As I was wrapping up my review of the Atlantic Technology AT-3 loudspeaker, I called SoundStage! founder Doug Schneider to discuss the forthcoming measurements—something I rarely do. As is usually the case, there were things I liked about the speaker and some things I would change, but as I said in the review, my biggest concerns were that getting the acoustic center of the speaker aimed at ear level was a fight, its sound seemed excessively bright from a normal seating height, and its vertical dispersion was . . . distinctive.
Needless to say, I’m not there when we measure speakers at Canada’s National Research Council (NRC). When these were in the anechoic chamber in Ottawa, I was down in Alabama banging out another review or working on the next episode of the SoundStage! Audiophile Podcast or something along those lines. I really, really wanted to be there the day the AT-3 was measured, though.
And I may as well admit it: the main reason I called to pester Doug was that I wasn’t quite sure whether the measurements would reflect everything I was hearing. I knew Doug asks for measurements at 45, 60, and 75 degrees off-axis. I assumed that was horizontal. Did we also go up and down off-axis? If not, could we? Otherwise, the measurements of these speakers were going to make my subjective impressions look like outright barking insanity. Doug assured me that he’d go the extra mile to make sure we got the individual graphs of measurements taken above and below the tweeter, which are components of our Listening Window measurements to begin with but which aren’t normally presented on their own.
A few weeks later, Doug was the one blowing up my phone. He was at NRC. The Atlantic Technology AT-3 was in the anechoic chamber. He was wondering whether I reviewed the thing with the grille on or off. “A bit of both,” I said after a quick check of my listening notes to make sure I remembered right. Which position was the high-frequency energy control switch in? “Mostly in the middle, with some listening in the ‘-’ position.”
Why am I telling you all of this? Because just like any other publication that marries subjective reviews with objective measurements (not that there are a lot of us), we at SoundStage! constantly strive to do a better job of both. We’re always looking to improve our understanding of how measurements and listening preferences correlate and how they don’t.
You might think that the measurements are mostly of benefit to us and our readers, but the truth of the matter is that manufacturers often benefit from our objective findings. Behind the scenes, Doug has told me plenty of stories about speaker designers who’ve tweaked their products after seeing our measurements from the NRC.
And it’s not just speakers, either. When bench testing the NAD C 3050 LE integrated amplifier I recently reviewed, Diego Estan, our electronics measurements specialist, discovered that the preamp section wasn’t delivering the intended 12dB maximum gain, which may explain why I had to crank the volume so much more than expected. (Looking back over my listening notes, I think these observations were made when I was using a line-level input and my iFi Audio DAC instead of the NAD’s own integrated streaming capabilities.) The beautiful thing is, the manufacturer was able to issue a firmware update to fix the problem on literally the same day.
And yes, as you may have ascertained, we do share our measurements with the manufacturer when we see things that we have questions about or where we think we might have made a mistake. Why wouldn’t we? The goal here isn’t some gotcha game; it’s to get a more complete picture of a product’s performance. And if there’s something we might have done wrong in the measurements or something we may have missed in our testing, we want to know that.
There’s one person who absolutely doesn’t get to see the measurements until just before they go live, though, and that’s the reviewer. When Diego told me about the gain setting with the C 3050 LE, as well as the firmware fix, my immediate response was to ask Doug if I could add an addendum to my review, alerting readers who might not click on the measurements page that this issue had been fixed. That request was denied. Once I’d crossed the Rubicon of seeing those objective measurements, my subjective impressions had to be forever locked in place to avoid the perception of cross-contamination.
But why? Why does SoundStage! have a policy of not involving the reviewer in the measurements process until the very end, past the point of no return?
That’s Doug’s policy, so for a complete explanation, you’ll have to ask him. But here’s why I think it’s exactly the right policy. As I’ve mentioned before, months after seeing the measurements of the Monitor Audio Silver 300 7G I reviewed, I was struck by the appearance of a little 1.5dB peak at 3.5kHz that I didn’t hear during my subjective evaluation. Had I seen those measurements first and then written my review, I can almost guarantee you I would have heard that tiny little bump—or at least I would have convinced myself that I did. But I didn’t.
I’ll give you another perfect example. In the conclusion for my review of the Musical Fidelity M6si integrated amp, I called it a “super-low noise” amp with “more power than you could ever reasonably need in most rooms.” That prompted a response from a reader who read Diego’s measurements and thought I should have heard the harmonic distortion and noise reflected in the measurements. In his conclusion, he said, “Only 89.9 dBA [signal-to-noise] at 186 watts? Either I am having a silly day, or there is a contradiction.”
