Speaker Review Anechoic Measurements- Grossly Inadequate and Why

[avanti1960]

On the way to try and understand why speakers with relatively inert non-resonating enclosures have such a good reputation, it was discovered that anechoic frequency response measurements are grossly inadequate for predicting how a speaker will sound in one’s room.
We knew that the room always dictated the nature of the sound regardless of anechoic open air frequency response measurements, but not to such an extent that missing the contribution of the cabinet resonance has on the outcome.
Based on two comparisons between speakers with inert cabinets vs. lively cabinets for anechoic, cabinet resonance and in-room response measurements the following conclusions are drawn:
1) Speakers with inert cabinets and anechoic measurements showing boosted bass have relatively flat in room response measurements.
2) Speakers with lively cabinets and relatively flat anechoic responses have very poor, uneven in-room frequency responses.

In the cases above “bad” anechoic response measurements that people would dismiss out of hand prove to be excellent speakers in-room because the manufacturer compensated for the lack of the enclosure resonance contribution by boosting the bass output.
The speakers with “good” anechoic responses that people seem to seek had extremely poor, uneven frequency response in-room because of the sound contribution of an energized vibrating enclosure. The manufacturer did not compensate for the enclosure resonance in the anechoic response.
The reason that anechoic measurements do not capture the contribution of enclosure resonance is because the cabinet response radiates in all directions and the sound is not reinforced in free air as picked up by the microphone, which hears the direct response of the drivers. Of course the cabinet response is easily reflected and reinforced in-room by room boundaries. This is why the in-room responses are so different between speakers with inert vs. lively enclosures.
Unfortunately relying on anechoic response measurements as a predictor for how they will sound in your room is much less valid than previously believed because of that fact.
In each case below the speakers with the more inert cabinets measure much better in-room than their anechoic response measurements would indicate.
The speakers with lively, resonating enclosures measure much worse in-room than their anechoic responses would have us believe.

EXAMPLE 1

EXAMPLE 2

 

[Ktacos]

It's all so wrong I don't even know where to begin.

 

[avanti1960]

It's all so wrong I don't even know where to begin.

Not a scientific study for sure but the idea that anechoic response measurements do not account for cabinet resonances (and that they have a significant impact on in-room response) is valid.

 

[olieb]

Not a scientific study for sure but the idea that anechoic response measurements do not account for cabinet resonances (and that they have a significant impact on in-room response) is valid.

Well, just measure other directions than on axis and you will see all the resonances that are there. Still an anechoic measurement.
EDIT: it is called "spinorama" and there even is a standard (ANSI/CTA 2034)

 

[BDWoody]

anechoic response measurements do not account for cabinet resonances

Ignoring the rest, why do you believe the mic only picks up sounds from the drivers, and not whatever sound is being emanated?

 

[Blumlein 88]

You might almost have a point, if you measure directly straight ahead in an anechoic chamber. However the whole standard which was developed by Toole et al took that into account. They measure anechoicly over 70 points. 360 degrees vertically and horizontally at 10 degree intervals. The directionality curve at the bottom of graphs shown by Amir here also will show resonances as wrinkles in that graph. One of the findings is resonances are fairly audible to people. Best practice is ditch the resonance and get the response right for best effect in room and anechoicly. You don't need to trade off one for the other.

 

[Scgorg]

The issue here is more with Atkinson's measurement regimen (which is not anechoic - it's gated measurements spliced with near-field measurements at LF).
If the loudspeaker cabinet's resonances are generating large dips in frequency response this would be evident in an actual anechoically measured frequency response (or equivalently one measured on a Klippel NFS). Even then, for a loudspeaker to have a cabinet resonance as bad as the one shown in the devore here is very uncommon, and itself a sign that the loudspeaker should probably be avoided.

Quick edit: the bass boost in Atkinson's measurements are not real either. He does uncompensated near-field measurements to splice together with the gated data, which means you'll be seeing as much as 6dB boost relative to what the actual far-field bass response is. Once again this is not a problem with a proper anechoic measurement or NFS measurement.

 

[avanti1960]

Ignoring the rest, why do you believe the mic only picks up sounds from the drivers, and not whatever sound is being emanated?

the sound from the enclosure resonance dissipates in an open air anechoic measurement.

 

[MaxwellsEq]

the sound from the enclosure resonance dissipates in an open air anechoic measurement.

It does not.

 

[avanti1960]

Well, just measure other directions than on axis and you will see all the resonances that are there. Still an anechoic measurement.
EDIT: it is called "spinorama" and there even is a standard (ANSI/CTA 2034)

agree about the spinorama measurement but it seems that the derived estimated in room response is not close to being accurate.

 

[avanti1960]

It does not.

without bouundary reinforcement, more so than the sound radiated directly from the drivers.

