Audio spectrum analysis can be a confusing topic for someone just starting out in music production. It sure used to confuse me – I remember that very vividly!
I wrote this article to give you some pointers on the different ways that I like to use spectrum analysis to aid me in the music production, mixing and mastering processes.
Spectrum Analysis is a Double-Edged Sword
Before we go into how to actually use spectrum analysis, I want to point out something very important.
The process of making music has become increasingly more visual in the past decade. Huge developments in computing power, screen technology and software (not to mention our natural bias for shiny new things) has made sure we’re being treated with ultra-precise visual feedback of most things we do inside a DAW now.
It’s a double-edged sword!
Why is it a problem?
In my mind there are three things contributing to it:
- We humans are very visual creatures by nature. We tend to prioritize our vision over other senses. Unconsciously, we prefer to ground our trust and decisions on what we see.
- Our brain has limited processing power. We are only good at properly focusing on one thing at a time. If you’re allocating most of your internal CPU on looking, you’re not listening as well as you could (and should). That is just how our brain works.
- In reality, sound (and especially music) is always a lot more complex than any visual representation of it. We are getting fooled by the fireworks. Most importantly, visual feedback doesn’t represent the emotional content of your music at all (which comes back to point #2 – if you’re looking too hard you forget to listen).
Here’s the punchline guys:
Don’t make music with your eyes.
You should only use visual feedback as a supporting tool to either confirm what you hear (or don’t hear), or to spot possible problems.
Make visual feedback work for you and not against you.
After all, all your audience has is what they hear (and how that makes them feel).
Using Audio Spectrum Analyzer in Music Production and Mixing
Now that we have the important basic premise established, we can begin to go deeper into the why, when and how to use audio spectrum analysis.
There are several situations where spectrum analysis can be extremely helpful. Let’s take a closer look.
Revealing Problems on Individual Channels and Instruments
This is the #1 thing that I use spectrum analysis for. I don’t always trust my ears completely, and my listening environment isn’t perfect. So spectrum analysis is a great help for indicating potential problems.
Here are the most common in-the-mix analysis situations for me:
- Checking every channel for unnecessary low end content (rumble or hum).
- Checking bass instruments for notes that are unintentionally louder or quieter than others.
- Identifying the exact frequencies of unwanted resonances or peaks (I then deal with them using a good surgical EQ – this is what I use).
- Identifying over- or underpronounced high end. Normally I should be hearing this too, but in long sessions we get used to hearing things a certain way. Be careful of ear fatigue! Constant referencing also helps here.
Example (problems in a vocal recording):
Pictured above is a screen capture of a vocal sound with three problems.
- The hum/rumble in the low end is clogging up the mix.
- The vocal sound is possibly missing some bottom end (of course this could be on purpose). You should be able to hear this without looking at the spectrum, too.
- There is a nasty resonant peak in the top end near 20 kHz. This peak is unrelated to the natural resonances of the voice that you can see in the 500 Hz-10 kHz range. The response should be rolling off neatly towards the high end.
Standing Waves and Working In a Compromised Listening Environment
There are a few fortunate people out there who get to work in perfectly tuned studios (if you’re one of them, congratulations).
As for the rest of us, we have to deal with less than ideal room acoustics – possibly even ambient noise.
Spectrum analysis can help us in dealing with this issue. First let’s look into the problem in more detail.
It’s fairly easy to improve the acoustics of an untreated room noticeably, but that will only take you so far.
Most often the hardest thing to deal with is an uneven bass response in the room. This is caused by the accumulation of so called standing waves between the walls, the roof and the floor. It plagues most rooms.
Low frequency sound waves are physically long and as they bounce around the room they cancel each other out at places and accumulate at others.
You can hear this as the bass gets louder and quieter as you move around the room.
The accumulation of standing waves depends on the dimensions of the room, the building materials (wood is good, concrete is terrible) and your listening position.
Standing waves are difficult to kill. The usual acoustic panels won’t help with low frequencies. It usually takes a lot of measuring and well designed heavy duty bass trapping to get the low end under control.
Standing waves can become anything from a minor nuisance (if you’re very lucky) to a major headache (more likely).
So long story short: standing waves cause you to hear some of the bass frequencies louder or quieter than they really are. The peaks/dips can sometimes be up to 30dB!