Here, in part, was my reply to that reader:
You’ve hit upon one of the main reasons why measurements are so essential to the reviewing process—but also why they sometimes reveal issues that won’t affect normal listening scenarios.
Let’s look at this graph.
See that knee between 150 and 200 watts, where the THD+N shoots up to the levels you referenced? There’s no way in hell I would be able to push the volume knob far enough to hit that knee in my two-channel listening room with my speakers. But look at how well the amp behaves below that knee.
Notice, too, the asterisks. Diego says that with the volume at the 1 o’clock position, SNR was 103dBA (that’s about 80dBA at 1W, 90dBA at 10W). Is that the single quietest amp in the history of ever? No. But would you be able to hear that little noise in the presence of any meaningful audible signal? Absolutely not.
What’s more, I could not stay in the room with the volume knob set at anything above 10 o’clock. And 1 o’clock would have blown my eardrums for good.
Let’s calculate what the SPLs would be if I did sit in my room with the volume knob turned to the level at which this thing starts to get significantly noisy. Let’s call it 150 watts for the sake of simplicity, since that’s right around where the knee is for an 8-ohm load.
With my speakers (Paradigm Studio 100 v5) at my seating distance in my two-channel room (about 1.67 meters), 150 watts of amplification would result in 109.5dB of sound, plus or minus—before adding the second speaker and before accounting for room gain. So let’s say between 113dB and 115dB just as a super-conservative back-of-napkin calculation.
You can tolerate sounds that loud for somewhere between 30 seconds and a minute before you run the risk of damaging your hearing.
Now let’s estimate how much amplification I’m actually using in normal listening.
I will, from time to time, push the volume knob such that, with some songs, I’m hitting infrequent peaks of around 99dB for brief bursts during the loudest portions of the song. With my speakers in my room, for those brief moments I’m drawing ~14 watts of power from the amp. How much THD+N are we talking about at those levels? I did some quick Photoshop work on Diego’s graphs to get a guesstimate.
Somewhere around 0.008%? I think that’s a safe estimate, just eyeballing it. That’s an average, mind you, and since my speakers dip down below 4 ohms briefly in the lower frequencies, you could theoretically double the wattage there and say, what, 30 watts? However briefly? Eyeball that on the graph and look at the 4-ohm trace, and the results for THD+N aren’t that much different.
And keep in mind, that’s at a listening level that I can tolerate for only a song or two at most.
I’m not saying to disregard any of this, and this doesn’t cast any aspersions or doubts on Diego’s excellent measurements. I’m merely saying that there’s no contradiction that I can see. I turned in my listening impressions months before this amp was measured. I didn’t see the measurements until literally the day the review was published. I, too, was surprised to see how the amp behaved at the extremes.
But there’s absolutely no way, with any speaker I own, in any room in my house, I could have pushed this amp hard enough to audibly detect the noise behavior we see in the measurements from the analog inputs. I can only do my subjective listening impressions in a way that reflects the average use case. If you have very power-hungry speakers and a cavernous room, the only way we at SoundStage! could detect the sorts of flaws like these that would affect your listening experience is on the test bench. And we did that.
Put my words and Diego’s graphs together, and you have a complete evaluation of this amp. I don’t see contradictions. I see a comparison between typical use and extreme use cases. But I do thank you for pointing out the seeming contradictions, because it lets me know I should get out ahead of situations like this and emphasize the limits of what one can hear with good three-way tower speakers in an average room at reference listening levels.
Something I wrote near the end of that email response really says it all: put the subjective listening experience together with the objective measurements, and you have a complete review. Or at least a more complete review. Because, as I was reminded with these Atlantic Technology speakers, neither the measurements nor the subjective listening impressions tell the complete story.
OK, so what can we learn from the AT-3 measurements?
Of all the unique characteristics of the Atlantic Technology AT-3 speaker that prompted this whole discussion, the one I was most interested in seeing reflected in the measurements was the excessive brightness I was hearing when my ear was higher than the speaker’s tweeter. I described it in the review as “stabbing me in the brain.”
Looking over the frequency response of the speaker taken on-axis, 15 degrees above, and 15 degrees below, though, I’m just not seeing what I would describe as “brightness” in the red trace (15 degrees up). To put a fine point on it, when I say “brightness,” I typically mean frequencies between, say, 2kHz and 7kHz or thereabouts. And if anything, these frequencies are actually recessed compared to the on-axis and 15-degree-down measurements. I still cannot square that circle, although if you can, please do write in and give me an ELI5.
