“Let our rigorous testing and reviews be your guidelines to A/V equipment – not marketing slogans”
Facebook Youtube Twitter instagram pinterest

Audioholics 2010 Subwoofer Shootout Measurements Per CEA 2010

By paul

The Purpose of This Subwoofer Shootout

The purpose of this subwoofer shootout is simple: compare products for performance using measures and techniques as defined by the CEA 2010 standard.   The CEA 2010 standard is a means for measuring the maximum output of subwoofers at the two lowest audible octaves, 16-32 & 32-63 Hz.   Unlike a traditional harmonic distortion number, the CEA number attempts to weight different harmonic distortion components based not only on their amplitude relative to the fundamental, but on their separation in frequency as well. 

 60 hz.jpg

Example of Product with Excessive 3rd Harmonic Distortion
Fundamental Frequency =60 HZ, 2ND = 120 HZ, 3RD = 180 HZ, 4TH = 240 HZ, 5TH = 300 HZ, ETC.

Above is a spectral plot of the output of a 12” subwoofer in a box tuned to 40 Hz.  While this author found this level of third harmonic distortion easily perceptible, I must note that this output would be considered well within the CEA third harmonic distortion limits imposed by the CEA standard.  One of those individuals who had taken part in the creation of the standard has personally voiced opinions to me that perhaps the CEA standard is too lenient.   I have also heard this comment from some of the vendors who submitted product for our own testing.  Those vendors, such as SVSound, who held themselves to a higher standard than the CEA published “acceptable” distortion limits, actually placed themselves at a disadvantage because they had the system “draw the maximum line” at a location lower in distortion than CEA allowed for what it considers to be acceptable at maximum usable output. One must bear in mind, that both an amplifier with a built in compressor, as well as a loudspeaker, do not have a brick wall limitation in performance.  (While technically possible to build an amplifier this way, the resultant distortion it would produce when it hit the threshold would be egregious.  Some of you already know it, and refer to this as “Hard Clipping”.)  When an amplifier is rated on power, the number is meaningless if not accompanied by a distortion figure, and whether or not the output is peak or RMS (which most of us think of as continuous instead of transient or short term.)  The same is true when testing subwoofers for maximum output.  You might be able to drive them past their intended maximum limit, but at what price in distortion?  The CEA standard not only attempts to define that limit, but does so in a completely quantifiable way.  Note the graphic below titled Signal Spectrum.

20Hz Fundamental.jpg

20 Hz Fundamental output within CEA allowable distortion limits

Please note the stepped red boundary which looks a bit like the side-view of a staircase.  That red line represents the CEA 2010 Sub-Woofer standard limit on distortion.  It does not represent a certain amount of distortion in absolute terms.  It does shows the acceptable upper limits of distortion as a percentage of the fundamental output. (The test frequency).  Note the center of the large “finger” shape in blue.  It is centered at 20 Hz, in this case the test frequency.  Notice a second finger centered at exactly twice the frequency, 40 Hz.  That is second harmonic distortion.  For this product, that is the dominant distortion when driven at 20 Hz to this output level.  (It may well change if driven harder towards its physical limitations).  Note also, the level of 0dB at the peak output centered at 20 Hz.  This is NOT a measurement of the output in dB relative to any particular sound pressure.  The program captures the signal output, and transforms the signal into a spectrum.  Whatever the sound pressure is at the driving frequency (in this case 20 Hz), this peak value will be considered the reference, or “0 dB” value.  It means that the other frequencies present are measured RELATIVE to the fundamental, which is why distortion can be expressed as a percentage.  Notice that as the harmonic frequency is farther and farther away from the driving frequency, the CEA limitation, (like your ear) is more and more intolerant.  The red stepped line means that the second harmonic distortion can be up to 10 dB less than the fundamental, the third an be up to 15 dB less, the fourth and fifth 20 dB less, and so on. This is weighting the distortion spectrum in a way that corresponds to how the human ear-brain mechanism responds.   Since the distortion components are all below the threshold of the stepped red line, (CEA standard limits) this speaker at this test volume passed. 

During my testing, I was sitting about 40 feet away (to the side) of the units under test.  While I cannot determine how different the spectrum of sound was at the microphone relative to my position, on a relatively small number of occasions when the software recorded the output as a "FAIL", or "PASS", I might have decided from listening alone, that should not be the case.  Despite that, I would argue that the CEA 2010 standard is as good or better than any other metric I can name for relating subwoofer distortion to usable output.  The vast majority of the time, my ears would draw the line at the same levels the software would report a "FAIL".  Let's take a look at the graphics shown to the test operator in the image below which shows a failure due to an excess of third harmonic distortion (60 Hz resulting from a 20 Hz input).


