Debunking the debunkers.

[Jon Lane]

Introduction. As many know by now, TAI provided various speaker samples for a recent weekend listening event in which the sound of a variety of loudspeakers was rated. We're proud to report that a number of them excelled in blind, level-matched listening in front of a panel of experienced listeners, in most cases beating out more expensive competing models, all of them excellent in our view.

On what first appeared to be valid technical grounds, a few members of the audio community, presumably fans of models from other brands and one a competitor himself, challenged these findings. One individual's assertions tacitly disparage TAI models with a claim that no consumer guidance could be drawn from the results of this lengthy blind A/B listening session.

Once this seed of doubt had been planted, other contrarians chimed in as if the premise were proven fact. Obviously, it isn't, as others have pointed out as well. I've declined comment until now.

TAI has never engaged in controversy with proponents of another brand. We have not criticized our competition. We do not typically debate these topics with anonymous online personalities. We feel this would justify arguments such as this one which, being faulty, are simply not in anyone's interest.

Today I'm going to make an exception. What are actually technically questionable and even highly subjective criticisms are being leveled and they should be rebutted. One example of these comments even condemns this post well in advance of my writing it.

Heck...we now have the reseller that is involved saying that there is something technically incorrect about the level matching issue and is going to post it on his forum. Plain ridiculous.

The remarks criticizing the results of this event lack merit but to dispense with them and to maintain the good name of our partners and team members, I'll address them once. There's nothing ?plain ridiculous? about preventing mistaken notions becoming common wisdom.

Level Matching. A competitor is quoted here as suggesting that the event did not include an acceptable level-matching device. From everything we can tell, it did, and I'll explain how shortly.

The competitor initially remarked:

Further research revealed that Craig was using a passive remote switchbox. This is a speaker selector box that is placed directly in-between the loudspeakers and the amplifier. This particular switch box is made up of high-power potentiometers, which are variable resistors. In order to level match, series resistance is used to pad down the power being received by one of the two speaker pairs. The higher sensitivity speaker must have series resistance applied to the signal in order to bring the output level down enough to match the lower sensitivity speaker (which would not have resistance applied).

Yes, speaker level matching is accomplished by inserting various resistances into the speaker's circuit ? a variable complex load is adjusted until measured broadband loudness between two or more speakers under test matches.

Here, the competitor claims a simple series resistance lowered the level of the more sensitive speaker. Note the use of ?complex resistance?, above. In this context it is our contention that an ?L-pad? was used. We are awaiting verification of this assumption but I can state with certainty that the odds are fairly overwhelming that the level matcher was an L-Pad and not the simple series potentiometer implied.

In an L-Pad, for every such series resistance placed between speaker and amplifier, a second and corresponding resistance is loaded across that combined series resistance and ground, effectively joining the two as effective level-changing circuit elements, but not appreciably changing the effective load. The L-pad thereby forms a pair of linked, adjustable of resistances, the one rising in resistance as the other is lowered and vice versa. There is an effect on overall circuit impedance, but it is slight.

The purpose of an L-pad is to adjust level. In fact, this particular level matcher is said to be inaudible by it's builder, a veteran industry professional and head of a large loudspeaker brand.

The arithmetic is simple:

1.For a series resistance (R2) added to speaker load (also a series impedance, R1) the formula looks like this (we'll leave aside the reactive components in the speaker load for the purposes of understanding how this circuit functions within the context of the claims made against the PA level matcher.):

R1 + R2 = R3, with R3 being the total combined load.

Example: An 8 ohm speaker with 4 ohms series resistance results in a 12 ohm load.

8 + 4 = 12, or 50% over the design impedance

2.For an L-pad complex resistance, one of it's elements (R2) is in series with the load (R1) and its shunt resistance (R3) is introduced to ground. We'll assume a 12 ohm L-Pad for our example.

( 1 / ( 1 / R3 ) + ( 1 / R1 ) ) + R2.

Example: 8 ohm speaker with an identical 4 ohm series resistance (and virtually identical speaker attenuation) plus the remaining 8 ohms shunted to ground results in an 8 ohm load.

( 1 / ( 1 / 8 ) + ( 1 / 8 ) ) + 4, or the actual design impedance.

(I'm dashing these off, so I am entirely open to correction or modification. If anyone has the details of this particular level-matcher, please email me.)