You can imagine how that can really screw up what you are hearing. It can cause you to make poor mixing judgements as you try to compensate by boosting/attenuating the bass in your mix.
Now, does it mean you are unable to obtain a good mix in a room with an uneven bass response?
Not at all!
First, the better you know your monitoring environment, the better you will mix in it. The more you work in an environment, the better adjusted your hearing will become to it. Measuring a room’s frequency response also helps a great deal. Get the data if you can (but make sure you do it right).
Second, spectrum analysis can be a great help.
Using a spectrum analyzer you can confirm visually wether a particular bass note/frequency really is louder/quieter than others in the mix, or wether it’s your room playing tricks on you.
Working With Less Than Ideal Monitoring
Most near field monitors do not reproduce the lowest end of the spectrum at all. Depending on the speaker, the low end usually begins to fade out anywhere from about 90 to 50 Hz.
So you get gradually less and less bass as you go lower on the frequency spectrum.
Even if your monitors go pretty low, they might not be very accurate in the lowest end.
The way around this of course is to use a properly tuned subwoofer in conjunction with your near field monitors. But that is not always possible.
Again, spectrum analyzer is your friend in this situation.
It pays off to fire up your analyzer every now and then to have a look at what is going on in the lowest end of your mix.
You might be surprised to find rumble and noise going on that your monitors are not reproducing. This is especially true with any live recordings.
Revealing Ear Fatigue
Ear fatigue is a silent killer. You don’t really notice it as it creeps up on you slowly but surely in long sessions.
The louder you monitor, the worse it gets (please, for the sake of avoiding hearing problems, do maintain moderate levels).
For me ear fatigue seems to present the biggest problem in the mid-high/high end of the frequency spectrum.
It’s easy to get used to a harsh sounding hihat, or the lacking of highs altogether. Once again this causes you to make wrong mixing decisions as you begin to compensate where you really shouldn’t.
We simply can’t trust our ears completely over long sessions.
Quite often I literally get snapped out of ear fatigue when I take look at the spectrum analyzer that always sits on my mix bus.
Looking at the analyzer there, I might see overpronounced peaking in the high end or alternatively not enough stuff going on. In the mix bus, these for me are signs that something might not be right.
Of course it doesn’t automatically mean something is wrong. But the case usually calls for further investigation and referencing.
Here’s a tip: the best way to combat ear fatigue is to take regular breaks away from your workstation. It’s also great for your brain (helps creativity and concentration) and your body (back problems, anyone?).
Here we are approaching a territory that is potentially very deceiving: using spectrum analysis for comparing a reference mix to your own.
You must not jump into conclusions when comparing the spectrum graphs of two different songs.
It is not possible to make your track sound the same as the reference mix simply by trying to make the spectrum look the same.
You have to listen and understand what is going on under the hood in both tracks in order to get there. The visual feedback can only point you in the right direction at best.
You have to consider the content, the entire dynamic chain of events happening in the mix and the reasoning that the person behind the mix has had for making those particular choices.
If your reference music is mastered you have to take that into account as well (the dynamics can be different to your raw mix).
When done right and in conjunction with careful listening, visual referencing can definitely shed some light on potential issues and help you stay consistent.
Because of this it is also a great learning tool!
When referencing it’s important to choose a track that not only sounds great. You should also use reference material you know well.
Try to use the same reference songs as much as possible so that you can really internalize their individual characteristics. This makes comparing your own material to them so much easier in the long run.
Here are a few things that I generally look/listen for when referencing:
- Is the frequency response balanced? I am always especially concerned about the bass levels. Note: balanced here does not mean even. It’s quite normal to have the high end of the mix gradually roll off, for example. Wether this happens or not depends on how the particular analyzer you are using is calibrated (look for a setting called “slope” or “tilt”).
- Are there any obvious gaps or peaks in the spectrum? If so, why? It could be for a good reason, or there could be a problem.
- Is the snare/kick punching through enough (depending on the material I might not be concerned about it)?
- How do the Mid and Side spectrums compare? I generally try to avoid getting Side content in the lowest end of the spectrum, for example.
I always have Magic AB sitting on the mix bus, ready to go with my reference music. After Magic AB in the chain I have my spectrum analyzer.
This way it’s very quick and easy to pop both of those open anytime and start referencing.
