A Podcaster’s Guide To Noise Reduction

Joe Nash
9 min readFeb 26, 2015

Speaking with a few people in the podcasting sphere recently, I’ve realised that many people find the aspect of audio effects processing to be mystifying. While I’m no expert, I thought I might write about what I have learnt during my time editing radio shows and podcasts.

So, here it goes, but I apologise in advance if I’m teaching my grandmother to suck on those proverbial eggs.

When starting a podcast edit for The Committed, the first thing I like to do is tidy up the files I receive from the presenters. I want to make sure their speech is centre-stage, and all background noises are cleared away the best I can. So, noise reduction is the first step.

A Brief History of Hiss

Ahh, the humble Boots cassette tape, juxtaposed against a background of classic iPods.

Noise reduction techniques have been around for many years, well before the advent of digital audio. In fact, back in the days of the good ol’ cassette tape, noise reduction techniques were paramount. The problem with recording onto tape was that, when played back, in the background you would hear a lovely, hissing sound. The hiss you heard was caused by the size and type of the magnetic particles used to make up the surface of the tape. On the more expensive tapes you could buy — such as chrome (chromium dioxide) or metal (purer and non-oxidised) — the particles would be smaller, and that meant less hiss on playback, as well as better dynamic range and frequency response.

Since magnetic tape was used all the way through the professional recording process, a whole industry spawned out of technologies dedicated to reducing that hissing noise. The two most prominent figures of this industry were — arguably — Dolby and dbx. They applied many clever techniques for getting rid of the “noise floor”, the name given to the volume at which the unwanted background noise sits. Without going too far down the audio theory ‘rabbit hole’, volume level is measured in decibels (dB). The noise floor of a cassette might have been about -40 dB, where 0 dB is the tape’s theoretical maximum volume level. That doesn’t sound like much, but decibel is a logarithmic scale, so 40 dB means a difference of 100 x between the loudest noise and the hiss.

Hiss noise floor at -40 dB, magnified in Sound Studio 4

So, how did Messrs Dolby et al create a system for removing that unwanted noise? The two main processes they used were audio compression and expansion. When recording the audio, the sounds at certain frequencies would be squished into the range above the noise floor using a compressor. Then, when played back, that range would be stretched back out using an expander, pushing the noise floor down to a much, quieter level.

Tape noise reduction: Certain frequencies in the recorded signal are compressed so they fit above the noise floor, then expanded on playback to hide the hiss.

Clever, eh? Well yes, very, and the techniques used to make that happen are pretty complex. Now we have nice, easy-to-use plugins that do the same technique in our digital audio software. But why do we need these tools, now that tape hiss is a thing of the past?

Back To Podcasts…

You might have a nice setup like this. I don’t (yet). I just edit…

Nobody’s podcast setups are perfect. You can have the best vocal microphone money can buy, but most of us have background noise in our rooms such as computer fans, or traffic noise coming from outside. Even if these don’t sound very loud to our ears, there’s still only so much of that ambient noise a mike will be able to cut out.

In this case, for noise reduction we only need an expander, not a compressor. This is because our recording storage (usually a WAV or AIFF sound file on a disk) doesn’t create any of its own hiss; we don’t need to worry about ‘preparing’ the sound before storage. All we are interested in doing is moving the background noise down lower in the mix permanently. So, once we’ve applied an expander, we’re done.

Tip: Since we’re dealing with noise introduced by the recording and not in the storage medium itself, we want to run the noise reduction filter on each individual recorded track. If you are using a DAW (Digital Audio Workstation) such as GarageBand or Logic Pro X, this means selecting the track and adding the filter there, rather than adding it to the master output channel. The reason for this is that each mike recorded will have its own background noise level, as some mikes will be in noisier environments than others. If you apply a noise reduction filter to the final output, you run the risk of filtering out too much vocal on some of the better-recorded mike channels.

Depending on your audio tool of choice, you may have a selection of effects processors that can be used for noise reduction. Expander, as described above, is my effect of choice. However, we should also give mention to the Noise Gate.

Noise Gate

Click on the play button to hear a sample of how the Noise Gate effect sounds in GarageBand. Note how the breath between sentences is cut off slightly.

The only option for noise reduction filtering in GarageBand is the Noise Gate. I won’t go into how it is applied here because others out there have already done a good job. At its most basic, a Noise Gate ‘opens’ to allow sound through above a certain volume threshold, and then ‘closes’ when that level drops back below. In GarageBand, the only control you have is over the level of the threshold. It’s okay, but it’s not great, because you can sometimes hear moments when background room sounds cut in and out abruptly between sentences. If you have quite a lot of background noise in the room and need to set the threshold level quite high to compensate, it can even drop out in mid-sentence or word.

