Home studio acoustics

My journey in home studio acoustics

 

Introduction

One challenge in mixing is to produce a mix that not only sounds good in your home studio but also translates well to different audio systems such as your hifi system in your living room, your car audio system, or even you mobile phone. This is not as easy as it sounds since your monitor speakers and, more importantly, your room acoustics play an important role in the perception of your mix (sound) in the studio environment. If your monitors or room acoustics are suboptimal then during the mixing proces you will probably be correcting imperfections of monitors and acoustics instead of working towards the ‘best’ frequency balance. Consequently, a song may sound good in your home studio but may sound awful when listening in another room or on another system.

My home studio, like most other home studios, are far from acoustically perfect. However, to some extend you may improve the situation depending on the time and money you want to invest. But also your willingness to sacrifice other functionality of your room, if any, in favor of acoustics. In this post I discuss how I approached the situation in my room and how I attempted to create a better listening environment.

One lesson learned is that (the theory underlying) acoustics is extremely complicated. There is much information to be found on the internet but also books have been dedicated to this subject. See for example a selection of references given at the end of this post and the links given throughout this post. It helped me to get a much better understanding of acoustics. Yet, I have a long way to go for a really thourough understanding of the matter. Yet, this does not imply that you should abandon the whole exercise of improving acoustics, but getting a solid understanding of acoustics, including the physics and mathematics, and how to measure it will certainly help.

The foundation of this post was laid through several excellent lessons from Sound Education Nederland (SEN), in particular by Ignace Dhont (owner and main tutor at SEN), Ben Bok (acoustician with a wide experience in the design and consulting of acoustical critical spaces), and Mischa Jacobi (builder of music studios and of the SPHERE). But also discussion on social media with other students of SEN helped to shape my thoughts.

Disclaimer

I am still learning about acoustics, its theoretical concepts, the measurements, and the interpretation of the measurements. Thus, I don’t consider myself an expert on this topic although I might have gained some more knowledge then the average (home) studio owner. Therefore, some of my interpretations below might now be fully correct. Let me know. Or I need additional measurements to back up some of my interpretations and conclusions. Let me know. Nevertheless, I hope you find this post useful.

 

My room

Figure 1.1 shows the floor plan of my studio, which I explain in more detail later in this post. Figure 1.3 gives impression of the room. This room is located at the attic and is far from symmetric partially caused by the ceiling (slanted roof with a dormer on the front and back side). Initially, there was no acoustic treatment. The speakers are not placed symmetrically in the room, resulting if different reflections (sound coloring) from the right and left side. The desk also is a cause of reflections.

It is, however, not clear to me if this asymmetry is beneficial or not. Ideally, the speakers should be positioned symetrically in the room (which is clearly not my situation). On the other hand, the asymmetry may perhaps reduce standing waves and interfering reflections. But this is hard to determine and would require many more measurements and a control situation to compare to.

Basically, this room has four functions:

• Mixing and mastering music
• Playing keyboards
• Office for my daily work.
• Fitness (home trainer and treadmill; upper left corner).

This gives various constraints for placement of all items and acoustic treatment. Without going into detail, it is, for example, not possible to have the desk at the other side of the room (to increase symmetry for the monitors) since the current optimal placement of my keyboards and X32 would have to be sacrisfied. It is also not really possible to place bass traps in the corners or to move the desk to get a more favourable listening position (instead of being at 25% of the room’s length). It is what it is. I am not complaining since I have much space. Giving this situation I measured and improved acoustics. However, a more ideal listening position is shown in Figure 1.3

Figure 1.1 Floor plan of my home studio. Note that the proportions in the drawing are not correct.

 

Figure 1.2 Impression of home studio

 

Figure 1.3 Ideal listening position. Symmetric room and listener at 38% of the room length.

 

Characterizing my room

There are different approaches towards understanding the acoustics of your home studio.

One approach is to use different types of audio files (sine sweeps, pink/white noise, harmonic distortation, etc) to test for comb filtering, standing waves, resonances, definition, etc. The required audio can often be generated easily from your DAW, or otherwise you can download them from the internet (e.g., (here]).  You only have to know how to use them and what to listen for but, again, lots of information is to be found on the internet. I have used such audio files but don’t report on it in this post. However, some observations:

• Using pink noise, I clearly hear the effect of comb filtering in the higer frequencies when moving a few cm from the listening position.
• Using a slow sine sweep, I can clearly hear the effect of the comb filtering even at lower frequencies. Sometimes the drop in volume is quite significant. I don’t hear any clear resonances at normal volumes.
• Checking for standing waves (with sine waves of lower frequencies) makes clear that my room is not free of them. Bass trapps probably would help.
• Using a MATT (Musical Articulation Test Tones) signal I tried to determine the presence of degrading effects of room acoustics on the audio signal.

