Distortion in Digital Audio Players

by Bill Machrone
Copyright © 2005 Bill Machrone

This Web page is a companion to my column, "Getting an Earful" in PC Magazine.

I've always looked at harmonic distortion when I've reviewed players, and it's very rare that a player doesn't make the grade. But I was only testing players with the EQ off or set to flat. Imagine my shock when I looked at a player that I'd just tested, and found a huge amount of harmonic distortion where none existed just minutes ago. The difference? I'd selected one of the EQ settings.

The player in question, an iRiver H10, comes from a company that has prided itself on audio quality, so I was surprised to see the distortion.

The distortion shown on the screens captures below is clearly audible. In some cases it sounds like you've stuffed cotton in your ears--the music is muffled and indistinct. In other cases it sounds artificially warm. A preponderance of even harmonics can do that, because they're "in tune" with the notes in the music.

Note: Since I first published the column and these graphs, iRiver released new firmware for the H10. I downloaded and installed it, and re-ran the distortion tests. As you can see from the additional graphs below, the new firmware solves the distortion problems.

	This octave band sweep screen shows the frequency response of the iRiver H10 in 1/3 octave increments. The peak-hold line at the top of the screen shows that frequency response is flat above 50Hz and only down a couple of dB at 40Hz. The test tone rises continually in frequency and repeats; only one column is active at a time, but the software smooths the response, so the trailing bars decay slowly. The bars in front of the peak wave are partially second-harmonic distortion and partially an artifact of the smoothing algorithm.
	The spectrogram of the same sweep tone shows the intensity of the sound at each frequency as a color, keyed to the vertical bar on the right. An ideal spectrogram has minimal fringing, although the fringing at low frequencies is normal. This central portion of the sweep shows that the H10 has very clean frequency response, with very low harmonic distortion.
	This is the same sweep, with one small difference--I selected one of the H10's EQ presets. Each parallel bar is another harmonic. You can see how intense the first through fourth harmonics are--definitely strong enough to be heard. The cathedral-like traces at the high end of the sweep are clearly audible as swirling tones that sound like tuning an old AM radio. In my column I attributed them to intermodulation distortion among the harmonics, but they're also the sound you get when a transistor is overdriven and "hits the wall." Since this pattern was clearly visible at all volume levels, the problem was not that the audio output stage was being overdriven, but either a DSP programming error or a badly overdriven transistor before the output stage.
	NEW: With the new firmware in place, the EQ does a proper job of boosting the bass, in this case, without introducing any additional harmonics. The other settings (all 30 of them) were equally clean. The new firmware solves a number of other problems, but the audio distortion was by far the worst of them. I would have warned people away from the H10, but now its audio performance is in keeping with its other high-quality attributes.
	The same sweep again of the H10's response with the EQ on, in yet another spectral view. This screen is expecting test tones to be of constant intensity and duration, but I was sweeping the frequency logarithmically, so it appears to tail off slowly through the range. More importantly, the harmonics are clearly visible. At this particular frequency, the right and left channels have diverged in intensity. At other frequencies, they match. Basically, the audio processing with EQ on is a mess. The H10 exhibited this kind of behavior on most of its EQ presets; it also appeared when you moved the graphic equalizer sliders more than a couple of notches.
	For comparison, here's the new iPod mini 6GB, which shows a nice, flat frequency response curve. Although it's still a bit weak in low bass, it's at least as good as other hard disk iPods, and noticeably better than the first-generation iPod mini. The question is, can you boost the player with its EQ presets to overcome the shortage of low bass without destroying the quality of the music.
	Here's the iPod mini 6GB in spectrogram mode. You can see a tiny amount of harmonic distortion; this amount is inaudible and insignificant. The player's total harmonic distortion is still less than 0.01 percent.
	Now let's turn on the bass boost. Notice how the red core extends down farther, and the bass region is fatter. There's a tiny bit more harmonic distortion, but it's at least 50dB lower than the signal, limited to a narrow range, and, again, insignificant. This is the way tone controls and EQ settings are supposed to work: boost the frequencies in question; leave the rest of the audio spectrum alone. The iPod had this excellent performance on every EQ setting, showing proper attention to the signal processing.