Notes by ff123
Here's another music clip, this time with high-impact snare drums (The Commitments, "Mustang Sally," MCAD-10286). I applied a brick-wall filter at various frequencies, then listened to find out where I stopped hearing differences from the original. I found that while I could hear a difference with the 11 kHz filter, I couldn't with the 12 kHz filter. Although I can hear tone sweep frequencies up to 16 kHz (see this page), the result here suggests that in real life musical situations, I probably won't detect the filtering used by mp3 encoders at 128kbs (which start operating at about 14 kHz).
Addendum (11-26-00): I can hear a difference between the 12 kHz filtered clip and the original if I listen carefully through high-quality speakers. I brought the test clip into my local hi-fi store, and the salesperson immediately said: "It sounds like the highs have been filtered," which of course they had been. I had to listen twice to notice a difference, and I didn't hear one at 14kHz, although the salesperson, who looked to be approaching 50, could tell the difference between the original with16kHz filtering. He says he can hear up to 18kHz, and I believe him.
I bought a pair of Grado SR325's, which has a brighter lower-treble than my Beyerdynamic DT990 Pro's. However, I am still not able to distinguish between the original and the 12kHz-filtered clip.
Download the various filtered files (filtered in 1 kHz increments from 10 to 19 kHz) for your own listening pleasure, available on my Audio Samples Page. They have been compressed using the lossless compressor, FLAC.
Here's what the 10kHz filtered spectrum looks like, compared with the original.
| Filter | Spectrum |
| 10kHz | |
| none | |
Here's what other people heard:
dculton: I stopped hearing a difference at 12-13khz (hardly heard a difference at lower), using headphones, and could hear no difference over my computer speakers. That explains why i`ve never been to concerned about the frequency cutoffs of an encoder (never noticed them) as much as the artifacts some make. Thats why i use EAC/Lame for it`s consistent quality, i use cbr @ 128kbps almost exclusively because i have a Rio500 and rarely ever hear any artifacts. I`ve heard that fraunhofers encoders sound better at this bitrate but i`m unwilling to pay for them to find out since i`m happy with lame. Plus now the speed isnt much of an issue to me the combination of 3.87mmx and my new 700mhz Duron processor (built me a new comp last week
) it`s now plenty fast for me, now vbr takes as long as cbr used to take on my other computer, so i will probably start using it more now.
Don
HansHeijden (original post is on the MP3.com > Message Boards): I hear the difference in brightness between the 15 and 18 kHz mustang files. Concentrate on the snare drum. I'm getting a cold now and by tomorrow I won't hear it any more. You should find some others for listening tests, artifacts are often in the high region..
