Two curves is enough - Axiom Message Boards

Axiom Home Page

I thought I’d chime in on which measurements matter. And it might be possible to distill goodness down to two composite curves. This is an expanded discussion of the 150+ anechoic measurements discussed in Amie’s blog:

https://www.axiomaudio.com/blog/the-family-of-curves/

First of all, it’s been known for three decades how sound coloration affects frequency response. This unattributed image has been circulated around Audio Science Review, and I haven’t found anyone in the audio field who’s refuted it’s validity yet.

https://www.dropbox.com/s/0bcbw2c7v7spqzb/1614637989068.png?dl=0

Using the audiophile press to make recommendations, particularly those that don’t use measurements nor A/B direct comparison with rival products, is audio foolery.

https://seanolive.blogspot.com/2009/04/dishonesty-of-sighted-audio-product.html

Axiom only provides three types of curves. The 2 way speakers usually have a frequency response graph (leaving us to imply that it’s an on-axis frequency response curve), and the 3 way speakers have a listening window and sound power curve. I’ve wondered why they chose to post these particular curves out of the 150+ measurements they’ve made.

These are the Axiom measurements of various models:

M5HP: https://www.axiomaudio.com/pub/media/catalog/product/m/5/m5hp-listening-window-and-sound-power.png
M60: https://www.axiomaudio.com/pub/media/catalog/product/m/6/m60_freq.gif
M80: https://www.axiomaudio.com/pub/media/catalog/product/m/8/m80_freq.gif
M100: https://www.axiomaudio.com/pub/media/catalog/product/m/1/m100_freq_graph.gif

LFR660: https://www.axiomaudio.com/pub/media/catalog/product/l/f/lfr660.jpg
LFR880: https://www.axiomaudio.com/pub/media/catalog/product/l/f/lfr880.jpg
LFR1100: https://www.axiomaudio.com/pub/media/catalog/product/l/f/lfr1100-sp-lw.jpg
LFR1100 active: https://www.axiomaudio.com/pub/media/catalog/product/l/f/lfr1100_active_graph_1.jpg

These are some competitors that have been acclaimed for both their measurements and their subjective audio performance.

JBL M2: https://www.erinsaudiocorner.com/loudspeakers/jbl_m2/
KEF Reference 1 Meta: https://www.erinsaudiocorner.com/loudspeakers/kef_reference_1_meta/
Dutch & Dutch 8C: https://www.erinsaudiocorner.com/loudspeakers/dutch_dutch_8c/
Genelec 8331a: https://www.erinsaudiocorner.com/loudspeakers/genelec_8331a/
Revel Performa F226be : https://www.erinsaudiocorner.com/loudspeakers/revel_f226be/

What they all have in common is a great listening window and gently sloping sound power curve. These curves are straight because straight means minimal colorations and resonances, and we are highly sensitive to low-Q deviations (very broad hills and dips). High-Q (sharp spikes up and down) look terrible on a frequency response graph but are psychoacoustically inaudible to us. Here are some deeply flawed (and expensive!) loudspeakers to contrast with the above. Can you say “buyer’s remorse”?

Klipsch Heresy IV: https://www.erinsaudiocorner.com/loudspeakers/klipsch_heresy_iv/
Bose 901: https://www.erinsaudiocorner.com/loudspeakers/bose_901_series_v/
McIntosh XR50: https://www.erinsaudiocorner.com/loudspeakers/mcintosh_xr50/
Wilson Audio TuneTot: https://www.audiosciencereview.com/...review-high-end-bookshelf-speaker.29219/
Magnepan LRS: https://www.audiosciencereview.com/forum/index.php?threads/magnepan-lrs-speaker-review.16068/

An example of what high-Q looks like:

IKEA Symfonisk: https://www.erinsaudiocorner.com/loudspeakers/ikea_symfonisk_lamp/

There’s a composite curve that appears to combine the listening window and sound power curve into something called the in-room response. Ideally, it should be straight and sloping slightly downward. I’m still evaluating its merits because some reviews with good looking in-room response have noted subjectivity issues if there’s deviations in the midrange and lower treble range.

Is frequency response king? If you don’t get it right, it’s hard to compensate with other redeeming qualities. Sean Olive listed some other factors that matter to perceived performance: spatial width, spaciousness, dynamics, and maximum sound pressure level:

https://seanolive.blogspot.com/2010/05/evaluating-sound-quality-of-ipod-music.html
https://seanolive.blogspot.com/2009/01/what-loudspeaker-specifications-are.html

At least there’s some agreement (for some of us) that science is a great thing in audio:

Don’t fear modern tech: https://www.audioholics.com/editorials/future-audiophiles
Why measurements matter: https://youtu.be/zrpUDuUtxPM?t=3433

The image of coloration that you shared is fabulous. I have never seen it before. I can't say I totally agree with it due to my lack of knowledge but I certainly agree with the parts I've personally experienced. I've downloaded it and will study it very carefully.

