Loudspeaker Design - from the perspective of a loudspeaker design engineer...

[Chris A]

Moving on to the next level of aggregation of requirements/characteristics, expanding on the transfer function:

 

 

The above next-level hierarchy is from the same two-year-old file I created to notionally talk about the type of transfer function-related characteristics (i.e., SPL and phase response...and group delay--the first derivative of phase).  Here you will begin to see some of the lower level characteristics that affect acoustic performance. It pulls in the box stiffness/resonances and "baffle step" (otherwise known as "flat-horn gain") that a gentleman inquired about above, but also picks up a few other items, such as Olive's low frequency extension (and tacitly, SPL response smoothness and absolute deviation variables). 

 

________________________________________________________

 

From this hierarchy, a set of performance measurements can begin to fill in typical values along the right-hand side hierarchy items.  For instance, you could measure typical garden variety and "good" drivers and convert their engineering performance values into "utility" values--to be aggregated together via a weighted hierarchy (just like Sean Olive was attempting) and then compare to listener subjective scores to develop correlation plots that can be dialed in to form some single-value rating for candidate loudspeaker configurations (i.e., a statistically significant set of blind-blind listener scores) until you can predict with high confidence how well a particular candidate configuration would score relative to a "brand X" or other existing company product line value.  This is the goal that Olive was attempting to achieve.  This is important from the standpoint of comparing different loudspeaker configurations, i.e., "apples, oranges and bananas".  Olive was attempting to quantify the aggregated levels of performance and turn those into single value (loudspeaker) utility scores.  The method works quite well.  (I used a much more involved process in another life before embarking upon this pastime--which I'm sure that no-one here is interested in...😉)

 

For their needs, they likely got the answers that Harman was wanting.  However, the fly in the ointment is that I believe that they limited their subjectively reviewed loudspeaker types to a few competitors and their current product line, but not an expanded set based on much more capable loudspeaker configurations, so therefore they didn't pick up some variables that the engineer that mentored me has been using.  Namely, they washed out distortion in their decision model--something that I believe was a critical mistake.  (Perhaps more on that subject later.)

________________________________________________________

 

Returning to the subject at hand--I believe that it's clear that the above "decision science" methods to develop locally optimal decision models for loudspeaker design could be used, but I believe that virtually all other manufacturers are using organizational (seat-of-the-pants) knowledge and local engineering and marketing culture to determine their future loudspeaker products.  This is most likely the case at the enterprise where my (former) mentor works.  I would not use their tacit decision models to make these kind of decisions.  But I believe the key to this is that each enterprise is using its own set of mental models, and the results that they produce reflect these differing organization decision models. My guess is that each organization prides itself on the ability of their decision models--and their internal crystal ball models of where the market demand is headed.  There are always winners and losers in this game.

 

Time for a pause.  More to come on loudspeaker distortion...

 

Chris

[semente]

Since you've mentioned directivity (lateral response family at 50", normalized to response on tweeter axis):

 

Dutch & Dutch 8C (wave-guided cardioid narrow baffle box with rear woofers)

 

Kef Blade 2 (wave-guided coax narrow baffle box with side woofers)

 

PSB T3 (wave-guided narrow baffle box)

 

Sonus Faber Stradivari (650mm wide baffle box)

 

Jamo R907 (open baffle)

 

Gradient Helsinki 1.5 (waveguided coax open baffle)

 

MartinLogan Montis (planar dipole above 300Hz)

 

Avantgarde Uno Nano (horns above 350Hz)

 

MBL Radialstrahler 101E Mk.II (omni)

 

source: https://www.stereophile.com/

[Chris A]

That's fine with the Dutch & Dutch approach, but note that the technique that is used is diffraction--the same approach used by line arrays (see the Danley paper on this subject).  To be honest, line arrays only work well if you're sitting in a certain place in the audience, and they really suck if you're not sitting in the right areas...

 

Danley is systematically replacing line array technologies and gaining significant market share in that market (i.e., when there was a fixed installation PA market before SARS-CoV-II hit the scene).  There's a reason why they are. 

 

Chris

[Chris A]

Note that I own no stock in Danley and have no other financial interests with them, nor do I own any stock in any other loudspeaker company or have any financial interests in any of them. I sell no products and charge no one for my services.

 

Chris

[semente]

4 minutes ago, Chris A said:

That's fine with the Dutch & Dutch approach, but note that the technique that is used is diffraction--the same approach used by line arrays (see the Danley paper on this subject).  To be honest, line arrays only work well if you're sitting in a certain place in the audience, and they really suck if you're not sitting in the right areas...

