We conclude our two-part examination into whether the traditional distinction between studio monitors and hi-fi speakers is justified. This month, the power-handling and compression characteristics of our four test monitors come under scrutiny...
The first instalment of our investigation into the differences between nearfield hi-fi speakers and pro monitors ended last month with the conclusion that, in terms of frequency response, there was little obvious distinction between two entry-level products from each market sector (see SOS June 2002). The 'hi-fi' B&W DM303 and 'pro' Dynaudio BM5 demonstrated response characteristics that appear pretty well balanced for nearfield duties, while the 'pro' KRK K-Rok and 'hi-fi' Wharfedale Diamond 8.2 appeared less well suited. If speaker engineers were asked to look at the response curves of all four with no knowledge of the identity or intended application of the speakers, I'm pretty sure they would be unable to sort the 'hi-fi' from the 'pro' other than by guesswork.
But of course frequency-response issues are but one aspect of speaker performance, and in this second instalment of the investigation, I'm going to look at some power-handling and compression characteristics of the four speakers — again with the aim of investigating the potential for the significantly less expensive hi-fi speakers to be used as nearfield monitors. If you're unfamiliar with what these terms mean with regard to speakers, don't worry too much, as I'll explain as I go along.
If we were to speculate on the power-handling characteristics that we'd want to be better on a nearfield monitor than a hi-fi speaker we'd probably focus on compression and distortion effects. We'd expect that both hi-fi and pro speakers would have good long-term reliability, but as nearfield monitors are probably more likely to be used for critical listening at relatively high levels over extended periods, we'd want them to be more accurate and consistent in that scenario.
Compression effects caused by speakers are likely to have a direct influence on how we record and mix. We all use audio compression in recording, but a monitor that adds its own level- and time-dependent dynamic and tonal aberrations is not going to be much use as a consistent reference. However, compression issues are very rarely raised by speaker manufacturers in promotional material. This is partly because a loudspeaker designed to be completely free of such effects would be significantly more expensive to manufacture than one that isn't, and partly because compression is a kind of 'stealth' distortion that very much influences our opinions about sound (and therefore decisions about how we record and mix), even though we're often not specifically aware of it happening.
Since a speaker with a higher quoted power handling might be expected to perform better in terms of compression, power handling is the one oft-quoted specification that might give us an idea how each speaker is likely to behave at high levels. However, there's probably just as much misunderstanding and confusion surrounding power handling as most other speaker specifications, so before getting onto the performance of our four potential monitors, it's probably useful to describe a little of what happens to power in a speaker, and where power-handling limits actually come from.
In very simple terms, you find out the power handling of a given speaker by turning up the volume till the speaker breaks. But just as frequency-response measurements can be misleading unless you know where the measurements were taken from, and with which mic, so it is with power handling. Power handling has also become a 'competitive' specification that suffers inflation from market pressures — and manufacturers, knowing the spec is hard to pin down, often play safe and pick a nice big round number. The way power handling is specified for our four guinea-pig speakers demonstrates this nicely (see below). Each figure is specified differently and, moreover, in the context of the highly complex mechanisms at play, means very little. In fact, the specs shown below are all pretty meaningless. For example, what exactly is the spectral balance of the "unclipped programme" that B&W refer to? Or how do we define the 100W of the still-vaguer "programme" that the Wharfedale will stand? Are we talking thrash metal or a recorder quintet? And for how long? Going one stage further, KRK simply specify a maximum amplifier power. Does this mean the speaker can never be damaged by an amplifier of less than 100W? I don't think so... Only Dynaudio, in referring to a "long-term power" for the speaker while at the same time recommending an amplifier rating, hint at the true complexity of the issue — and even then, the spec has lost all real value through simplification.
Explaining the "true complexity" of the power-handling issue and how it ties in with compression effects isn't easy, but if you have a grasp of electrical knowledge and basic physics, you shouldn't find it too heavy going, and my article on understanding how monitors work from SOS October 2000 should also help. Essentially, any loudspeaker comprising moving-coil driver units in a box will exhibit two distinct mechanisms that limit its power handling, and the limit that's reached first will be dependent on the frequency balance of the programme material, and perhaps the length of time it's been playing.
