Optimizing your music for digital distribution – Sonnox’s Fraunhofer Pro-Codec in the studio

Visit Us

As the discovery of new music increasingly moves to downloading and streaming, taking the time to ensure your online music sounds as good as it can takes on ever greater importance. For many independent artists, a new listener’s first impression will be formed by hearing some type of compressed audio file.

Properly encoding your files for digital distribution is definitely not a one size fits all solution, as different sites and different processes use different bit rates. The bit rate is the number of bits that are conveyed or processed per unit of time, and in simple terms, the more bits, the better the sound quality. Streaming radio portals such as Pandora and LastFM generally utilize lower bit rate coding schemes (64-128 kilobits per second) than download sites such as iTunes and Amazon (256 kbps).

So in addition to creating killer mixes for your WAV and AIFF files for CD, it might be worth optimizing your mixes to conform to the nuances and specs of the various digital distribution channels.

Sonnox Fraunhofer Pro-Codec
While at the NAMM show last January, I got an interesting demo of a software plug-in from Sonnox called the Fraunhofer Pro-Codec. It’s designed to offer a recording engineer the tools to evaluate and choose between various codecs to optimally compress your mixes for digital distribution.

A codec is a computer program that can encode or decode a digital data stream or signal – the word codec is a combination of “compress” and “decompress.” A codec encodes a data stream or signal for transmission, storage, or encryption – or decodes it for playback or editing. So whether you’re prepping mixes for iTunes, Amazon, or SoundCloud – or streaming from your own website – you can optimize your mixes for the various codecs, bit rates, and sound qualities that the various distribution channels require.

Human hearing is pretty much limited to the range between 20Hz and 20kHz (20,000Hz), though sound frequencies extend above and below that range. A quality audio recording stores a tremendous amount of data that you don’t hear, because your recording equipment (microphones, guitar pickups, etc.) captures a broader range of sound than does the human ear. The basic principle behind MP3s and other compressed music files is to trim away the information that can’t be perceived by humans, while also using algorithms that make use of the principles of psychoacoustics.

The Fraunhofer Pro-Codec is a plug-in that runs within your digital audio workstation (DAW) program. It’s at home in the Pro Tools, Logic, Cubase, Nuendo, Wavelab, and Sequoia environments, with a Mac or PC. The breakthrough feature of the Fraunhofer Pro-Codec is that it allows you to work in real time to audition, encode, or decode your mixes for online distribution. So instead of selecting one codec and bit rate, encoding your track, then playing it back to see how it came out, with the Fraunhofer Pro-Codec, you can select up to five different codecs, and seamlessly hear the differences switching between each as your song plays in real time.

The plug in supports five key MP3 or AAC “lossy” codecs, as well as two “lossless” codecs. Lossless codecs use lower compression ratios and include all the audio information in the original sound file; lossy codecs use much higher compression ratios and may toss out as much as 80-90% of the original signal.

In the Studio
I visited our campus recording studio where engineers Shane Pilgrim and Jeff Crawford had graciously set up a demo of the Pro-Codec for me to listen to. I asked them to use a track that I knew like the back of my hand, Steely Dan’s classic “Peg,” for our first test. We started with the track as a 16-bit, 44.1k uncompressed WAV file and decided to compare encoding using MP3 and AAC, two of the most commonly used codecs.

Shane opened the stereo master of “Peg” in Pro Tools, then selected the Pro-Codec as a plug-in. When the Pro-Codec screen opens, it offers you the option of choosing up to five codecs to audition simultaneously. For each, you can select the bit rate, mode, and quality using the menu box. We chose “AAC-LC / 256k / Constant Bit Rate” and “MP3 / 256k / Constant Bit Rate.” In essence, we were simply comparing the codecs themselves, since we used the same bit rate for each.

We started off playing the original, uncompressed audio file to get the tonal characteristics and dynamics as our baseline by simply engaging the “bypass” switch. After we established our aural baseline, we then engaged the plug-in, and began to see and hear in real time how each of these standard codecs affected the original audio file.

