In the article on audio filters, we mentioned the use of spectrum analyzers as a support tool for our decisions when filtering, equalizing, and to ensure that we can obtain transposable mixes to different sound systems.
While spectrum analyzers are a potent tool whose management will do much to achieve better results, the fact that the mix “looks” right in the spectrum analyzer does not mean that the blend is artistically correct.
The correct use of the analyzers will give us a guideline of where we are standing regarding level by frequency. Let’s see what this is about:
What are they?
They are measurement devices that allow us to graphically see what happens with frequency in time, regarding level, with a specific signal. Since it is a frequency analysis, they are called spectrum analyzers.
Recall that the spectrum in the audio is the equivalent to the frequency content of a source or mix.
What are they for?
They serve to have an objective measurement of what happens with our audio signals regarding frequency. The real value they have is that the analysis is not affected by what happens in the room, with our monitors, or in the position that we are inside it since it is an electrical measurement of the signal.
They also give us crucial information about the noise that we may be having in some of our insertion processes, either with plugins or with hardware processors. Many times the simulation plugins of analog circuits have a noise that can be ignored or not heard; however, when using an analyzer, it facilitates the location of where our noise sources can be and therefore improves the noise level of our mixtures.
That is why they are an essential tool to support when mixing, mastering, or making live sound/sound reinforcement. With a spectrum analyzer applied in the master fader, for example, we can see the behavior of all the signals that make up our mix separately, by listening to them on their own and making filtering decisions, EQ knowing what is happening.
How do they work?
Spectrum analyzers work by dividing the spectrum into groups of frequencies, then applying a mathematical process called Fast Fourier Transform or FFT (Fast Fourier Transform).
In the case of audio, what it does is take the signal that is in the domain of time that music would be in time and transforms it into the different frequencies that compose it.
The result is a given voltage value for each frequency or group of frequencies, which is plotted in the program and shows the rate or spectral distribution of the music or signals that make up the mix in real-time.
Where are they placed?
In the context of mixing, they are usually placed as an insert in the last place of our chain of the master fader if it had processed; the same applies for mastering.
It is usually used in this way because if we want to see what happens with a particular channel of our mix, we just have to place it on its own and how the master fader receives signals from all the channels it would be analyzing.
How to interpret a spectrum analyzer
One of the most important things when it comes to making a correct reading of a spectrum analyzer is to know what you have to show us. In the FFT type analysis, we will see the electrical part of the audio signal, so we have to understand how to interpret what happens to get the most out of the tool.
We then have to understand how our ears work regarding frequency first. The ears do not have a “linear” perception of the spectrum, that is to say, specific frequencies, the low ones, it is more precise, and small changes of rate are perceived. However, in the acute frequencies, it does not have the same precision, and it is difficult to distinguish small changes in those areas.
To exemplify this, try to generate in your DAW a pure tone, or sinusoidal signal, for example, 100 Hz and a 110 Hz. Memorize the frequency hop and create a signal of 1000 Hz and a 1010 Hz, listening to what is happening. What you will notice is that the jump between 100 and 110 Hz is notorious; however, when we go to 1000 Hz, the same frequency jump almost is not perceived.
Example: In the audible case, we can first hear a sinusoidal signal of 100 Hz followed by a 110 Hz, then a 1000 Hz, followed by 1010 Hz. Note the frequency jump between the first two signals, low frequency, and the increase between the last two barely perceptible.
What we have is that the ear has a perception that resembles a logarithmic response as far as the frequency is concerned and, more specifically, the ear associates the rates by frequency octave intervals. The octave corresponds to twice the previous frequency.
For example, if we start at the lower limit of hearing, the 20 Hz, the next octave are 40 Hz, the next 80 Hz, 160 Hz, and so on up to the 20 000 Hz corresponding to the upper limit audible. Let’s see the frequency octaves and the frequency falls between them:
Frequency octaves in Hz:
20 40 80 160 320 640 1280 2560 5120 10240 20480
Frequent jumps between octaves in Hz:
20 40 80 160 320 640 1280 2560 5120 10240
What has all this to do? The answer is a lot. As we saw before, the ear relates the frequencies by octaves with what at low frequencies the jumps are small; for example, from 20 to 40 Hz, there is a jump of 20 Hz (if we put it differently, there is a jump of 20 frequencies).
In contrast, from 10240 to 20480 Hz, there is a jump of 10240 frequencies or Hz, significantly more. This is the reason why the ear is much less sensitive at high frequencies, and it is because it associates a large number of rates and does not distinguish very much what is between them.
