Disclaimer: I didn’t include high-end DAPs such as the AK380/240 because whilst borrowing them, I didn’t conceive of this simple metric. But you can be sure that they, too, would suffer the same fate as the superbly performing MS-AK100 and AK Jr.
Finally, while I didn't write it to be click bait, I suspect that this article, like my Fujifilm GX680III article, is.
Clarification: This article has been updated numerous times for grammar, spelling, and to shoehorn in the occasional missing link. I’ve also done away with some of the fedora tipping in the original version of this article.
Prerequisite reading
Prior to reading this article, please check out Lachlan’s critique of 24-bit audio. In it he distills the important bits about Hi-Res bits.
Before we begin, let me be clear: there is nothing wrong with believing that Hi-Res audio sounds better, or that you get a better experience out of it. No, stair steps do not represent discrete sampled signals, and no, you won’t ever hear music at 120dB of dynamic range. But if you receive joy from listening to music recorded or mastered or merely upsampled into Hi-Res audio, good on you.
Let’s get going
Beyond theoretical benchmarks and personal anecdotes, there is no evidence that hi-res 24-bit audio delivers any audible benefit to the ear, especially at human-safe listening volumes. As a semi-objective reviewer, I call out audio gear that fails to perform up to its hi-res marketing promises. Largely, that is the purpose of the audio/data portion of this website. The latest dedicated audio device to fail is FiiO’s M6. The most famous to fail the test is Sony’s NW-ZX2.
Now, if listening to 24-bit audio makes you feel better, or if a particular 24-bit re-master is verifiably better mastered than its 16-bit analogue, have at it. Express to the world how much you appreciate hi-res audio. After all, music should be enjoyed not analysed.
But blanket claims that 24-bit audio is better than 16-bit audio are useless without first defining what you mean by better. Internet fora brim with bold claims, praise, and stupefyingly misleading recommendations punctuated by the insincere axiom: sorry about your wallet. The de facto opinion is that 24-bit audio is better, and that Hi-Res sounds better. Evidence for both assertions is, for lack of a better descriptor, mystical, at best.
For my part, I am sorry. I have failed to consistently remind followers of mine at TouchMyApps, Headfonia, Headfonics, ohm, Head-fi, and now Fauxtaku Lounge, that on paper, 24-bit audio delivers incredible performance. But in actual use, 24-bit music is restricted by too many external variables to be of much, if any, non-marketing or post-recording use.
Again, if you have doubts, watch Lachlan’s and Xiph’s videos.
Then, consider your listening habits. Do you listen to music at low or high volumes? Do you listen to it in quiet or loud environments? And, because duds do exist (I’m looking at you, Sony NW-ZX2), through what equipment do you listen to it?
I’m an IEM guy: Ultrasone IQ, Grado GR10, Noble K10, FitEar MH335DW, and the like, might be in my ears on any day of the week. When Tokyo finally cools to end-of-summer Canadian temperatures (around November) I might even sneak out with a portable headphone on. I really like the Oppo PM-3 and the Audio Technica ES10. No matter the headphone, I set my music to safe listening levels, which I measure at below or around 90dB. I do this because I’m a music lover and I want to keep healthy ears to enjoy music for a long time.
The main theoretical advantages of 24-bit audio over 16-bit audio are dynamic range, noise, stereo separation, and distortion. Typically, each of these is described in the logarithmic, on a decibel scale.
Files with higher bit depth may possess higher dynamic ranges (this is completely dependent on how and where the original instruments were recorded). Theoretically, a 44kHz 16-bit audio file will possess a dynamic range of 96dB; and a 44kHz 24-bit audio file will possess a dynamic range of 133dB. In order for a recorded signal (or music) to reach either theoretical maximum, it must first possess such dynamic range to begin with. And, in order to fulfil the requirements of 44kHz 16-bit audio, an instrument must output and sustain 96dB above the loudest ambient noise in the recording studio. For 44kHz 24-bit audio, that value rises to 133dB. In a perfect universe, it would be possible for instruments to sustain those values above the loudest ambient noise level, but our universe doesn’t do perfect. And we can’t forget about the noise introduced by the recording and mastering equipment, your playback equipment, and your listening environment. Even in a perfect universe, in order to truly experience 16-bit audio, your music would have to output a mean volume level of 96dB, a volume at which the government of Canada would limit your daily maximum exposure to 2 hours, a volume at which, regular, and sustained exposure, is thought to cause hearing loss.