The Fraunhofer Pro-Codec display screen showing the MP3 audition for "Peg." (Click image to enlarge.)
The visual display uses FFT metering to accurately show program levels across the entire frequency range. We started with the AAC codec audition, represented by the red frequency display, which shows what the encoded file would look like. Directly above the AAC display, a yellow frequency display represents the original input signal so you can easily see how much audio will be squeezed out in the encoding process.

The key benefit is that we were listening to what the actual encoded signal would sound like in real time – that’s different than other encoders which only allow you to hear the result after you’ve encoded the track.

The AAC encoded track sounded pretty good to all three of us, so we did a pass listening to whole song through the MP3 codec. Right away, the visual display indicated that the difference between the input and the MP3-encoded signal was greater than with the AAC audition for this song. Rather than leaving this analysis solely to your eyes, the Pro-Codec includes a “Difference” button that allows you to switch between monitoring a particular codec’s output signal with the “difference” – the frequencies that have been removed in the compression process.

Depending on how much the audio is being compressed, the difference sounds like a distorted speaker, but in comparing the two difference-only audio signals, we noticed that more of the original high frequencies were being removed in the MP3 codec than with the AAC codec at the settings we had selected. This correlated to our experience listening to the preview of the compressed file, as Jeff commented that he was hearing more highs in the AAC version than the MP3.

Another key feature of the Pro-Codec is the metering that measures the Noise-to-Mask (NMR) ratio as you audition each particular codec. Represented by a series of green horizontal lines at each frequency, the NMR metering shows us when and at what frequencies codec-induced noise may be audible. When it’s all green, any noise generated by the encoding process is masked by the music. However, if the noise will be audible after encoding, a red vertical bar appears, alerting you that the codec is not performing optimally. You can then adjust the amount of compression using the trim control (in real time) for each codec to ensure you won’t experience any clipping in your encoded file on playback. This feature is a real time saver and may be worth the price of the plug-in on its own.

We decided to encode using the AAC/256k setting, which is done by arming the “Record Enable” button. First, we set the destination for the encoded file, which is done by clicking the “Export” button. When the “Export Setting” window pops open, you chose a target file folder for the encoded file.

Shane pointed out that we could also compress and export up to five different encoded files – each using a different codec – in one pass. After processing, all five would appear in the same destination folder. We could also customize the name for each file prior to export, as well as what information relating to the file we wanted appended to it. For instance, we could include the codec name, bit rate, mode, and exact date and time. This is a handy feature that proved very helpful that night, as we ended up with a few dozen compressed files by the end of our test session, each of which were identifiable by the unique information we tagged the files with.

Once you have armed recording, hit the space bar to play the source audio and start encoding. At the end of the fade out for “Peg,” Shane hit the space bar again to end the encoding process. If you don’t hit stop, you will be encoding silence and adding to the duration of your file. After the encode is completed, the file appears in your target folder, which in this case resulted in 7.7 MB file, compressed from its original 41.8 MB file size. More importantly, the sound quality of the AAC-encoded file was judged to be very good by all three of us in the room.

Next we decided to test out the plug-in using some well-recorded classical music, in this case, a three-minute excerpt from a Vivaldi concerto. Our original file size was 35.3 MB. Again, we started by orienting our ears to the 16-bit, 44.1k WAV file. This particular recording featured a lot of room tone (natural ambience) as well as a gorgeously recorded solo violin with the texture of the bow and instrument clearly audible, creating the impression that the performer was there in the room with us.

This excerpt would pose a much different challenge to the encoders than the pop song, as this passage features dramatic changes in volume and a full frequency range. Shane chose the AAC-LC, but at a slightly higher bit rate of 320k. We stacked that up against MP3-LC, also at 320k bit rate. For this test, we experimented with the Pro-Codec’s A-B comparison switch. Using this feature, we were able to seamlessly switch between the two codecs in real-time to better hear the differences.