The first mechanism is thermal. The laws of physics are not kind to loudspeakers — typically, less than one percent of the electrical energy thrown at them is converted into acoustic energy. So around 99 percent of the power an amplifier pumps into your monitors does nothing but heat things up. As the internal components of the drivers warm, thermal stresses begin to build and, if the heating continues unchecked, things will eventually smell, smoulder, break, melt or even burst into flames.
Testing to establish the point at which a driver will fail thermally, and therefore the thermal power handling, is pretty straightforward if the test signal is a nice, consistent one with a defined spectral balance, such as pink noise, or even noise 'profiled' to resemble the spectral balance of some average, imagined music (and believe it or not, there is such a thing intended to do just that — it's called IEC 268 noise). But real music is, in test-signal terms, as far from predictable as it's possible to get. The spectral balance varies continuously, as does the instantaneous level.
So, because the variable nature of a speaker's input impedance means that power at different frequencies will have different heating effects, you can perhaps see that specifying a 'thermal power handling' that's relevant to real programme material is all but impossible. Or, to put that another way, if you're a manufacturer, you're pretty much free to think of a 'music programme' number, double it if you wish, and still remain pretty confident that nobody out there will really be able to prove you wrong. I mean, what if the music programme you decide to base your spec on is John Cage's '4 minutes 33 seconds'? What's the 'music programme' power handling if the music programme is silence ("for piano or any instrument" as Cage specifies with delicious irony)?
The second power-handling limit mechanism is predominantly mechanical, with some acoustics and fluid dynamics thrown in for good measure. Those same laws of physics that stop a moving-coil speaker from efficiently converting Volts into noise require that the volume of air a speaker has to move to generate a specific sound level increases as frequency falls (one major reason why woofers are big and tweeters small).
- B&W DM303 "Maximum amplifier power 100W, 100W unclipped programme".
- KRK K-Rok "100W maximum amplifier power".
- Wharfedale Diamond 8.2 "100W programme".
- Dynaudio BM5 "Long-term amplifier power 100W, recommended amplifier power 200W".
Now, it's pretty easy to calculate how much air a speaker can move — you just multiply the area of the cone by its maximum travel. But unfortunately for loudspeaker designers, a power-handling trade-off applies here. You can have any two of the following: a speaker with good bass extension, a speaker that goes very loud, or a speaker with a relatively small bass driver. What you can't have is all three at the same time, and some pretty nasty things can happen if you try. The first problem is that well before a bass driver hits its end stops, it'll begin to produce high levels of harmonic distortion. Play a bass guitar bottom E through your monitors (but don't blame me if you damage them!) and you'll hear the tone start to get dirty as the level rises. Its pitch may even start to become more hazy as high-order harmonic distortions begin to dominate the fundamental frequency of the note.
Problems two and three are the exclusive of reflex-loaded speakers (ie. those with a port to increase the perceived amount of bass they can produce — there's more on this in that SOS October 2000 article). These days, however, it's getting hard to find any loudspeakers that don't have such ports! Problem two is that the necessarily tiny reflex ports used in small speakers behave in a far-from-linear fashion at anything but low volume levels. There's only so much air that can be pumped through a port before it becomes noisy and a significant source of distortion.
Problem three is related; the non-linearity of the port at high levels can begin to cause the bass driver to oscillate away from its nominal zero position. Given that one end of the port terminates in a small volume of air inside a mostly enclosed box, while the other terminates in the outside world at atmospheric pressure, it's hardly surprising that it doesn't behave the same way for air moving in as it does for air moving out. Once the driver's zero position begins to drift, if it has any significant non-linearity inherent in its magnet or suspension system, the drift may be all that's needed to 'lock' the cone (potentially terminally), against either of its end-stops.
Maybe you can now begin to see how truly uninformative the average manufacturer's programme material power-handling spec is. If you need convincing further, I should also point out that that it's very difficult to accurately to measure the true power of a programme material signal anyway. And, of course, along with the completely unpredictable nature of musical material, you should also consider that the playback systyem you use to hear the programme material will respond in completely different ways at different frequencies!
OK, so that's how the power-handling limits are reached (and how power-handling specs can be 'invented'), but what has all that got to do with compression? Well, long before the failure limits are reached, the very same thermal and mechanical mechanisms that will eventually send a speaker to the service department are responsible for compression and distortion effects. Let's look at the thermal side again first.