On the other hand, when you look at the measurements taken at 0, 15, and 30 degrees off-axis left/right, you can really see what I meant when I talked about the “problem frequencies . . . from 500 to 1000Hz, with some frequencies getting swallowed and some tearing through the mix like chainsaws in a bad 3D horror movie.”
Here’s what’s perhaps most fascinating and informative for me, though. If you look at the left/right off-axis plots, those frequency-response anomalies between 500 and 1000Hz are quite consistent, which I would attribute to resonances. Compare that to the on-axis/15-up/15-down plots, though, which I’ve zoomed in on here.
Could that explain what I heard that prompted me to write, “As I was standing up to make more adjustments to speaker positioning, the sound went . . . lenticular”? Perhaps. I’d love to sit down with someone far more knowledgeable about speaker design and measurements to learn more, and I will after I’m done writing this piece. But the whole point of this exercise, for me at least, was to document my reaction to these measurements in real time, as I experienced them, not to provide all the answers.
There’s one more little anomaly worth discussing, just briefly. When Doug sent me the measurements for the AT-3 (after my copy was laid out, loaded, and locked in place for perpetuity), he couldn’t resist making one observation about the bass, especially the weirdness centered on 200Hz or thereabouts. “That should be audible,” he told me. “It should make male vocals sound chesty.”
Looking through my review again, I don’t see anything about chesty male vocals. But digging through my listening notes, I found some observations about Anaïs Mitchell’s “Why We Build the Wall” that didn’t make it into my final review but which seem relevant here.
Keep in mind that although I only reference two or three songs in any given review, I generally listen to dozens of tracks, if not more. Often the same tracks with every review. I’ll take notes on probably ten or 15. Then I’ll pick the three or four that best represent what I want to convey about the product, and those are the ones that end up in the review.
“Why We Build the Wall,” to put things in perspective, is a track from the Hadestown concept album (long before it became a hit on the stage) with vocals by Greg Brown, whose voice is gravelly and bone-shaking, with fundamentals that must be below 100Hz, easily. If there was chestiness to be found in male vocals, certainly it would be found here. Looking at my notes, though, I was positively blown away by the AT-3’s delivery of Brown’s portrayal of Hades. I wrote in my notes that these were the best vocals I heard with this speaker. If my memory serves me well, the speaker also delivered one of the best renditions of these vocals I’ve heard in quite some time.
All of which is to say that here the measurements seem to contradict my subjective impressions yet again. If I had the chance to go back and listen to that song on those speakers once more, would I hear the chestiness that the measurements indicate I ought to be hearing? Maybe, maybe not. It could be that there’s some masking effect going on here, or that my attention was simply drawn to something else. Who knows? Oftentimes, what we think we’re hearing is not what we’re actually hearing. A good (although perhaps not entirely relevant) example is that some people hear “brainstorm” and some hear “green needle” when listening to this audio clip. I actually alternate between the two.
I sort of feel the same way about these AT-3 measurements. In some ways, the measurements vindicate me. In some ways they certainly seem not to. I honestly think a few extra measurements at, say, 30 and 45 degrees above and below the tweeter might have given a clearer picture of the lobing issues I heard in my listening. But in the absence of those measurements, I have to concede that they might not, and that what I think I heard wasn’t necessarily what I actually heard.
Either way, I think my listening notes combined with the measurements give a more complete evaluation of the AT-3 than either would in isolation. As Doug said to me in an email, the Atlantic Technology speakers at $3629/pair measure better in some respects than the $18,500/pair Vimberg Amea that Diego loved the sound of, so there’s obviously a lot to like about them. And even though they don’t necessarily conform to the sort of sonic signature I prefer, especially in the midrange, there’s nothing to say that you wouldn’t love them. Again, they do a lot of things very well.
Looking at my words alone, you might form some impressions about the speakers that don’t stand up to scrutiny, because I have my biases, as does any audio writer. And if you merely looked at the measurements alone without reading my words, you might form some impressions of the bass that I feel are misleading. Reducing any product to one technical measurement is a mistake, to say the least.
Put them together, though, and I think you get a better sense of what this speaker does and how it does it. Consider the sum of objective and subjective evaluations, and you might think, “Yep, that’s exactly the speaker I’m looking for!” or, conversely, “Nope, not my jam!” And that’s all that really matters in the end. It shouldn’t make a hill-of-beans difference in the end whether I or Diego or Jeff or Doug or anyone else on the team likes the sound of a speaker. What’s important is that we give you enough information to decide whether or not you might like it enough to audition it. And proper measurements are an essential part of that, whether they confirm, deny, or seem wholly tangential to our listening impressions.
. . . Dennis Burger