Now if only they'd offer Lossless encoded songs and I'd never buy a single CD again.
Despite what some self-proclaimed audiophiles love to maintain, 128kbps AAC is fundamentally indiscernible from dithered 16-bit Linear PCM (CD Digital Audio, aka "Red Book").
I have seen all kinds of flawed so-called A/B comparisons but never any substantial discernible differences from a rigorous double-blind test.
Also, it helps to understand that PCM is not really a perceptual coding schema like AAC. The bitrate requirements to achieve equal fidelity are determined by the efficiency of the encoding algorithm. AES has stated that 128kbps AAC is in fact indistinguishable from 16-bit dithered LPCM.
There are various aspects of digital encoding which can reduce bitrate requirements. The first of these is, of course, filtering out frequencies above the Nyquist limit that might otherwise result in aliased frequencies distinctly off from the fundamental. Quantization interval throttling and recording of only the change in amplitude from one sample to the next, instead of the absolute value, as in an ADPCM system also further reduce bitrate requirements without any perceptible loss whatsoever. Dithering is yet another technique, specifically used with lower rate bitstreams (e.g. 16-bit CD Audio) to add an iimperceptible level of noise that carries enough voltage oscillation to correct quantization interval errors.
I would say there is a marked and perceptible difference between AAC and 24-bit Linear PCM, even between 16-bit PCM and 24-bit PCM. However, it should be noted that the increase in amplitude quantization intervals between 16-bit PCM and 24-bit PCM is exponential... 65,536 possible values per sample at 16-bit versus 16,777,216 possible values per sample at 24-bit. The result is a tremendous increase in dynamic range from 96.3dB to 144dB! Most listeners will really have to squint their ears just to perceive a difference in AAC vs. 16-bit PCM. However, as I can tell you from trying to tell if I can see the hair-thin scratches on my lenses.
The more you strain your senses, at some point you have to ask yourself if it's not the squinting that is straining me and causing you to perceive things that aren't really there.
In the case of 16 vs. 24 bit it's a glaring difference on an order of magnitude many times larger than the difference between 128kbps AAC and 16-bit PCM
*... and even then some folks still cannot tell the difference between 16 and 24!
Note that the dynamic range of AC-3, which as a perceptual coding schema is a direct predecessor to AAC (co-developed by Dolby), is about 103-105dB, assuming 192kbps for stereo and 448kbps for 5.1 surround. This is quite a feat and a good illustration of how perceptual coding schemas can efficiently reduce the data requirements to produce perceptual transparency.
* 16-bit stereo PCM at 44.1kHz sampling has a rate of 1.411 Mbps whereas 24-bit stereo PCM at 48kHz sampling has a rate of 2.306 Mbps... The difference between the two (note I'm comparing identical encoding methodologies, not AAC and PCM where bitrate comparisons are not apples to apples) is greater than the bandwidth of the entire 16-bit stream!