Dyed-in-the-wool 24-bit audio supporters would have to listen to sustained volume levels of theoretically perfect recordings louder than that to notice any difference to red-book 16-bit; or, they would have to listen to theoretically perfect recordings louder than 120dB to notice a difference against the theoretical limits of dithered 16-bit audio. At those levels, you would lose your hearing in short order. You probably already have a certain amount of hearing loss.
Loudness theory as it relates to dynamic range is best discussed elsewhere. For my purposes, I wanted to provide a preamble to a few descriptive RMAA results I recorded some months ago.
RMAA is useful in illustrating qualitatively objective differences from device to device against theoretical limits. It won’t tell you how a device sounds, or even if you will like it. It will show how well a device conforms to the limits of a suite of theoretical tests.
For the following RMAA results, I matched the below devices to the maximum volume at which I am comfortable listening to Aphex Twin’s Selected Ambient Works 85-92 through an iPhone 4s and a pair of Audio Technica ES7 headphones. Because that album is particularly quiet, I had to bump volume up more than I would a newer recording. On the iPhone 4s, that volume was seven steps from maximum. Then, using a calibration signal, I matched the volumes of an iPhone 6, an AK JR, and an MS-AK100 through a Lynx HILO. Getting every device to match in the meters of my ADC was tough. The iPhone 6’s volume scale has the same number of steps as the iPhone 4’s, but at maximum volume, it is louder by several decibels. After ensuring a mean output difference between all devices of no more than 0,5dB, I tested the same 24-bit audio signal through each device.
Note: the following results are unloaded. Volume settings are indicated in the lead column for each device.
16 and 24 bit TARGET files should be on the left. I'm sorry for the error.
As you can see, not even the exemplary MS-AK100 comes close to hitting the 16-bit ceiling. It is also interesting to note that no device gains a definitive upper hand across the board despite massive differences when tested at maximum volume. Unlike recorded music, whose loudness and dynamic range vary with the passage, RMAA tests only for theoretical upper limits of the sampling frequency and bit depth at which the signal was recorded.
At normal listening levels, my target 24-bit audio file (TARGET 24-44), whose potential dynamic range is 133dB, dropped to tested maximum of 90,5dB. This is far below the theoretical dynamic range limits of 16-bit. Even in a perfect world where perfect instruments were recorded in vacuum by exemplary hardware, and then played back through equally perfect devices, 24-bit audio would still have to be enjoyed at volume levels way beyond my threshold for pain. This of course raises the following question: with ears ringing could I even distinguish the difference between 16 24-bit?
End words
24-bit audio is a wonderful marketing tool whose theoretical limits push the development of better and better playback and recording equipment. But it offers no benefits to the listener beyond placebo.
Our brains are wired to believe narratives. It is part of our evolution. But willing patterns of self-deception can be problematic and dangerous. Suffice it to say: your favourite 24-bit music isn’t a record of the theoretical limits of the format. And, it won't be listened to at volumes which demonstrate those limits.
Finally, of particular note, Hi-Res audio players playing back audio at comfortable listening levels offer no demonstrable benefit against a good smartphone, at least on a test bench. I tend to prefer a good high-pass filter, which most smartphones lack at the DAC level, so there is that. But then again, there is bound to be an app that just about approximates one.
Noise-related resources.
1. Noise - Occupational Exposure Limits in Canada
2. Decibel (Loudness) Comparison Chart