Now with this last in mind, let’s take it to the operation of the analyzers. Since we measure the electrical part of the signal, we measure electrical power. What matters is that when we double the power, the level is increased by 3 dB.
The latter acquires meaning because if we look at the frequency falls, we mention; in each octave, we go up, we double the number of wavelengths, therefore the amount of power. So every time we go up the octave, there is a 3 dB increase in the perceived level.
What this means is that if we see a straight line in the analyzer, what we are going to hear is a very high-pitched sound, since we saw that increasing in octaves increases the level by each octave by 3 dB upwards.
White noise and pink noise
The type of signal that is going to be represented as a straight line is the white noise, which, as a character, has the same amount of energy in each frequency. That is the same energy in 20 Hz, in 21, 22, etc. Therefore, this noise will have an increase of 3dB per octave and will be heard with a lot of energy at high frequencies.
Therefore white noise is a signal with all the spectral or frequency content and with equal energy by rate, which makes it an electrically flat signal. While this noise is useful in audio applications, it will not be helpful to refer our mixes to it. For that, we are going to support ourselves in the pink noise.
Example: White noise, notice that it is heard with many highs.
The pink noise is a noise that has the same amount of energy per octave frequency and therefore is heard flat; it is understood with an equal or similar amount of energy in low, mid, and high frequencies. In essence, this noise is the white noise to which a low pass filter with a slope of 3 dB per octave is applied.
What we are going to see in the graphical representation of the spectrum analyzer is a line that from the low to high frequencies has a slope or fall of 3 dB per octave.
What we have to do is look for our mixtures to resemble in the content or spectral distribution to the pink noise, to obtain a smooth mixture. Finding that our mix is flat or neutral is very important because it is what will allow when we bring to another system the music is always heard well, without excess or lack of energy in the different frequency areas.
Parameters of the spectrum analyzer
We are going to take into account an analyzer of the FFT type, and we will break down the parameters that we need to know. For example, we will see the RND XL Inspector analyzer, let’s see:
- FFT Size: This parameter controls the accuracy of the analyzer. The higher the number of the FFT, the more accurate the analysis is and especially in low frequencies, which are the most crucial. Usually, the size of 8192 samples is sufficient to handle and analyze music. When increasing the size of the FFT, the processing required by the analysis is greater. In other programs that contain spectrum analyzers such as the Izotope Ozone, you have to go to options, spectrum and select 8192 in the window size, for musical analysis. The same happens if we are, for example, in the Ozone Insight. To access the preferences, you have to go to options and spectrogram and then select the FFT Size of 8192 samples.
- Peak and Peak Hold: This function allows us to analyze sounds with high transient/transient content, such as acoustic drums or percussion. In this case, you have to take the analyzer to the peak display mode. In this mode, we are presented with a much faster view of the spectrum, and we can see what happens with this type of signals, especially at low frequencies. With the peak hold function, it maintains the maximum value reached by the signs.
- Average: This is the function that we must activate when we want to analyze material with arms content mainly. For example, voices, guitars, pianos, basses, and all instruments with long envelopes. What the graph shows us is what happens in a certain period, the average of the signal. In general, when activating this type of measurement, the message has an integration time, and then the visualization is presented.
- Logarithmic and linear view: This type of analysis will serve to analyze the low frequencies since it resembles the graph to the response of the ear or as we listen. If, on the contrary, we want to see what happens in high frequencies, we have to spend at sight in linear mode, which gives us a more excellent definition of the graphic in high rates, since it provides more space for visualization because it divides the frequency linearly.
- Weighting: Some analyzers allow us to use contour curves A, B, or C to visualize the measurement of the spectrum. When we activate any of these curves we will be able to see what happens at a low level, Curve A, at medium level B and high levels of sound pressure C. For standard measurements it is a good idea that there is no contour curve and the Analyzer in its flat response.
Spectrum analyzers are a potent tool for visualizing the frequency spectrum of the sources that make up our mix and the entire mix.
They serve as support for our filtering, equalization and to visualize the timbre response of different instruments or sound sources. By using an analyzer, we can know for sure what portion of the signal we are mixing is low-frequency noise and what is necessary to conserve.
On the other hand, it is crucial to understand the operation of the analyzers and especially those of FFT, and to know how a mix has to be seen if it has a flat response. For this, we help with the pink noise that is the noise that corresponds with extended listening.
As we mentioned at the beginning, using a spectrum analyzer correctly does not guarantee that our mix sounds good. If not, it is a tool that allows us to make decisions with better criteria, especially if we are not sure of our monitoring system or the acoustic response of the room.
As we say, always try this tool to see their different functions and draw their conclusions to start using them in their productions.