Ian has told me Axiom has both curves for all v4. He was going to ask Andrew to publish them but likely other matters took priority. I know from listening experience the M2 and M3 image and stage like the M5 and the M2, within its specified band, is more linear than the M3 and M5.

With regard to the curves Axiom publishes, the sound power is a composite of all curves using weight coefficients that Axiom has selected. You could call the sound power curve the weighted average of all curves. How Axiom selects the weights is largely what makes the "Axiom sound". We don't know how Axiom selects those weights - and they won't divulge the methodology - but it stems from experience, science and listening. I believe they've nailed the formula. Once the formula is nailed, it makes speaker engineering easier because the formula constrains or maybe informs the design. For example, if the sound power curve comes out "too lumpy" at say 200 to 300Hz, the curve weighting factors for that band won't be questioned. Rather, it will point to a deficiency in the design. The designer can then examine the mess of curves in that band to determine how the lump should be mitigated. The challenge is that to mitigate the lump, one or more lumps or dips may be introduced elsewhere. Would that be better or worse? The sound power curve for a modified design helps to answer that. If the new lumps and dips are within a specified tolerance, then the modified design is good to go. This is why for a speaker like the LFR1100, there may be thousands of curves to inform the final design. Those thousands of curves also include the curves for each driver.

The listening window and sound power are the only curves lay people need to judge how the speaker will sound in a room. The science says when these curves ride on top of each other, audio Nirvana is reached. Why? Because the direct and reflected sound fields become fused and the speakers acoustically disappear.

With regard to the slope of the curves, the science says a slope of 3dB/decade out to 15Khz or beyond will make the speakers sound natural in a room.

These two curves inform how wide and spacious the speaker will sound in room. In fact, what you listen to provided you're not sitting in the near field of a speaker is sound power.

The last factor is SPL. Ideally, we should have a family of listening window and sound power curves for various SPLs. If these two curves "deform" at higher SPLs, that spells trouble. I personally have not detected this trouble in v4. I can turn them up ridiculously loud, and they will distort, but the stage and imaging remain intact.

It's also interesting to note that some (all?) Bryston speakers have flat listening window curves. Was this done for marketing reasons or do Bryston owners have more dampened rooms that demand a flatter curve rather than the 3dB/decade attenuation I described in the last post?

BTW, the Family of Curves is not unique to audio. There are for example families of curves for flow rate vs. pump pressure for various RPMs. There are families of curves for chemical sensors. Even transistors, other semiconductors and passive components. What's unique about audio though is that the curves need to be interpreted through the lens of subjective descriptions provided by trained listeners.

In principal, yes. But what is the definition of "straight" and "look good"? There may be trade-offs between those two. Brystons may be "straight" but the wife doesn't like the aesthetics. So now you have to look elsewhere.

There is also the issue of how much power is required to achieve the necessary SPL. And how low you need the speaker to go.

What makes the Middle T better than an M60? A good exercise may be to look at the specs, including curves between the two, and try to identify the answer.

Originally Posted by Mojo

In principal, yes. But what is the definition of "straight" and "look good"?
What makes the Middle T better than an M60? A good exercise may be to look at the specs, including curves between the two, and try to identify the answer.

Those are good questions, and I don't know what the audible threshold is between one companies squiggly line and another companies less-squiggly line. It might be that spending a little more on better cabinets provides better measurements, and I certain think Bryston has the appropriate high-end look. But I don't think it's unreasonable to be satisfied with any of the well measuring speakers.

Heck, it would beat having this $11,000 CDN stinker:

https://soundstagenetwork.com/index.php?option=com_content&view=article&id=2629:nrc-measurements-bowers-wilkins-805-d4-loudspeakers&catid=77:loudspeaker-measurements&Itemid=153

Let's analyze, using published specs, if the Middle T is better than the M60v4.

1. The T goes 3Hz lower (33 vs. 36). This is assuming that the Middle T is an anechoic spec like the M60 and not in room.

2. The impedance of the T is 4 Ohms vs. 8 Ohms for the M60. Better have an amp that supports a 4 Ohm load.

3. The SPL is identical assuming the T spec is for anechoic and RMS power. Note there is no dynamic SPL given for the T. The M60 dynamic SPL can be calculated from the dynamic power and sensitivity to be 117 dB. This just means it won't melt or rip apart at that SPL.

