Mixing MP3's And Compressed Audio Files


DJ Mixing : What format?

I wrote a list of 20 reasons back in 2013 on why wav's were better than MP3's and Audio Animals reposted it to their website. It was quite a quick blog, and looking back on it I would probably change a few things but the fundamental point was that I wasn't and still am not fond of MP3's.

Recording a DJ mix and getting it right is really important and recording straight to your computer is the quickest and most sensible option. If you start with an MP3, save it as a MP3 and then it's streamed as a MP3, I'm going to explain why this is not good for the audio and why using WAV's or lossless files all the way through the mix to record to end process is vitally the best process.


What file size should I use to start my mix?

So expanding on the question of MP3's and WAV / FLAC file mixing, let's first look at file sizes.

There are 2 types of audio quality. Lossless and lossy.

Lossless formats keep all of the information from source, including WAV, AIFF, FLAC, APE and Apple Lossless (ALAC).

Lossy formats for regular listening, saves a ton of space and if saved at a high enough bitrate, you generally can't hear the difference through your £5.99 headphones. You will start hearing the difference on high end audio equipment. Lossy formats include MP3, AAC and OGG. A MP3 is compressed audio, so loses audio quality.


MP3's use "perceptual coding" to compress audio. I wrote about this in 2013 here however, further to this it's important to understand that the encoder analyzes your music and removes pieces of it that doesn't think you can hear. An example is at 128kbps, the encoder will remove anything above 16kHz as many people can't hear beyond that point, but there is something that I try to remember which is about perception hearing and psycho-acoustics - both are very deep subjects, which goes on to BWE, but my main argument is that even if we can't "hear" lower than 20khZ or higher than 16kHz, we can still feel it through bone induction, body vibes and sound waves bouncing off each other create other sound waves, so it does have a definite effect.


So what format should I use for creating a DJ mix?

Ideally, record in a lossless format, and end in a lossless sound. By doing this, you need to record one of the above, WAV for example, save the mix to a WAV and then it's as close to the original source.


Should I get my mixes mastered?

Ideally, yes. I run my end guest mixes through a limiter and compressor. You can certainly master an MP3 but mastering engineers will request a  wav as lossy formats have issues or AIFF because you can't replace information that has been lost. Therefore, start with a WAV and end with a WAV.


How do I create the cleanest recordings possible?

Avoid distortion, start with lossless audio (WAVs / AIFF) and export in a lossless format.


Should I upload my mix as a MP3?

Websites such as DNBShare or Mixcloud do not give you an option to upload a WAV file. You also have constraints on file sizes or upload limits.  In these instances, make sure you only ever compress the audio once, images at the bottom of this article explain why. Having an MP3 music file and then compressing it as a MP3 creates noise.






When should I not be converting a MP3 down to a MP3?

Remember that a great sounding mix starts with great sounding audio. So when I record my guest mixes for professinal radio shows, I start with lossless wav files. I then master it and save that mix as a lossless WAV recording. I send that mix to the radio station or DJ's or MC's in WAV format to use as a recording to play. If they play it on radio which compresses it down to 320MP3 stream, at least it's started at the highest possible audio point.

The problem that happens with not thinking about the audio you are playing, is when you start with a 320MP3 file, play that in a mix of other 320MP3 files, and then save that mix as a whole 320MP3 lossy file. You are doube encoding an MP3 and raising it twice raises the noise floor. 


So where is the evidence that playing a MP3 then recording it to MP3 reduces quality?

So after a lot of discussions today, my friend Matt Fresha did a scientific test with Wavosaur (a free PC sample editor) and generated a 50Hz sine wave at 16bit and 24bit wav. Spectrum checks show pre and post MP3 export, using the Lame codec so nothing fancy and here is the result.

The raw sine wave shows the clean 50Hz spike you would expect. The MP3 first pass is adding noise, but it's low around the -100dB mark so nothing untoward there.

However, once you render the MP3 to a new MP3 file, the noise does increase, and after 4x passes, the noise creeps upwards.

The green is pass 1, red is pass 4 just to show the differences through more passes.

This is only a 50Hz sine. More frequiences in a mix set will cause the signal to noise ratio to decrease and get worse on only 2 passes. A 2 pass example may be you recording from a 30 minute 320MP3 selection set to then saving it down to a 320MP3.

