The first few times I looked at a frequency response of a room measurement I thought something must be broken.
That can’t possibly be real. It literally looks like a mess!! It looks like the aftermath of war!
That’s what I’m hearing??
I’m in serious trouble…
How can I possibly make any sort of valid judgment based on THAT?
I thought my speaker was flat!!
Well, measurements can be brutal.
Don’t panic though: Interestingly enough of all the things you can measure the frequency response doesn’t actually tell you all that much about what you are hearing. It’s just one small piece of a much bigger puzzle.
In this article I’ll debunk some of the myths around the frequency response.
By the end you’ll know:
- What exactly the frequency response shows you in the first place.
- What a good frequency response looks like in a typical home studio so you know what to expect.
- How what you see translates into what you hear.
- What matters about your room’s frequency response and what doesn’t.
The bathtub model
Imagine a bathtub filled with water. Drop a pebble into the water at one end and watch the waves travel over the surface. Soon the wave fronts reach the side of the tub, get reflected back and start combining with the waves reflecting from the other sides.
The water surface gets chaotic pretty quickly.
Now imagine holding a small stick into the water somewhere, and measuring the water level at the stick as the waves pass by it. Some waves are high, some are small. Some waves are close together, some are further apart.
Let’s say you count the height of, and distance between these waves passing by the stick for 3 seconds. Then you collect this data in a graph. For each wavelength (ie distance between wave peaks) you show the average height of the wave that you encountered during those 3 seconds.
That’s basically the equivalent to a frequency response measurement of one single speaker with a single microphone placed randomly in a room.
One pebble, one stick.
As you can already tell, that information is pretty limited in its usefulness. It only gives you a glimpse of what is happening to the waves in that bathtub.
And if you throw 2 pebbles in the water instead of one (stereo sound baby!) it gets even more complicated to deduce anything. So although there are certain things you can tell by measuring both speakers simultaneously, I prefer to measure just one at a time.
Of course in the case of sound, all this happens in three dimensions unlike the bath full of water which functions across only two dimension. Instead of waves on water, sound is more like quickly expanding bubbles around the sound source.
But for your measurement, you still only get to look at a single point in space with your microphone. You still only have that one stick.
A tiny window into a huge world
This constraint is the basis for a first important lesson. A frequency response graph only tells you what is happening at the exact spot where you place the microphone for any particular speaker position.
So if you want to know what is happening at your listening position, that’s where you need to put your microphone, and your speakers should be set up for stereo listening.
There are other microphone/speaker positions that can tell you valuable information about a room, but for the sake of this article I will focus purely on the listening position, measuring one speaker at a time.
What a real home studio frequency response looks like
Here are the frequency responses of the left and right speakers from a recent build, measured separately. The room is a rectangle of 5.8m x 4.15m with a ceiling height of 3.07m. So around 24m2 with a volume of 74m3. Pretty standard. Windows running along the entire right wall.
Each response is smoothed to 1/48th octave. Notice the 5dB scaling per line on the vertical axis. The entire plot only spans a range of 40dB (from 60dB to 100dB), just enough to show all the data. That way we have maximum resolution on screen.
We found the listener position using the Bass Hunter Technique and that’s where I placed the microphone for the measurement.
In this case we placed the sweet spot to the left of the center axis of the room. There simply was no choice as a balcony door in the front wall was supposed to stay accessible. So much for left-right symmetry. Sometimes you just have to compromise.
What do you actually hear in these messy graphs?
Well, simply enough, you hear the parts where there is energy (the peaks) and you don’t hear the parts where there is no energy (the dips).
You brain will connect the dots, and that’s what you end up perceiving.
In practice this means you can imagine a line running across the top of the peaks. That is the actual frequency response „visible“ to your brain.
So I call it the perceived frequency response.
I first learned this from Merlijn Van Veen during his level 1 seminar on sound system design (who btw you need to check out. King of the sound nerds). I remember sitting there thinking “mind blown” and “duuuuh” at the same time.
In terms of energy balance in your room, this is what actually matters. It determines if you’ll describe your room as “missing low end”, “a little aggressive in the mid range” or having “too much sizzle in the high end”.
Note that the effect on what you are hearing is similar to using EQ on a mix.
For example, missing low mid energy will make the high mids sound aggressive although they might be perfectly fine. Or missing bass will make everything sound very mids and highs accentuated, although they are actually fine.
In this case, going down from the top, we’ve got a slight dip at 2kHz followed by a gradual build up to around 125Hz, with another small dip at around 650Hz. From there it gently drops off to around 25Hz with another dip at 75Hz.
At least on the right speaker.
The left speaker adds a slightly more erratic low end bump at around 45Hz.
But in both cases we’ve got usable energy down to about 26Hz. Fantastic. No need for a sub.
There’ also a general trend for more low mid energy in the response. As a consequence you might describe the sound as somewhat “beefy”, perhaps even a little “muddy”.
So what can you tell from looking at this frequency response?
We’ve got A LOT of comb filters. Basically that’s why it looks so jagged.
Comb filters on top of comb filters on top of comb filters.
