What makes a High Efficiency Speaker

As you've probably noticed, Punkuk, there's a wide diversity of opinion as to what exactly consitutes a "high efficiency" speaker.

One thing you should keep your eye on is exactly what's being specified by the manufacturer. Typically manufacturers of speakers with an impedance below 8 ohms specify the 2.83 volt sensitivity, as that gives more impressive looking numbers. And typically manufacturers of speakers with an impedance higher than 8 ohms specify the 1 watt efficiency, again because it looks better. Here's how it goes: 2.83 volts into 8 ohms is 1 watt, but 2.83 volts into 4 ohms is 2 watts, and 2.83 volts into 2 ohms is 4 watts. Going the other way, 2.83 volts into 16 ohms is 1/2 watt, and 2.83 volts into 32 ohms is 1/4 watt. When you turn the volume control knob on your preamp, you're changing the voltage output (and the wattage output changes along with it). When you leave the volume control setting the same and switch to lower impedance speakers, the voltage output stays the same but the wattage output increases (assuming the amp doesn't clip).

Let's do an example of converting sensitivity to efficiency. Suppose you have a "93 dB sensitive, 4 ohm" loudspeaker. Sensitivity refers to 2.83 volts input which into a 4 ohm load is 2 watts, so the speaker is really doing that 93 dB with TWO watts input, so the efficiency is only 90 dB with ONE watt.

Unfortunately, some manufacturers use the word "efficiency" when "sensitivity" would be more appropriate, and vice versa. And unfortunately some manufacturers use an "in-room" rather than "anechoic" or "simulated anechoic" measurement, which inflates the specification by 2-3 dB due to the added reverberant energy. Using an "in-room" measurement is not without justification, but does complicate things when you're trying to make an apples-to-apples comparison.

In my opinion efficiency is a more useful specification than sensitivity because wattage input (not voltage input) is what relates directly to voice coil heating and thermal compression. Then if you want to you can take into account how much power the amplifier you have in mind puts out into that particular speaker's impedance. Most solid state amps increase their power output into lower impedance loads, but often there's a sonic trade-off, as many amps distort less into a high impedance load.

I'm not qualified to say where anyone else should draw the line between "high efficiency" and "not high efficiency". Personally, I consider 91-94 dB as moderately high efficiency; 95-98 dB as high efficiency, and 99+ dB (all with a 1 watt input) as very high efficiency. But in my book those are only ballparks, and the lines are blurry.

I've focused on measurements, but your question could be approached from other angles as well (how is a high efficiency driver constructed, what are the techniques commonly used to build a high-efficiency loudspeaker system, what are the challenges involved in building a good one, why would you want one in the first place, etc.) But I'll save the muddying of those waters for another day.

Duke

Swampwalker - thanks for your kind words! Regarding line source loudspeakers, maybe I can offer a few thoughts.

From a point source, radiation intensity falls off at 6 dB for every doubling of distance, and from a line source radiation intensity falls off by 3 dB for every doubling of distance. With a point source the radiation is expanding in all three dimensions, but with a true line source (which would extend infinitely in one dimension) the radiation is only expanding in two dimensions - hence the more gradual falloff with distance.

In practice how well a speaker approximates a true line source depends on the height of the radiating line or line array, the height of the room, the wavelength being reproduced, and the distance from the line-source-approximating speaker to the listener. As a general rule of thumb, line source characteristics will hold up quite well out to about four times the height of the line-source-approximating radiating element(s), then begin transitioning to point-source characteristics.

There is yet another type of source - a planar source. With a true planar source (infinite extension in two dimensions), there is NO falloff in SPL with distance! Up close to a large elecctrostatic panel, planar source characteristics dominate - but we're talking distances of only a few feet at the most.

Getting back to line sources, I once measured a point source speaker and a line source speaker at 1 meter and again back at 8 meters (practical limit in my room). Anechoic theory predicts the point source speaker's radiation would fall off by 18 dB over that distance, and it fell off by 11 dB. That extra 7 dB came from the reverberant sound field. Anechoic theory predicts that the line source speaker's radiation would fall off by 9 dB over that distance, and it fell off by 4 dB, with the reverberant field contribution making up the difference. As you can see in a real-world room the line source speaker's radiation fell off by 7 dB less over that distance (1 meter to 8 meters) than the point source speaker's did. That's a significant difference. But in this case the point source speaker was still more efficient than the line source speaker even measured back at 8 meters.

Opalchip - very good explanation of Hoffman's Iron Law. Having scratched my head about the Druids quite a bit, I think that they are exploiting undamped pipe resonance to extend the bass deeper than it would normally go, but the tradeoff is the deep 150-hz notch revealed in the SoundStage measurements (indicative of a roughly 90 inch long pipe). Having built undamped pipe speakers myself I'll say that they measure worse than they sound - the ear is surprisingly forgiving of that deep notch.

Let's look a little bit more closely at the Zu specs though, and give 'em the benefit of the doubt just for kicks. The Druid is a 12 ohm speaker, so what if when they claimed 101 dB "sensitivity" (implying 2.83 volts input) they really meant "efficiency" (implying 1 watt input)? If so, then translating that to 2.83 volt sensitivity we'd come up with about 99 dB. SoundStage came up with 97 dB/2.83 volt sensitivity. That's pretty close, especially if we let Zu use an "in-room" rather than "anechoic" or "simulated anechoic" measurement.

Next let's look at the claimed bass extension. Zu claims a "bandwidth" of 38 Hz to 25 kHz. We're used to thinking of the bandwidth as the -3 dB points or maybe -6 dB points, but that might be a mistake on our part. In prosound use, the -10 dB point are often given as the limits of a speaker's bandwidth.

Now let's go back and look at the SoundStage measurements. See that little response bump at 38 Hz? I think that's the 1/4 wavelength pipe resonance (though can't correlate it with the impedance curve). Relative to the 97 dB sensitivity determined by SoundStage in a free-air measurement, that 38 Hz bump is 19 dB down. With the reinforcement of three room boundaries, we'd add 9 dB and then we'd be only 10 dB down. And -10 dB fits the prosound definition of bandwidth.

I don't know if this is how Zu arrived at their specifications, but it might be. I welcome correction from anyone who knows the real story.

One final comment on the perceived loudness of the Druid's bass. If indeed the Druid's enclosure is an undamped or minimally damped resonant pipe, then it will sound louder in the bass than it measures. This has to do with human hearing - if two fairly short-duration sounds are exactly the same frequency and sound pressure level, but one lasts longer than the other, the longer-lasting sound will be perceived as louder. So if we have relatively slowly-decaying bass resonances in the Druid's enclosure, the bass will sound louder than it measures. As to why the Druids don't sound resonant and boomy, I think it's because the shape of the frequency response curve keeps us from perceiving it that way. If so, then the Druid's bass loading system (based on a patented automotive muffler design) would not work well with a speaker that measures "flatter" in the bass region. The height-off-the-floor tuning, which apparently is critical to getting the bass to sound right, mght be adjusting the decay of the muffler-like pipe resonances. It may well be that Sean and Adam and the crew of Zu are really pretty darn advanced in their design work if they've taken advantage of psychoacoustics to this extent.

The above paragraph is purely speculation on my part, and once again correction is welcomed.

What Zu has done is build a speaker whose perceived bass extension is much better than I ever would have thought possible in that box size and efficiency combination. I don't think the Druid is without sonic issues, but it is at the very least an intriguing design.

Duke