It's the similar effect of the old tape to tape recording of yesteryear. We used to record on a tape deck and every time you recorded the audio got progressively worse; or take a hired VHS movie - you copied that but the quality was a bit lower but acceptable. In the background each copy degrades each time. Back to the C90 tape example, you give the tape to your friend, he then duplicates it (pass2) but the built-in hiss of the tape you supplied gets copied in the whole mix, and then his tapehead applies hiss as well as naturally re-recordings your hiss - in turn raising the noise floor.


So is a MP3 converted to the same MP3 losing quality?

Yes. The main sound signals in the MP3 conversion to a MP3 conversion are not lowering, but the floor is raising which affects the quality of the intended audio.

WAVs don't have this issue at all, as the noise floor is way way lower. Using WAVs keeps everything clean. Your MP3 recording would be pass 1 with minimal MP3 noise but if you choose to save it as a mix it's then at pass 2. Pass 3 comes when your mix is played then on a radio or uploaded to a 192 bit rate stream where the noise is then apparent.

Recording from WAV music files to WAV instead could be called a "Pass Zero." No noise added and no algorithm bashing.

Rule - start with a WAV music file or lossless format and save as a wav, and only ever compress to a MP3 once at the highest format if you have to.

This left image shows 8 samples all at +1 (0dB) in a WAV file. 

The image on the right is the MP3 version of it.  The algorithms that are used in the codec shows it is longer. The new MP3 file is actually 1152 samples long compared to 8 samples in the WAV. It completely changes the way the music is going to come out of your speaker.

Of course, if you are playing an MP3 in a club or bar, or at home to listen to, none of this really applies.

It does however matter when you're starting to play music in a club or bar, or at home to an internet radio station, where the MP3 you're playing is then recorded for promotional uses or you are maybe recording it to send out. It's then at pass 2. That could be then played on radio as a guest mix or uploaded further to mixcloud, it's then at pass3. 

Each time, a different codec is going to be used. If you want to get really geeky and technicalyou can read how the MP3's are based on psycho-acoustics and how the spectral information is based on the way our hearing interprets audio. Many of the ecoders use it to remove the psycho-acoustic information instead on operating on the sampled waveform.

This all comes back to my earlier point, that we are removing the realness of the sound, the warmth and the vibrations that were in the original WAV file. 

Again, WAVs don't have this issue at all. The argument that you can't hear the difference between a high quality MP3 and a WAV may be upheld on cheap audio equipment and a bad pair of ears, but you can't escape the evidence above that if an MP3 is played and then continually re-recorded through compression, that noise reducing the quality will be heard.

If you want to test this yourself, it's simple. Record a MP3 mix set, save it as a 320 MP3file. Play that file and re-record it again and re-save. Do this 4 times and you will hear. The important thing to understand that is while it takes 4 passes for an untrained ear to recognise, a trained ear can hear at one pass on high end audio equipment, that is why mastering engineers won't touch MP3s and music production is not used with MP3s. On top of that, high end systems at festivals and music events, will amplify good sound as much as it will amplify bad sound.

So the MP3 is not only being made redundant by the Fraunhofer Institute, who is terminating the licence because they have stated "There are more efficient audio codecs with advanced features available today." This is because the AAC format can compress vast amounts of data without a large deterioration in quality. 

Gizmodo wrote about the MP3 : "The audio quality is trash by modern standards and some research suggested that it's compression reinforced perceived negative emotional characteristics in music instrucments to the detriment of positive emotional charagteristics."

The question I ask you is:

If the MP3 is being slated by so many people now, and explainations to you are simple that codecs aren't good for music, with these companies steering closer and closer to lossless codecs - why are you still using MP3s or compressed formats?

Worth realing up about HD-ACC or MPEG4 SLS in future (allows for having both a lossy layer and a lossless correction layer similar to Wavpack (WAV) Hybrid). 

For now, I'll sit here loving off my WAV music collection. I remember having an army of people on Twitter 5 years ago publically shaming me, rewteeting what an idiot I was for speaking out on MP3's saying you can't hear the difference, tweeting to me I didn't know what I was on about. I'm just glad now that evidence has surfaced with even the company who own the MP3 licence now despising it. 

While I'd like to post a middle finger photo up, I'll just stay thankful to those who acknowledge this article and who continue to strive for amazing production techniques and great music. Huge shout outs to all the DJs and MCs who play music at the highest quality possible to reach the audience at end format sounding pristine - honouring the true vibrations we all love most. 

Thank you for reading.

20th June 2017 by Missrepresent


Missrepresent Facebook

Missrepresent Soundcloud

real<br />
time web analytics View My Stats