Before you seed a show you may want to take a look at some of the wav files to make sure that what you see is consistent with the type of source you are supposed to have. In particular, you're looking to make sure that the version of the show you have wasn't put through lossy compression (such as mp3, mp4, or aac) at some point in its lineage; and that the transfer was done properly. (So this is worth doing even if you taped the show yourself, or got it from the taper - a bad transfer can do as much damage as lossy compression.)
Introduction. Lots of people get carried away with looking at frequency and spectral analysis. But evaluating a source is more an art than a science - there's only so much this type of analysis can tell you.
And while you can sometimes prove a version of a show is lossy, you can't really prove that it isn't. So,
looking at these images is no substitute for knowing the lineage of your source.
Lots of factors can affect the spectrum and frequency profile, including mics, transfers, tapes and other source issues. A source may also have
[clipping] (oversaturation of the recording level at the show). This variation in show sources is why True Audio / Aucdtect isn't usable on live shows - rather, those kinds of programs are designed to identify pirate copies of commercial CDs. For example, see this
[hydrogenaudio thread] and this
[hydrogenaudio thread] on the topic, particularly the comments by M about older shows in the second thread. So there's no magic program that's going to substitute for human judgement.
Also, remember that looking at these images tells you
nothing meaningful about how good a show sounds. For example, sometimes MD tapers get a better capture than DAT tapers. Lots of things besides the recording device affect whether a show sounds good - location, mics, and luck for example. The only way to evaluate sound quality is to listen to the tape.
Analyzing a source. Most audio editors, as well as EAC, will let you look at a spectral view and a frequency analysis of the audio data. For an introduction, see
[Detecting MP3 Sourced Audio].
A few notes in addition to what's at the above site - it's useful to
- Look at both the frequency analysis and the spectral view. While the frequency analysis may give the first indication of whether there is a problem, ultimately the spectral view is the more useful of the two.
- Make sure you select a section of music - not speech, applause, etc. - when doing either frequency or spectral analysis. This of course means you don't want to analyze a whole track, but only a section of a track.
- Select a few short ranges of music (30 seconds, 10 seconds, 1 second) and for each one look at the frequency analysis. Compressed audio often sharply reduces the frequencies in the higher ranges, so you may see a dropoff at 12-18khz. But, don't expect to see this dropoff at the same level - it varies with different encodes and encoders, and you might not see a dropoff at all in some cases.
- Look at a few different time spans in the spectral view - 30 secs, 5 seconds, 1 second. Use the controls in the audio editor to select and zoom in on that range.
- In the spectral view, look for squared off or unnatural shapes or cutoffs, rather than the random look that uncompressed audio has. (One note on FM sources: broadcast technology cuts out the frequencies above 15-16khz, but you should see no sign of compression below that.)
- Try to look at a couple tracks from each disc of a show, if you have any reason to think they are from different sources. It's not at all unusual. And if there are tracks from more than one show or date, check some of each.
But this kind of analysis is only half the story, and is hard to fully interpret without some idea of the kind of source you're supposed to have. Find out what you can about the known sources for a show and which one you're supposed to have. You'd expect to see very different results for different sources. At a bare minimum, knowing the date of the show can provide some clues.
The real question here is whether the analysis you see matches with what you'd expect to see for that type of source.
Microgaps. Audio editors are also handy to look for signs of mp3 source, and to spot sector boundary errors that have gotten embedded in the wav files by a conversion from audio editor > wav > cdr > wav.
- Zoom way in on the beginning and ends of the tracks - you're looking for silences of 0.01 to 0.03 seconds.
- Look for digital silence, aka the "dreaded flatline." On a live show, you should typically see no digital silence.
If you hear audible microgaps between the tracks of a show, mp3 source or sector boundary errors are the likely explanation, with mp3 source being the most likely.
Some specifics for Audacity:
[Audacity] is an open-source audio editor; it will do both spectrum and frequency analysis on Mac, Windows or 'nix.
You need to decode a track or two to wav first, using the shn or flac programs. Open a wav file with File, Open.
To see the spectral view, go to the Track menu (pulldown above the Mute and Solo buttons) and choose Spectrum.
To see the frequency analysis, you have to first go to the Track menu and choose Split Stereo Track. Then select a musical section (not applause, silence, etc.) of one channel by clicking and dragging the mouse. Using the top menu, select View, Plot Spectrum to get a basic plot.