I'm recording messages to be used on a VoIP system. Due to software contraints of the phone system, the finished files must conform to CCITT u-Law, 8.000 kHz, 7kb/sec. Bad for the stereo but good for a phone.
The files play back fine within the quality constraints and the voice sounds ok over a phone, but silence is not silent. It sounds like static, white noise.
I'm including a link. These are perfectly silent as a regular 16bit, 44.1kHz file. The 1.5 secs at the front of the recording are the problem. This lead time was necessary due to the playback function of the phone server.
http://www.wentwoodstudios.com/clients/cfe/AfterHours7bitFrontSpace.wav
What can I do to make the silence really be silence at this quality?
I'm using Pro Tools LE 7.3, bouncing to 16bit, 44.1 and using Microsoft's, Sound Recorder in XP to convert to the finished format.
Any help would be appreciated.
Topic Tags
Comments
Thanks, Bos. I'll give these idea a try. This is the first t
Thanks, Bos. I'll give these idea a try.
This is the first time I've had to do anything with u-law.
There was a bit of sibilance to begin with, but it's been magnified a hundred-fold. I may just have to record it again.
I'll get this noise elimitated first.
Thank you for your technical advice.
(1) Something's gone badly wrong in the u-law encoding. Do a ver
(1) Something's gone badly wrong in the u-law encoding. Do a vertical zoom on the waveform and look at the baseline - you've got less than 20dB DR there.
(2) The sibilance is unpleasant. Did the original recording have so much, or did it creep in from aliaising as part of the down conversion to 8KHz?.
(3) There's too much low-frequency energy in the processed waveform. You need vicious high-pass filtering at 250 - 300Hz for these jobs.
I suggest you go back to the original PCM and re-filter and re-compress it. You need to make sure you are fully normalised before hitting the u-law encoder. You could also do an aliaising check by adding a 5KHz tone at 0dB on the end of the original recording. The tone should be completely eliminated by the down-sampling. If it isn't, or remnants of it come out at 3KHz, fix the down-sampling first before trying to push the speech through.
It could just be that MS Sound Recorder is not the right tool for professional use...