Page 3: The Yamaha NS10 Story

NS10 Listening

Why have I included a frequency-response curve here? I mentioned earlier that the frequency-response curves in a sales brochure are typically meaningless in terms of providing information that's useful to an end user. Actually, though, I'd go further than that, and suggest that in many respects making any judgment about the worth or likely value of a monitor by examining its frequency-response curve is not far short of pointless. I often read opinions on the SOS Forum arguing that to be of any value monitors require a 'flat frequency response', but numerous recordings made during what many would consider the golden age for musical sound quality (the '60s and '70s) were monitored on speakers that were all over the place in terms of frequency response — and I don't know why recording engineers seem to believe so strongly that a monitor should be anechoically 'flat' when so much end-product evidence suggests that this isn't particularly important.

A frequency-response curve appears to tell you if a monitor is going to reproduce different elements of the audible bandwidth at the same level, which intuitively seems vitally important. But a simple frequency-response curve tells you no such thing, and the psychoacoustics of human hearing is more about the time domain than the frequency domain.

When we measure a monitor's frequency response in an anechoic chamber, the microphone 'hears' the output at just one position in space. However, when we listen to a monitor in a room we hear a combination of the monitor and its interaction with the room boundaries (and big items of furniture). Reflections from the walls, floor and ceiling are integrated over time by the brain, to create a composite tonal balance. When I design a typical 'box speaker', I've learned through experience (and reading Dr Floyd Toole's work on the subject) that a frequency-response curve taken at between 20 and 30 degrees horizontally off-axis is likely to be most representative of an appropriate target tonal balance. For a speaker tonally voiced for domestic free-space mounting (not up against a wall or sat on a meter bridge), this off-axis anechoic curve should be reasonably flat up to around 2kHz and then fall slowly at around 3dB per octave for the rest of the range. This is a long, long way from 'flat', but it will sound neutrally balanced in a typical domestic room at average playback levels.

And speaking of 'average playback levels', in addition to the room effects that influence our perception of tonal balance, listening level plays a significant part too. The brain's perception of tonal balance is level dependent. At low levels we're far more sensitive to mid-range than bass and treble — hence the 'loudness' button beloved of '80s Japanese hi-fi amps. So, again, expecting a frequency-response curve measured at one position in space and at a single arbitrary level to reveal the full story on the worth of a monitor is to simplify reality to the point of nonsense.

Moving swiftly on to the second assertion I made a couple of paragraphs ago, we humans have evolved to respond more to the transient than to the tonal elements of sound. Try a little experiment: find a sample of something like a clarinet and a flute, each playing the same continuous note, drop them onto two tracks in your DAW and listen to them in turn. It's very easy to tell which is which. Chop the first, say, 500ms from the front of each so that the characteristic beginnings of the notes are suppressed, and listen again. They'll sound much more similar: the brain uses the characteristic transients to differentiate the instruments, and without them it struggles. Now, go back to the un-edited samples and apply the same severe EQ to each and listen again: despite the EQ, you can still differentiate them. A similar illustration of the use the brain makes of transient rather than tonal information is that a familiar voice remains familiar in wildly different acoustic environments — environments that imprint different tonal characters on the sound. So, concentrating on the 'flatness' of frequency response is to miss a hugely important point: if a monitor handles transients accurately, its frequency response is much less important than you probably think.

But It Sounds Horrible!

Before I wrap up this epic (and promise never, ever to write about the NS10 again), there's just one more issue that probably deserves to be kicked around a little. If the NS10 is so good, why do people so often express their dislike of listening to it? I suspect that there are both practical and emotional answers to this conundrum.

First, the emotional. Thanks to its time-domain accuracy and mid-heavy balance, the NS10 is an extremely revealing speaker that takes no prisoners. In other words, if the recording is poor, the NS10 will tell you in no uncertain terms. You have to work harder to make things sound good on the NS10 not because it sounds bad but because recorded music, even today, is often a poor approximation of the real thing, and the NS10 reveals it. I found a familiar comment on the SOS Forum that reads: "If it sounds good on NS10s then it'll sound good on anything." Again, that's not because the NS10 is inherently poor, but because it is effective at revealing the fundamental compromises inherent in recorded music. If you've worked hard on NS10s at a mix and overcome those compromises, or perhaps cleverly disguised them, the mix will translate well to other systems because it is a good mix. Put another way, the NS10 better enables you to get to the nub of a mix by more accurately reproducing its fundamental time-domain information — and it is this which can make the task of mixing seem more challenging.

And the practical? Well, it's certainly true that the NS10s have a mid-heavy balance and little bass extension. This is especially so if they are not mounted close to a suitable boundary — such as a big desk or a rear wall — to provide low mid-range reinforcement. They're also just as revealing of any shortcomings in the monitoring chain as they are of the mix, and they don't take very kindly to being driven loud. While Newells and Holland showed they have very low levels of distortion, they do suffer from thermal compression, which will not only cause wide-band dynamic attenuation in response to high levels of drive, but will upset the characteristics of the crossover filters as the voice-coil resistance of the drivers increases. As temperature rises, the bass/mid low-pass filter frequency will increase significantly (and the tweeter high-pass filter frequency will reduce), and begin to give prominence to the resonances at the top end of the bass/mid driver's response. When NS10s are driven too hard by a poor amplifier, fed by a sub-standard monitor output, and mounted without any boundary reinforcement, you might well find that they sound horrible to the point of being unusable.

And Finally...

Where does all that leave us? Why do we still use that old monitor? We use it because it does a job, even if it sometimes doesn't sound very nice while doing the job, partly because, if it's installed or driven inappropriately, it will reveal such shortcomings without mercy, and partly because it sometimes reproduces elements of our work that we don't particularly want to hear. But we also use it because nearfield monitor manufacturers seem to have suffered a 20-year blind spot and failed to identify why the NS10 works and remains so popular. Go figure. 

Thanks to: FX Rentals for the loan of a pair of NS10s; Acoustic Energy for permission to use their NS10 data; Phil Knight for doing the original measuring; the SOS Forum members whose words I've borrowed; and to Chris Binns for advice that (hopefully) ensured I've not written anything really dumb.