Good Sound (Part 3) – What Matters?

Gary Gesellchenblog

In the previous entry in this blog series I described how I go about designing for good sound.  The design process involves the three key elements of modeling , measuring , and listening.  So far, so good.  But what exactly should you model or measure?  And what are you trying to hear when you listen?  In short, what actually matters?

Fortunately there has been a lot of research into what matters in terms of perceived loudspeaker quality.  And of course, separate from the research, there are things that matter to me personally that might not appear in research papers, but do appear in products! 

(Note: I have added a list of sources at the end of this blog so as to streamline the main body.  Anyone with a technical interest in loudspeakers or who is interested in what loudspeaker anomalies sound like is highly encouraged to investigate the sources!)

My list of the things that matter most, in rough order of priority are:

“Sweet Spot” Frequency Response – Frequency response refers to how loud the speaker plays at different frequencies.  Ideally every note or frequency should be produced at the same loudness as every other note or frequency (this is what it meant by “flat response”).  The “sweet spot” refers to the position directly in front of the speaker(s) where a listener would normally sit if they were intently listening to music.  It is measured by placing a microphone directly in front of the speaker, or sometimes by averaging measurements from several locations around a line directly in front of the speaker.

Tight Bass – I cannot listen to a system with boomy, flabby bass.  It will cause me to leave the room.  I’ll take quality of bass over quantity of bass in nearly every situation.  Boominess is normally caused by a system that does not have enough “damping” (think “shock absorbers” in a car), and is often done on purpose to give the illusion of extra bass.  While somebody, somewhere may like it, no Vanatoo product will ever have boomy bass!  Tight bass is quantifiable with modeling and measurement, but is ultimately decided by listening for realistic bass reproduction.

Freedom from Resonances – Resonances tends to smear the sound of transients and muddy the overall musical presentation.  Resonances frequently occur in the drivers, cabinet, port, or in the air inside the box.  Resonance control is at the heart of most megabuck loudspeakers and their extremely stout cabinets.  The normal tools for uncovering resonances are investigation of the input impedance and  of the Cumulative Spectral Decay plot of the frequency response.  

Bass Extension – This obviously must vary with the overall size of the speaker, since it is a more-or-less iron clad rule that a larger speaker can go lower than a smaller one.  Still, in my experience it is necessary to have bass extension to well below 100Hz for music to have enough heft and weight to sound decent.  Response down to 50Hz covers most of the musical spectrum and is satisfying with most material.  The ClearBass™ technology used in the Transparent One gives response down to 48Hz, which makes for much more realistic sounding music than the ~75Hz cutoff more common in speakers their size.  Bass extension is easily measured using a microphone.

Off-Axis Frefloyd-toole-book.jpgquency Response – What you actually hear when listening to music in a room is a combination of sound that comes directly from the speaker (direct sound) along with a lot of sound that is bouncing around the room (reverberant sound).  It is important for good sound in a room that the frequency response of the reverberant sound is similar to the frequency response of the direct sound (“sweet spot” above).  Most of the reverberant sound comes from sound radiated by the speaker outside of the “sweet spot”, or what is known as off-axis.  The off-axis response is measured by a microphone at locations all the way around the speaker, both sideways and up/down (normally the microphone is held still and the speaker is rotated).  Per the techniques pioneered by Toole (see below), 72 measurement points will give good correlation between measurements and the listening experience.

Maximum Volume – Just about every music system will start sounding crummy if the volume is turned high enough.  It is important that the system is able to play satisfyingly loud before it starts to sound crummy.  This level is obviously a matter of personal taste, and will vary a lot depending on the distance from the speakers and the size of the room.  The listening tests in the sources are often run at less than 80dB.  An average level of 85dB is fairly common in recording studios.  My design goal is to make sure any Vanatoo system can produce a level that is at least in the low 90dB’s when used as we expect before it starts to sound crummy.  This is louder than most people will sit and listen to for an extended period of time, but it is not quite dance party levels.  For reference, the Transparent One will hit about 94dB in my family room (13 x 19 feet with one long side open) when I sit ~10 feet away before it starts giving-up.


Dr. Floyd Toole of Harman International and the Canadian Research Council (now retired) literally wrote the book on perceived loudspeaker quality.  His book, “Sound Reproduction – Loudspeakers and Rooms” is required reading for any serious speaker designer.

Dr. Wolfgang Klippel has more-or-less revolutionized the understanding of how speakers really behave.  His measurement systems are the industry standard.  His publications can be found here.

Joseph D’Appolito wrote the book on “Testing Loudspeakers”, and has summarized his findings in a set of papers published in Audio Express title “Which Measurements Matter”.  You can download them here and here.

John Atkinson, best known as the Senior  Editor of Stereophile magazine, has measured hundreds of sets of speakers and correlated his measurements with the Stereophile reviewers comments in his paper “Measuring Loudspeakers”.

Dr. Toole’s work lead to a suite of listening tests being released by Harman International.  Harman uses this test suite (“How To Listen”) to train listeners and to quantify how well different listeners are able to judge different sound anomalies.  You can read about the test suite in Dr. Sean Olive’s blog and download it here.

Dr. Klippel has a very interesting set of listening tests available on his website that allow the user to quantify how well they can hear different types of distortion and to compare their results against others who have taken the tests.  You can access them here.