Is it possible to cover the whole spectrum, high SPL, low distortion with a 2-way?

There's a fun paper from my university on the topic of off-axis response and listener preference. It's done using simulated impulse responses of a pro audio array system, in an anechoic environment, but the results are still very interesting in light of the objective 'badness' of the off-axis responses versus the on-axis ones:
(PDF) Temporal Distortion Audibility of Directional Loudspeaker Arrays

The full paper is free access.

Thx for that :)

Must admit though, it seems like one giant leap of faith to give much credence to comparative line array sims played through headphones.
 
Thanks for the paper.

I'll read / chew on it later.

Not that we are worried about it here, but people hated my full range driver focused array (when i was obsessing over time/phase/full range drivers).

Below is one someone else built.
 

Attachments

  • p1000104m0rbesuskm.jpg
    p1000104m0rbesuskm.jpg
    135.5 KB · Views: 316
Thx for that :)

Must admit though, it seems like one giant leap of faith to give much credence to comparative line array sims played through headphones.

People often think that, but listening tests need to be done in a controlled condition for the proof of the null hypothesis. Using (ideally) the same set of headphones for everyone, in a controlled room, is how the majority of these tests are done.

While it would be nice to repeat the test with the same hangs as simulated, in a big anechoic room or outdoors, you then introduce lots of room for error. Does everyone stand in exactly the same spot? Has the environment changed in temperature or wind speed? Have the drivers warmed up, or the mains voltage dropped? What about the impact of 'sighted listening'? Etc etc.

There are of course limits in the model used. It doesn't include any boundaries for the cabinets themselves, and the author doesn't state what model parameters were used for the calculations.

Ray traced geometric acoustics are also only good from 100 Hz to 10 kHz. Given how much of our subjective preference is dominated by low frequencies, and how much of the 'error' of line array waveguides is present in the upper HF region, there is obviously room for further study.

Even then, I'd argue that capturing 3D balloon impulse response data in an anechoic room from say a studio monitor or hi-fi speaker, and doing good quality convolution should be sufficient to repeat the test for home-use conditions.
 
There's a fun paper from my university on the topic of off-axis response and listener preference. It's done using simulated impulse responses of a pro audio array system, in an anechoic environment, but the results are still very interesting in light of the objective 'badness' of the off-axis responses versus the on-axis ones:
(PDF) Temporal Distortion Audibility of Directional Loudspeaker Arrays

The full paper is free access.

Thanks! The off-axis was mentioned being 10m away and on-axis varies to where response drops -6db. I'm not sure if the level difference was compensated? probably. There is only few magnitude and group delay responses shown and they are pretty wild.. as you said, the off-axis in this case was perceived less bad :D I'm not sure the outcome is applicable outside of large spaces with huge line arrays. There were interesting bits on the beginning though, the phase distortion stuff.
 
You should measure >1meter or few millimeters of from the dustcap/throat. If you measure at grill it will pick all kinds of poop, resonances give destructive interference etc. Which is why for example el. guitar sounds similar to what it actually sounds miced a bit further away where the stuff averages out. The standard sm57 at grill doesn't sound anything like what the sound actually is to the guitar player.
 
Thanks! The off-axis was mentioned being 10m away and on-axis varies to where response drops -6db. I'm not sure if the level difference was compensated? probably. There is only few magnitude and group delay responses shown and they are pretty wild.. as you said, the off-axis in this case was perceived less bad :D I'm not sure the outcome is applicable outside of large spaces with huge line arrays. There were interesting bits on the beginning though, the phase distortion stuff.

I'd presume that the files were normalised for the listening test, as that's a common thing to reduce the number of variables. There's a bunch more plots in the appendices, but they're kind of hard to discern anything from.

The sims were done with all the boundaries as fully absorptive, so equivalent to anechoic. While the arrays were quite large physically, the principle should hold true for line arrays used in the home.

It would definitely be interesting to repeat the test method using hi-fi speakers and/or studio monitors. Unfortunately I don't know of any that supply balloon data in GLL format, and I'd have to pay to take something into an anechoic chamber where I can use a FourAudio ELF arm or pair of motorised turntables to make my own.

Any volunteers?

Vx15hp, measurement taken 1" away.

The density of the ripples in the upper region suggests quite a lot of reflection in that data. Particularly noticeable in the phase trace. What windowing is applied?

For some reference, the latest B&C 5" coaxial is reviewed here with measurements:
Test Bench: B&C Speakers' 5CXN44 Pro Sound Coaxial Driver | audioXpress
 
I just took a look at my axi+350 horn measurements at 1m vs 1".... They are different...The difference between crest and troughs ain't nothing to write home about, actually gets choppier with the 1m measurement in some areas....I"ll take a 1 meter measurement outside later. I don't expect to reach +/-3 db but maybe I'll get lucky.
 
Last edited:
...
It would definitely be interesting to repeat the test method using hi-fi speakers and/or studio monitors. Unfortunately I don't know of any that supply balloon data in GLL format, and I'd have to pay to take something into an anechoic chamber where I can use a FourAudio ELF arm or pair of motorised turntables to make my own.

Any volunteers?

... [/url]

For sure, what we have to do? Is it possible to construct GLL data with the home semi anechoic measurements?:D Wouldn't be very accurate but very cool experiment!

