[proteinchemist]

This week I was listening to a piano recording that didn’t sound much like a piano and I began to wonder about the relative importance of two issues of speaker accuracy.  I am naïve about high fidelity systems so let’s assume for the moment that all amplifiers are perfect and that we are evaluating an amplifier / speaker system.

If I look at a speaker review and see an excellent SPL vs. frequency curve,  that curve must be measured at some specific input voltage going into the preamp.  So how different would that curve be if it were measured at half that the input voltage?  Do the sound characteristics of some systems differ as you change the volume control?

Secondly, if I set my volume control appropriately during a pianissimo passage, how accurately is a speaker system going to reproduce the dynamics of the concert hall?  If I plotted SPL vs time during a concert and then tried to reproduce that with a given speaker system in my home wouldn’t that be problematical with many systems? 

Considering these issues, instead of the usual frequency response curve, would it be useful to instead use a 3 dimensional surface with frequency on the X axis, some parameter related to the voltage going into the preamp on the Y axis, and SPL on the Z axis?  What is it going to look like?  Do engineers already do that in their labs?

Or am I straining at gnats and swallowing camels?

Wes

[*Scotty*]

proteinchemist,
Question number one.
The theory is that a correctly designed driver should respond in a linear fashion to an increase in voltage resulting in a directly proportional increase in SPL.  When the driver approaches the upper limit of it's power handling capability it is not uncommon to encounter compression effects as well as other forms of distortion. As the power is decreased less distortion produced as well as less spl's ,once again the behavior is assumed to be linear.
For the most part most loudspeakers are fairly linear in their response to voltage input within their rated power envelope. All loudspeakers have a rising distortion characteristic proportional to power input. Depending on the speaker design the sound characteristics could change as a function of power input.
Question number two. That depends on how much compression was applied to your recording,how linear your particular loudspeaker responds to increases in power,its efficiency and peak SPL capability and the power rating of your amplifier.
 All most no loudspeakers intended for home use are capable of reproducing the dynamic range of live music.
I don't think adding a third dimension to this sort of  measurement would generate that much more information compared to a 2D plot.
Someone actively engaged in speaker design might disagree with this viewpoint however.
Scotty

[Duke]

Well to keep things interesting I'll toss out a slightly different set of answers...

Reply to question number one:  The tonal balance of a speaker system can indeed change with input level.  Let's consider a 6.5" two-way.  Neither the woofer nor the tweeter are completely free of power compression; in other words, for a 10-fold increase in power neither one will give you an honest 10 dB increase in loudness.  High efficiency and/or prosound drivers do better than most mainstream "audiophile" drivers in this regard.  So let's say that your tweeter will give you 9 dB increase in SLP for a ten-fold increase in input power, and your woofer will give you an 8 dB increase in SPL (although the tweeter's voice coil is much thinner and has less thermal inertia, much more power goes to the woofer than to the tweeter so its voice coil heats up faster).   Let's assume the designer picks 80 dB as his reference level, so at 80 dB the speaker is perfectly balanced.  On 100 dB peaks the tweeter will be 2 dB louder than the woofer, but then on 60 dB quiet passages the tweeter will be 2 dB softer.  So the speaker sounds bright at high volume levels and dull at low volume levels.

The primary power compression mechanism is thermal compression.  As the voice coil heats up, its resistance rises.  Assuming a voltage source amplifier, the amp will put out less wattage (or power) into the hotter voice coil.  This is one reason I prefer specialty tube amps that do not behave as a voltage source - they have better dynamic contrast.  Magnets can also heat up which results in a reduction of magnetic flux and loss of efficiency.  Magnets heat up slowly, but voice coils heat up virtually instantaneously (imagine touching a 100-watt soldering iron to a heavy magnet and then to a thin voice coil).

Another factor that comes into play is the ear's non-linear perception of certain types of distortion.  At high volume levels, some distortions become more audible and objectionable, imparting a harshness to the sound.  Also, at high output levels amplifiers are more likely to produce harsh-sounding odd-order harmonics, and then of course there's clipping (in other words, we shouldn't cling to the assumption that all amplifiers are perfect for too long, especially if we're using them near their limits).

We're not done yet.  If the midwoofer is driven beyond its linear mechanical limits, it will lose clarity and articulation at the upper end of its range (which will usually cover most of the vocal region).  Also, magnets are subject to flux modulation by interaction with the induced field of the voice coil.  This degrades clarity, and becomes worse at high power levels.  I'm sure there are other level-dependent nasties that I'm overlooking at the moment. 

