective impressions.
OK, I am finally home and here are my results. Remember though, if anything I say is wrong, I will be blaming it on the dogs!
96 Khz: it is as you say.
44 KHz: All there except the last click is exceptionally quiet. I had to significantly turn up the volume to hear it.
Let's investigate the difference. Here is the last part of the 96 Khz file:
Look where the red vertical line is, and to the right of that you see the last audio sample represented by a dot. Eyeballing it is at -4 db. Remembering that 0 db is full amplitude, this is way, way up there as far as amplitude. Since there is no other sample after that, you go from -4 db to -infinity (or the noise floor of the system) which creates an absolutely sharp vertical line. This is a classic "impulse" which means it has infinite bandwidth (limited by the sampling rate) and will sound like a click or pop as Arny mentions.
Now let's look at the down sampled 16/44:
I have zoomed in more on this waveform but otherwise it is the same as the 96 Khz one.
Again, look at the last audio sample on the right. The value is now is at -50 db. We once again go to silence after that since there are no more audio samples. But there is a big difference here with respect to the 96 Khz file: our impulse/sharp drops from -50 db to 0 as opposed to -4 db. We still get our full spectrum of waveforms as a result of the impulse but the total energy is far, far lower in the 16/44 file. It is like pushing a swing with your little finger versus your whole body. The former will not move the swing nearly as much as the latter would.
At the same level that I was listening to for the 96 Khz file, the click at the end for the 44 Khz one was not audible to me. I had to boost the levels to max to hear it. The above is the reason.
Now, turns out you could hear a much louder pop in the 16/4 than what you should have. When you play an audio file and you get to the end, the audio channel is "closed." What the sound hardware does at that point is "undefined." Ideally you would want it to go to silence as mine did but it may not. It may actually change the electrical (DC) characteristics of the DAC and cause a pop by itself. So if you are hearing a loud pop, that is the reason. It is not due to the two waveforms being similarly situated.
Given the system dependent nature of this last pop, the better test would have been to fade to zero and have both waveforms have that zero value. If a click is desired, it could be inserted in a controlled manner from max amplitude to zero as opposed to where it happened to land in the current test.
Anyway, probably way more information than anyone wanted to read first thing in the morning
As the soup nazi would say, "no IM distortion for you" in regards to my HP laptop.
Thanks as always for your hard work Arny.