 

Danley is systematically replacing line array technologies and gaining significant market share in that market (i.e., when there was a fixed installation PA market before SARS-CoV-II hit the scene).  There's a reason why they are. 

 

Chris

 

High performance stereo listening is a solo activity.

[Chris A]

1 minute ago, semente said:

 

High performance stereo listening is a solo activity.

Not in my house.

[semente]

6 minutes ago, Chris A said:

Not in my house.

Who gets to sit in the center?

[fas42]

Just worked out the link ... I pointed to @Chris A's post about his rig a month ago,

 

So, sitting in the centre ain't necessary ... 😉.

[Odder_Sea]

Thank you for taking the time to organize and explain these principles.

 

I've personally found that there are Three "Big D's" that limit the performance of the loudspeaker:

 

Distortion, Diffraction, and Directivity. 

 

Of these three, the first is the only one that gets much attention from designers and buyers, as the remaining two usually require designs that are too unwieldy, unsightly, or difficult to either produce or be seriously considered by most buyers. As such, with few exceptions, we keep getting a new batch of squirrel and monkey coffins every year, with slightly better cones n' domes than last year's models, with little to show for it sonically.

[PeterSt]

12 hours ago, Chris A said:

12 hours ago, semente said:

 

High performance stereo listening is a solo activity.

Not in my house.

 

FWIW, I explicitly tune for enjoying the system from almost everywhere in the room (I don't even have a listening seat). Besides, people may recall me saying that this is without any room treatment; all what it requires is "best quality reproduction" (from really all elements involved in the chain, and which are all 100% under our (design/production) control. Have one element "off" a little, and standing waves of one type or another are our share).

 

Peter

[PeterSt]

12 hours ago, Chris A said:

More to come on loudspeaker distortion...

 

Chris, I hope you can qualify this with some numbers. I mean, I can claim to be without standing waves whatsoever, but such a claim should come with numbers of some kind (doable, but quite tedious to set up in honest fashion).

Loudspeaker distortion is easy to measure, but I don't see manufacturers do it (or show the numbers 😌).

[semente]

Here's a good set of measurements (not to be confused with a set of good measurements):

 

Avantgarde Acoustic Uno XD

https://www.fidelity-online.de/avantgarde-acoustic-uno-xd-messungen/

 

 

 

 

 

 

 

 

 

 

 

 

 

[semente]

Have a look at Bowers & Wilkins 800 D3's measurements for comparison:

 

https://www.fidelity-online.de/bowers-wilkins-800-d3-messungen/

 

[Chris A]

Expanding the distortion hierarchy... (We'll get to directivity next after the distortion hierarchy is discussed.)

 

 

Here, the story begins to get a little more interesting-in my view.  Notice that I've "double dipped" on phase/impulse distortion here vs. the transfer function hierarchy, but perhaps choosing one place or the other would be the next step in finalizing the rolled-up hierarchy.  The light blue background indicates that these performance capabilities are correctable using signal processing.  There have been a few attempts to using DSP to correct harmonic and other transient distortion (David Gunness being a notable author), but general purpose applications and hardware to accomplish significant performance enhancements in this area are not generally available--nor free as of this writing. I'll put that discussion on the "we'll check on that later" pile.  In general, to get lower harmonic distortion, drivers and waveguides (or cabinets housing the drivers in the case of direct radiating) need to be selected carefully.  Same thing for modulation distortion. (Note that modulation distortion is ~16-20 dB lower in well-designed waveguides vs. using the same drivers in direct radiating mode--most people get that one wrong.)

 

Interestingly, most DIY loudspeaker designs that I I've seen usually pay lip service to harmonic distortion, and virtually no one that I've seen even acknowledge modulation distortion figures. (Note also that Klippel NFS systems produce AES standard reports for modulation distortion.)

 

Here also, I've captured the voice coil heating here in compression distortion.  Passive crossover network heating is not captured here. There are a technology-dependent factors shown here (waveguides, reflex ports) but are not exhaustive.

 

Chris

[Chris A]

39 minutes ago, semente said:

Here's a good set of measurements (not to be confused with a set of good measurements):

I'm really not sure what your needs are, but I'm pretty sure that what you intend and what I'm currently working on are not coincident.  I would recommend that you start a separate thread on the exact subject (which looks like random measurements).  (As the OP) this thread is clearly focused on something other than your expectations. 

