Drive Units (Part 2)
How to prevent Drive Unit damage.
Ben Duncan continues his look at speakers with an investigation of how they become damaged and how to avoid this.
Scenario: too much power, or too little? Frantic with excitement, you switch on the tape. The bass line sounds good. The musician beside you begins to beat out the heavily accented rhythm - a clenched fist against an open palm. You edge up the faders. The bass sounds much funkier now. With your fingers still edging up the faders it sounds very, very tasty. Oh dear! we seem to have lost all the bass....
The problem here is something to do with pushing the speaker - or the amplifier - too hard, but without knowing more, it's difficult to pinpoint why this (or any other) speaker has failed.
How long is a piece of string? A power rating is sadly not a simple indicator as to how far you can push a speaker; every manufacturer has (albeit justifiable in the eyes of the designer) an idiosyncratic method of balancing the information involved in arriving at a power rating. Even if there were a definite power input for a particular speaker, above which failure would always occur, and below which damage would be impossible, two factors would still intrude:
1) Certain manufacturers tend to conservatively rate their speakers. Others quote the highest power rating they dare to put in print. This in itself leads to confusion - the same conjectural driver rated at, say, 75 watts in the first instance, would be given, perhaps, a 200 watt label by another manufacturer. Even if the higher rating is justifiable there's still a credibility gap. Any given loudspeaker, it would seem, can have a range of power ratings.
2) Loudspeaker power ratings habitually involve hidden assumptions, for instance, (i) that the amplifier is never driven into overload, or (ii) that the acoustic power at low frequencies is limited.
For any speaker, the primary effect of too much power input is high temperatures in-and-around the voice-coil. Thinking laterally, loudspeakers are not 100% efficient, and the surplus power mostly appears as heat. Looking back to the Control Room article in the November 83 issue of HSR, it's easy to figure out that a 100% efficient speaker would have a 120dB @ 1 watt @ 1m sensitivity. Every electrical watt would dissipate 1 acoustic watt in the air.
Acoustic watts are loud! In real life, efficiencies lie in the 0.15% (92dB @ 1W @ 1m) region for ordinary direct-radiator speakers and even the most efficient horn-loaded systems only manage 5% to 10% at best - say 107dB @ 1W @ 1 metre. Clearly, most of the power applied to any speaker is thus thrown away as heat.
When a voice-coil heats up, there are two effects. First, its resistance rises. Under normal conditions (ie. assuming the amplifier isn't limiting), a rise in resistance spells a pro rata drop in power input. Eventually, the audible level will drop, and the rise in temperature will be curtailed. To an extent then, all moving coil speakers have inbuilt protection in the shape of thermal feedback.
This event is best known as thermal compression. It's often very audible, the sound becoming coloured and compressed as the voice coil approaches its maximum temperature.
A second effect determines this maximum temperature: the adhesive which binds the wire to the former begins to soften and then melt. It may even catch fire. Working in a minute gap, circa 1/100", any slight swelling on the surface of the voice-coil will rub against the magnet. If the damage isn't further aggravated, we're left with a scraping voice-coil. The speaker will probably work, but any sound will be marred by unpleasant distortion. More often, the driver dies outright as the voice-coil wires become detached from the former and are smashed against the magnet, or worn down by the friction.
The temperature above which adhesive softening imperils the voice-coil, largely governs the absolute maximum power rating. Modern adhesives, and formers can withstand temperatures of 250°C to 300°C. However, temperatures of this order come hand-in-hand with an approximate doubling in voice-coil resistance (depending on whether the winding wire is aluminium or copper), so the nasty audible effects of thermal compression can intrude first, offering an aesthetic safety barrier to overdriving your speakers.
Other aspects enter when we consider a voice-coil that's already hot. After a few seconds at high level, let's assume the voice-coil temperature has attained, say, 190°C. Now, for bass drivers, a strong asymmetric waveform from a synth will take a large area of the voice-coil out of proximity to the magnet. This is particularly true for long-throw speakers: the types with a springy, rubber surround.
Normally, the entire coil area is adjacent to the magnet. The bulk of the magnet helps to cool the coil, as too does the highly turbulent airflow in the gap. Once out of this zone, the tip of the voice-coil will suffer a rapid rise in temperature.