As loudspeaker driver voice-coils warm, their electrical resistance increases. Primarily, this increase in electrical resistance reduces the sensitivity of the driver, by making it harder for the amplifier to push current into it. There are attack and release times associated with the mechanism too (defined by the driver's thermal time constant), just like the dynamic-range compressors we know and love. However, there are also two more subtle thermal-compression effects that don't so much affect overall volume level as they do frequency response.
Firstly, a speaker's low-frequency roll-off is defined by a number of electrical and acoustic parameters, an important one of which is driver voice-coil resistance. So again, as volume increases, the low-frequency performance changes. Similarly, in passive speakers, the frequency response of the crossover filters that divide the audio spectrum between drivers is very sensitive to voice-coil resistance. So there's another mechanism that potentially changes the frequency response and the sound of a speaker as the volume level changes. I've already touched on how the dynamics of reflex ports at low frequencies can alter the overall response shape. But while these effects constitute one of the power-handling limits, their effects can begin to influence the sound of a speaker, even at relatively low volume levels. And in a monitoring situation, anything that dynamically influences the sound of the speaker without us being aware of it is an effect that we're likely (erroneously) to take into account when we record or mix. Have you ever noticed mixes that work at one volume level yet seem nothing like as good when you change the volume? It's not all down to your monitors — your ears play cunning tricks too — but your choice of speaker could well be part of the problem.
Now, having got through the theory, on to the practical. If we take our four guinea-pig speakers and carry out a few measurements designed to reveal how well they behave in terms of some of these compression mechanisms, will we see any obvious differentiation between 'pro' and 'hi-fi' types? The first simple thing to check is the voltage sensitivity of each speaker. Sensitivity is a measure of a speaker's output volume level for every volt of input and it's potentially important in the context of compression effects because, all things being equal, for a given volume level, a higher sensitivity speaker should be less prone to compression — less power needed means less dissipated as heat. However, as it turns out, there's not a great deal to choose between them.
The least sensitive speaker of the four is the Wharfedale, with an 86dB output level (measured at a distance of one metre) for an input of 2.83V (2.83 Volts is nominally one Watt into an 8Ω load). The most sensitive is the KRK, which produces 88.75dB of output under the same conditions (a surprising 3.25dB shy of the claimed 92dB). The B&W and Dynaudio sit in-between, at 88dB and 87.25dB respectively. So with all four speakers within 1.5dB of the group average, we don't have to worry too much about differences in sensitivity distorting comparisons of compression effects over comparatively large changes in level. Nevertheless, it's worth bearing in mind that the higher the baseline sensitivity, the louder the speaker will play before compression hits.
The first comparison of compression effects on the four speakers was a simple measurement of bass/mid driver voice-coil temperature rise with a noise signal of increasing level. The change in voice-coil resistance equates directly with a reduction in driver sensitivity — ie. compression by any other name. This is quite a tough test for a small speaker. The final input level of 20V RMS is pretty loud, but it was important to drive the speakers hard in order to reveal any differences without any ambiguity.
|SPEAKER||VOICE-COIL TEMPERATURE||VOICE-COIL RESISTANCE RISE||SENSITIVITY DROP|
|B&W DM303||195ºC||176 percent||4.9dB|
|KRK K-Rok||125ºC||149 percent||3.5dB|
|Wharfedale Diamond 8.2||120ºC||147 percent||3.3dB|
|Dynaudio BM5||100ºC||139 percent||2.9dB|
The directly measured voice-coil temperature rise, calculated voice-coil resistance rise (from knowledge of the voice-coil winding material and its temperature coefficient of resistance), and calculated sensitivity drop over a 20dB increase in input level (from 2 to 20 Volts of IEC 268 noise) for each speaker are as shown in the panel above.
There are three points of interest here. Firstly, the KRK and Wharfedale are very similar in behaviour, suggesting that they have bass/mid driver voice-coils of similar construction — so there's little differentiation between 'hi-fi' and 'pro' there. Secondly, the B&W DM303 shows fundamentally higher levels of thermal compression. This is most likely because the B&W has a voice-coil construction based on a polyamide former rather than the more usual aluminium. A polyamide former does have the significant advantages of being light (in terms of weight) and non-conductive, and it may well be that the use of this former significantly contributes to the B&W's exceptional frequency-response performance — but it clearly gets hot quickly. Polyamide simply disperses heat less effectively than aluminium. Lastly, the Dynaudio suffers the least thermal compression. This is an expected result, as its voice-coil construction is very different and features a significantly more massive aluminium component that can dissipate more heat.