 

Chris

[Chris A]

The last major category of loudspeaker acoustic requirements (as opposed to non-acoustic requirements) is directivity:

 

 

The biggest issues that I see with directivity requirements/capabilities can be grouped under the three major subheadings above: directivity at higher frequencies (generally above 1-2 kHz, but this can extend down to ~500 Hz), and low frequency directivity (below the higher frequency directivity regime). 

 

Additionally, there are issues with coverage: 90 degrees horizontally (-6 dB) above the room's Schroeder frequency--typically taken to be ~200 Hz seems to be a sweet spot.  This typically limits the vertical coverage to about 60 degrees in order to avoid pattern flip at lower frequencies, without introducing massive amounts of diffraction into the mix.

 

Lastly, there are issues regarding the smoothness of directivity response (i.e., frequency-dependent directivity).  There are often large directivity mismatches at the crossover interference bands, and these play havoc with the design of effective crossover schemas and with (calibrated) subjective assessments of sound quality--as detailed in Toole's and Olive's articles. 

 

In practice, directivity issues interact with other performance issues (as do most other performance parameters highlighted above).  The discussions that I plan to pursue at this point will be related to how these performance requirements/capabilities interact with each other and are therefore loudspeaker configuration-dependent and are highly affected by non-performance considerations, such as cost and form factor/size.  Too often, available acoustic performance is traded away without much discussion because of fears of meeting appearance thresholds from non-engineering functions within large corporations/enterprises--notably marketing/sales organizations who are often not driven by the acoustic performance of the loudspeakers, but rather their appearance.

 

Next up: discussion of loudspeaker requirements/capabilities interactions.

 

Chris

[semente]

When you mention waveguides, are you referring to gentle directivity control devices or to narrow dispersion horns (or both)?

[Chris A]

I use the term "waveguide" mostly because so many audiophiles here have trouble with the term "horn".  I think that Bjorn's book helps to bridge some of that divide:

 

https://hornspeakersystems.info/

https://www.parts-express.com/high-quality-horn-loudspeaker-systems--500-032

 

 

It's a good update on Beranek's "Acoustics" and Olson's "Acoustical Engineering" texts.  The theory portion of the above book (i.e., the last ~40% of the book) looks good and is useful in terms of theoretical considerations and current analysis developments, but is a little thin in terms of current high performance horn design which I think Bjorn is trying to market himself as well as other companies with their own offerings. 

 

The front portion of the above book is a comprehensive look at the development of horn-based loudspeakers over time since the beginning of home hi-fi loudspeakers appeared in the 1920s (and before).  Note that horn design developments over the last 20-30 years, i.e., since 1990-2000, have eclipsed the performance of those horns designed before that time.  There are a lot of poorly performing horns out there that exist from before that time.

 

But this thread is not about horns/waveguides exclusively...it's about loudspeaker requirements as viewed from a large company loudspeaker engineer's perspective.

 

Chris

[Chris A]

Some process-based information on what takes place in loudspeaker development at larger corporations:

 

Generally speaking, most DIY loudspeaker efforts involve buying the drivers and putting them into DIY boxes, then generating passive crossover networks that stitch the overall SPL response together (for better or worse), sometimes going as far as adding notch filters (attenuating) to smooth the SPL response. Usually little is done to minimize growth across the crossover interference bands, unless using a third-party deconvolver upstream of the amplifier to flatten the on-axis phase response.

 

As far as a loudspeaker engineer is concerned (and depending on the capabilities of the enterprise they work for), drivers can either be selected like catalog items (with significant developmental testing required as well as incoming inspection QA lots to ensure that the properties of the drivers don't change with production).  So as such, many/most loudspeaker engineers not involved with driver development, and select drivers based on their needs and on measurements (i.e., the source of the performance defects are not investigated).  Parameters measured include:

• SPL and phase response (high and low drive levels)--looking especially for resonances in the drivers or the box
• Directivity vs. frequency (horizontal, vertical)
• Harmonic and modulation distortion, compression distortion
• Input electrical impedance response
• Power handling and voice coil/passive crossover heating/temperatures
• Box/enclosure design -- perhaps using miniature accelerometers to measure bending/resonance modes, etc.
• Subjective listening trials to catch anything that wasn't caught in the above list of performance capabilities

The thresholds for each capability is usually company dependent and price-range dependent, i.e., lower priced loudspeakers generally having less stringent performance thresholds.