Already given these results it makes sense to check the mix on a head phone once in a while and to listen to the mix from different positions in the studio.

In addition, you may want to have a look at:

• Falstad, which is simulation software to acquire a much better understanding of sound waves (free)
• NorFlag, which is software designed for material acoustic performance estimation (but very expensive)

 

In this post I report measurements performed with Room EQ Wizard (REW). Many tutorial can be found on YouTub, including this nice introduction:

 

Room EQ Wizard

The measurements shown below were done with Room EQ Wizard (REW). REW is free software for room acoustic measurement, loudspeaker measurement and audio device measurement. The audio measurement and analysis features of REW help to optimise the acoustics of a listening room, studio or home theater and to find the best locations for the monitors, subwoofers and listening position. It includes tools for generating audio test signals; measuring SPL and impedance; measuring frequency and impulse responses; measuring distortion; generating phase, group delay and spectral decay plots, waterfalls, spectrograms and energy-time curves; generating real time analyser (RTA) plots; calculating reverberation times; calculating Thiele-Small parameters (define the low frequency performance of a loudspeaker driver); determining the frequencies and decay times of modal resonances; displaying equaliser responses and automatically adjusting the settings of parametric equalisers to counter the effects of room modes and adjust responses to match a target curve.

I used REW to measurement my current room acoustics. During measurement I have the ARC system turned off (basically, I only use ARC within Cubase but more about this below).

REW is free but, unfortunately, not open source. Based on my knowledge of the underlying mathematics, I made a started implementing some concepts (impulse responses, discrete fourier transforms, etc) as a Python program. This will be insightfull but is a long term project on which I will work if time allows.

For some of the initial measurements I used the Behringer ECM8000 microphone (uncalibrated) until it ceased to work. However, for most of the more important later measurements, I used the Beyerdynamic MM1, which comes with a calibration file that was used by REW during the measurements. I also tried the MEMS microphone from IK Multimedia but, surprisingly, this gave very different results compared to the ECM8000 and, therefore, I decided to initially continue with the  ECM8000.

I use the Behringer X32 as soundcard. Prior to REW measurements I made a calibration measurement of the interface’s frequency response according to REW instructions. This calibration measurement was used to correct all subsequent measurements. The MM1 and X32 calibration profiles are shown in Figure 2.1.

 

Figure 2.1. Behringer X32 (dashed line) and Beyerdynamic MM1 (dotted line) calibration profiles over the complete measurement range. Note the scale of the vertical axis.

 

All measurements were done with a logarithmic sine sweep as an alternative to an impulse (see also [here]). This sweep takes the same time to double in frequency, thus it takes the same time to go from 40 to 80Hz or 4kHz to 8kHz as from 20 to 40Hz.. The Measurement Sweep signal (length 256k samples corresponding to 5.9 seconds) consisted of a logarithmic sweep from half the start frequency (0 Hz) to twice the end frequency (44kHz) to provide accurate measurement over the selected range. Since the start frequency is below 20Hz the signal begins with a linear sweep from DC to 10Hz, followed by a logarithmic sweep from there to the end frequency. This sweep is shown in Figure 2.2:

Figure 2.2. Logarithmic sine sweep used in the REW measurements.

 

Note that for most measurements I did not calibrate the sound pressure level (SPL) and, consequently, precise comparison of the absolute levels of the various measurements is not possible.

Measurements were done before and after acoustic treatment. Also note that the control measurements (no acoustic treatment) were done using  the  Yamaha HS5/HS8 monitors and subwoofer (which was not turned off).  After acoustic treatment of my home studio, I also replaced the HS5/HS8 with the Kali IN-8 2nd wave monitors and no subwoofer was used. Consequently, the REW measurements performed after acoustic treatment show the effect of the acoustic treatment and a change of the monitoring system. The individual effects cannot be disentangled. For me this is not per se a problem since I only want to know if acoustics has improved. On the other hand, I won’t be able to make a hard conclusion on the effect of the acoustic panels.

 

Improving acoustics

Studio monitor placement

One improvement was to move my monitor speakers from my desk to monitor stands from Gravity behind my desk. In this way they are 125cm from the listening position, and the distance between the speakers is 135cm (because I have two computer screens in between). Alsmost the preferred equilateral triangle (equal angles and sides).

 

Polyester wool

I constructed 12 acoustic panels from polyester wool (50mm/40kg/m3/1200x600mm; from Zilenz).  Regarding health-related properties of this material, it is generally found to be safe (which is my final interpretation) but there are other considerations (see for example [here]).

 

Definition: sound absorption (copied from Wikipedia)

Sound absorption is the conversion of sound energy into heat. The sound then disappears into the material. The sound waves (sound pressure) deform the material, which costs energy. Depending on the type of material, the energy will partly be returned in the form of reflected sound. The energy that is not returned is stored in the material in the form of heat (mass spring system with damping). The degree of damping determines the degree of sound absorption. Thus, effectively, absorption reduces the amplitude of the sound.

