September 23, 2011
“We Said What…?!?” or Measured vs. Calculated Maximum Output
On Facebook today, our fan Jez Minter asked about maximum output specifications. He had read an article from our friends at Mackie about how some manufacturers generate this statistic, and, through no fault of his own, had come away with the basic message that calculated maximum output specifications are inherently flawed and intended to mislead. Given the context for which the article was intended, that’s mostly true.
His question to us was, “Why does EAW publish a calculated maximum output specification?” It’s a very good question, and we had quite the discussion in the engineering hallway. Naturally, we got to the bottom of the controversy and want to set the record straight.
What the Mackie article says is more-or-less correct – more so when dealing with Mackie’s competitors in the guitar store PA market, but less so when dealing with higher end manufacturers like EAW.
When we review the basic assumptions of the article, and we’ll quickly see where things have gone wrong. EAW does not – nor would we ever – follow the calculation process as described in the article. Nor would Mackie. But Mackie’s competitors do use that process, and the article was written about their calculations, not ours.
The Context
The backstory is that EAW worked with Mackie to develop the HD range of powered loudspeaker systems. We fully support that product line and believe it delivers exceptional performance at that price point. The HD Series used the same rigorous measurement and documentation process that EAW loudspeaker go through, and its maximum output was calculated as are all of EAW’s loudspeakers.
Mackie’s competitors, however, used a different approach to come up with a higher maximum output statistic – one that overstates the system’s capabilities. Thus, in the measurement comparison, the delta between calculated and measured performance is much smaller for the Mackie product.
But it raises the obvious question: why is there any delta at all between the published specification and measured performance? We’ll deal with that question last.
First, let’s look at what the article says and what it doesn’t say. The article takes an unnuanced approach to the discussion of calculating a maximum output specification. Specifically, it gives the impression that this stat is calculated as opposed to measured in order to mislead. While that might be true of some companies, it’s not what we do at EAW.
The Nitty Gritty
The article rightly points out the flaw in taking _peak_ axial sensitivity as the starting point for calculations. That is a worst practice and would only be used to produce a overblown specification. That much is accurate. But it also implies that that is the _only_ way to calculate this spec. That’s not true.
In fact, EAW uses a very different approach to this calculation. We use the _broadband average_ sensitivity as our starting point. This means that our calculated maximum output will be significantly lower than if we used the other approach, but it will be more accurate. The graph above – hand drawn by EAW engineering director Nathan Butler and annotated by yours truly – shows how an unscrupulous company would use peak sensitivity as a starting point to calculate a bogus max output vs our approach of using the average sensitivity to calculate a solid max output. (Yes, bogus and solid are the technical terms.)
Axial sensitivity is the core statistic for a transducer. It measures output at a very close range (1m) and a very low current (1 Watt, nominal). We say nominal because given continuous voltage, Wattage varies with impedance, and a transducers impedance varies with frequency. Therefore, there is no such thing as a 1 Watt broadband measurement. At EAW, this test uses “voltage that would produce 1 Watt at the nominal impedance”.
Splitting the hair four ways…? We want to be completely transparent about what we do to develop our specifications.
So, if you get a good sensitivity measurement _and_ use a rigorous approach to generate a broadband average, you should be able to add the SPL equivalent of a given power handling statistic (long term broadband measured with pink noise) to calculate a maximum output specification that can be verified in real world measurement.
Why Not Just Measure?
The answer to this question – and to the question of why the measured output of the loudspeakers in the article fall short of the published specification – is that measuring very high output levels is difficult to impossible. That is, there is an upper limit to the capabilities of measurement instruments. Very good testing microphones go into the 130s and a very small set reach 140 dB SPL. Beyond that, it’s a guessing game.
For EAW, so few of our loudspeakers have maximum outputs below the mid-130s that it’s functionally impossible for us to measure maximum output. Our stuff is just too loud! But even if measurement microphones went to infinity, you’d still need a very large space or an anechoic chamber. EAW has The Pit, but it’s not big enough to measure very high outputs.
The reason is that as output levels increase, so does reflected sound. Lower volume tests let us “window” the data to eliminate reflections, but this is not possible in a maximum output measurement that averages output over a period of time. High levels of reflected sound that appear as “late arrivals” will cancel out direct output at relevant frequencies, so your measured output can actually decrease with increasing direct output. (Crazy, right? Welcome to our world.)
Long story short: for EAW loudspeaker systems, measured maximum output would be less useful and less accurate than our calculated output. Our goal is to deliver the most accurate, most useful data possible, and we believe we do so.
Then Why the Difference Between Calculated and Measured Output?
[Note: This is how you know you're dealing with a straight shooter - they say they don't know when they don't know.]
We don’t know, nor can we ask the questions that would help us unravel the mystery. The responsible engineers are no longer with us. But several factors suggest themselves.
Obviously, “power compression” would account for some degradation of maximum output. Power compression is the phenomenon in which a transducer’s voice coil heats up due to the high voltage passing through it. Heat increases impedance, so the more power you deliver, the more impedance increases above the nominal specification. More energy radiates as heat and less radiates as sound. That said, the 9 dB difference indicated in the article would be an extraordinary amount of power compression. Something else is awry.
First, we might want to go back and examine the original sensitivity data to be sure that the broadband average is an accurate one. Given the rigors of our process, we expect it to be, but you never know until you look.
More likely, though, is that the issues associated with measuring very high output in a closed space and/or the specifics of the measurement techniques used have produced unreliable data. Until we know more about exactly how these tests were conducted, we couldn’t develop a meaningful response.
The Bottom Line
A 9 or 10 dB difference between published and measured maximum output is unacceptable. That competitors have an even bigger difference is irrelevant; our data must be accurate and useful. EAW intends to look into this issue and develop a more thorough response.
Again, we need to thank our Facebook fan Jez Minter for bringing this to our attention.
