A radio program with various contributors like WINGS receives files saved with various codecs, so in order to avoid multiple audio losses, it's best to convert all files to a lossless format like .WAV for editing, so you don't do "codec chaining" and lose information several times. (My favourite editing software only recognizes 16-bit, 44.1 mHz files, so I convert to that.)
Even codecs that are ostensibly the same - have the same name, but are programmed by different companies or people - can have different aural effects. So, it's good to use your ears to help decide not only the file format but also what computer program to use for conversion.
As an example: When I edit with Soundforge 4.5, for which I bought the mp3 plugin, I prefer to save my shows as 112 kbps mp3s, because I find the 128 kbps mp3s made on that codec sound too hissy. But now that I have Audacity (a free and open-source editing program) with a Lame mp3-type converter installed, I usually drag my edited .wav files into Audacity and export them from there as 128 kbps mp3 files. That does a nice clean job.
Here are the discussion items about codecs that I've copied from the listserve of the GRC (Grassroots Radio Conference):
dateNov 13, 2007 1:36 PM
subjectRe: [grc] Anyone using flac audio codec?
I like to second this view. Every time you import an mp3 into any editor it is converted to an approximation of the original PCM (or wav)file. Every time you save it as an mp3 it is throwing our over 3/4ths of the data, you edit again and you just threw out a different set of 3/4ths of the data. Pretty soon there is not much left of the original signal. After reading about how they make mp3's I was amazed that it
even works.
All editing should be in full PCM format, wav, flac, or whatever apple calls it. This includes segments that are being shared. You can save space by using a lossless compression like Flac (there are others such as Shorten, Monkey's Audio, and WavePack, but I think Flac is the best, at least it's the most used).
Flac is the default PCM format for my Ubuntu box at home, which is handy because it takes up less then half the space. And most of the Linux sound editors use it directly. I just wish there was a plug-in for Sound Forge to give it Flac support so you didn't have to manually convert it.
BTW there are some awesome conversion tools at: http://www.rarewares.org
I like the drop utilities Flacdrop, OggDrop, and LameDrop, it's just a panel on your desktop you drop the file into it, it converts it, and very fast too. Oggdrop even works in Linux under Wine, I haven't tried Flacdrop.
I agree that this is overkill for the final product. Last step should be to convert it to mp3 for distribution.
Any original music recording should always be archived in a LOSSLESS compressed format, such as Flac. And in case you can't tell, I'm one of those DeadHead taper dudes that Scooter mentioned.
As far as lossy formats go it should be noted that ogg files sound better and make a smaller file then mp3's or wma's. But only some players support them.
(I have a song running around in the back of my head, it's called "MX'd by baby Jesus")
Dave Willard
Pseudo-Engineer
Radio Free Moscow
Here's a reply from Brian Shirutski:
dateJan 6, 2008 1:11 PM
subjectRe: [grc] Anyone using flac audio codec?
BBC paper on cascaded codecs (september 2005):
>Broadcasters have experienced significant problems with cascaded
> audio coding in the broadcast chain following the introduction of
> digital transmission. It has been found that cascading different
> codecs can result in an overall degradation in sound that many
> listeners find objectionable. A comprehensive investigation of this
> problem has been conducted by members of the EBU project group B/
> AIM. This paper describes typical cascades of codecs found in radio
> broadcast chains, and aims to identify the most critical
> combinations. The intent is to guide broadcasters in deciding which
> codec combinations should be avoided to maximise sound quality...
>
> ...An extensive, thorough, and time-consuming investigation has been
> conducted by members of the EBU B/AIM project group into cascaded
> audio coding. A model of a broadcast chain consisting of 5 cascaded
> codecs was assumed. From the thousands of possible combinations of
> codecs, a subset of the more likely ones was tested for audio
> performance using objective and subjective methods.
>
> Objective testing using PEAQ was successfully employed to reduce the
> number of combinations to be subjectively tested. The subjective
> testing was performed using the MUSHRA test method, with the subset
> of codec combinations being divided amongst a small number of test
> laboratories. Some codecs were tested by all sites for comparison
> purposes. The results clearly show that the cumulative effect of
> cascaded audio coding can be highly detrimental to audio quality,
> even when each stage in the chain makes only a small reduction in
> quality.
>
> The comparison of objective and subjective results showed a good
> correlation between scores. Caution should be exercised here because
> the scales and descriptive terms associated with the two test
> methods used are quite different. The objective and subjective test
> results were both analysed to try to identify codec performance that
> was significantly better or significantly worse than expected. It
> was found that none of the combinations showed any unusual
> behaviour. This should simplify the selection process for users of
> low bit rate coding - it implies that choosing the best codecs will
> give the best results.
<http://www.bbc.co.uk/rd/pubs/whp/whp-pdf-files/WHP118.pdf>
No comments:
Post a Comment