The amount of absorption is a property of a material and is expressed by the absorption coefficient. The absorption coefficient of a material is the fraction of the incident sound power that is absorbed. The rest of the sound is reflected. The absorption coefficient depends on the frequency of the sound, and is usually measured at any octave band between 125 and 4000 hertz. The absorption coefficient has a value between zero (no absorption, all sound is reflected) and 1 (complete absorption, no sound is reflected – open window).

 

Absorbance of polyester wool

Figure 3.1 shows the absorbance coefficient of the polyester wool panels I used (green line).  The absorbance seems comparable with glass wool (see References below). The general recommendation is, however, to have 100 – 200mm of thickness. The fibreboard itself has only a very low absorbance capability. It has been experimentally determined that peak absorption of a frequency occurs when the material thickness is about one-quarter the wavelength of the wave (see [here]). Thus, for 50mm thickness this corresponds to a full wavelength of 200mm, which corresponds to a frequency of 1715 Hz (see [here]). Below this frequency the absorbance drops off. According to the graph below (green line) the peak absorbance of polyester wool is around 500 Hz, much lower than 1715Hz. Not sure why this is the case. Probably, in some way, this graph gives a far too optimistic representation of the absorbance. Having in mind the general rule of having 200mm thick panels, this corresponds to a full wavelength of 800mm, corresponding to a frequency of approximately 429Hz, which is more in line with the graph. 

Figure 3.1. Absorption coefficients of polyester wool for different sizes and mass.

 

Building acoustic panels

I glued the polyester wool panels to medium-density fibreboards (MDF) of 1210 x 610 x 4 mm (which has a very low absorbance; see here and the absorbance table in the References below; Figure 3.2). To several panels I attached small chains in each corner such that I can easily hang the panels to the wall. Finally, I wrapped the panel in cheese cloth, which is very loosly woven and, therefore, does not prevent the air to enter the material. Except for the two half panels in the slanted roof (which are glued directly to the roof) none of the panels are permanently attached and, therefore, can easily be moved. See photos below for the panels.

 

Figure 3.2. Construction of the absorption panels using polyester wool, medium density fibreboards, and chees cloth. 

 

In hindsight, glueing the polyester wool to the MDF was a big beginners mistake! Instead of glueing the polyester wool to the fibreboards, I should have assembled them in a wooden frame such that sound can pass through it (see for example here). The way I constructed the panels the MDF may actually serve as a reflector.  This is shown in Figure 3.2, where MDF creates much more comb filtering.

 

Figure 3.2. Upper left: measurement with polyester panel that does has not been fixed to MDF but has about 70cm of air behind it. Upper right: measurement with MDF. Lower left: frequency response (logarithmic frequency scale; no smoothing). Lower right: frequency response (linear frequency scale; 1/48 octave smooth window). The lower right graph clearly shows the increase amount of comb filtering of MDF compared to polyester wool. 

 

Intermezzo

Note that cheap microfoon reflection screens show the same problem!   Figure 3.3 shows a measurement with a microphone reflection screen that I used in the past (unfortunately). These reflection screens are supposed to reduce the amount of reflection during vocal recordings.  Clearly, the thin layer of foam is not capable to doing obsorbing any frequency in the vocal range and mostly will absorb top-high frequencies. To make things worse, the metal backside of the screen acts as a reflector! A REW measurement directly reveals this. Here, I put the measurement MM1 microphone inside the screen and placed my HS5 speaker in front of the screen at singing distance. Looking at the frequency response (linear frequency scale), the comb filtering is clearly present. If I remove the screen, the comb filtering is largely reduced. Thus, even in my problematic room I would better record vocals without the screen that I own. Probably there exist better screens but this one can be forgotten about.

 

 

Figure 3.3. Comb filtering produced by a microphone reflection screen. Red (without) and purple (with) the reflection screen. Smoothing window is 1/48 octave.

 

Additional acoustic panels

During initial measurements with the 12 acoustic panels placed in my studio (see below) I was not really satisfied with the result in terms of improvement nor with the fact that I placed two large panels to the left and right of my desk which was for several reasons rather inconvenient (Figure 3.4).

 

Figure 3.4. Initial placement of two acoustic panels to the left and right of my mixing desk.

