Visual comparison of lossy audio encoders is utterly pointless. You aren't "seeing" the audio, you are hearing it. Psychoacoustic phenomena also don't translate to the visual realm in a meaningful sense. You may be able to see something on a graph, but in no way be able to hear it because of various masking effects or the like. Another issue is that usually you don't view the graphs at a high enough resolution to see common encoding artifacts, and what's more, most people don't even know what they would look like. Freq analysis is totally useless. Spectrogram analysis
can be useful, but only if you are looking for something you have already heard. For example, it's good for pinpointing what a particular artifact you are hearing is doing to the audio. It's mostly useless for comparison.
These kinds of tests are very misleading and are the single biggest reason why so many people have massive misconceptions about the way these encoders are designed to work. They see these types of graphs on sites like this and *ahem* "
other" ones and then go off talking about the "frequency response" of various encoders and other nonsense.