Audio encoders are interesting- in order to get the data rate down to relatively few kilobytes a second, the encoder must throw away information. Which information? That's what these analyses will tell you. A background in high end audio concepts is useful but not required- I will try to explain things as they come up. The basics are as follows: Frequency and Time are the things we are measuring. In the visualisations you'll see, if you see a waveform and 'sonogram' (picture of the frequency content of the wave), time goes from left to right, and frequency goes from bottom to top, with high-pitched frequencies at the top of the sonogram. In the charts below the visualisations, there are two little graphs. The one on the left is the extent of frequency response level distortion, and the one on the right is the extent of pre-echo distortion and over-ring. Both are plotted against frequency response- the middle of the graph (sometimes with a little notch) is 1 kilohertz, the left side goes down to about 9 hertz, and the right side goes up to 22 kilohertz. Frequency response level distortion is measured by masking off areas of the sonogram that aren't part of the original sound, suppressing distortion components in the time domain. It's plotted between the minimum amount of any test, and the maximum amount. The distortion can be increased or decreased level for that frequency: if it is extreme, it is probably decreased level, as with encoders that throw away frequencies over 16K completely. Pre-echo distortion and over-ring are measured by masking off the inverse areas of the sonogram- ignoring the errors in direct reproduction and weighting the measurement towards the areas that are supposed to contain no frequency content. Pre-echo (most notable on BladeEnc) is heard as a sort of whooshy blur that veils transient attacks. Over-ring at high frequencies (most notable on high bit rate Fraunhofer) is heard as a metallic, unpleasant edge on the sound. Over-ring at low or mid frequencies is more likely to soften and blur the sound, acting like a cruddy, tiny reverberation. Different encoders have markedly different effects, and there is no single 'right' encoder. The choice depends on the musical content, and the biases of the listener. A fault (such as Blade's severe pre-echo) tends to be balanced out by other strengths (Blade has very extended frequency response, and unusually low coloration). When you have only so many bits to give to the music, it's always going to be a trade-off.   That said, let's jump in and begin looking at the encoders, starting with the most nasty problem- low bit rate encoding at 32 and 64K. This will introduce the distortions and sonic problems we're inspecting, in a serious way- although higher bit rates are not a guarantee of cleaner sonics! Click here to begin.

page created Mon, Oct 23, 2000 last modified Fri, Oct 27, 2000 Send Email back to airwindows The Airwindows Studio High Bit Rate 160K and up- can you get ideal sonics? Low Bit Rate The encoders at low bit rates of 32 and 64K Medium Bit Rate 128K and VBR- mainstream for mp3s