In a previous installment of this series in Issue 126, I discussed measuring the most important component of an audio system, which is the ear. It is most important because no matter how expensive or brilliant the rest of the system is, we need be able to hear it to appreciate it. In the rest of the series, I will discuss what I do in terms of measurements to improve the sound of my system. It is by no means an exhaustive guide, and I have limited technical expertise, but other audiophiles with the same or higher level of technical competence (which is pretty much everybody) could benefit from this as well.
In my experience, the most problematic part of an audio system is often the room acoustics. Room acoustics cannot be looked at in isolation without considering the interaction with the loudspeakers. Therefore, it is perhaps more accurate to call this loudspeaker-room interaction. I am lucky since I have had expert help in this regard; I recruited a friend who is an architectural acoustics expert to design my sitting room. But even if this is not an option for most, there are often maneuvers that one can do to correct at least some of the problems.
The basic characteristics of the loudspeakers play an important role; the frequency extension, especially the low frequencies, and whether they can work well in a room with a particular set of dimensions, must be taken into consideration on initial purchase, since it is difficult to correct problems later that are related to overloading a room with bass. The dispersion pattern of the mid and high frequencies should also be considered. In general, the more controlled directivity of horns has less interaction with the room than the omnidirectional dispersion pattern of say, the MBL Radialstrahler omnidirectional driver. Often, the improvements audiophiles hear after changing speakers are the result of better integration with the room as opposed to the installation of better speakers per se. Having at least some understanding of these fundamental principles could save a lot of dollars, time and frustration.
Loudspeaker-room interaction is multi-factorial and is too complex to tackle without the aid of measurements. The first gadget an audiophile should acquire is therefore a tool to do acoustic measurements. In the past, such tools were expensive and usually only available to professionals. However, one highly useful tool is now available to everyone. REW, which stands for Room EQ Wizard, is a freeware application with the power and the features of professional acoustics software worth thousands of dollars. I would urge everyone who finds the software useful to make a donation to support its continual development. You will also need a laptop computer (PC or Mac), and a calibrated USB measurement microphone with stand (or an analog mic and external audio interface).
I use a Steinberg UR22 (now obsolete) USB sound interface with a Behringer ECM8000 microphone. More expensive measurement microphones come with their own individual calibration files that enable the REW software to compensate for the microphone’s non-linearity, whereas the ECM8000 has a generic file, but it should be accurate enough for our purposes.
The setup is simple. In my case, using an external audio interface, the interface is first calibrated by feeding its input from the output. The software measures its frequency response and generates a calibration file to correct for any anomalies. The calibration file of the microphone is also uploaded. The file either comes with each individual microphone, or can be downloaded from the manufacturer’s webpage. For making actual measurements, the input of one channel is fed from the output in a loop as a timing reference. The input of the other channel is fed from the microphone, and the output goes to an input of your preamplifier. You also need to have an SPL meter to calibrate the output of the microphone. A good old Radio Shack meter will do (at the time of this writing there’s one on Reverb for $30), and there are apps for your phone such as the AudioTools suite by Studio Six Digital that will do the same thing.
Pink noise is played through your system to set the sound level, and it is recommended to set the loudness at the measurement point to around 75dB. You can use a higher volume level if the background noise in your listening room is high, but the software will limit the level based on the headroom of the setup.
The software has an automatic measurement function that relies on a logarithmic sine wave sweep. The length of the sweep can be adjusted, and longer sweeps will allow more samples to be taken and improve accuracy. The sweep goes from 0 Hz to 22 kHz (for a 44.1kHz audio interface) or 24 kHz (for a 48 kHz audio interface). The software captures the measurements and performs a Fast Fourier Transform (FFT) to generate frequency and phase response plots. Then an inverse FFT is performed on the data to generate an impulse response, which mimics a short burst of sound such as a gunshot. This allows the user to see how the sound changes in the time domain.
This is very useful for me as I use an electronic crossover in my system, and the frequency response plot gives me a basis to set the levels of the individual drivers, which can then be fine-tuned by ear. The software also has an equalizer function, which gives the user data points to enter into their DSP equalizers (if they have one).
It is important to understand what causes frequency response non-linearities. Peaks and dips in the bass and low-mid frequencies are often the result of standing waves due to room modes (the sound waves bouncing back and forth between two walls, and between the floor and ceiling), which cause areas of frequency cancellation and reinforcement (nodes and antinodes) in the room, especially in the bass region. As such, moving the speakers and/or the listening position can ameliorate some of these problems by moving either or both away from the areas of cancellation and reinforcement.
Adding sound absorption will damp the room’s peaks. The REW software has a room simulator; you can enter the dimensions of your room, the positions of the speakers and the listening position, and it will predict the peaks and dips in the frequency response. You can play around with the speaker and listening positions to try and optimize the frequency response. Note that using equalization to tame the peaks might work for a narrow spot in the listening position, but will worsen the non-linearities elsewhere. The higher the frequency that needs equalization, the narrower the sweet spot.