The Vivaldi concerto being auditioned using the AAC 320k encoder. (Click image to enlarge.)
We started with the AAC/320k, and saw immediately that unlike the pop track, there were many more distinct frequency peaks visible on the FFT metering. As to the sound quality, Shane commented that he heard less of the ambience in the encoded file, which had given the original its very live feeling, and Jeff suggested that the high frequencies seemed harsher than the original.

We decided to compare the AAC/320k codec to the MP3/320k to see what differences would be discernable. Using the A-B switch, Shane flipped instantly between the two, with no click or delay. All three of us commented that there was little difference between the audition signals of the two codecs. However, when we used the A-B switch on the “difference,” we did notice that the MP3 signal appeared to have a good deal more sonic data loss, creating a hash of high frequencies that were being subtracted from the original signal. The AAC difference signal had significantly less, meaning it was maintaining a greater amount of the original input signal’s integrity.

Final Thoughts
As we were getting ready to leave the studio, Jeff said that he was impressed with the program’s sophistication, and suggested it has a few more features than a basic home recording studio needs. While that’s true, and free codecs such as the one available in iTunes work reasonably well when set properly, the Pro-Codec is a worthwhile investment for anyone who wants more control and is serious about ensuring the best possible quality for their digitally distributed music.

Jeff also pointed out that although we used pre-existing, master stereo mixes for our test, the Pro-Codec would really shine in a mixing situation when you’d have the option of customizing individual mixes for each distribution platform’s delivery requirements. We all agreed that the advanced features, such as the A-B switching, the NMR metering, and the ability to listen only to the difference signal, were extraordinarily useful tools.

All in all, the $470 Sonnox Fraunhofer Pro-Codec plug-in significantly raises the bar for performance and features against any other codec I’ve seen or used.

Keith Hatschek is a contributing writer for Echoes and the author of two books on the music industry. The Golden Moment: Recording Secrets of the Pros and How to Get a Job in the Music Industry. He directs the Music Management Program at University of the Pacific in Stockton, CA.

Story Links
Sonnox Fraunhofer Pro-Codec – includes an excellent short video overview of the product, as well as a link to download a free 15-day trial version of the program.
Wikipedia reference to various digital audio file formats and compression schemes.
Wikipedia entry on codecs.
Article on psychoacoustics. Provides a deeper explanation of psychoacoustics and the way we perceive sounds.
Article on the MP3 codec.

10 thoughts on “Optimizing your music for digital distribution – Sonnox’s Fraunhofer Pro-Codec in the studio

  1. Yeah, thinking of where I have control (few and they want what I send them in MP3 format) and I know of no sites that have ever asked me for an AAC (Apple) formatted file and once you’ve given control to an aggregator (CDBaby, etc), it is out of your hands.  I guess, if it was cheaper, it would be useful to see what it will sound like encoded, while you are mixing and you might adjust for MP3’s, which I DO hear the diff and hate.  But then again, I’m a recording engr and audiophile, not the avg listener with avg equipment.  Not worried about the earbud listeners lol.

    In answer to the best settings for iTunes…I think their encoder is very good, sounds better than most, even with MP3s, not sure if that is player math or their own encoder (as opposed to Lame, etc).  I use VBR as it intelligently and dynamically alters the bit rate tailored to the content through time for the best audio while still getting reasonable compression.  This is a purely subjective, although educated, audio opinion. Google VBR for more info.

  2. This is all well and good, except that Discmakers customers do not get any choice about how the files are encoded when they are distributed, and CDBaby just rips the CDs for their files. This information is theoretically interesting but basically useless to me.

  3. Great article! What do you recommend for highest quality AAC conversion using the iTunes settings? i.e. should I be using constant bit rate instead of VBR? What stereo setting, low cut filter, etc…. Big thanks!

  4. this is dribble, the average listeners will not be able to distinguish between any of this overkill, as long as you stick with the nyquist theorem everything will be fine, besides maybe you want a LoFi overall sound.

Leave a Reply

Your email address will not be published. Required fields are marked *