The noise floor also affects the two measurements I usually take to assess a digital component's DAC linearity error: the amplitude error as a dithered 500Hz tone fades to zero (fig.5), and the waveform of an undithered 1kHz tone at exactly -90.31dBFS (fig.6). But even with the noise overlaying the plots, the iPod's behavior suggested good DAC performance on these tests, with linearity error remaining below 2dB down to -110dBFS on the former, and the three voltage levels described by the latter clearly visible.

Fig.5 Apple iPod, right-channel departure from linearity, 16-bit AIF data (2dB/vertical div.).

Fig.6 Apple iPod, waveform of undithered 1kHz sinewave at -90.31dBFS, 16-bit AIF data.

At low frequencies and high levels, the iPod's distortion spectrum featured primarily second and third harmonics (fig.7), with other harmonics below -90dB (0.003%). The second was the highest in level, at 0.1% (-60dB). The overall distortion level was lower at higher frequencies, with the second harmonic—still the highest—at -73dB (0.022%) into a 100 ohm load (fig.8), though the upper harmonics are still visible. However, intermodulation distortion is low in level, even at high playback levels into 100 ohms (fig.9).

Fig.7 Apple iPod, spectrum of 50Hz sinewave, DC-1kHz, at -3dBFS into 150 ohms, 16-bit AIF data (linear frequency scale).

Fig.8 Apple iPod, spectrum of 1kHz sinewave, DC-10kHz, at -3dBFS into 100 ohms, 16-bit AIF data (linear frequency scale).

Fig.9 Apple iPod, HF intermodulation spectrum, DC-25kHz, 19+20kHz at 0dBFS into 8k ohms, 16-bit AIF data (linear frequency scale).

Finally, I used the Miller Audio Research Jitter Analyzer to look for word-clock jitter-related spuriae in the iPod's analog output signal. The diagnostic signal, as usual, was a high-level sinewave tone at one quarter the sample rate, over which has been superimposed an LSB-amplitude squarewave at approximately 229Hz. Both signal frequencies are exact integer fractions of the sample rate, so the signal is free from quantizing artifacts. Any spuriae that appear in a player's output are therefore a result of something it is doing wrong.

The Miller Analyzer performs a narrowband spectral analysis of a player's output signal, then searches for symmetrical sideband pairs around the 11.025kHz fundamental. The result for the iPod is shown graphically in fig.10: while the noise floor is around 6dB higher than is theoretically possible from a 16-bit system, only a few sidebands can be seen above the noise. Those circled in red and/or indicated with red numbers are spaced at multiples of 229Hz on either side of the central peak and are therefore data-related. Those indicated with purple circles and numbers are due to other, unknown sources of jitter. The overall result is superbly low, at 225 picoseconds peak-peak.

Fig.10 Apple iPod, high-resolution jitter spectrum of analog output signal, 16-bit AIF data (11.025kHz at -6dBFS sampled at 44.1kHz with LSB toggled at 229Hz). Center frequency of trace, 11.025kHz; frequency range, ±3.5kHz.

The iPod's measured behavior is better than many CD players—ironic, considering that most of the time it will be used to play MP3 and AAC files, which will not immediately benefit from such good performance. But if you're willing to trade off maximum playing time against the ability to play uncompressed AIFF or WAV files, the iPod will do an excellent job of decoding them. Excellent, cost-effective audio engineering from an unexpected source.—John Atkinson

<< page 1