When you're putting together a music or audio collection, you'll inevitably face the question: what format should you use to store your recordings? This choice affects not only sound quality, but also disk space and compatibility with your devices. Making the right choice isn't easy, because each option has its own strengths and weaknesses.
In this article, we will look at three popular formats: the good old MP3, the more modern AAC, and the lesser-known but interesting WavPack. We will not delve into technical details, but will simply and clearly explain which situations each format is suitable for.
Ultimately, there is no such thing as a perfect format for all occasions. It all depends on what is more important to you: maximum quality, space savings, or versatility. Let's figure out together which format will be the best helper for your specific tasks.
MP3: avalanche-like distortions
Let's start with the most popular format. MP3 is a monster from the Fraunhofer Institute that has taken over the world. Because of it, no one today would think of using pure WAV to record sound. Even if you rip mangled audio from YouTube, it is still converted to MP3, and at a lousy bitrate of 128 kbps. We won't do that, and for the test we'll use the latest version of the LAME 3.100 encoder with the insane preset and a bitrate of 320 kbps.
In the very first figure, it was clear that the spectrum in MP3, as expected, experiences fluctuations in the high-frequency range and is finally filtered out at the 20 kHz boundary. Of course, this is the limit of the synthetic test; in a real music signal, it will certainly be even lower. The dynamic range in the MP3 file has not changed compared to the original. That is, the LAME 3.100 encoder at a bitrate of 320 kbps does not add any noise of its own to the recording.
Distortion of the 1 kHz signal shape when encoding in MP3 compared to the original WAV
Conversion of a single 1 kHz signal to MP3 showed the appearance of many small harmonic distortions. And although formally their share is small (0.0009%) — i.e., one and a half to two times less than in the output of a good DAC — in the dynamic spectrum of a real phonogram, their number will grow in an avalanche-like and unpredictable manner. Also, the “thickening” of the base of the narrow 1 kHz peak in the original indicates certain problems, the accumulation of parasitic oscillations. This feature is clearly illustrated by the “square” 100 Hz wave after its conversion to MP3. As you can see, its contour loses clarity on the horizontal axis. All this ultimately has a negative effect on hearing fatigue when listening to MP3, alas, even at the highest bitrates.
A 100 Hz square wave after conversion to MP3 (top) and AAC (bottom)
AAC: increase noise but maintain clarity
The AAC algorithm, which Apple actively uses, works in a more accurate way, and it is not the only one. Digital TV broadcasters work with this audio codec, and AAC is also included in the MPEG-4 container package.
After conversion to AAC, the square wave retains its shape, although distortion of the base and harmonics around the 1 kHz peak also occurred, albeit less noticeably than in MP3. At the same time, AAC demonstrates a 1 dB higher measured noise level. What could this mean—intermediate recording on tape, perhaps? No, surely the AAC algorithm uses something like noise shaping — a great invention that reduces quantization errors by mixing in a pseudo-random noise signal. Again, this is not just drowning out distortions below the noise threshold — more sophisticated mathematics is involved here.
To illustrate this, let's look at the artifacts around the so-called jitter test at a frequency of 11.025 kHz. Why this frequency? Because the harmonic multiple of this peak falls exactly on the upper limit of the spectrum from a digital stream with a sampling rate of 44 kHz, and all the others will be outside its limits. Parasitic small peaks, especially those that are symmetrical to the fundamental tone (modulation products, “sidebands”) — these are the grains of jitter.
Stability in the jitter test AAC (top) and MP3 (bottom)
As we can see, the MP3 codec maintained a low noise level but generated more high-frequency jitter (most noticeable to the ear), while AAC slightly increased the noise but avoided interference in other parts of the spectrum. But the WavPack codec performs even greater feats with noise shaping.
If you need to compress audio without losing quality, WavPack is the best choice. It preserves all the detail of the original, just like WAV, but takes up less space. If file size is more important than perfect quality, MP3 and AAC are better choices: AAC generally sounds better at the same speed, especially at low bitrates, and works well on phones and players. Ultimately, what you choose depends on what is more important: quality, size, or device compatibility.