Cancellations due to the reflected sound from the loudspeakers arriving out of phase to the direct sound will also result in frequency dips. The reflections usually come from the side walls and the ceiling, and the reflected wave creates areas of cancellation with the primary wave, causing the phase shift. The software allows “windowing” of the impulse response (looking at different time periods after the impulse), which enables the user to see the effect of the direct sound that arrives first, and the effect of the reflected sound that arrives later. The software can then generate minimum-phase plots that show the regions in the frequency response that are susceptible to phase cancellations. These dips are not amenable to equalization, since increasing the level of the direct sound will also increase the level of the reflected sound. Therefore, one would need to use room treatments such as diffusion to disperse these reflected sounds.
Another measurement I find most useful is determining the time delay of each driver (the time it takes for the sound from the driver to reach the listener). Any difference in the time delay between drivers results in phase cancellations where the frequencies between the drivers overlap (a phenomenon known as comb filtering, since the graph looks like the teeth of a comb). At 7,000 Hz, the crossover point between my tweeter and midrange, the wave length of sound is 48 mm. That means a 24 mm misalignment would result in cancellation of this frequency. However, a smaller amount of misalignment distance still results in phase shift, and recent psychoacoustic experiments have shown the brain to be very sensitive to modifications in the phase spectrum1. The scale of perception of phase shift is equivalent to a 2 – 4 dB difference in amplitude, and affects the frequencies one octave higher and lower. Moreover, since our perception of the position of a sound source in three-dimensional space depends on the brain analyzing phase differences (arrival-time differences) of the sound that reaches each ear, poorly-aligned drivers could affect imaging.
The time delay measurements allow me to physically time-align the drivers of my speakers down to a resolution of 1 mm. Unfortunately, most commercial loudspeakers do not allow users to change the alignment of the drivers. Don’t assume the drivers are already time-aligned; they often are not. I had conversations with Greg Timbers, the long-time loudspeaker designer for JBL, who was unceremoniously let go by his private equity overlords in 2016 after 43 years of service. He was responsible for such classics as the 43XX studio monitors, the L250, the K2 Series and the Everest Series, as well as the design of numerous transducers. He told me that when he was there, the engineering department at JBL was often overruled by the product design department for aesthetic reasons. After my recording partner removed the tweeters of his Everest DD67000 speakers, Greg Timbers’ masterpiece, from their cradles and time-aligned them with the midrange drivers under Greg’s instructions, the imaging vastly improved.
For those readers who are open to digital sound, the use of a digital crossover with EQ functions makes a lot of sense. You can customize the crossover frequencies, add filters and introduce time delay for driver alignment, all without introducing phase shift, while getting rid of the passive crossovers (my biggest pet peeve) of your speakers. In my estimation, you have a much better chance of making improvements to the performance of your system this way than by “upgrading” your speakers.
REW has many other useful functions. It can generate a waterfall plot of the speakers’ frequency response that is derived from the impulse response. This is very useful for evaluating the effects of room modes, as well as identifying some design flaws of your speakers that perhaps you would rather not know about. It can also calculate the reverberation time of the room (the RT60, or the time it takes for the sound source to decay 60 dB), although this only applies to larger spaces. This is an important parameter to measure for recording venues.
REW also has a signal generator function that can produce sine waves, square waves, standard dual tones, tone burst, pink noise, periodic noise and sine wave sweeps. It has a real-time analyzer (RTA) that one can use with pink noise to do frequency analysis down to a resolution of 1/48th-octave. You can also measure the total harmonic distortion (THD) of your whole system. With conventional dynamic speakers, the THD would be in the several percent range (the THD of horn compression drivers and electrostatic panels are generally an order of magnitude lower), which makes the less than 0.1% THD of electronics pretty insignificant.
In addition to acoustic measurements, the software can also be used for electrical measurements, such as determining the Thiele-Small parameters of loudspeaker drivers and measuring the frequency response and distortion of electronics, The REW software also has an oscilloscope function. All this for the grand price of $0. The greatest bargain in audio. Who says good things don’t come cheap?
For those of you who want something a little more portable and don’t mind paying around $1,600, I would recommend the Phonic Audio Analyzer, model PAA6. It is a self-contained analyzer with two microphones, two balanced line-level inputs and a signal function generator. The PAA6 has internal memory as well as an SD memory card slot and a USB output, so that data can be easily transferred to a computer. It has many of the functions of REW, albeit without as much resolution. It has a 1/6th-octave RTA, an FFT function and can measure RT60. It cannot do inverse FFT though, which means one cannot do time domain measurements.
The PAA6 has a digital oscilloscope, and it has a signal polarity-checking function that is helpful to ensure that speaker drivers and cables are connected at the correct polarity. It also has a phase meter that measures the difference in phase between two inputs. This is useful when setting azimuth for tape heads and phono cartridges. The PAA6 can measure THD in real-time, which is also useful for phono cartridge setup.
At home, it also comes in handy if you want to find the best listening position for a smooth frequency response. You can just walk around with the RTA function switched on until you find the perfect spot, as indicated by the frequency response readout. I also use it to tune the EQ on my tape head preamp. I also find the PAA6 useful when I have to make a recording in an unfamiliar venue, in lieu of hauling around the laptop, audio interface and microphone required for using REW. Readers who put up demos in show venues and customers’ homes will find it useful in getting a read on an unfamiliar acoustic environment.