 

If the company produces cone/dome type drivers and/or compression drivers, etc., the situation is different in that there is usually a team doing custom drivers and a separate team doing the balance of the loudspeaker development. 

 

Many companies design their loudspeakers such that one driver is custom, and perhaps the other drivers are catalog items or other off-the-shelf non-developmental items.  The custom driver parameters are usually picked to preclude DIY or third party assembly of the loudspeakers (in order to assure that the consumers do not try to bypass the higher prices charged by the company for the finished loudspeaker models. 

 

[This is especially true now that very good freeware or other low-cost measurement applications and low cost calibrated measurement microphones are available, as are low cost/high performance DSP crossovers that can be used to eliminate the passive crossovers used on most consumer-grade loudspeakers (i.e., trading the requirement of mono-amping the loudspeakers with bi-amping or multi-amping). In particular, this practice has been creeping steadily upward in recent years, awaiting a more general realization that DSP crossovers and multi-amping provide much higher in-room fidelity than stock passive crossovers.]

 

If the company produces its own drivers, then many more tests are performed by the integration team (the loudspeaker designers) on the received developmental drivers in order to provide feedback to the driver design team(s) to iterate their designs. 

 

Lastly, the loudspeakers can be sent off to get EASE-type data gathered from a third-party testing firm anechoic chamber or Klippel NFS-type system.  This is especially true of loudspeakers built for commercial use. 

 

Chris

[Chris A]

Beginning discussion on loudspeaker requirements/capabilities interactions:

 

This is an area that I believe results in the different types of loudspeakers we see in home hi-fi and commercial marketplaces, and (candidly) this is where I had intended to focus the discussion when I started this thread.  There were preliminary discussions on this subject, above, but were necessarily brief in order to also fill in some conceptual gaps in knowledge that I believe most hi-fi enthusiasts simply do not see or haven't considered. 

 

First, let's start with the biggest constraints on acoustic performance:

1. Small size and expected shape of the loudspeaker enclosure overall (i.e., "WAF" and "MAF" [M=man]), and
2. Cost constraints on the finished product--which turn into price categories for offered loudspeaker models.

Note that these two design constraints actually work to significantly and severely limit the acoustic performance of the loudspeakers in home hi-fi applications today. 

 

If you were able to transport yourself back to the late 1940s when home hi-fi was kicking off for the first time, if you were to show most of today's loudspeaker models to loudspeaker engineers practicing in the 1940s, they'd laugh at them as "atrociously undersized and entirely the wrong shape and configuration for hi-fi performance".  This is useful to remember when talking about hi-fi loudspeakers today. Contrary to the general thought that "we know a lot more about loudspeaker technology today", most of the physical and measurable psychoacoustic factors that govern loudspeaker design today were largely already known by 1949.  What's changed today is buyer expectations...in terms of size, shape and cost (and willingness to severely compromise loudspeaker acoustic performance in order to meet the two constraints mentioned above even if they have accommodated to the sound of these ultra-small boxes).

 

If you look at the highest-priced "high end" loudspeakers--they are large and most of them are also not shaped like shoeboxes.  Instead, they are usually require similarly large rooms to be placed into--to function at their intended levels.

 

Unlike virtually all other discussions of loudspeakers within the home hi-fi enthusiasts marketplace, for the purposes of this particular discussion, it will be assumed that the starting focus or emphasis here is on acoustic performance over form factor and miniature size (the two constraints shown above).  Additionally, the cost of the loudspeakers is not something that you would likely find in a lineup of Audio Science Review forum tests (i.e., the lowest cost models available from the major brands).  Loudspeaker performance discussed here will begin toward a larger size and higher performing in terms of lower distortion (distortion as defined above).  As the subject evolves further down the road, we can talk about smaller size (but not until higher fidelity sound reproduction loudspeakers are first discussed in detail.

 

So the for purposes of requirements interactions, I'll repeat a figure from the top of this thread for reference:

 

 

Note the three major areas identified here: transfer function (i.e., SPL and phase response), freedom from distortion, and directivity. 

