Syntrillium Backup Forums :: View topic

EVEykel · Location: USA

Ok, I've searched around a bit and haven't found an answer to what is probably an easy question. The other day I was tooling around with some .wav files that I didn't record. I brought one of them into CoolEdit and did a frequency analysis on it and discovered that there is a huge dropoff after 16,000 Hz. In fact, all the other sound files that I analyzed had this characteristic. The ones that I recorded, however, are full up to the top of the frequency analyzer. What is the purpose of this, if any, and what step of the recording process is it done at? My recordings sound pretty decent without this drop, I'm just curious as to it's purpose. Thanks.

kylen · Posted - Fri Jul 25, 2003 10:20 pm

What is the source of the wav files with the 16K drop off ?

CD, mp3, your friends mix, etc.

I've heard of an 18K dropoff from the early harsh days of digital that is probably more or less a fable in this age of digital converters. I'm not a converter guy though so maybe one will walk by pretty soon. Or maybe there's another reason for a 16K slope.

I recently took a look at a CD sold by a band in a bar that had a narrow but deep (about 9db) 13K or 16K notch that looked pretty mechanical - no slope, just notch, so who knows. It didn't sound bad but it sure looked funny in the wave editor!:)

kylen

EVEykel · Location: USA

The wav files that I'm looking at are from CDs that I've bought. I guess it could be something that the computer does when ripping the audio.

MusicConductor · Location: USA

You may be on to something with that last comment.

Real Player and WindowsMedia Player, among other ripper/encoder/library managers, are notorious for working in the background with settings that you've specified and long forgotten -- or never had occasion to examine at all, ever. Many times this includes ripping directly to mp3-encoded files, even if just to temporarily park it on your drive so you can make a copy! What makes it even worse is that some Windows system settings are linked to this as far as ripping quality goes.

All that to say that there is one and only one thing that makes a very sharp shelf at exactly 16KHz, and that is an mp3 file. So I'd venture that whatever's on those CDs either got mashed through the mp3 format, or your computer did it when ripping.

Do you have CEP2? Why not rip directly in CEP and compare the two? That would at least answer one of the questions.

Craig Jackman · Location: Canada

I work with MP3 files all day in radio. The sharp 16k drop off you describe is characteristic of 192kbs MP3 files. They usuall sound pretty good, so if you got them as WAV files you'd be none the wiser, but at one time - somewhere - they were MP3 files.

... unless of course there were originally 32kHz sampling rate files at one time. They'd certainly have a drop off at 16k ...

EVEykel · Location: USA

I guess that this happens with .wma files as well? Thanks for the input. I'm a lot less confused now Smile

MusicConductor · Location: USA

WMA does this, but not quite with the same symptoms.

See, mp3 encoders (at around 100 Kbs and up) almost universally employ a 16KHz shelf unless specifically overriden by the user (CEP's excellent encoder allows manual higher settings -- but often noticeably at the expense of overall sound quality). High bitrate mp3s will often retain a 16Khz shelf above which only the loudest high frequency sounds are retained.

WMA also uses a low-pass shelving, but where that lies varies greatly with the bitrate of the file. Furthermore, one of WMA's selling points is better frequency resolution, so at higher bitrates there's no obvious shelf at all (this is done somewhat at the expense of temporal accuracy). At medium and low bitrates a shelf might be visible, but what I think you're more likely to see at any bitrate is a spectral view with lots of lower amplitude high freqs "carved out" and eliminated altogether -- sort of like having run noise reduction on it. Only above a certain threshold do the high sounds get passed through.

I'd recommend taking a couple of short clips, saving it in various bitrates of mp3 and WMA, and then re-open the compressed file and have a look in Spectral View. For anyone whose never done this and compared it with the original, I promise you it will be educational!