 

Therefore, I decided to buy two smaller panels (600mm x 600mm x 150mm; Woody Kid) directly from Ekustic (Czech Republic) to replace these two side-panels and to hopefully get better readings on the acoustics. Ekustic is a small company that manufactures hand-made panels, and they were very responsive and helpfull. Their panels use Envizol (see document in References) to absorb sound. Envizol is nonwoven thermally bounded textile from recycled polyester fibers and BICO co-polyester binder fibers (Figure 3.5). Ekustic provided my with their measurement of a 100mm thick panel (Figure 3.5), which seems to do (much) better compared to the polyesterwool (Figure 3.1). But again, this does not correspond to general rule of peak aborbance at a quarter wavelength.  They informed me that from the experience of their customers, the 15 cm thick panels can comfortably treat frequencies even lower than 150 Hz. The efficiency at 100 Hz should be around α = 0.5 and that the full absorption is somewhere around 200 Hz.

 

Figure 3.5. Handcrafted Woody panel from Ekustic (front and back) with a thickness of 150mm. The absorbance material is Envizol (top left). Absorbance (from a 100mm panel) is shown in the graph.  

As a bonus, Ekustic provided me with some socks (against standing waves?)

 

Placement of the acoustic panels

I placed the acoustic panels at the places where I think it would best reduce early reflections (Figure 3.6). In the end I removed the acoustic panels to the left and right of my desk altogether and placed the Ekustic panels in the back of the room. Below the floorplan of my room with green indicating the location of the acoustic panels. Due to the way I attached the panels to the walls and ceiling there is a small layer of air behing them although this doesn’t help since I glued the panels to MDF! But leaving space behind the absorbance panels should generally improve the acoustics somewhat further (see [here]). Initially, I also tried to place two half panels on my desk to prevent the desk reflections but this is just too inconvienient.  

Two observations:

• If we would assume that my room is rectangular with dimensions 645 x 296 x 220 cm (l, w, h) then this gives the ratio of 1: 1.35: 2.93 (h, w, l). See [here] for calculations.  See also [here]. Figure 3.7 shows that the yellow dot (my room) is still in an acceptable area. However, due to the asymmetry of my room this calculation is likely to be very inaccurate. Indeed, using 350 cm for the width to account for the left part of the L-shape directly moves the yellow dot to a white space.
  • These calculations were according to Cox recommendations: Cox et al (2004) Room Sizing and Optimization at Low Frequencies, Audio Eng. Soc., 52(6), 640. (article at end of this post). These recommendations differ from the older Bolt recommendations: Bolt et al (1946)  Note on the normal frequency statistics in rectangular rooms. J.Acoust.Soc.Am. 18(1) 130-133.
  • Cox: 1.1W/H <= L/H <= (1,32W/H) + 0,44; W < 3H, L < 3H; 5% protective rule (values within +/-5% of integer values should be avoided).
  • Bolt: 3/2(W-1) < L-1 < 3(W-1); 2 < W+L<4.
• The 38% rule by Wes Lachot (a world class studio designer) suggests that the best listening position in rectangular rooms is at 38% of the length of the long walls [here]. If you can’t position your chair and desk 38% from the front of your room, you could try different positions at 36% to 44%, but not at 50% or 25% . Thus, this places me directly in a room node, which I can also measure (see below). Shifting the desk more to the back is not an option.

 

Figure 3.6. Floorplan of home studio with placement of  mixing desk, monitors (Kali IN8 and Avantones), instruments, and the Behringer X32. Note that the proportions in the drawing are not correct.

 

Figure 3.7.  Acceptable room ratios according to Cox.  Dark areas on diagram background correspond to room sizes with the smoothest modal characteristics; consequently, these corresponding proportions are more acceptable while building music-rooms and control rooms. “Best” proportions areas are shown in black color, gray color is given to areas of acceptable proportions, and others are shown in white.

 

Figure 3.8 gives an impression of the location of the acoustic panels.

 

 

 

 Figure 3.8. Placement of the mixing desk, acoustic panels, and keyboards. 

 

Finally, the REW measurments 

Below I show a selection of measurements that I have performed and a selection of the corresponding results. I believe these show the most important information about the acoustics of my room. I don’t show any phase plots since I still don’t understand how to use them in practice. Nevertheless, the group delay/excess delay plots below are based on the phase of the signal and have a much clearer interpretation.

 

Figure 4.1. Comparison of old situation with HS5/HS8S monitoring and no acoustic treatment to Kali IN-8 monitoring with acoustic treatment. Smoothing window 1/3 octave. No SPL calibration. between the measeruments that were taken at different days and different measurement microphones (HS5/HS8: ECM8000; IN-8: MM1). 

 

From Figure 4.1 I make the following observations:

• The HS5/HS8 peak around 70Hz is probably caused by the subwoofer being too loud (which I didn’t use in conjunction with the Kali IN-8).
• To account for the unevenness in the IN8 frequency response I determined the EQ parameters (frequency, gain, Q) with REW up to 1kHz. It doesn’t make much sense to correct above 1kHz since the small wavelengths (1kHz ~ 34 cm) makes this very position dependend. If have set the EQ on my X32 and measured again. EQ settings are shown below.
• There was no EQ proposed by REW for the dip at 67 Hz (probably caused by a null?). I tried to eq this as well but without success.
• The large peak at about IN-8 is probably caused by a standing wave and the fact that my listening position is at the most worst position (25% of the room).Indeed, when measuring 50cm behind the listening position, this peak disappears (data not shown).