Summarizing these findings tells us that if the level matcher is a proper L-pad, the kind of which is far more likely and almost certainly the more customary for the event's supplier to have provided (that supplier being, we're told, Ian Colqhoun at Axoim Audio) it's net sonic effect is negligible, if not completely inaudible.

I should emphasize that this is a simple analysis, that it disregards source/load impedance mismatches, ignores reactances, and that it assumes a 12 Ohm L-pad, which is a reasonable assumption. However the net result remains largely the same: An L-pad is not a simple series resistance. It behaves entirely differently in terms of total circuit impedance, the kind of which would upset a speaker's designed response.

I'll repeat: Valid speaker level-matching, the kind of which very likely occurred here, has no significant effect on the net circuit impedance presented to the driving amplifier. It does not become a frequency-selective filter to the speaker, interacting with it's reactive load to unbalance it's response. It cannot be ascribed even a fraction of the importance needed to radically change the speaker's amplitude response. As the series resistance changes, the shunt or parallel resistance changes in inverse, restoring the overall circuit to close to original impedance for the purpose of the test.

Regardless of the precise configuration of the level-matcher, the measured results on-site show that no speaker suffered any degradation of designed response whatsoever, thereby disproving the claim that a series resistance ruined any particular speaker's response.

Manufacturer Claimed Sensitivities. The second finding is ironic and it's far simpler to bring to light: In one of the more controversial comparisons offered by our critics, the claim made is that ?their? speaker was severely attenuated, thereby, under their incomplete level-matching analysis, suffering substantial sonic damage.

In fact, our model was attenuated (it being more sensitive) by about four tenths of a Decibel, and the favored competing model was run without any attenuation whatsoever. As it turns out, no speaker in this event was level matched more than one dB, a level presenting a truly miniscule circuit impedance variation to the speaker.

Other Effects of Level-Matching. Third, an added minor phenomenon relates to the L-pad actually linearizing the speaker's load. In the case of the speaker attenuated most, that speaker being the alleged victim, the level-matcher's shunt resistance slightly minimized that speaker's reactive load, a phenomenon routinely exploited by designers, including myself, to ease both the crossover-to-driver and amplifier-to-speaker interface.

In this case, the speaker our competitor thinks may have been impaired by faulty load-matching actually ended up becoming a very slightly easier and less reactive load to the amplifier, effectively making that speaker ? the speaker with deeper attenuation so as to lower it's slightly greater initial sensitivity ? the technically slightly advantaged speaker of the set under test. Whether this is audible is highly questionable, but there are the facts.

In short, any claim that a more sensitive speaker suffered either by way of excess series resistance or by the presence of the level matcher at all, is disproved. The attenuated speaker had no appreciable and certainly no audible disadvantage, and it's attenuation actually may have enhanced it's load presented to it's amplifier, at least in theory. Just as there was no net change in impedance in this particular test of anything approaching the magnitude needed to radically disadvantage one particular speaker, there is none in tweeter circuits containing properly designed L-pads to change the speaker's average impedance magnitude.

Logical Fallacies and Other Mischief. As I mentioned, planting seeds of doubt is the first step to nullifying what are otherwise potentially useful findings. The 4th issue bandied about in that interest relates to the environment: The room was somehow inadequate, the participants were not suitably motivated, the speaker grille cloth used to obscure the samples from view was acoustically intrusive, and any number of other unusual variables prevented anything approaching a valid set of sonic witnessed data being condensed from this event.

One can respond in kind: Let's not forget too that the participants were all male (women are widely documented to have superior hearing.) It was perhaps too cold. Or too warm (the speed of sound changes with temperature.) Too humid; too dry (ditto.) The speakers were designed at different altitudes than heard at this event. The sources were not perfected, speaker by speaker. No analog sources were present. No line conditioners were present, or if they were, surely had a negative effect on pet models. The loudspeaker and/or interconnecting cables were unknown and therefore inferior. There was alcohol present. There were jumbo shrimp present!

Or maybe the choice of music advantaged some speakers and disadvantaged others (when it was not disadvantaging some and advantaging others.) There were too many people in the room. There were too many chairs in the room. The room had too high/low a ceiling. The dimensions were unacceptable, at least to the favored models. The wall voltage was either over or under 120vac. Or was it 115vac?

The point that needs to be made is that all speakers present were subject to the same exact set of conditions. The playing field was absolutely level.

