Are old MP3s worse than new ones?

David Lee

17.10.2023

Translation: Jessica Johnson-Ferguson

The quality of MP3s mainly depends on the bit rate. But the software you use also comes into it. And this has improved over time. I tried out encoders from the last millennium and compared them to the quality available today.

Many people form an opinion on a something once and never change it again. Here’s one of those opinions: if you can’t tell MP3 from uncompressed music, something’s wrong with your ears.

This isn’t true, of course. At least not in such generalised terms. MP3 doesn’t equal MP3. First of all, it depends on the bit rate used. At the highest quality of 320 kilobits per second (kBit/s), MP3 is considered transparent. This means there’s no audible difference to the uncompressed original. Blind tests proved this years if not decades ago. Funnily enough, it’s actually the other way around. Only people with a hearing impairment can hear a difference in the highest MP3 quality.

Where does this belief come from that MP3 is generally acoustically offensive? My guess is it’s rooted in the last millennium. I remember my first experience with MP3 being negative. This was probably due to the low bit rates common at the time. 128 kBit/s was the top notch. And that’s already in the critical range. Anything below that sounds clearly different to uncompressed music.

The interesting thing is that there were also MP3s out there that sounded terrible at 128 kBit/s. This might have been because MP3 encoders used to be lower quality. This combination likely sealed the lousy reputation of MP3 forever.

In search of a bad encoder

To find out if my assumption’s right, I want to create an MP3 both with an old and a new encoder and then compare the two. My new encoder’s a current version of LAME. It’s been the standard for a long time and is considered high quality.

As for the old encoder, I need to see if I can even get one to work. I use a Mac, and Apple doesn’t think much of backwards compatibility. After a few failed software installations, I get my geriatric 2002 Powerbook out of the cellar and use it to create MP3 files using iTunes. Like everything on this device, this takes an eternity.

A female voice says «Encoded by N2MP3» at the beginning of each track created by iTunes. This must be the name of the encoder my old iTunes uses.

Turns out, 2002 must already be too up to date. At least, I can’t make out a clear difference in quality compared to what the new encoder’s produced. That one’s at 128 kBit/s, I can’t select a lower quality.

Next, I take a look at Windows software – after all, Windows runs on my M1 Mac. I soon come across an interesting subject for my experiment: AudioCatalyst 2.1. This CD ripper is from 1999, but still runs on Windows 11. It uses the MP3 encoder Xing, and is perfect for what I have planned. Why? Because Xing is considered very fast, but also unsatisfactory (link in German) in terms of quality.

What I hear – and what I don’t

Whether you can hear a difference also depends on the music. I can make out quality losses best in the high frequencies produced by drums and percussion. Or if there are large volume differences in a track. For this test, I use five pieces that cover pop, rock, funk, jazz and classical music.

To my surprise, Xing generates pretty decent MP3 files at 128 kBit/s. I can’t reliably tell them apart from the files LAME produces. In a blind test, I’m right about the rock and funk tracks, wrong about the classical one and simply can’t hear any difference with the jazz track.

Next, I reduce the quality to 64 kBit/s. Nobody wants to listen to MP3s in this quality, no matter which software produced it. But at least this should allow me to hear differences between the encoders’ results. And yes, I can clearly point them out at this rate. The Xing encoder files sound consistently muffled – Xing simply cuts out the high frequencies at 64 kBit/s to save data. The new encoder has kept some high frequencies. Not that this always sounds better. Cuts need to be made to the data elsewhere to make do with the extremely low bit rate. As a result, the typical MP3 artefacts are more audible.

The differences are so blatantly obvious, I can even show them to you in a YouTube video – in spite of YouTube compressing the audio again. The artefacts are particularly glaring in an MP3 I exported from the GarageBand software.

So this much is clear: even with the same bit rate and the same original file, not all MP3s sound the same. It depends on the encoder. However, things only become really audible at very low bit rates.

Guess which encoder

But this doesn’t answer the initial question. Why did old MP3s sound so rubbish? I have a few old U2 tracks from «unofficial distribution channels» encoded at 128 kbps. For copyright reasons I can’t show them to you, but I swear by my cochlea: they sound abysmal! Which encoder do you think was used to create them?

The command line tool mp3guessenc analyses MP3 files to find out. How reliable is it? Not a clue. But I’ll give it a try nonetheless.

For the U2 tracks in question, mp3guessenc states Xing (old) as the encoder. For the files I created with AudioCatlyst 2.1, mp3guessenc says Xing (new).

1999 is actually considered new by Xing. The software was developed from 1995 onwards and sold to RealNetworks in 1999. This means there’s only one thing for it. I need an encoder that’s even older for my research.

The Xing encoder from 1997

So I download Xing encoder 3.0 from 1997. It’s a small, 83 KB command line tool for Windows. Now you can clearly hear differences. Most clearly in the pop track Look at This. The drums sound duller throughout the whole track compared to the MP3 created with the modern encoder. In addition, it’s also lacking dynamics. Even worse, there’s the odd dropout.

Conclusion: only ancient encoders are bad

An MP3 with the same bit rate sounds different depending on which encoder it was created with. With a 128 kBit/s bit rate, however, I have to go back to 1997 to find an encoder that generates files that clearly sound worse. In the tested encoders from 1999 and 2002, the differences are already so minimal I can’t clearly tell them apart in a blind test. By contrast, at 64 kBit/s, the differences are generally very audible. However, this has no significance in practice, as nobody listens to audio in such lousy quality.

Your choice of encoder may have played a role at the very beginning of the MP3 era. And there could be a number of different explanations for the occasional abysmal quality of MP3 files available on file-sharing networks. Bit rates were generally lower than today. What’s more, MP3s weren’t always created from original CDs. I’ve been known to burn CDs from MP3 files myself. If you rip MP3s from this type of CD, a file that’s already been compressed is compressed again. The same applies to music recorded from internet radio stations.

After this test, I’m still convinced MP3 got its bad reputation in the early days. But the encoders would’ve probably been a minor issue at the time.

David Lee

Senior Editor

David.Lee@digitecgalaxus.ch

My interest in IT and writing landed me in tech journalism early on (2000). I want to know how we can use technology without being used. Outside of the office, I’m a keen musician who makes up for lacking talent with excessive enthusiasm.

These articles might also interest you