Excess Third Harmonic Distortion as per CEA 2010 Standard

As we can see from this example, the distortion of this speaker under test is predominated by third harmonic distortion (60Hz), and it rises in excess of what the Red stepped curve would allow as a maximum limit.  At this point, we have crossed the line (literally) and are “in the red”, so the SPL produced by this speaker was not considered “usable”.  That is, in a nutshell, how this program works.  It produces ratings of Pass or Fail only.  The only two discernible frequency components in this upper graphic are the fundamental driving frequency (test pulse centered at 20 Hz) and the 3rd harmonic distortion, centered at 60 Hz.  During testing, the frequency center is changed from 20 to 63 Hz in 1/3rd octave steps (20, 25, 32, 40, 50 & 63 Hz).  Unlike a traditional tone burst, which consists of a single frequency, this stimulus has a 1/3rd octave width.  Traditionally 1/3rd octave is considered to be “critical bandwidth” of the human ear, meaning we lack the ability to distinguish tones that are within this close to one another in frequency.  What does that mean?  Well, take a look at the first graphic (#1).  You can take note of the single vertical blue line which looks like it is centered at or about  550Hz.  This is a very narrow-band noise, which likely has less to do with the speaker under test than it does to noise being emitted by a nearby factory.  This line is how a traditional tone burst might appear, very narrow in comparison to the 1/3rd octave pulses used in testing.   The advantage of 1/3rd octave signal pulse for testing, is that it will tend to average a speakers strengths and weaknesses over that 1/3rd octave.  It is ENTIRELY possible for a speaker to have a big HUMP centered at EXACTLY 40 Hz, while at 35 Hz, it might have a huge and narrow notch.  A traditional 40 Hz tone burst would show a huge output yet would be blind to the notch.  Since 35 Hz is not a standard frequency center, this defect might go unnoticed by tone burst testing stepped in a similar way, 1/3rd octave at a time.  With the 6.5 cycle burst from Don Keele's CEA test software, running on the graphics program Igor Pro, the nature of the signal is to be DOWN -3dB at a 1/6th octave increment on either side of the center frequency.  (Meaning these pulses are 1/3rd octave wide). This is in fact quite similar (though not identical) to the kind of signal proposed by Siegfried Linkwitz to be used for testing on his site.

(http://www.linkwitzlab.com/sys_test.htm   Scroll down to the section titled shaped tone burst generator). 

Why do we care?  Is this just to make my job easier? (As if anyone really cares!) No, this is important as it is a useful way for us to collect data in a way that corresponds to how we hear, without having thousands of data points to represent all the single frequencies we can perceive.  It is also exciting the speaker with a signal far closer to music, as instruments tend to make sounds of a complex nature, containing many frequencies simultaneously.  It is common to test a speaker with a single frequency, either stepping or sweeping it up and down to “connect the dots”, yet this kind of excitation is essentially unheard of in music.  Musical tones are full of harmonics and generally have a wide spectrum looking completely different from traditional test tones.


Confused about what AV Gear to buy or how to set it up? Join our Exclusive Audioholics E-Book Membership Program!

Recent Forum Posts:

KEW posts on February 17, 2011 14:41
Paul_Apollonio, post: 792889
(Just like my puppy!) LOL

What the heck is that thing? (aside from creepy!)
Paul_Apollonio posts on February 17, 2011 14:30
pbc, post: 792921
Also, would it be possible to post what the sweep levels were that correspond to each of the colors on the graphs?