Overall conclusion: coomparing the HS5 and IN8/EQ curves shows little effect of the acoustic treatment. The IN-8 shows a deviation of the average response of about 5-10dB, which is not too bad for a home studio.

 

EQ settings

Left monitor:

• Filter  1:        Fc   154.4 Hz  Gain  -7.00 dB  Q  3.147
• Filter  2:        Fc   244.0 Hz  Gain  -5.10 dB  Q  1.000
• Filter  3:        Fc   642.0 Hz  Gain   1.90 dB  Q  1.000
• Filter  4:        Fc   823.0 Hz  Gain  -4.60 dB  Q  3.980

Right monitor:

• Filter  1:        Fc   148.0 Hz  Gain -11.30 dB  Q  3.618
• Filter  2:        Fc   224.0 Hz  Gain  -6.00 dB  Q  2.367
• Filter  3:        Fc   795.7 Hz  Gain  -4.30 dB  Q  5.339

 

Figure 4.2. Comparison of old situation with HS5/HS8S monitoring and no acoustic treatment (green line middle)) to Kali IN-8 monitoring with acoustic treatment (red line top), and IN8/EQ (purple line bottom). Only left speaker is shown. Linear frequency scale up to 1.8kHz to show comb filtering. No smoothing. The measurements have between shifted apart for clarity. 

 

From Figure 4.2 I make the following observations:

• Although perhaps difficult to see, in the case of the old setup with the HS5/HS8 monitors and no acoustic treatment (green line), there was more (deeper troughs) comb filtering.
• The many dips in this graph indicate that the delay of the reflection is relatively long (see [here]).
• The EQ vs non-EQ are about the same although the EQ introduced troughs at other frequencies (for example around 1kHz).
• Figure 4.3 show the comb filtering for the left and righ speaker with a smoothing window of 1/24 octave, and also doesn’t show drastic differences between the three situations.

Overall conclusion: acoustic treatment and/or change of monitors have some effect on the comb filtering in favor of the new situation. However, the improvement is (very) minor?  From Figure 4.4 (1/3 octave smoothing) we might even conclude that the comb filtering is not really audible?

 

 

Figure 4.3. Comparison of old situation with HS5/HS8S monitoring and no acoustic treatment to Kali IN-8 monitoring with acoustic treatment, and IN8/EQ. Left and right speakers. Linear frequency scale up to 1.8kHz to show comb filtering. Smoothing window 1/24 octave.  No SPL calibration between the different measurements. 

 

Figure 4.4. Comparison of old situation with HS5/HS8S monitoring and no acoustic treatment (green line middle)) to Kali IN-8 monitoring with acoustic treatment (red line top), and IN8/EQ (purple line bottom). Only left speaker  is shown. Linear frequency scale up to 1.8kHz to show comb filtering. Smoothing window 1/3 octave. 

 

I wondered whether there were regions that are ‘minimum phase’  and are open for correction by EQ without introducing artefacts. This can be more clearly determined from the Excess group delay graph (Figure 4.5). When zero, the region is open for correction although this doesn’t make much sense for higher frequencies because of the sensitivity for listening position (short wavelengths).  Surprisingly, there is a large difference between the left and right speaker, which is probably caused by the asymmetry of my setup but is not immediately clear from the frequency response (Figure 4.1). For the left speaker there are some regions below a few hundred hertz with zero excess, which can be treated with EQ without doing too much harm. For the left speaker these correspond to the EQ settings proposed by REW (see above).

Figure 4.5.  Minimum group delay (light grey) and excess group delay (dark grey) for both Kali IN8 monitors. Modal frequenties at bottom can be neglected. Not that this graph starts at 2 Hz but is not really reliable for the very low frequencies. 

 

Considering the Kali IN-8 left speaker it seems that the frequency correction made by the EQ reduces not only the peak around 150Hz but it also reduces the ringing over the entire frequency range (Figure 4.6)! Thus, EQ correction in minimum phase regions does some goed for the sound. For these spectrograms I followed the guidelines from Ethan Winer:

• When using Spectrograms for comparisons you must uncheck “Normalize to peak at each frequency.” Also uncheck “Match top of scale to peak” in Appearance Settings. Otherwise the upper and lower SPL limits might change when you switch from one measurement sweep to another. In order to fairly compare decay times the SPL range must of course be the same for each measurement.
• The spectrograms are set for a time resolution of 600 ms, which equates to a frequency resolution of 1.7 Hz. Such a high resolution is equivalent to not using averaging on a frequency response graph.

