Table of Contents
What is Noise Floor?
The Noise Floor is the continuous background noise in your audio recording. Within the frequency spectrum, noise can be broadband or narrowband. Broadband noise can be described as a “hiss” noise and takes up a large range of the frequency spectrum. Narrowband noise can be described as a hum or whine and takes up only a section of the frequency spectrum. A hum is a low frequency and a whine is a high frequency.
What Causes the Noise Floor?
Room Tone is the background noise of the room you are recording in. It can be characterized by whatever continuous sounds are commonly audible in that space, such as the noise from an air conditioning unit.
You can reduce the level of the room tone to reduce the Noise Floor. This can be achieved with increased sound insulation to reduce the level of sound bleeding in from outside the room. This can be cheaply done by closing any windows and doors of course! I’ve been in a studio that automatically turns on air conditioning inside when the door to the studio is closed, which increased the Noise Floor. We had no access to the air conditioning controls and the surroundings were quiet, so actually it was best to keep the door slightly open!
Using an omnidirectional microphone can increase the Noise Floor. If possible, use a type of cardioid microphone that can be pointed directly at the sound source without capturing surrounding noises. The surrounding noises can reduce the quality of your sound source.
Microphones also have a Self-noise; this is the internal noise produced by a microphone, which can be defined by a microphone’s Signal-to-noise Level. The consensus is that a Signal-to-noise Level of 79 dBA is good enough for most applications, although more is always better.
Increasing the gain on the preamp to compensate for a low-level sound source, because the source is positioned too far from the microphone, for example, can increase the noise. This is because by increasing the gain you are increasing the electrical noise and room tone in your recording.
Although increasing the gain will make the sound you’re trying to capture louder, instead simply positioning the source closer to the microphone will eliminate the issue. If you’re recording a podcast, a condenser microphone should be positioned approximately the distance of one clenched fist from the person’s mouth.
Noise can occur in cabling through the pickup of external noise for multiple reasons. Cabling can be found with shielding to reduce noise. Copper and aluminum shielding is effective against the electric field and can reduce the noise effects of neon lights for example. The magnetic field caused by electrical devices can be reduced through the use of twisted cabling.
What is the Ideal Noise Floor?
You want the Noise Floor to be as low as possible otherwise the quality of your audio signal will be degraded. A low-level Noise Floor is essential to capture a clean recording as a high-level Noise Floor can be unpleasant, can reduce speech intelligibility, and can make your podcast or content come across as less professional.
The Signal-to-noise ratio (SNR) can be used to measure the Noise Floor. This is essentially the difference between the noise and the signal you’re recording. The SNR can be calculated as a difference in dB by calculating the difference between the dB of the noise and the dB of the signal. The SNR can be calculated by representing this as a ratio.
In a DAW, dB is represented using the dBFS scale. For SNR calculation in dBFS, it is slightly different, but an example is given below. In the images shown, the Noise Floor measured in Audacity is approximately -48 dB and the signal is approximately -12 dB. So, the SNR can be worked calculating -12 minus -48, meaning the SNR is approximately 36 dB. This can also be described as a ratio of 4:1, calculated by dividing the noise by the signal (-48 / -12).
Of course, the noise is likely a constant dB level, but the signal can vary. It’s probably better to monitor the levels and find an average dB for the signal or, to be on the safe side, use the dB measured at a lower level signal, to ensure a good SNR.
It’s always recommended to listen and consider yourself if there is too much noise but, as a rule of thumb, take a look at the table below for some recommendations based on some listening tests I have done.
|60 dB||Noise Floor sounded quiet but was still noticeable which some may find distracting.|
|70 dB||Noise Floor was barely audible. I’d say this was a sufficient SNR for a professional podcast.|
|80 dB||Noise Floor was inaudible. This SNR is ideal.|
How to Check the Noise Floor
You should check the Noise Floor and consider the SNR to test the quality of your signal. As discussed, it could be a range of reasons causing Noise Floor. I’d recommend that if you find your Noise Floor is high that you swap out a piece of equipment one-by-one and see if there’s an improvement.
The following tutorial uses the latest version of Audacity (at the time of writing), version 2.4.2, although similar versions are likely compatible with this tutorial. Audacity is free and, if you don’t already have it, can be downloaded from here: https://www.audacityteam.org/download/
- You may be recording to an SD card or directly to Audacity; either will work for this tutorial. Using whatever setup you’d like to check the Noise Floor for, first, record a room tone. Simply press record, try to stay quiet and stop recording after approximately 20 seconds.
- If you weren’t recording directly into Audacity, then skip this step. Import the audio to Audacity by dragging and dropping the audio clip into the Audacity window. This can also be done by selecting File > Import > Audio. From here, you can browse and select the file.
You may notice that the audio file you want to import is grayed-out, like below.
To fix this, simply select ‘File type’ at the bottom of the window and select the appropriate type for your audio file. In this case, I’m using a WAV file, although if you’re not sure about your file type, you can select All files.
You can now see that my file is selectable.
Selecting the file and pressing Open will present you with your audio file in Audacity.
- Next up, we’re going to efficiently check the levels and frequency content of the Noise Floor. Double click the audio so that it is highlighted.
Next, select Analyze > Plot Spectrum… This will open the Frequency Analysis window.
- This window shows the dB and the Noise Floor distributed over the frequency spectrum. As can be seen on my recording, the Noise Floor is most noticeable at approximately 250 Hz and 1 kHz
How to Reduce Noise with Audacity’s In-built Noise Reduction Feature
There are two common methods in Audacity to reduce noise. Of course, reducing the noise source before the post-production stage is ideal; it will save you from having to do the following and can result in a cleaner audio signal in comparison. But the following methods will show you how to reduce it if you’ve already recorded.
The Noise Reduction feature is a quick and easy way to reduce noise, although could reduce the quality of the sound signal you want to keep. If you’re short on time, feel free to use this method, but I’d advise you to use the EQ method in the next heading.
- Import a normal recording using the same setup, maybe from a podcast, and use the Frequency Analyzer on the entire audio as discussed previously.
- The above analysis still includes the Noise Floor from the previous frequency analysis so you can see here that some of the human speech is present above 2500 Hz, although some is also below this frequency. Close this window. Highlight a portion of the audio that is just the Noise Floor and select Effect > Noise Reduction…
- This brings up the Noise Reduction window. You can preview the period of the Noise Floor by selecting Preview. Press Get Noise Profile to store the Noise Profile. The window will close.
- Highlight the entire audio by double-clicking and open the Noise Reduction feature again. Now, you can apply the Noise Reduction. The Preview button should still play the audio sample selected previously for the Noise Profile. I found the default settings worked okay for this audio, although it’s always best to try different parameters for different audio. Press OK.
How to Reduce Noise with EQ
If you’re finding the signal quality is reduced too much while leaving too much noise, you can try to EQ-out the noise.
- As has been shown, we know that there is a lot of noise around 250 Hz and 1 kHz. This can be reduced using the Graphic EQ. Import some audio with a Noise Floor and someone speaking.
- Double click the audio and select Effect > Graphic EQ…
- Drag down certain areas of the spectrum to reduce the levels of different frequency bands. The following EQ has certain frequencies reduced based on a number of factors:
- There is little low-frequency speech content and little as can be seen from the analysis no frequencies above 9 kHz, so both have been reduced completely.
- The Noise Floor is prominent around 250 Hz and 1 kHz, so these frequencies have been reduced.
Happy (noise-free) podcasting!