The sweep levels I used varied from one sub to the other, as there was no standard output power for any given voltage input. (The input sensitivities are not standardized to volume knob position). What I did was this; I started at a low level where there was no question the sub was linear, and drove it up by 2db per curve (input voltage step) until the compression effects became visible to show that frequency response is a dynamic and not a static phenomena. When I first ran the HSU Sub, I had the volume knob set to maximum. It had so very much voltage gain that this caused the problem of having to use really LOW voltages on the input, so much so, I stopped, turned the gain to 12 noon (straight up) and reran the test. (ALL HSU CURVES ARE RUN ON CLIO WITH VOLUME GAIN SET TO NOON) Since the amp voltage gains, input sensitivity, and input sections differ from one sub to the next, standardizing the voltage input used in this test is not meaningful. MOST of the time, the red curve is 0.2 Vrms input. I have the ability with Clio to save 9 curves at one time, and display 10 (last one taken). Since the sweep is 14 second long, the power at the top curves is maxed out. What the power REALLY IS cannot be known without taking the system APART and measuring both the impedance of the woofer and the voltage at the terminals. (Which often results in the product being chipped or scratched meaning the vendors are getting back B stock, and no one wants that, especially me!) Since the mike is quite close to the box, and since some systems (Like the HSU if you look carefully) limit the input voltage at the highest frequencies, each system has been taxed a different amount of time before the power supply runs down to its final level of output. Still 14 seconds is a long time to pull maximum power, so below 100 Hz, this curve shows you what kind of frequency response to expect when you drive the product to the limits, clipping it frequently. As for getting a reference SPL, it would not be a fair contest at a mere 1 meter distance, since that number will be higher for smaller boxes than larger boxes when compared to a much great distance, as the smaller system will follow the inverse square law rule more closely than a really big box (like the Hsu) will. Another really good reason why that data was not collected is frankly, a swept sine signal is invariant in its demands on the subwoofer. We can collect that data, but for a power amp designed like the “ICE”, it will be a lousy indicator of what kind of SPL's you would enjoy with music. (Unless you play it like a DJ not an audiophile). Subwoofer bandwidth signals have a VERY high crest factor, so the CEA test is a FAR BETTER determinant on what kind of peaks you can create than is the steady state test. Adding an additional set of numbers for the consumers may have been simply adding confusion to the mix. Even CEA suggests taking those 1/3rd octave numbers, and lumping them together into bigger groups, under the assumption that the 1/3rd octave numbers (what WE presented) is already too much information for the consumer to digest. Frankly despite the length of this reply, I am just scraping the surface here, there is much more to be said on this subject. Maybe later when GENE stops bugging me by SKYPE….. I hope that shed SOME light. - Mr Paul
MinusTheBear posts on February 17, 2011 13:38
gene posts on February 17, 2011 12:09
E.g., Each sweep was at an increasing level so that one could gauge at what levels compression started and at what frequency? I.e., in the above graph one could see that the sub started to compress when a 105db sweep was run compared to the 100db curve?

Yes but Paul is extremely anal about accuracy. I personally show SPL levels for continuous sweep tests but Paul's argument is that actual SPL data may not be 100% accurate as it may not reflect losses from power supply sag, driver thermal compression and other issues. It's quite a complex topic to comprehend and I'd hate to put words into Paul's mouth. Perhaps down the road he can write up an article about it.

The max SPL on continuous reverse sine sweeps really doesn't accurately represent real-world output capability, which is the whole reason CEA-2010 was developed. But the data could nevertheless be useful from an academic standpoint. I personally publish it and compare all the subs I test like I did with the Emotiva sub.

The reason why I rated the Emo sub a 4.5 in performance wasn't b/c of its extension capabilities but b/c it was able to play at MAX SPL under extreme stress and the woofer never complained. Yes its extension is quite limited, but its also a very small sub and very conservatively designed. You can't break it.

Continuous tests are really useful in separating the men from the boys in subwoofers IMO.
pbc posts on February 17, 2011 11:41
gene, post: 793062
The SPL calibration was performed for a different set of hardware and NOT for the CLIO. We post SPL data per CEA only for these subwoofer review series and and use freq sweeps to show product linearity under continuous testing. We stand by the calibrated SPL PEAK numbers from the Don Keele/Igor ro/Soundcard/mike Preamp setup.

If you want to extrapolate RMS data from CEA #'s subtract 3dB. If you want to relate it to in-room SPL add between 6-18dB depending on subwoofer location, frequency and room gain factors.

These SPL #'s are short term per the burst tests and don't take into account heating effects of the driver voice coil or power supply sag in the amp section. Heating effects during sustained output can account for addition compression of up to 6-10dB some of which can be seen in the CLIO curves.

Sorry Gene, not following possibly as I'm not familiar with the CLIO curves.

Are the sweeps in the above graph not something similar to what Illka was doing here …

E.g., Each sweep was at an increasing level so that one could gauge at what levels compression started and at what frequency? I.e., in the above graph one could see that the sub started to compress when a 105db sweep was run compared to the 100db curve?
Post Reply