It's not just overall sensitivity that suffers when a voice-coil warms. A speaker's low-frequency alignment will change too, and the severity of the change will depend not only on the level of heating but the sensitivity of a specific alignment to variation in voice-coil resistance. Figures 1 to 4, shown throughout this article, display the mathematically modelled base-line low-frequency performance of each speaker, with red overlays generated by the same mathematical model with raised voice-coil resistance. The values for voice-coil temperature rise, and therefore resistance rise, were taken from the same IEC 268 noise test data, but not for the full 20V. These temperature rises correspond to 10V input.
There's actually not a great deal of difference between the way each speaker behaves in this respect — or at least there's nothing that surprises in the context of what we already know about each one's ability to disperse heat. The B&W again suffers the most, changing from a well-judged low-frequency alignment to one with a mild 1.5dB peak at 100Hz. The KRK gets a little more resonant and loses some extension, as does the Wharfedale. The Dynaudio is again the least affected, reflecting its voice-coil's better ability to disperse heat. While these changes in response generally look pretty subtle, remember that at these frequencies they represent the speakers becoming more resonant.
Talking of resonant, the last set of measurements aimed at sorting the men from the boys looks at reflex-port compression and distortion. Figures 1a to 4a show the output of a microphone (a B&K omni measurement mic) placed directly at the port mouth of all four speakers while the speaker is driven with a sine wave at its port-tuning frequency. Using a close mic and driving the speaker at its port-tuning frequency means that the bass driver itself makes negligible contribution to the measured signal. I carried out the test for each speaker at two levels, one 18dB higher than the other (2V and 16V), but the higher levels (the red overlays) in graphs 1a to 4a have been lowered by the same 18dB. If the speakers showed a complete lack of port compression, or noise and distortion increase, the two curves would overlay exactly.
The first thing to note is just how much harmonic distortion a reflex port generates, even at low levels, and even when extravagantly flared and dimpled like the B&W's (such design features are supposed to minimise the effects of the port). Harmonics all the way up to 12th-order or higher are easily identifiable. The second thing to notice is that there's significant wide-band noise between the distortion peaks — this is the audible chuffing that ports are prone to produce. The third point of note is that the B&W and Dynaudio ports get noisier as they work harder, more so than those of the KRK and Wharfedale — the base-line noise levels between the peaks increase more markedly. This is probably because they are tuned lower in smaller cabinets and their ports are necessarily of a smaller diameter. Compression at the port-tuning frequency (the lowest peak) can be seen on these curves — although the result here will incorporate some thermal element that is hard to remove from the mix. The Wharfedale comes off worst in this respect, showing a 4.3dB loss, while the KRK is top dog, with a loss of only 1.8dB. The K-Rok's significantly higher tuning frequency probably helps — although, harking back to the frequency measurements, the high-tuning frequency also results in a resonant bass characteristic.
So have any of these two sets of measurements revealed a fundamental difference between 'pro' and 'hi-fi' speaker types? Well, the 'pro' Dynaudio, thanks probably to its voice-coil construction, is best suited to being driven hard, and it has a well-controlled and appropriately balanced frequency response. The 'pro' KRK, through its higher sensitivity and reasonable compression performance, is also more suited to operation at high levels than either the B&W or the Wharfedale. So maybe there's a glimpse of a distinction here. But then the KRK is, in my opinion, let down by the questionable frequency-response characteristics I noted last month. The Wharfedale is pretty similar to the KRK in terms of its thermal-compression performance, but loses out through being around 3dB less sensitive. Its frequency response isn't really the stuff of a nearfield monitor either. Lastly, the B&W is not so good at elevated levels, but its exceptional frequency response performance still argues pretty strongly in its favour. In truth, there are a range of abilities across the four different designs, and it'd be a brave man to say that any group of two is obviously inappropriate.
If pressed to rank the four speakers for nearfield duties in, say, a small project studio handling a wide variety of material, I'd go for the Dynaudio first — if I had the cash. But if I knew that really high-level monitoring was going to be a rare occurrence, I'd plump for a hi-fi speaker, the B&W, on the grounds of its response accuracy — and I'd maybe put my savings towards a microphone. In the right environment, with the right type of material, any of these four speakers could be used successfully — countless wonderful recordings have been made using monitors worse than anything here. So, can you use hi-fi speakers as nearfield monitors? Well, as I wrote at the start of this article, it depends...