Google seems to find EASE data, is there anything for example here to construct a headphone test with? JBL Professional Loudspeakers

I don't know what to look for, first time heard / read about it from your post.

edit.
RCF seems to post GLL files Product Detail - RCF
QSC Loudspeakers - Software and Firmware - Resources - QSC

I suspect many pro audio / installation manufacturers have made GLL data available.
 
Last edited:
I just took a look at my axi+350 horn measurements at 1m vs 1".... They are different...The difference between crest and troughs ain't nothing to write home about, actually gets choppier with the 1m measurement in some areas....I"ll take a 1 meter measurement outside later. I don't expect to reach +/-3 db but maybe I'll get lucky.

The 1 meter is simple rule for small speakers, but must be bigger for larger speakers to get out from the nearfield. I don't remember what the rule of thumb was, maybe 3x the largest dimension of the DUT or something like that. This is not possible at home with big speakers since the refletions will limit very short window.

You can use VituixCAD or any other calculator to get time for first reflections or just look for the reflections in the measurements. Inside home with ~2.4m typical (european) room height this can be about 3-4ms or less if the walls are even closer :)
 

Attachments

  • window.jpg
    window.jpg
    52 KB · Views: 272
Last edited:
People often think that, but listening tests need to be done in a controlled condition for the proof of the null hypothesis. Using (ideally) the same set of headphones for everyone, in a controlled room, is how the majority of these tests are done.

While it would be nice to repeat the test with the same hangs as simulated, in a big anechoic room or outdoors, you then introduce lots of room for error. Does everyone stand in exactly the same spot? Has the environment changed in temperature or wind speed? Have the drivers warmed up, or the mains voltage dropped? What about the impact of 'sighted listening'? Etc etc.

There are of course limits in the model used. It doesn't include any boundaries for the cabinets themselves, and the author doesn't state what model parameters were used for the calculations.

Ray traced geometric acoustics are also only good from 100 Hz to 10 kHz. Given how much of our subjective preference is dominated by low frequencies, and how much of the 'error' of line array waveguides is present in the upper HF region, there is obviously room for further study.

Even then, I'd argue that capturing 3D balloon impulse response data in an anechoic room from say a studio monitor or hi-fi speaker, and doing good quality convolution should be sufficient to repeat the test for home-use conditions.

It's good to see innovative ways of researching both preference and accuracy. (i think the two are getting very confused with CEA2034, MUSHRA, etc, but that's a complete digression off topic)

I understand the desire to offer repeatable science that gives statistically valid results. All too often, such experiments seem to get distilled down to such a narrow set of conditions, that i think it is very dangerous to extend their findings beyond the actual set of conditions. (which folks do rampantly..again i digress :eek:)

For instance, I'd say the paper's results are valid for just what the test was....impulse sims through headphones hopefully replicating the arrays, using a couple of different test signals (pink and snare hits.
For me, I can't logically extend the results much further.

One major shortcoming i believe has existed in virtually all phase/group delay audibility experiments is the use of headphones.
Imho, headphones do not reproduce the bass dynamics and vibrational impact of loudspeakers. Headphones do well at steady state low freq, but simply cannot produce any sense of bass feel...which for me is a strong part of the auditory experience.

When a point source is taken to flat mag and phase outdoors, and bass is appropriately strong and low,....... I believe that would make a much better platform for assessing either speaker accuracy or preferences.


Anyway, my 2c...
 
I agree with the comments about headphones vs bass....I think it is neglectful to omit haptic feedback. Bass obviously transitions from auditory to a more touch based experience as it gets deeper. Where the point of touch is no longer relevant, Im not sure, but I've felt sensation from my horn at 350hz when played loud enough. In a room you might find a null/peak thats been compensated for using eq, and even though, you may reach your target FR, the vibrational signal may no longer line up with what you hear and it can play tricks on perception.

Then again, in this instance, looking at pro line arrays, at a distance, the haptic part of the experience, may be trivial
 
Last edited:
One major shortcoming i believe has existed in virtually all phase/group delay audibility experiments is the use of headphones.
Imho, headphones do not reproduce the bass dynamics and vibrational impact of loudspeakers. Headphones do well at steady state low freq, but simply cannot produce any sense of bass feel...which for me is a strong part of the auditory experience.

This right there is worth more than 2c to me :D.
 
I think this guy knows what he is doing. Let me quote
"Many shy away from headphones because they miss the visceral impact that big loudspeakers deliver, the shaking of the floor and body, which no headphone can reproduce. No debate here. In fact, that is why I have been using subwoofers with my headphone listening for almost thirty years. Just limit the subwoofer’s high-frequency bandwidth to 70Hz to 100Hz and adjust its level to the point where it is only felt, but not heard with the headphones playing. "

Aikido Headphone Amplifier
 
Auxiliary Bass Radiators (ABRs) and transmission lines only work on steady tones and are unable to reproduce transient waveforms. The only low frequency technology which can reproduce the input waveform accurately is the sealed enclosure.
https://mangeraudio.com/en/discover/about/acoustical-reality
The 4th order bandpass is a combination of vent and sealed enclosure...Results must fall somewhere in between, as well, for them.