Figure 4.6. Spectrograms of the Kali IN8 left speaker  without and with EQ correction of the peak at 150Hz. Top: 20-400Hz region. Bottom: full frequency range. 

 

I also had a look at the total harmonic distortion (THD). Interestingly, the EQ seems to reduce the THD in the range from 100 – 300 Hz, which is the area in which the EQ was applied (Figure 4.7). Apparently, the Kali now has to work less hard in this area, resulting in a reduced THD.

The THD is determined from the harmonic distortion components up to the ninth harmonic. These distortion components appear at negative times, behind the main impuls (in the Impulse Response graph; not shown).

Figure 4.7. Total harmonic distortion Spectrogram for the Kali IN8 left and right monitors. The grey line indicates the regions where THD could not be determined sincethe all harmonics were below the noise floor.

 

Next, I considered the reverberation time (RT60) which is a measure of how long sound takes to decay by 60 dB in a space that has a diffuse soundfield, meaning a room large enough that reflections from the source reach the mic from all directions at the same level. Domestic rooms (i.e., my home studio) are usually too small to have anything approaching a diffuse field at low frequencies as their behavior in that region is dominated by modal resonances. As a result RT60 is typically not meaningful in such rooms below a few hundred Hz. For the low frequencies it is better to considered the spectrograms (see above), Decay or waterfall plots.  According to https://www.powerestudio.com/reverberation-time-in-control-room-rt60/ the RT60 should be somewhere between 200 en 500ms, and should ideally be similar across all frequencies. Figure 4.8 shows the reverberation time in my studio, which actually has improved compared to the initial situation (HS5/HS8 without acoustic treatment).  It now even falls within the lower part of the recommended range, i.e., around 250ms. I am happy too see that the acoustic panels have at least some beneficial effect.

The lower reverberation is probably also the cause of the higher clarity after acoustic treatment as reflected by the C50 values (graph not shown). C50 represents the early to late energy ratio in dB, using the first 50ms for the early energy.

 

Figure 4.8. Reverberation time (RT60) for the old versus new situation. Values for the lower frequencies are difficult to interpret. 

 

Some thoughts

• After spending quite some money and effort in the absorption panels, their effect on the acoustics is kind of disappointing (apart from the reduced reverberation). I probably need much more and/or thicker panels (not glued to MDF) to see more significant results.
• I can probably improve the acoustics at the listening position by changing the location of the listening position and/or speakers. However, at this stage I am not willing to sacrifice other functionalities of the room that would be the result of that. Bass traps may also help but I don’t really have the space (yes, that is relative) to accomodate a sufficient number of them.
• EQ’ing seems to help to reduce this peak in the Kali IN8 frequency response without introducing large new artefacts. In fact, it seems to reduce the ringing over the complete frequency range, and to reduce THD in the area that was distorted. Probably, because the correction for the left speaker was made in a minimum phase area.
• The Kali IN8 are (hopefully) better monitors compared to the HS5/HS8S that I previously used. However, given the acoustic issues in my room, it remains to be seen to what extend I can benefit from this. We will see with the translatability of the next mix I make.
• Acoustics never behaves as in student text books. It is very difficult to get a grasp on this.
• All these efforts have learned by a lot about acoustics and a little bit about the underlying physics and mathematics. I am trying to fully grasp the math of some of the calculations done in REW in an attempt to write some software that demonstrates for simple (sine) signals the principle of (minimum) phase. It is and was fun. To be continued…….

 

Room calibration software: does it work?

In an attempt to improve the frequency response of the speaker and/or room, different companies developed software (plugins) that are claim to accomplish this task. Thus, similar to the EQ that I applied during some of my measurements to account for the peak between 100-200Hz in the Kali IN8 response, and which seemed to have a beneficial effect.

Examples of such software are SoundID from Sonarworks and ARC3 from IK Multimedia, which I used in the past (see [here]). Information from their websites show that ARC3 is positioned as software to correct room acoustics, while Sonarworks is more carefully positions its SoundID software for speaker calibration. In fact, they mention “SoundID Reference is not a substitute for acoustic treatment. We highly recommend treating your room before calibrating your speakers“. However, information provided by Sonarworks is strongly suggestive in the sense that acoustic problems are corrected with SoundID.  Also the fact that measurements are performed at the listening position (and not very close to the monitor) supports the notion that such software aims to correct both the speaker and room frequency response. In fact, there is no way to disentangle the effects of the monitor and room acoustics. This, and similar software, is widely use by (professional) mixing engineers in the field. Yet, there is a lively discussion whether or not such software can work.