 

Just like Hofman's Iron Law, if we put a particular emphasis on transfer function response (usually focused on-axis only), we typically give up low distortion--especially modulation distortion which is all inharmonic in nature and therefore most objectionable to listeners)--and directivity control.  We tend to end up with small direct radiating loudspeakers of the "monkey-coffin" variety, of which there are literally thousands of extant example models (and companies based on this requirements emphasis).  The curious thing about this loudspeaker performance area is that this one responds the most easily to DSP correction methods and IIR/FIR filtering techniques.  If we combine small size with loudspeakers designed for transfer function linearity, we get the vast majority of direct radiating loudspeaker designs found today.  But we also get one other trade off: low efficiency (via Hofman's Iron Law) that limits the allowable dynamic range of the loudspeaker to significantly below that of most real music found today, particularly that of multitrack-recorded (i.e., built on the mixing board) amplified music.  Even music performed via large unamplified ensembles, such as orchestras and wind symphonies have too much dynamic range for small loudspeakers to reproduce without severe and limiting modulation and compression distortion severely limiting the reproduced dynamic range and clarity.

 

Some other loudspeaker types that prioritize transfer function linearity are dipole radiating film-type loudspeakers, such as electrostatics and dynamic-membrane types (e.g., "Magnepan").  The usually trade away low modulation distortion and controlled directivity.  (In fact, dipoles have the added fun of the owners having to deal with a backwave radiating field via their listening room design in order to rectify the phase relative of the backwave at lower frequencies to that of front-wave in order to avoid significant listening position cancellations.)

 

If we instead first focus on low distortion, particularly modulation (inharmonic) and compression distortion, then the size of the loudspeakers typically grows to that which is required to easily avoid the nonlinearities of small diaphragms moving air relative to their wavelengths, and in particular, the use of "acoustic transformers": acoustic waveguides or horns to reduce moving mass effects and other sources of modulation distortion, i.e., the sources coming from low acoustic efficiencies of direct radiating drivers without acoustic transformers.  Higher efficiencies are the flip side of avoiding modulation distortion at realistic playback levels.

 

Lastly, if we focus on full-range loudspeaker directivity in-room (as evidenced via Toole's and Olive's own performance factors), we find ourselves again in the realm of acoustic waveguides/horns to control the polar acoustic output (otherwise called "power response" or integrated SPL over output angles). 

 

 

So for the above, all other things being equal (and loudspeakers are not severely undersized via major constraint #1 detailed above), the approach to create both low distortion and controlled directivity loudspeakers is via use of waveguides/horns.  This is, in fact, the way that electrodynamic loudspeakers were first designed--for cinema use (starting approximately in 1930 via the WE horns, then the Shearer Horn system, then later the Voice of the Theater (VOTT) systems from Altec-Lansing.  These had the added bonus of not requiring large amounts of electric power, which was expensive during the tube/valve amplifier era. 

 

So, depending on where you put your emphasis (threshold requirements performance levels), you get entirely different loudspeaker designs. 

 

Next up: the effects of DSP transfer function linearization/crossover filtering and acoustics computer modeling on opening up the design space of hi-fi loudspeakers.

 

Chris

[semente]

According to this admittedly limited (some of Harman's research suffers from overly small sampling) study by Olive (4) untrained listeners preferred excessive bass and quite a bit of treble (the infamous boom-tizzz sound) and perhaps we can also agree, from observation, that loud rates high in their preferences.

 

 

 

 

Could this perhaps justify different performance targets or "voicing" across the range, from a manufacturer's perspective?

 

Surely if a company can manufacture a flagship with flat on-axis and smooth off-axis response they can also do it with an entry-level design...

[Chris A]

41 minutes ago, semente said:

According to this admittedly limited (some of Harman's research suffers from overly small sampling) study by Olive (4) untrained listeners preferred excessive bass and quite a bit of treble (the infamous boom-tizzz sound) and perhaps we can also agree, from observation, that loud rates high in their preferences...Could this perhaps justify different performance targets or "voicing" across the range, from a manufacturer's perspective?

I'm not sure how we got from my above discussion to here, I'll address your questions...

 

In most of these instances, upturned bass and treble SPL response is a reaction to typical mastering of music (i.e., making the tracks sound louder via attenuation of bass response below 100 Hz) and also a response to narrowing of polar response of the tweeters, and broadening of polars below 1 kHz.

 

Additionally, if you draw a straight line across the untrained listeners curve, you will see a drop out in response in the midbass (generally 100-300 Hz) and midrange (300-1000 Hz) that is typical for those listeners trying to self-correct in-room midbass decay issues ("muddiness") and loss of polar control via direct radiating bass (woofer) loudspeakers.  This is partly due to the music tracks themselves (and what they are listening to) and part loudspeaker power response issues.  If you've never heard a loudspeaker that can control its polars below 1000 Hz in-room, you might not have experienced the effect of full range controlled directivity in-room and the effect that has on desired loudspeaker EQing.