4. The sensitivity is identical.

5. The cross-over frequencies are different. The T is 160 and 2.3KHz vs. 500 and 2 KHz for the M60. I find it really hard to believe the M60 woofer is crossed to the mid at 500Hz. I think that's a mistake. Regardless, I really have no way of knowing if crossing so high in the mid-range on the M60 detracts from the M60 compared to the T. Ditto for the tweeter.

6. The T woofer is 8" vs. the M60 6.5". This tells me nothing about performance.

7. The T cabinet encloses more volume. Likely this and the 8" woofer is why the T goes 3Hz lower.

8. The T is almost double the mass. Hmmm...might this imply less audible resonances and hence improved acoustic disappearance? Maybe.

9. Now we get to the amplitude response:
a. The T does go lower.
b. The T has a noticeable depression between 70 and 200Hz of around 4dB. I have no idea how that might affect what I hear.
c. The T listening window is flatter. Maybe that will make it brighter. I don't know.
d. Both sets of curves are well-behaved and more similar than different.

So what can I deduce from all this? Not much other than the 3Hz difference down below and beefier amp needed for the T. As for the difference in mass, there's nothing in the curves to suggest a more massive T is better.

Based on all of this, I can only conclude from the published specs, the additional money for the Middle T will not result in improved performance over the M60v4 other than the 3Hz in bass performance.

Listening may reveal more differences. But if so, where in the curves and specs are those differences reflected?

Anyone who wants to geek out on these two measurements should go to Pierre Aubert's collection of Spinorama measurements. It's an incomplete list, but you see a pattern on whom takes their R&D seriously, and the fact that good measurements are unicorns in a sea of mediocrity.

https://pierreaubert.github.io/spinorama/index.html

I would be curious to know the source of that ASR graph on sound qualities vs. frequencies. Certainly there are some truths to it, common knowledge (like the general range of human hearing), but some subjectivity on how something like upper/lower midrange is defined. Still, an interesting graph and from my experience, using the frequency response graphs and relating it to my sound preference, i know i am not a fan of big bass humps, have heard what a small bass suckout sounds like, and know that i like neutral speakers but with a small hump (or minimal rise defined as no more than around +2-3dB relative to lower frequencies) in that upper midrange-ish area which sounds more clear, detailed to me.

There is so much data out there on the topic, i'm sure there's more in publications along these lines. I try to defer away from people's blog sites unless they are known experts in the field (such as Sean Olive, Floyd Toole who have more publications than many over many years).

An interesting topic though Hamb.
Lots of searching went into this one!

Originally Posted by chesseroo

An interesting topic though Hamb.
Lots of searching went into this one!

Actually, it took less than an hour!

That graph shows there's only one way to make a good all-around loudspeaker, but countless ways to make a bad one. I was deeply dissatisfied with Audio Science Review and Erin's Audio Corner. Most of their data is noise, and I wanted to know what the best sounding and ranking speakers had in common that the poor sounding ones didn't.

After rewatching some of the Axiom videos, I noticed Ian and Andrew kept harping on the listening window and sound power. So I gathered up all the acclaimed reviews on those two sites to look at those two measurements. Lo and behold, every single one of them have a straight listening window and straight sloping sound power. It tells you EVERYTHING you need to know about who wins a double blind listening test. All the other stuff (on axis, step response, phase response, dispersion map, etc) is just noise.

Even going back to the early 2000's with the loudspeakers Harman used in their research papers about the state of the art (which are never mentioned, but are the Infinity Prelude MTS and IL60), and you see the same pattern.

I can still be proven wrong, but it looks like this has legs.

Two curves? No. Great thread idea!

How dynamic does the Middle T sound vs an M60? How exciting? Can an FR curve tell you that?

I agree that the new adoption of klippel by more “open source” enthusiasts is awesome. But I think the more data the better.

Comparing the polar response plots of speakers allows some prediction of their sound intensity. Probably the most ignored and important metric when it comes to the subjective impression of bass in a room.

The JBL 4367 breaks the rules of the other speakers you mention, yet it is the speaker Erin would own if he could. He is campaigning pretty hard for one. I hope Harman gives him a pair!

The polar plots are incredibly important. Those and directivity index.

Originally Posted by TrevorM

Two curves? How dynamic does the Middle T sound vs an M60? How exciting? Can an FR curve tell you that?

A family of FR curves can tell you that. The family would consist of the listening window and sound power at various SPLs.

The plots below are for the Model T Signature. The top plot contains the LW and SP at 1W input. The bottom plot is the LW at two different power levels. The bottom pen is at 1W input and the top pen is at the power threshold where compression (thermal and/or dynamic) begins. It would be reasonable to expect the LW and SP curves are all similar between the two pens shown.