Enough already.

Summary. We have high regard for our competitors. We do. But we also have a lot of regard for the scientific method as opposed to sowing doubt based on one's agenda, such as it may be. We trust these non-issues will be given precisely the attention they actually deserve.

[Chris_Brunhaver]

Knowing some details about the Acculine construction and speaker measurement / rating practices, I have a few things to add.

Firstly, while sensitivity of a speaker is quite important, depending on the application, it can also be confusing and lead to "specsmanship" (not that that is necessarily the case here). While two speakers can be rated as 8 ohms nominal, actual impedance can be quite different, leading to different voltage sensitivities and current demands. To compound this, many times frequency response is perfectly flat axially and so it can be hard to meaningfully distill the frequency response into a single number. There is some good debate and information about this on the Porsoundweb.com forum, as sensitivity ratings are possibly more contentious in pro audio.

Overall, I think the sensitivity ratings on the Acculine are pretty Accurate, if not a hair conservative. With the BG Z-1 (a more expensive cousin to the A-1) that was the case, as seen in these third part measurements. http://stereophile.com/standloudspea...bg/index3.html

I'm not intimately familiar with the Ascend stuff but have seen some measurements form Dennis Murphy (on the Madisound discussion forum) and the NRC Soundstage measurements of the CBM-170. Ascend posts some measurements of their own which also seems to agree well with these other sources.
http://www.ascendacoustics.com/pages...t340mmeas.html
However, they don?t list measurement voltage but rather say dB / watt (8 ohms @ 1 meter). I assume this means 2.83 V. The CMT-340, by their measurements is ~88.5 dB at 1 watt. They call it 90 dB and then add 2 dB for an ?in-room? sensitivity.

Contrary to a A-1, their speaker?s impedance looks pretty low for an 8 ohm load, with the impedance dipping to ~5 ohms.

This appears to also be the case with the Sierra, the measurements show 85 dB sensitivity with impedance dipping below 5 ohms (and presumably a 2.83 V measuring voltage), probably putting the true 1 watt SPL somewhat lower than specified. However, I don?t doubt that 5.25? Tang Band woofer has good bass extension in that pretty big cabinet.

I think this may explain some of the discrepancy seen in the get together versus published specs. Still, this is a good debate to have as, philosophically, especially for HT use, good sensitivity is rather important for the required dynamics, dynamics being a key to the emotional you-are-there experience.

[Jon Lane]

Thanks for stopping by, Chris, and I hope we hear much more from you as the Acculine project progresses.

(Chris is one of the Acculine engineers and a good friend from BGC. He's also an Audire alum and one of the most savvy speaker gurus I know -- we'll be chatting about some quite interesting and esoteric new topics in the months to come. Meanwhile I hope Chris keeps us on the straight and narrow on all things Acculine... )

Chris also brings up a good point: Speaker sensitivity and speaker efficiency are not the same thing. Efficiency is sound level versus power -- typically the output level we get when we set up a test that delivers precisely 2.83 signal volts across an eight ohm loudspeaker and then measure it at precisely one meter.

2.83vac across 8 ohms is one watt.

8 ohms is close to the customary loudspeaker load. However there are many speakers in the field rated 8 ohms when in fact they're closer to 4. Loudspeaker impedance is neither constant with frequency nor is it standardized where 8 = 8 and 4 = 4. It's all in the manufacturer's specmanship.

Sensitivity, however, may show what happens when we use a current-source amplifier (many performance solid state receivers and amps, the kind we typically recommend) that simply raises it's power when the load diminishes. In other words, it delivers 2.83 signal volts across four ohms as well, which adds 3dB to the speaker's output by comparison to the 8 ohm version.

The correct and customary way to rate a speaker's output power per watt is simply to never drive it with more than one precise watt. If the speaker is a 4 ohm load, this means reducing the level to measure 2.0 volts across the speaker under test.

To summarize, efficiency is the measure of a speaker's ratio of input power to acoustical output, typically the one watt output, measured broadband at 1 meter.

Sensitivity is typically the measure of a speaker's output when driven, regardless of it's impedance, with what would have been 1w of power had we sent it into a 8 ohm load. Sending 2.83v into 4 ohms gives us in effect two watts, skewing the sensitivity figure up by twice, or 3dB.

2.83vac across 4 ohms is two watts.