Sound can be corrupted in dozens of ways. Field recordists constantly battle intrusive sounds, noise, buzz, static, hum, and other sonic enemies.
It’s not always possible to avoid them. Thankfully, editors have an arsenal of weapons to fix damaged sound. That’s the good news.
The problem is that there are so many options for fixing sound it’s hard to know where to start. Should those sharp ticks in a theatre crowd be fixed with a de-crackler, de-clipper, or de-clicker? Should the noise be stripped from the file first, or be left to the end?
Sequence of Audio Restoration
I found the following chart while browsing the depths of noise repair software manufacturer iZotope’s blog articles. This one explained how to choose the right sequence of “modules” (e.g., repair tools) to fix sounds with the most success. Click on the image to view a full-sized version:
For people new to mastering, the chart goes beyond merely listing module titles and shares sonic problems with examples beneath each problem type, for instance:
Steady Noise: Stationary or slowly changing broadband noise like jet flyovers, HVAC, or white noise.
It presents a top-to-bottom, left-to-right sequence for using each tool in iZotope’s RX restoration software: once one tool fails, move to the module to the right until something works. Once that’s complete, move down the list to fix the next problem.
The repair order iZotope suggests is:
- Problematic stereo files.
- Clicks or crackle.
- Steady noise.
- Variable noise.
A Noise Repair Workflow
I also found another workflow chart, which you can find in the current RX7 manual. It’s essentially the same chart, however the alternative language adds a bit more detail.
The Idea Behind the Processing Steps
Of course, you don’t need an iZotope product to benefit from the flowchart. The idea is that certain types of damage are best dealt with in sequence, whether you use a Sonnox, Waves, Accusonus, or other repair plug-in.
While the original blog post doesn’t mention the rationale for the steps, browsing the chart seems to suggest that simpler, broader errors are fixed first (clipping, stereo image), and tasks that require more fine tuning (hum, noise) follow after. The earlier tasks are more serious do-or-die problems, which is likely why they’re listed first. After all, if you can’t fixed a clipped sound, it won’t matter if you can reduce broadband noise.
It’s interesting that spectral repair is the last step for a number of problems on the chart. For field recordings, I find I use spectral repair far more than any other tool.
The visual layout is also a good reminder of how to think about noise repair; when fixing damaged audio, a scattershot approach is less effective than a focused methodology of diligently fixing problems one by one.
Processing step images courtesy iZotope.