A DC offset arising from a faulty amplifier can have the same effect. An offset of 5 volts isn't enough to qualify as a 'DC fault', and it won't harm the speaker under quiescent conditions, or at low levels, but around 20% of the voice-coil area will be displaced substantially 'out of the gap', making it vulnerable to premature overheating, if it doesn't hit the endstops first.
The most important secondary factor is amplifier overload - or clipping; the point where the signal hits the amplifier's own endstops.
For simplicity, I've illustrated this (Figure 1) with a pure sinewave (a tone), but real music signals are more complex. In particular, they're much more peaky, and frequently asymmetrical. Nevertheless, the sinewave shape will help us get to grips with the basic dangers of overloaded amplifiers. Here's some more lateral thinking: The term clipping stems from the apocryphal teacher, who, anxious to put across to his students the concept of 100% distortion in feedback amplifiers, in a dramatic fashion, used garden shears to lop the top off a cardboard model of a sinewave! More mundanely, scissors are often used to change not just the shape, but also the area of a piece of card. Looking now at Figure 3, do you see how a clipped signal encompasses a greater area than a pure tone? If you did any science 'A' levels, you'll perhaps recall the significance of 'The area under the curve'. Perhaps you used it as an 'at-a-glance' check on the validity of the answer to a tricky integration problem. For our purposes, no abstract mathematics is needed to suss out that a clipped waveform involves more power than a pure tone - for the 'area under the graph' represents power in this instance.
To an extent, this is stating the obvious. Even though the signals' peak voltage level is curtailed, most of us know in practice, that the power output of an overdriven amplifier is greater than the power of the same amplifier when it's just on the threshold of clipping. But by how much?
Checking the area under the curve, we can arrive at some figures. Driving an amplifier only 1.5dB into clipping spells a 20% increase in heat dissipation. So, if the voice-coil temperature is already 200°C, it will rise to around 238°C and with a 3dB overload, power dissipation is doubled resulting in a hypothetical doubling of temperature to around 360°C. Strictly, this applies solely to pure sinewaves, but it does demonstrate the general trend. In particular, it underscores the importance of accuracy in the overload indicators on power amplifiers, if they're to be of any help at all. A 2dB error renders metering of this type next to useless.
Amplifier clipping has other ramifications, apart from taking an already hot voice-coil into the danger zone. Looking at Figures 2 and 3, note how the clipped waveform stops suddenly - it has none of the graceful curves at the top of the original sinewave shape; tracing a clipped waveform for the driver is rather like accelerating from 0 to 800 mph into a brick wall. Generally, this is most traumatic for compression drivers and tweeters. Typically, the diaphragm or voice-coil assembly falls apart under the massive g-forces, or at the very least, the delicate connecting wires snap.
For bass drivers, mechanical over-excursion per se is more significant; this can happen when we add 20dB of EQ at 30Hz to an already fulsome bass sound. The distance the cone moves is inversely proportional to frequency, viz. the largest excursions are on the lowest notes, for any given sound level. With 20dB of boost, and assuming the amplifier doesn't clip, the cone will move 10 times the original distance, and even at medium sound levels, hitting the endstops is a distinct possibility. The loud 'crack' on some Thiele-loaded speakers is perhaps the most familiar example of this event. Cone-flap is aggravated at the lowest frequencies, where the driver isn't properly loaded against a stiff cushion of air.
Back to clipping: the chopping-off of a waveform generates copious quantities of harmonic energy. In a 2-way speaker system, a clean 800Hz tone won't normally be audible on the tweeter. With an overloaded amplifier though, the 3rd and 4th harmonics at 2.4kHz and 3.2kHz will be passing a significant proportion of the total power over to the tweeter. In other words, overloaded amplifiers have the ability to translate the frequencies at which power is concentrated.
Put in the context of a system playing loud music, and with the tweeter at 80% of it maximum temperature, the extra power dissipation, thanks to harmonics arising from mid-band overload, can prove altogether too much for the tweeter.
Turning for a moment to active speaker systems, if the midrange (or bass) amplifiers clip, the tweeter won't see any of the harmonics - or any clipped waveforms at all, for that matter. So once again, the absence of harm comes hand-in-hand with clean sound.
Next time, I'll put together these disparate factors into a tidy list of do's and don'ts, and bring the power amplifier's rating into the discussion. We'll also look at how speaker ratings are arrived at.
Feature by Ben Duncan
Previous article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!