The ultimate goal of speaker and/or room correction is to improve the translatability of a mix (or master) to different audio systems. Yet, there is no documented evidence (as far as I know) that the software actually helps with this. With evidence I refer to a test in which, for example, a mix was made with and without room correction (ideally incorporating different studios, mixing engineers, etc). In fact, my experience with ARC3 is that it didn’t improve translatability and I still needed endless iterations between different playback systems before arriving at a good mix (which is why I went to all the trouble to try to improve my room acoustics this time). On the other hand, there is documentation (i.e., REW measurements) that this software doesn’t work in the sense that is does create new artefacts in the sound. See, for example, the measurements of Ethan Winer ([here] and [here]), and the REW manual ([here] and [here]).

 

Yet, many (professional) mixing engineers claim that Sonarworks really helps. Some initial thoughts on this:

1. If you bought the software then you are probably already biased and you fool yourself in thinking that it indeed help (otherwise you just wasted your money)
2. I also tried SoundID (more about that below).  When I turn it on, I like what it is doing to the sound, that is, more bass and more high. And that is what we like. Similar to the loudness switch on your hifi. Thus, the temptation is to switch it one, but will it help you in doing a mix?

 

There was on interesting article in SoundOnSound about Sonarworks ([here]). Some quotes and observations from this article:

• Every 3dB of extra gain (to correct for dips in the frequency spectrum) implies a doubling of amp power, and of the monitor in terms of bass driver cone movement and port linearity. With the speakers that were used in this test they show that the second and third harmonic are increased with 20 and 27dB respectively; thus, increasing distortion.  In the end they conclude “I’m not entirely convinced that correcting for room acoustics, by distorting the monitors’ frequency response, is always a good thing”.
• Furthermore, “Sonarworks compensates for room effects by putting significant response anomalies in a monitor’s frequency response. and those anomalies will be imprinted on the sound that reaches the ears first” ……”there is little doubt that the first arrival is vital”.
• Yet, this article also mentions “Sonarworks, could be a live saver”, and “Recommended? Yes, but use with care”.

 

The observations from Ethan Winer can be added to these SOS measurements: trying to improve the frequency response introduces more ringing (longer decay times at other frequencies) thereby minimizing the definition of the audio. Therefore, to add to my two points above:

 

3. Even if Sonarworks (or other software, or manual EQ’ing) improves the frequency response objectively, it will almost certainly create other artefacts and different frequencies in your audio.
   • These artefacts are likely easily to miss as they cause ‘smearing’, less clear transients, distortion (perhaps only audible at higher volumes).
   • Thus, perhaps you want to take these less audible artifacts for granted in favor of the more audible changes in frequency response?
   • However, we should keep in mind that any artefact introduced by Sonarworks does not end up in the mix. Thus, if Sonarworks allows to make better frequency judgements then other artefacts might be of less importance.

 

In the REW manual there is a chapter devoted to the subject ([here]) explaining the issues involved in trying to correct your room acoustics. Just to quote their final paragraph:

“Given all the limitations we have uncovered, and with the problem of non-minimum phase on top, we might wonder whether the equaliser is any good to us. All is not lost, however. The non-minimum phase behaviour of the room is connected to the dips in the response. It means we are even less able to deal with them, but there wasn’t a lot we could do about them anyway, so we are really not much worse off. On the plus side, the peaks of the response are caused by features that lie firmly in the region our minimum phase equaliser can handle. We can use the equaliser to help tame the peaks, and the lower down they occur, the better the results we are likely to get – a nice complement to our acoustic treatments, since they start to struggle (or we start to struggle with the size of them!) at low frequencies. EQ is a useful tool to keep handy when trying to fix our acoustical problems, but it can only ever be a small part of the solution“.

I can go along with this suggestion. Thus, don’t aim to correct the dips and don’t overdo the correction of the peaks.

 

Some insight the I acquired while studying this subject is the following: audio has a frequency and a time (i.e., phase) component. EQ’s cannot correct phase but do change it. Consider the following example, a pulse (shotgun) and a sine sweep both contain all frequencies. In fact, they give identifical (flat) frequency responses.

pulse —>  frequency response

random noise –> frequency response

The sine components of the pulse have identical (zero) phase, while the sine waves that make up the random noise have random phase. Now suppose we make some changes to the frequency response with an EQ. The EQ doesn’t now if the source was a pulse or sine sweep. Thus, the EQ will make changes to the frequencies and at the same time make changes to the phase components. The effect of these phase changes are difficult to predict but certainly are not directed towards the reconstruction of the original phases (i.e., they introduce artefacts in the time domain (audio).