GarageBand’s Noise Gate settings are very limited, but do provide a starting point.

Logic Pro X’s noise gate takes things a lot further, as you can imagine. There are probably many other inexpensive audio editors out there with better noise gate effects than GarageBand, but without using them I couldn’t tell you which ones to try. What you want is a noise gate that also has the controls attack, hold and release. Attack and release speeds allow for the volume change to start and end more gradually, while the hold time will stop the level from dropping for a certain amount of time so the whole thing doesn’t sound so jilted. Some also have an extra setting, level reduction. With level reduction, rather than replacing the quieter sounds with silence, you can just lower them further in the mix, keeping some ambient noise and therefore a more natural sound.

Logic Pro X’s Noise Gate settings probably have more options than you really care about, but some are very useful.

I tend to have the decay set to at least half a second, especially if there is a fair amount of background noise, so it fades away nice and smoothly. As for attack, keeping that very short but not exactly zero ensures you don’t get a sudden volume change at the beginning. Some dynamics effects also have a look ahead setting. This allows the attack to happen actually before the sound has crossed the volume threshold, something only possible with digital editing! If yours does have this setting, you can increase the attack time slightly (up to 0.1 seconds or so) so the fade in is more subtle.

Expander

Although a decent noise gate with the above options will do a fine job, my filter-of-choice for noise reduction is the Expander. As mentioned, not all sound editors and DAWs provide this option. Some have a generic Dynamics Processor that do compression and expansion in a single filter.

As discussed earlier, compressors and expanders stretch and squish different volume levels; I’ll discuss the use of compressors in another article. But using an expander at low volume levels (also known as a downward expander) is very similar to using a noise gate, but with slightly more control. The level of an expander (or compressor) is measured in ratios. For example, an expander with a ratio of 1:2 will spread out volume levels to twice their original levels. If this happens below, for example, -40 dB, then the volume levels drop off to lower levels, like a smoother noise gate.

Graph showing how a downward expander drops the volume off below a specified volume threshold.
The original sound file, a voice recording from The Committed podcast.
After applying an expander with a low threshold (about -35 dB). The louder sounds are unaffected

You will find the controls to be very similar to Noise Gate, such as attack and release times, as well as threshold for where the volume drops off and ratio for how quickly it happens.

Destructive versus non-destructive

So, now you know how to apply the effects. If you apply them in a sound editor like Sound Studio, the change to the file is permanent. However, if you add it to a channel in a DAW like GarageBand or Logic, the filter is applied but can be removed if necessary. This is known as non-destructive editing, because the original data is unaffected. Common sense would tell you that this is the better way, because you can always go back and change it if you get it wrong. And if you’re starting out, this is probably the best way. Having said that, it’s not usually the way I work.

Podcasts are usually pretty long, and so each presenter’s sound file can be an hour or longer. In this time, things can change to their sound quality. For example, a presenter may move to a different position on his or her chair, or decide to move the computer into a different room altogether. If I run a non-destructive expander in Logic, it assumes that I want to apply the same level of effect to the entire length of that file. But life isn’t always that simple. You could use track automation to tweak the settings of the filter along the timeline, but that is a real ball ache! We don’t want to spend hours tweaking each file to the nth degree. So, I prefer to destructively edit each file beforehand, normalising the volume levels wherever needed, and applying the expander before taking each file into Logic — making sure I create a back-up of each file first, of course.

An extra added benefit to doing it this way is that running Strip Silence in Logic works a lot more smoothly. Without destructively editing the audio beforehand, Strip Silence will work better in some places than in others, resulting in a lot of faffing about with the track data afterwards.

Logic Pro’s Strip Silence tool works really well, as long as audio levels are fairly consistent throughout.

And we’re done…

So, that’s pretty much it for noise reduction. Obviously, this is my way of working, and it works for me; you may have your own methods, and I’d be happy to hear them. I like to think I’m always learning!

You may wish to make other tweaks to each file before importing into your DAW, such as equalisation (EQ) if the microphone needs to sound crisper, or to remove low frequencies if someone knocks the table or types on their keyboard — a common issue! But once you’ve found the “sweet spot” for each file, putting them together will sound so much more professional. Better still, you can then apply some crafty compression to the final edit to give it some vibrance. But that’s for next time.

Thanks to Kirk McElhearn for his help on this article.
If you found this useful, please click “Recommend” below.
This will help to share the article with others.

I edit The Committed, a weekly tech podcast, and produce and edit
Uniting Friends Radio, a monthly show for people with learning disabilities.

Follow me on Twitter: @jowie

--

--

Joe Nash

iOS software engineer, editor for @TheCommittedShow, pianist & DJ, biker in hibernation, all-round great guy.