Note: Before routing to the final mix bus I often have my mix go through another bus in Pro Tools. This allows me to set up the spectrum graph of my mix and the one of my reference music to display in the same analyzer window using the sidechain function in FabFilter Pro-Q2 (some other analyzers have this feature as well).
Pictured above I am comparing my mix to some reference mixes I have running from the Magic AB plugin.
Fabfilter Pro-Q2 is set to show the original in red (via sidechain) and the reference mix in grey graph.
I am seeing three areas that call for some further investigation and possibly crafting:
- The sub bass is almost 6 dB louder in the reference mix.
- The mid/mid-high range is louder in my mix.
- The reference mix has more top end. The top end on my mix resembles a roll-off you would normally get on vinyl, which isn’t necessarily bad – just different.
Please note that in reality you should never compare static snapshots. Instead, see how the spectrums behave live. A loud snare or sub bass hit can drastically alter the curve from one moment to another, for example.
In a referencing situation it helps to set the spectrum response to a fairly slow speed so that it displays an average reading over a few seconds. This gives you a more balanced behavior and a better overview of what is really going on.
It is also extremely important to match the levels of the two tracks as close as possible. RMS meter can help there, but always use your ears to adjust.
Look at the example picture above to illustrate this. After matching levels by ear the reference mix (B) is showing less RMS than the track I’m working on (A). This is because the reference track is a very loud mix and has been mastered so I’ve had to bring it down quite a bit. The frequency balance of the two mixes is also different as you can see from the spectrum, which also affects how the RMS meter behaves. The RMS readings are different but judging by ear, the two are playing back at roughly the same level in my listening environment.
Measuring Room & Loudspeaker Frequency Response
You can use spectrum analysis to help reveal the real frequency response of your listening environment.
Warning: You may not like the results. It’s very common to have large bumps and gaps in the low frequency response.
For this I recommend using a great donationware application called Room EQ Wizard (available for PC, Mac and Linux). You will also need an SPL meter that has been calibrated for the software (REW comes with calibration files for some common and affordable SPL meters).
I bought a basic cheap Radio Shack meter and it works great (again, make sure there is a REW calibration file available for meter you are getting).
REW will guide you through the measurement process. The basic idea is very simple – a controlled sine wave sweep from 0 Hz to 20 kHz is played through your speakers and the SPL meter records the sound pressure level at different frequencies. REW will then create visual data that reveals the frequency response of your room in all it’s glory.
Note that the results of your measurements depend very much on the position of the meter/microphone. Make sure to measure exactly at your listening position.
If you don’t want to go that far, you can get some indication of your room’s response by simply playing a sine wave sweep and listening for peaks and dips in loudness.
Spectrum Analyzer Software
Here’s some audio spectrum analyzer software and a couple of other tools that I can recommend.
This is what I am currently using as a spectrum analyzer (and, of course, as an excellent EQ). I love the clean and large analyzer display, I love how easy it is to zoom in and out and the fact that you can even make it full screen.
The EQ functions are a whole different topic altogether. But the spectrum grab function has to be mentioned here. It allows you to grab peaks of the spectrum and simply pull them up or down to create EQ nodes. Great for killing resonances.
This is the holy grail of visual metering. I’ve never afforded to buy it myself, but if you want the best and have the budget: there it is.
This is a great free spectrum analyzer. It is very configurable to your specific needs and tastes. Span also has a very valuable loudness meter that can display dynamic range (just turn on the Density mode in the settings).
There’s lots of free spectrum analyzer software out there but frankly, with Span, it’s not worth to bother with anything else.
This plugin has made referencing so much easier for me. It allows you to load up to 9 reference tracks at a time to do quick A/B referencing between them and your mix. And you can save presets to quickly load different sets of reference tracks.
I always have Magic AB sitting on my mix bus, followed by a spectrum analyzer. This allows me to also inspect the spectrum of my reference material and compare to the mix I’m working on.
REW has been a great help in figuring out the frequency response in my studio rooms for almost a decade now.
It’s donationware. If you have not measured the frequency response of your workspace yet, you should definitely get to it.
How Are You Using Spectrum Analysis?
Any questions? Is there something I’m missing? How are you using spectrum analyzers when making music?
I would love to hear your thoughts. Let me know in the comments.