 

Some other quotes from the internet for your consideration:

• If EQing speakers would make them better, why haven’t speaker companies caught on on that one? Like make the crappiest speakers ever and just “fix” them with EQ.
• Also if EQing speakers would make them better, why haven’t speaker companies caught on on that one? Like make the crappiest speakers ever and just “fix” them with EQ.
• Dollar for dollar, you will get exponentially more improvement from properly treating your room acoustics and properly positioning your speakers than you will gain from any upgrade to speakers or electronics, and most definitely more than you will gain from trying to EQ the problems away.
• Seriously . . . attempting to use EQ in this way to create any kind of “reference” system is a fool’s game. That’s not gear snobbery or technical snobbery . . . it’s simply scientific fact.
• EQ cannot raise a dip (which is caused by either a modal null, or a non-modal null that is caused by comb filtering off a nearby surface — i.e., a wall, ceiling, desk, mixer surface, etc.). What EQ CAN do is to lower some stubborn peaks a bit (in the same way as lowering the overall volume will reduce the amplitude of that same peak) . . . but it CANNOT stop the modal ringing at the frequency caused by the modal peak. The modal ringing is indeed just as important (and in some cases MORE important) as the amplitude peak at a given frequency!

 

Some testing with Sonarworks in my home studio

Figure 5.1 shows the SoundID calibration curve measured by Sonarworks. It has added lower and higher frequencies, and also corrected the peak between 100 and 200Hz (which, previously,  I did manually as explained above). So far so good. I used linear phase to avoid introducing additional artefacts due to different phases for different frequency components.

Figure 5.1 SoundID calibration curve obtained from measurements at the listening position and the Kali IN-8 monitors. 

 

One first test that I did was using the MATT (see above) signal to test for the audio definition with and without Sonarworks turned on. Although the effect is subtle, turning on Sonarworks decrease the audio definition. Thus, although it might have improved the frequency response, I am left with less defined audio.

 

Next, I did some measurements with REW.  The RT60 measurements with and without Sonarworks (not shown) show small changes but overall nothing to be worried about. In Figure 5.2 the frequency response of for the left speaker is shown.  The REW measurement shows two peaks (one between 100 and 200 Hz, and the other close to 300 Hz). This is not obvious from the Sonarworks calibration curves (left speaker) but changing the smoothing window to a full octave reproduces the Sonarwork curves. Thus, the many measurements of Sonarworks are in quite good agreement with REW.

Figure 5.2. Frequency response for the Kali IN8 left monitor. Red line: without Sonarworks. Blue line: with Sonarworks. Smoothing window is 1/3 octave.  Note that in these measurements the SPL levels can be compared.

 

Next if we have a look at the comb filtering for frequenciess up to about 2kHz we see something strange. For some reason the Sonarworks frequency response (blue) shows, overall, a more strong fluctuations (more comb filtering?) especially between 800 and 1.2 kHz.  I consider this an artefact introduced by Sonarworks, but it remains an open question for me if this is serious enough to abandon Sonarworks?

Figure 5.3. Frequency response for the Kali IN8 left monitor. Red line: without Sonarworks. Blue line: with Sonarworks. Smoothing window is 1/3 octave. Linear frequency scale. Note that in these measurements the SPL levels can be compared.

 

Figure 5.4 shows the total harmonic distortion (THD). In agreement with the SoundOnSound article mentioned above I also see an increase of the THD over the full frequency range up to an increase of about 10dB around 80Hz. In particular at the low and high frequencies where the frequencies were boosted to a larged extend the distortion is also larger.  However, this distortion is still 45dB lower than the audio signal and there might be heard only at higher volumes.

Figure 5.5. Total harmonic distortion for the Kali IN8 left monitor. Red line: without Sonarworks. Blue line: with Sonarworks. Smoothing window is 1/3 octave. Note that in these measurements the SPL levels can be compared.

 

Figure 5.6 shows the spectrograms for without (left) and with (right) Sonarworks turned on. It is clear that this introduces more ringing (longer decays) over the entire frequency range but in particular between 50 and 150Hz.

 

Figure 5.6. Spectrograms for the Kali IN8 left monitor. Left: without Sonarworks. Right: with Sonarworks. 

 

Final verdict Sonarworks

It is clear that SW introduces artefacts. If these artefacts outweigh the ‘corrections’ in the frequency response remains to be seen. I am not convinced that Sonarworks helps with translating a mix but I will give it (again) a try. Perhaps the general rule should be to lightly correct only peaks in the lower frequency range? To be continued…..

 

 

References

• The Beginners Guide to Acoustic Treatment (SoundOnSound)
• Studio SOS Guide To Monitoring & Acoustic Treatment (SoundOnSound)
• Bob McCarthy (2007) Sound Systems: Design and Optimization, First Edition, Focal Press, Amsterdam
• Don Davies, Eugene Patronis, Jr. (2006) Sound System Engineering. Third Edition. Elsevier Focal Press, Amsterdam.
• F. Alton Everest, Ken C. Pohlmann (2021) Master handbook of acoustics. Seventh edition, McGraw Hill, New York.
• Room EQ Wizard (website, documentation)
• Control Systems.
• Sonarworks Reference 4 Studio Edition (SoundOnSound)
• Sonarworks (SoundID Reference) (SoundOnSound)