Reply to question number two:  Compression comes into the recording chain before the speakers of course, but a loss of 2 or 3 dB to thermal compression at the speaker end will audibly reduce the liveliness of the presentation.  Musicians use dynamic contrast to convey emotion, and when it's reduced the music is less engaging.  A primary attraction of most high efficiency speakers is the preservation of dynamic contrast.

The foregoing probably makes it sound like high efficiency is an automatic panacea.  It's not.  There are tradeoffs all along the road, whether that particular road takes you to loudspeaker nirvana or loudspeaker hell or somewhere in between.

Duke

[richidoo]

Cool question, and great answers!! Thanks
An alternative to increasing efficiency is decreasing background noise and electronics' noise floor. Reducing that by 20dB on the bottom end allows you to listen 20dB quieter and gain the linearity and lower distortion without the cost of equipment that can play it clean and loud. Quieting the listening space is an art in itself. Your brain adjusts easily to the volume difference, and 85dB becomes the new 105dB. When compared to the cost of clean sounding, loud playing gear, Silence is golden.
Rich

[Daygloworange]

This week I was listening to a piano recording that didn’t sound much like a piano and I began to wonder about the relative importance of two issues of speaker accuracy.  I am naïve about high fidelity systems so let’s assume for the moment that all amplifiers are perfect and that we are evaluating an amplifier / speaker system.

Welcome to AudioCircle proteinchemist. Great first post. 

There are many problems along the chain, from an event (live performance) to final reproduction (your speakers playing back).

Right from the onset, piano's are generally recorded from an improper perspective, with the results sounding unnatural. Recordings of piano are often done with ambient (far field) and spot (nearfield) miking. The spot miking is where things get askew. They often aim (one, sometimes 2 mics) aiming down at the soundboard in an effort to highlight the piano better. The only person that hears the piano close to that, is the player, not the audience.

Do the sound characteristics of some systems differ as you change the volume control?

Yes, most definitely. I think Duke's post  is a great explanation, and I can tell you that in practicality, the SPL level you mix a recording at, has a lot of bearing on how an engineer will shape the sound due to exactly what Duke explains is happening as SPL increases.

Room interaction (room acoustics) also play a huge factor. Comb filtering and room gain affects on bass response as SPL increases are important factors to the overall tonality of a speaker.

The ear itself has non linear characteristics with frequency vs SPL. Fletcher/Munson plotted curves many years ago on human hearing, and there are other more current studies.

A little know fact is that many engineers have bad hearing from years of exposure. A major industry publication years ago did hearing tests of recording engineers under strict anonymity, and the findings were alarming. Many of the engineers had severe losses at particular mid frequencies.

As far as compression added during the recording, in the case of a classical recording, it would use limiting and not compression. Limiting (hard or soft)is compression above a certain db threshold (usually high, in order to prevent overload or clipping), whereas compression is when you compress at a low db threshold and decreases the db gains (at a user adjustable preset ratio, ex: 4:1) of the entire performance. Compression as properly defined, is more of a pop recording (and it's abuse has become a popular sound shaping )technique, and would not be used (for the most part) in any classical recording.

There is also tape compression. As the db level approaches oversaturation levels, tape begins to compress the signal. At oversaturation, tape distorts and compresses severly.

richidoo is absolutely right about the importance of a low noise floor. It's an easy way to increase low level resolution, and avoid SPL related non linearities in the audio system.

Or am I straining at gnats and swallowing camels?

No, most definitely not. Very good questions.

Cheers

[miklorsmith]

That is the best first post I've ever read.   

IMO, dynamic contrast is among the most important and overlooked elements of speaker design.  By this I mean micro- and macro-.  And I don't mean speaker designers don't account for it, rather that these skills are placed low on the sheet in favor of other goals.

The way it sounds to me - low efficiency drivers and power robbing crossovers shrink the heights of musical sinewaves.  Some do pretty well on the macro level, very few do well on small shadings.

High efficiency speakers and no/simple crossovers do better at maintaining natural scale between loud and quiet.  This is my chosen path for this reason more than any other.  Unless egregious, I'd rather listen through frequency anomalies than compressed dynamics and unnatural tonality.

However, most high-eff systems have quirks and lots of folks can't abide them.  They are generally limited in output and can get screechy.  Most horns have all kinds of other issues.  As with most things and nearly everything in this hobby, it's a series of compromises where knowing what you prize and can live without is pure gold.