WavPack: keep the frequency, change the bit depth
In general, to put it briefly, WavPack's encoding math is currently one of the most flexible and impressive protocols for audio enthusiasts, no joke. Unlike FLAC, it supports 32-bit arithmetic (I recommended it for creating lossless vinyl rips). Moreover, you can even pack a DSD file into WavPack without converting it to PCM. The size of such a file will be much smaller than the original dsf file. But we'll talk about lossless WavPack some other time. For now, let's look at the unique principle of the lossy WavPack codec.
In one of my reviews, I showed that in some cases, when compressing with loss, it makes sense to reduce not the sampling frequency, but the bit depth of the signal itself (i.e., below 24 or 16 bits), carefully mixing in dither (i.e., a special noise profile to reduce quantization errors). WavPack has taken this glorious path, leaving the sampling rate and frequency untouched, but changing the bit depth, which is now a dynamic value describing the signal volume level. It's somewhat reminiscent of the DSD principle, isn't it?
It is noteworthy that when converting to such a lossy WavPack, you can additionally save a parallel “correction” file, which can be used to completely restore the original, down to the last bit. However, you won't save any disk space in this case, since the size of such a pair will still correspond to the lossless original. Nevertheless, the functionality of the protocol is still impressive.
The bitrate of our test file was set to 320 kbps to match the maximum of our MP3 and AAC files, but theoretically, it can be set even higher in WavPack. This is especially useful for high-resolution files. I will show you how below — and you won't need MQA anymore!
Comparison of harmonic distortion in the original (green) and the file encoded in lossy WavPack (white). The noise level has increased slightly and there is almost no distortion.
The speaker diaphragms are made of time-tested materials: impregnated fabric (tweeter), Kevlar (midrange), and cellulose (woofers). The head sizes are 25, 130, and 2 x 210 mm (1, 5, and 8 inches), respectively. The wide tweeter suspension smooths the frequency response, and the waveguide in the flange slightly corrects the off-axis frequency response and directivity.
Everything here is done in the classic style. The slender rectangular prism-shaped body is detached from the floor by four height-adjustable steel cones. Three frequency bands are handled by four Davis Acoustics speakers, one each for the upper and middle ranges, and two for the lower range. The low-frequency design is a phase inverter; the port is located at the front, at the bottom of the front panel. The design is clearly mainstream.
The series is named after the French painter of the last century, Balthus, who belonged to the circle of Cocteau, Picasso, and other greats. The top model in the line, the Balthus 90, is the largest and most powerful of the sisters.
The song “White Wood” by Lush will help you evaluate the degradation of the high frequency range. Someone on the forum joked about my loyalty to shoegaze, but it is precisely this kind of fragile, yet spectrally rich fabric that is very sensitive to the bulldozer of lossy codecs and crappy equipment. If you give in even a little, all the guitar choruses with their coppery plates will turn into tinny garbage. The second track, “Deep Sleep” by the B-52's, is an example of adequate mastering with a rich sound, studio effects, and a preserved dynamic range.
As you can see, there is no universal winner in this choice. It all depends on what is more important to you in a particular situation.
If you need maximum compatibility with any device and are willing to sacrifice slightly higher quality for a smaller file size, MP3 is the right choice for you. For most everyday tasks, such as listening to music with headphones or storing your personal collection, its capabilities are more than sufficient.
When file size is critical, but you want to maintain better sound quality than MP3, consider AAC. This format has become the modern standard for streaming services and mobile devices, offering an excellent balance.
If you are an audiophile, archive your precious recordings, or work with sound professionally, then WavPack is the option for you. It provides complete, bit-for-bit preservation of the original, while saving space without any loss of data. This is the choice for those who value accuracy above all else.
Ultimately, the best format is the one that perfectly suits your goals, equipment, and ears. Don't be afraid to experiment and compare the sound yourself.