 

58 minutes ago, semente said:

Surely if a company can manufacture a flagship with flat on-axis and smooth off-axis response they can also do it with an entry-level design...

So, by my response above, it's usually not that simple of an "on-axis SPL response" by itself.  It has even more to do with the off-axis response and the music tracks being played (and the genre of music, i.e., its instrumentation, etc.).  Most audio enthusiasts miss this point: the full-range loudspeaker directivity effect on listeners desired on-axis SPL response. 

 

Chris

[Chris A]

Here is a typical mastering EQ curve used on a particular album (Siamese Dream by Smashing Pumpkins, 1993):

 

 

As you can see, even the mastering/mixing people can significantly unbalance the on-axis response of the loudspeakers in-room to achieve the effect they are seeking with the instrumentation they're working with.

 

Chris

[Chris A]

Perhaps a few comments on the subject of distortion sources and interactions might be illuminating:

1. Harmonic distortion (HD) is related to modulation distortion (MD), of which there are both AM and FM sources of modulation distortion. 
   
   Many people are not aware of the sound of modulation distortion, but it is in fact the sound of opaqueness or  muddiness that you hear when loudspeakers having very small diameter direct radiating woofers have when played at higher volume levels.  The harmonics generated (as seen in the figure below) by the loudspeaker are also turned into both AM and FM distortion sidebands around the higher frequencies being simultaneously produced ("f2" as shown below). 
   
   
   
   AM distortion comes from the nonlinearities of direct radiator drivers playing both high level bass and midrange-treble frequencies at the same time, where the electrical motor (voice coil beginning to leave its magnetic field at the ends of its travel.  The suspension also begins to stiffen (like a spring with spring factor "K"--a stiffening spring). The net effect is modulation sidebands on either side (in the frequency domain) of the higher frequencies being produced.  This is the dominant modulation distortion type in low frequency drivers.
   
   All modulation distortion sidebands are non-harmonic to both the lower and higher frequencies produced, so their "just detectable distortion" (JDD) levels are typically much lower than harmonic distortion (HD). 
    
2. FM distortion, also called Doppler distortion, in loudspeaker drivers produces additional sidebands on either side of the higher frequencies being played, just like AM distortion, except that the sideband frequencies that are produced are generally different than AM distortion, shown above.  One interesting source on FM distortion audibility comes from Keith Howard.  This is the dominant modulation distortion type found in higher frequency drivers, and is directly proportional to the bandwidth of the frequencies handled by a driver.
   
   
    
3. The sources of modulation distortion in acoustic drivers can be seen in the following Klippel-generated table--some of which is a little more advanced in terms of this discussion. "HD" is harmonic distortion, "IMD" is FM distortion, and "AMD" is AM distortion.
   
   
   
   A good discussion on the sources of harmonic and modulation distortion can be found here (Klippel).
    
4. Group delay (GD)...the first derivative of phase with respect to frequency:
   
   Group delay is a noted distortion source that is almost completely ignored in manufacturer's data (except perhaps some studio monitor manufacturers), but when combined with full-range controlled directivity loudspeakers and control of early reflections from within the first 2-3 feet from the loudspeakers, combine to form a different audible presentation from the loudspeakers: increased perception of bass, elimination of harshness (especially for non-multitrack recorded recordings, such as jazz, classical, folk genres), and much greater listener involvement in the recordings.  Below is a notional group delay audibility threshold plot vs. frequency gleaned from another audio forum that's useful for comparing against your own REW measurements in-room (because you're probably not going to see these specifications from manufacturers--you have to measure them for yourself):
   
   
   
   As can be seen there is good agreement in audible perception levels between 500 and 5000 Hz, but is more speculative below 500 Hz.  The plot above closely follows personal experience in terms of GD audibility levels.
    
5. Phase distortion: Little is published in the realm of phase distortion audibility (i.e., total phase shift vs. frequency) under conditions of controlled early reflections and/or full-range loudspeaker directivity, it has been shown that total phase shifts exceeding 90 degrees over the 200-10,000 Hz band can be sensed in detracting from the clarity of the loudspeaker's presentation.  ref: Griesinger presentation on Clarity (especially slides 16-19).  Noted sources of phase distortion include ported bass bin cabinets and crossover filters, particularly higher order crossovers.

There are other forms of distortion not discussed above (notably compression and power/directivity distortion) but note that the choice of loudspeaker type/configuration tends to lock-in these other forms of distortion.

Chris