The Future of Mastering Is Here: Behind the Technology of iZotope Ozone 10
Take a deeper look at the groundbreaking mastering capabilities in Ozone 10 with Bill Podolak, iZotope Product Manager. Learn about the digital signal processing behind the new soft clipping features, Impact and Dynamics Match, and more.
Mastering with iZotope
Ozone Advanced
Ozone 11 Advanced
Jump to these Ozone mastering features:
Follow along with this feature deep-dive with your copy of iZotope Ozone 10.
Ozone Magnify Soft Clip
While many plug-ins will opt for modeling analog electrical components to saturate an audio signal, we decided to look forward and apply advanced math for our new Magnify Soft Clip feature within the Ozone Maximizer module. Borrowing from graphics processing, the research team applied a technique of high-order polynomial interpolation in order to achieve the smoothest possible transition between the soft clip part of the transfer curve and the hard clip. The result is a more controlled spread of harmonics and a smoother, less harsh tone.
This is the harmonic profile of the soft clipper with a 100 Hz sine wave. Notice how the harmonics are being rolled off. This is the result of increasing the order of the mathematical formula. Most soft clippers will look like the order=4 frame with the most shallow diagonal pattern of harmonics. The steep harmonic roll off of the Ozone Soft Clipper is largely responsible for its smooth tone, along with high-quality oversampling to prevent aliasing.
Lastly, we opted to expose Wet/Dry percent as the main control of the Soft Clip processing. The harmonics generated from a full 100% wet soft clip can often be too intense for mastering, so this control enables a precise amount of boost and saturation.
Here's what Magnify Soft Clip sounds like in an audio example.
"What Are We Fighting For" Audio Example
Impact and Dynamics Match
At first glance, the new Impact module might seem like a Transient Shaper or one-knob compressor, but it actually uses a novel form of signal processing that directly adjusts microdynamics and unlocks a new paradigm for matching the microdynamics of one signal to another.
The term "microdynamics" is relatively new in the audio engineering vocabulary, so it's worth defining: Macrodynamics refer to loudness changes over a long time-scale, like the relative volume of a verse to a chorus. Microdynamics, on the other hand, are the short term changes in loudness between the transient and sustain of a note or drum hit. Records with high levels of microdynamics will be perceived as "punchy" or "snappy."
Our new processor builds upon research into perceived measures of microdynamics. Studies have shown that "Loudness Dynamic Range" (LDR) matches well with how humans perceive microdynamics. This measurement is defined by the ratio between a short and long term amplitude envelope. By using a ratio, this relative measurement is unaffected by the macrodynamic changes in loudness between various sections of a song.
The new processing that powers the Impact module directly adjusts that relative measurement of microdynamics via a single user-facing parameter. This enables a more intuitive control over the expansion or compression of the dynamic range of the signal in each band. This is paired with a powerful auto-gain that combines static and dynamic balancing of output to match the input. Since the Impact module is multiband, the auto-gain ensures that each band stays level and the tonal balance of the signal is not significantly adjusted.
Finally, this new module allows the Master Assistant to perform a groundbreaking multiband microdynamics match between a user's audio and a target. By measuring the microdynamics of the source and knowing the microdynamics of the target, the Impact module can be set to match the source to the target. The target can be any reference audio file on the user's computer, or one of our pre-made genre targets, which were derived from analyzing the latest chart-topping hits.
While this overview should hopefully give a good insight into the tech here, any scientifically-minded readers should definitely go check out our paper that was published in the recent International Conference on Digital Audio Effects.
Recover Sides and Width Match
The Imager is a well-loved tool in the Ozone toolkit, but we observed that it was primarily used for boosting the width of a master. While reducing width may be useful for taming problem areas like a wide sub-bass, it can generally be undesirable because side energy and information is lost as width is reduced. Let's set up a visual metaphor:
The stereo image is visualized as the large white rectangle between the two speakers. When you turn down the width of the full audio band using the Imager, it's as if you are fading out everything except for the mid signal.
Recover sides solves this problem by mixing any reduced side channel information back into the middle as width is decreased. While intuitively you might think that side information can simply be collapsed into the middle with a tool like the Imager, it's actually more tricky than that, so we needed to introduce new DSP that uses phase rotation. Here's a way to visually represent reducing width when Recover Sides is enabled:
The stereo image is still being collapsed, but rather than fading out everything except the mid, the image is being squeezed into the center. The idea is that all the information in the original stereo image is still maintained when it has been collapsed to mono.
While making the entire stereo image mono is not recommended, the Recover Sides feature can help your song's playback on mono audio systems. By reducing some width and recovering that signal into the mid, you can actually maintain stereo information in the mono playback. To hear exactly what signal is being recovered, definitely try the "solo" button attached to the Recover Sides controls.
Ozone Master Assistant
Finally, I'll touch on the Master Assistant's new ability to set the Imager module to match the stereo width of the user's song to a target. This was much more straightforward than Dynamics matching because the Imager is a linear process, while dynamics processing is inherently non-linear. Master Assistant will analyze the user's audio to measure the ratio of the power of the mid signal to the side signal in four separate bands. The Master Assistant will then set the Imager to match that ratio to the target, as defined by a reference file or our average of chart-topping hits in each genre.
The Master Assistant overview page shows a visualization of the Tonal Balance target, but does not show the target for the width or dynamics matching. We were interested in showing this, but did not have time. Since it's interesting, I'll show a chart that represents the width matching targets for each genre:
The Y-axis is the mid-to-side-ratio, so a higher value means more mono. The four bands have the preset crossovers at 120 Hz, 2 kHz, and 10 kHz. There are a couple interesting things to observe:
First is the width of the Cinematic target, which is far more wide than any of the musical genres. During beta testing, we heard from many users that none of the width matching targets were appropriate for the film scores they were working on. Because of this, we developed the new Cinematic target by analyzing the scores of the highest grossing box office movies.
Next, it's interesting to see which genres are more mono-focused. The Rap and R&B genres in particular are the most narrow. The chart-topping songs in this genre tended to heavily focus on the kick, bass, snare, hi hats, and vocals coming straight down the middle of the mix. While a narrow master may sound less impressive on stereo speakers or headphones, it's important to note that a lot of people will be listening on mono systems like phones or bluetooth speakers. When the playback is also loudness normalized like on almost all streaming services, these more narrow masters may be louder on mono systems.
Last, it's notable how narrow the bass bands are across genres. While you can see that the targets are more narrow below 120 Hz, none of the targets are completely mono in the bass region. This means that if the Master Assistant analyzes a user's track and measures that the bass is very mono, it will set the Imager to boost the width of the low end. We saw in many reviews that this kind of mastering move flies in the face of conventional wisdom that bass should be mono. Based on this feedback, we've iterated on Width matching in our Ozone 10.1 patch update. Now if Master Assistant measures a very mono bass band, it will not attempt to adjust the width of that band.
Start mastering with iZotope Ozone 10
We wanted to pack as many innovative new features into Ozone 10 as we could, so there's actually not enough room in this article to include everything. Some honorable mentions are the new Stabilizer module with its Shape and Cut modes, our new and improved method of setting EQ curves, and the new Maximizer threshold learning that utilizes an integrated LUFS measurement.
Thanks for reading! We hope that iZotope Ozone 10 will provide an invaluable tool for beginners and pros alike and help everyone make great-sounding audio masters. If you haven’t already, get your copy of Ozone and start mastering with the new innovative features.
Already own a previous copy of Ozone? Log into your iZotope account for special loyalty pricing.