# Can spectrograms tell us why one song is "more dynamic" than the other?

I have two songs shown here in a Spectrogram. The first one is significantly more dynamic than the second one. The song feels like it has more dynamic range, more mixture of different sounds/instruments, more energy, more lively. But judging from the graph, I am not able to tell the pattern that leads to this.

## Analysis in Numbers

Simply put, dynamic range is the difference between the lowest and highest amplitude values of a given audio signal. - You could say it's the difference between the "quietest and loudest parts", but to amateurs, loudness is quite subjective and to engineers, it can have a definite meaning with a unit, which is not always reflective of the waveform's amplitude. For example, the "common" compressor (Peak) compresses the dynamic range using a threshold that is set as an amplitude value and attenuates any signal that passes the threshold (with a customisable in/out ratio and trigger envelope). So you're not directly affecting "loudness", that's just a consequence of changing amplitude values. So, in this way, it's technically more precise to say "lowest and highest amplitude values". That's the actual height of the wave, so there's no ambiguity.

So anyway, with the jargon laid out, we can now look at the displays and see that we are given amplitude values as a colour spectrum. So, although it's not an "ideal" situation, if we could accurately recognise these mapped colours, we could not only calculate the dynamic range, we could also calculate its loudness properties and anything else that requires not only frequency, but amplitude and time values to calculate. It all depends on the accuracy (and resolution) of the display.

So what values can we, as humans, generally read from audio signal graphs?
(Low to high numbers = most to least accurate/easily read in general)

Spectrograms (Frequency domain)

• Displays interpreted frequency spectrum changes over time (via a window function). (Can be adjusted for greater temporal OR frequency precision)
• uses a colour map to represent amplitude values in a time/frequency display
1. Frequency
2. Amplitude
3. Timing

Frequency spectrum analysers (Spectrographs++) (Frequency Domain) - Kind of like a temporal slice of a spectrogram.

• Displays an interpreted frequency spectrum in a given finite interval of a signal (via a window function).
1. Frequency
2. Amplitude
3. Timing >((Finite) Time Interval)

Waveform displays (Time domain) - The most common and temporally accurate audio display.

• A (usually interpolated) representation of a sequence of digital amplitude values, creating a display that is analogous to a sound wave.
1. Amplitude | Timing | Phase/Polarity
2. Frequency

So we can use different signal analysers depending on what values matters most to us. And while spectrograms are not the most accurate display types for interpreting dynamic range values, we can certainly get a rough estimate of the relative differences in dynamic range between two signals using these spectrograms, especially if the same spectrogram settings were used to display both signals. These values will ultimately tell you why one seems to be "more dynamic" than the other.

## Analysis by Eye?

Although a logarithmic scale would be better, I can still tell by just looking at these images that the first should sound "more dynamic" than the second (I'm not just talking dynamic range as a measurement here, but how the track feels, dynamically). How? Maybe it comes from experience, maybe I just understand what the characeristics of dynamics look like in a spectrogram, just as I know what it looks like in a sound wave display. Maybe it's because the image itself is more dynamic and the harmonics are more defined, which is directly reflective of the track's dynamic amplitude properties. In the end, It could be a mix of all of these. But I'm sure you can learn to see it, too. Just recognise the colour scale, and imagine how the relative depth of different sections and the definition of harmoics, gaps, fades, decays and highlights would probably sound. Recognise the level the elements fall back to. The frequency properties are there, too, for you to imagine, Albeit in a linear scale.
So analysis by eye is a difficult one to explain and is a little subjective, I suppose. Maybe it's one of those instinctive things you just have to learn with experience.

# In Summary...

So although you could interpret the dynamic range by studying the colour map and numbers of a spectrogram, it's not a very practical or accurate way of doing it. If you wanted to recognise different dynamic characteristics by eye, it is possible to do so. You may be able to recoognise variance scales and limitations by viewing the dynamic colour depth of the image, but this kind of instinctive analysis by eye usually has to come with experience.

Note: There is also a 3D spectrogram. I saw one of these for the first time in an iZotope analyser plugin called "Insite". But it can safely be classified as a spectrogram in this case. The only difference is the amplitude values are displayed as height rather than colour.

++ I'm sure the term "Spectrograph" was used to technically refer to a spectrum analyser graph, but I can't find any references to this online, so I may be wrong, maybe someone with better knowledge of the past could clarify this :)