As for the bottom plot, it only exists for the Model T. No other Axiom or Bryston. So it's impossible to compare the dynamics across Axioms or Brystons.

Originally Posted by TrevorM

Two curves? No. Great thread idea!

Researching this topic is fun. Buried in the data (all of which are available in the public domain, kudos to Harman, Axiom, and Sausalito), it's clear from the NRC research that everything points to the amplitude response as being the most important factor in selecting a loudspeaker (which includes deep bass). Dynamics and SPL matter, so that's why every manufacturer has multiple sized offerings. The other stuff like spaciousness and spatial clues are room and placement dependent.

What's problematic is that it took me years to distill this wisdom. Even up to last month, I thought Axiom's talk about neutral loudspeakers was marketing fluff.

https://youtu.be/igzzXKAqmi8

There is so much contradictory voices out there that say you can't have audio goodness unless you have this next thing (active crossovers, active EQ, AMT, etc). I partly blame EAC, Audioholics, and ASR for the confusion: they throw a lot of data at you while ignoring the two measurements that actually matter; and they spend most of their energies evaluating products that don't solve a real problem (DAC's, interconnects, amplifiers).

That JBL has directivity issues in the midrange between 500-1000 Hz. I hope this thread taught everyone reading it the skill set in identifying it immediately just from the listening window and sound power.

I should probably quit now, a minute discussing audio is a minute not enjoying music.

Oh, and the user interface issues of those review sites is easy to fix. Assuming that our assumptions are correct, just place a graph of the listening and sound power against a band indicating audio perfection (and it's likely a band +/- 1.0 to 1.5 dB wide). Then talk about company, construction, price and market position against rival products. Leave out the other graphs and subjective impressions or put them on the last half of the review.

After all, you wouldn't clutter up your dashboard with oil pressure, clock, fuel gauge, and compass meters that are the same size as your speedometer and tachometer, would you?

Yeah. Its trash. Bummer.

I checked my hypothesis with ASR's Olive scores. The straight lines of all the Axiom and Bryston models (except the Mini A) imply an Olive score of 6.5+. Every score between 5.0 and 6.5 have small but noticeable deviations from straightness. Anything below 5.0 and you can see major flaws.

And we seem to have broken Bryston's site.

Sean Olive on Predicting Loudspeaker Sound Quality and Listener Preference
https://www.audiosciencereview.com/...d-quality-and-listener-preference.26815/

Within the ranks of Harman there was dissention between Toole/Olive and other designers.

A recent interview with Greg Timbers touched further on the idea of sound intensity as a factor of reproduction not tested by the Harman scientists, nor weighted in the preference score. Its a pretty enlightening vid if you have the time.

Ultimately Great measuring speakers can be entirely boring and disengaging. I altered my curve in my setup from “perfect” for this reason. It sounded flat and non dynamic. Honestly, I measured the in room response of my M5HPs at the mlp in my 2ch room and translated that room curve to my theater mains with Dirac. Bingo!

There is a happy middleground I’m sure, but as Greg mentions in the interview reducing a speakers sound to a graph or number is not possible. Troels Gravesen agrees, and he is lauded as the “best” open source designer on the planet.

Dynamics cannot be measured with distortion or compression plots. Those only reflect max spl before bad things happen.

In a post earlier this year i did provide a paper written and published by Olive that demonstrated the concept that "all" listeners fall into a category of preference for a flat response curve, but also that there are small variations within the range (as to be expected for confidence intervals in any dataset) where preferences included small bumps in treble and bass off the flat response. When you look at the frequency response of really good speakers, those subtle differences in bumps or more humps/valleys in certain frequency ranges and to a certain extent (not too much or too little) can set one's personal preferences specifically to a speaker brand/model OR through the use of sound engineering like Dirac or Audyssey.

In short, i have no doubt that a best sounding speaker is one that is at it's basis 'neutral' and generally flat, but that with small tweaks, people find what they feel sounds best to them (at least to those who aren't trying solely to shop with eyes and wallets as the metric for quality).

Originally Posted by TrevorM

I had a listen to that interview by EAC. I can see how heat soak affects dynamics, and why large drivers, enclosures and voice coils are the way to go if you wanted high SPL.

It's bizarre that of the measurements that we have, only about a third to half of the Revel and JBL models measure well by NRC standards. I take it that good industrial design and having a lot of models is the priority, rather than have a handful of models in a variety of form factors.

I also get the impression from multiple interviewers of alumni and current researchers that Harman is a Darwinian organization ruled by a culture of fear. haha