Each reflection path from the room creates a comb filter. When a reflection interacts with the direct sound at the microphone membrane, the two sound waves interfere and sum together or cancel out. The exact frequency and magnitude of the effect depends on the travel time of the reflected energy relative to the direct sound.
And since there are A LOT of reflection paths you end up with A LOT of comb filters destroying your frequency response. They all just pile up on top of each other. Often the peak of one comb filter covers up the dip of another. In effect you’ll be hard-pressed to tell them apart in any meaningful way.
The important thing you need to note is that the cancellation between the direct and reflected energy doesn’t happen in the reproduced source material (i.e. what your loudspeaker puts out), it happens at the membrane of your microphone or your ear.
This phenomenon has an interesting consequence. Comb filters cannot be corrected with EQ. Any change in EQ to your source material will affect both the direct sound AND the reflected sound in equal amounts. But the interference of the two at your ear won’t change and so the resulting comb filter won’t be affected.
EQ cannot compensate for comb filter effects on your frequency response.
To reduce comb filters you need to reduce the reflected energy. And for it to have a noticeable effect on your frequency response you need to get rid of pretty much all of the reflections that happen in your room. Like by turning it into an anechoic chamber or a diffuse field. That’s what Non-Environment type rooms and Studio C at the Blackbird Academy try to achieve respectively.
But for most of us that’s way out of our league. And so it’s not something you need to pay much attention to at this point. As you will see, the measurement of the treated room looks similarly jagged. And it’s not a deal breaker either because of how your brain connects the dots and actually perceives the sound.
Are you going to get good stereo?
In comparison both traces overlap well down to about 90Hz.
As a result the sonic character from both speakers should be quite similar. Any difference in the response affects what one ear hears in relation to the other. That would reduce the detail of the stereo image.
The traces are also at the same volume so there should be no strong stereo image shift to the left or the right.
What’s going on in the bass?
Both speakers have dips and peaks in their low end that span almost 25dB.
In this case you can easily tell that you are looking at a room mode when the peaks and dips appear at the same frequency in both speakers, only at a different magnitude.
Note that you can only tell this because the responses for the left and right speaker aren’t the same. If you didn’t have that difference in speaker position relative to the room, you wouldn’t be able to deduce what is a room mode.
But reflected energy can also cause dips in the low end. They are typically very sharp and very deep. Above you can see one right around 100Hz.
You generally have to differentiate between two types of interference induced cancelations in the low end:
- Standard Comb Filters
- Speaker-Boundary Interference
Depending on the dispersion pattern of your particular speaker and it’s position, the most obvious dip is often caused by the reflection from the floor. In a typical near field monitoring setup this happens somewhere between 100Hz and 120Hz.
This is a standard comb filter. The reflected energy interferes with the direct sound at the membrane of the ear or the microphone.
As with any comb filter, the effective frequency of the cancelations is dependent on the distances of travel between the speaker and the microphone or ear. So both their positions in relation to the room are the determining factors.
Speaker-Boundary Interference on the other hand is independent of the microphone and its position. The reflected energy interferes with the sound at the speaker instead. It directly affects the output sound power and so in this case what ever is played by the speaker!
The effect of Speaker-Boundary Interference on the frequency response is the same, no matter where you position the microphone or listener in the room.
Although these two interference patterns are different in their nature they are still in effect the same thing: a low passed comb filter.
A comb filter because the effect is based on reflected sound interfering with direct sound, causing peaks and dips.
And it’s low passed because the reflected energy only carries low frequencies. The directional nature of the speaker simply does not contribute any high frequencies in the direction of the reflecting surface.
And, as with any comb filter, these low frequency cancelations cannot be corrected with EQ.
Going back to the perceived response, the difference between the left and right speakers is also strong in the bass region. At 45Hz there is a difference of about 8dB between left and right. At 50Hz it’s around 10dB! The same goes for the dip at 75Hz.
It’s this difference between the left and right speaker that is actually more concerning than the effect of the room modes on the individual responses.
Bass trapping will reduce the effect of room modes, but the difference in volume in the perceived response is due to the asymmetrical positioning of the speakers and sweet spot in the room.
More than anything, it is the position of the sweet spot and speakers in relation to the room that affects the overall balance of the frequency response. So positioning is the best way to solve problems in the perceived frequency response.
But in this case the asymmetry was a practical necessity, so it is simply something that the user of the room will have to live with.
What You Cannot Tell From The Frequency Response
Even pre-armed with all this knowledge, without any further investigation, you still cannot tell decisively what acoustic effect is causing a particular peak or dip in the response.
Every effect that I explained above, I deduced by looking at the frequency response in combination with other data like a Waterfall Plot or Energy Time Curve. And some effects I can only explain because I’ve done countless experiments and certain patterns kept repeating themselves.
The frequency response measured at the sweet spot simply does not give enough information on it’s own to positively deduce the cause of any problems.
You need more information. You need time related information. And that’s the one thing that isn’t shown at all in the frequency response.
That’s why any acoustic effects based on time (unfortunately pretty much all of them), cannot be diagnosed from the frequency response alone.
And that’s the main reason you shouldn’t get hung up on the frequency response. As a tool to diagnose problems, it actually tells you very little. It’s only in combination with other data that it becomes useful.