 

[a4eaudio]

Please explain the physics of why sound pressure waves dissipate quicker in the air when derived from different sources.

 

[Blumlein 88]

Please explain the physics of why sound pressure waves dissipate quicker in the air when derived from different sources.

I think what he means is off axis sound that in a real room reflects to the listener with delay will be absorbed in an anechoic chamber.

Which still does not validate his premise here.

 

[olieb]

agree about the spinorama measurement but it seems that the derived estimated in room response is not close to being accurate.

Sure, it cannot be accurate for a particular room, as each is different. But measurements in your room will probably be much less accurate for my room, MUCH less.
EDIT: And above ≈500 Hz it is actually quite accurate for most (usable) listening rooms.

 

[dfuller]

Not a scientific study for sure but the idea that anechoic response measurements do not account for cabinet resonances (and that they have a significant impact on in-room response) is valid.

They actually do, especially if they leak out via a port. Look how the directivity just disappears where that resonance leakage occurs.

 

[Matt_Holland]

The issue here is more with Atkinson's measurement regimen (which is not anechoic - it's gated measurements spliced with near-field measurements at LF).
If the loudspeaker cabinet's resonances are generating large dips in frequency response this would be evident in an actual anechoically measured frequency response (or equivalently one measured on a Klippel NFS). Even then, for a loudspeaker to have a cabinet resonance as bad as the one shown in the devore here is very uncommon, and itself a sign that the loudspeaker should probably be avoided.

Quick edit: the bass boost in Atkinson's measurements are not real either. He does uncompensated near-field measurements to splice together with the gated data, which means you'll be seeing as much as 6dB boost relative to what the actual far-field bass response is. Once again this is not a problem with a proper anechoic measurement or NFS measurement.

Underrated post ^^^

 

[Scgorg]

agree about the spinorama measurement but it seems that the derived estimated in room response is not close to being accurate.

This topic has been discussed to death on this and other forums multiple times before.
The predicted in-room response is highly accurate above the sparsely modal region of the room. It cannot predict the frequency response at <200Hz, or even <400Hz in most cases, as that will be dominated by a very sparse set of room modes and specular reflections/boundary interferences which are almost entirely dependent on the room, not the loudspeaker. The larger the room, the lower in frequency the predicted in-room response will give a decent approximation. This is a limitation of the model, and most people understand this (I hope).

In the first place, your examples here are not very good, because you're only looking at the on-axis frequency response. Stereophile does not calculate predicted in-room response (their measurement set is both too small and not accurate enough for it). Compare the PIR of a loudspeaker to the in-room response above 500-1000Hz and you'll generally see very strong agreement.

the sound from the enclosure resonance dissipates in an open air anechoic measurement.

Sound luckily behaves in a fairly neat and predictable manner when you're in the far-field of the source. Any contributions to the frequency response stemming from cabinet resonances will dissipate at the same rate as the loudspeaker's direct sound (6dB/DD), as long as you're in the far-field, the distance at which you measure does not matter for the resulting frequency response (exceptions for air absorption of high frequencies). If the resonances have particularly strong contributions at certain off-axis angles, then that will be captured in the anechoic off-axis measurements, and weighed in the calculation of the predicted in-room response.

 

[kemmler3D]

All agreed that bad cabinets contribute to a screwed-up (or sometimes oddly fixed) in-room response. However, these are missed or understated by virtue of measuring nearfield or on-axis and not because the measurement environment or style is anechoic.

The mic doesn't care where the sound is coming from, and so on the klippel it picks up cabinet resonances as well as anything else.

 

[MAB]

It would be good to see actual acoustic measurements of the contribution of cabinet resonances to the overall sound of a speaker. They actually contribute very little.

Most studies I’ve seen either change too many variables simultaneously to tell what is going on, confuse internal standing waves with vibrating cabinet, make apples to oranges arguments (like first post in this thread), or use indirect measurements (like accelerometers with no attempt to measure the change in sound. This is one careful study that shows the actual changes when bracing and cabinet are modified:

Differences in Klippel NFS Measurements Based on Enclosure Thickness and bracing

I plan to compare measurement differences using the Klippel NFS based on wood thickness and the presence or absence of bracing. Three different boxes will be used for the test. The drivers to be tested are the SB Acoustics SB17nbac and the test boxes are currently complete, and measurements will...

www.audiosciencereview.com

These are tiny improvements. If you have a great speaker, it will become slightly greater. If you have a bad speaker it will still sound bad.

 

[Geert]

I had a look at the first example, the B&W805's versus the Devore 0/96's. The in-room response of these speakers is measured in different rooms; Art Dudley's versus John Atkinson's room! And the speakers have a totally different directivity and form factor (important for SBIR's and room modes).

Case closed.