What a Good Frequency Response of an Untreated Room Looks Like
Ideally you want to see two traces for the left and right speakers that closely match. This means that both speaker “see” the same room (one being a mirror image of the other) and that the reflection patterns and room mode effects are the same for both speakers.
Overall the perceived response is smooth. It may show some gentle hills and valleys.
It can span a range of around 10dB starting at the very high end down to the very lows. Often you will see a gradual build up of energy starting at the top and ending in a smooth roll off in the bass that starts at around 100Hz.
Some more recent speaker designs put out a lot more energy in the low end and so may not exhibit any noticeable bass roll off. It just cuts off somewhere in the subs.
I call this a balanced frequency response.
In terms of energy it includes everything you need to reliably hear the entire spectrum of music. Ultimately this is what you want to see from a frequency response measurement. No less and no more.
A balanced perceived frequency response is the minimum you can get away with.
Once you’ve reached this point you can call your sweet spot and speaker positioning successful. Improving it further will now be down to acoustic treatment.
What About Speaker Calibration Software?
“Why don’t I just slap on some room correction software and be done with it?”
Let me address the elephant in the room first:
Speaker calibration systems work by adjusting frequency, not time. So they are inherently incapable of fixing any time related acoustic issues. And again, unfortunately pretty much all acoustic issues are based on time.
Room EQ can only cover up problems, not remove them. And it will do so for exactly one spot in your room while making everywhere else sound worse. Granted, the sweet spot is the primary position you should care about. But if our goal is a great sounding room, that certainly is a step in the wrong direction.
Second: Like I mentioned, dips in the frequency response caused by interference between the direct sound and a reflection cannot actually be corrected with EQ.
The software will most certainly try, but it will be very aggressive in the process, introducing nasty phase shifts, and robbing your amplifiers of headroom. Ultimately it will result in higher distortion which, on the path to critical sound, is the opposite of what you want.
Of course if your frequency response is radically unbalanced to start with, room EQ may get you in the ballpark. But then why not use positioning instead? It’s free, avoids the problems mentioned above, and sets you up perfectly for the next steps in treatment.
Not to mention the lasting benefits of learning about your speakers and your room in the process of doing the involved listening tests.
“Ok, but what about EQing the more gentle hills and valleys that remain even after proper positioning?”
The ultimate nail in the coffin for room EQ is, in my opinion, auditory memory. Since it is so short (under 5s), you are like a gold fish in a glass bowl when remembering what your room sounds like. Your brain will get accustomed to what it is hearing within a few seconds, which then becomes the new “normal.”
Switch on an EQ and within seconds your brain will forget the difference to what it sounded like before, and for all intents and purpose you are left with the same information that we had before.
Of course, no large chunks of information should be missing. But as long as the perceived frequency response is balanced enough to get us in the ball park, our brain will do the rest.
The one scenario where I can justify the use of room EQ is the simple matter of taste. If using EQ will better align the sound at your listening position to your personal taste, which will in turn let you enjoy the sound more and so have a net positive effect on your workflow, I’m all for it.
This will also involve broad strokes of EQ. But it won’t be for the purpose of creating some ideal frequency response with the unrealistic promise of accurate judgment and perfect translation. It will simply cater to taste. And who can argue with that.
So What Changes With Acoustic Treatment?
Surprisingly little! At least in the frequency response.
First and foremost you will see the effects of bass trapping. Bass traps dampen the resonant systems that make up your room modes. As a result you’ll see dips come up and peaks come down. Their tops and bottoms will also round off.
Simply put, the more bass trapping you apply to your room, the more it will even out your low end.
But apart from that comb filters will still dominate the entire response. You may even get “new” dips and peaks where cancelations used to be hidden by peaks from comb filters that you’ve removed through treating the first reflections.
The floor reflection dip will still be as obvious as before.
And if you are using a desk (who isn’t) you’ll still have a nasty comb filter starting in the lower mid range caused by the reflection coming off of it.
Summing up, here’s what we learned:
- The next time you’re freaked out by your room’s frequency response calm down and breathe. You good? Ok.
- Remember: It’s only showing you what is happening at the exact spot where you placed the microphone.
- Set your smoothing to 1/48th octave. 1/24th is ok too. Scale your vertical axis limits so you just cover the data. A range of 40dB should be enough.
- Expect to see lots of comb filters. But don’t worry, it’s not what you need to focus on right now. EQ won’t help.
- Instead “Connect the Dots”. The perceived frequency will tell you what you are actually hearing. It’s ok to show some gentle hills and valleys.
- Positioning is your friend.
- Compare the left and right traces to know if you are set up up for stereo.
- In the bass it’s normal to see peaks and dips span a range of 25dB. It will reduce with bass trapping.
- If you are wondering how the f*ck to tell from the frequency response what you need to treat, here’s the truth:
You can’t. Not without looking at time related data.
So the next time someone tells you to do “this” or “that” to your room simply by looking at the frequency response, you can safely ignore it. Because actually it’s impossible to tell what needs to be done from it alone.
 Toole, F.E., 2008. Sound Reproduction: Loudspeakers and Rooms. Focal Press.