Dynamic compression and distortion vs. SPL

[adam2434]

I posted this same topic on AVS, but thought that folks here might have some good input.

 

Some speakers are able to reach higher SPL with less dynamic compression, distortion, and frequency response change vs. other speakers.  Some speakers retain their dynamics and tonal balance better as the volume is pushed.  Other speakers sound compressed and lose their composure as volume is pushed.

 

Typical specs provided by speaker manufactures do not provide an indication of how well a speaker performs as SPL level increases.  

 

Isn’t this a critical area of performance, especially if one has a large room and/or likes to listen at fairly loud levels?

 

Seems like manufactures that are superior performers as SPL increases would generate and publish the data to tout this benefit.  

 

Thoughts?

[derrickdj1]

Well, you are in the home of high sensitivity speakers that have horns for lower distortion, better dynamic range and clarity.  What more do you need to know, lol.

[Schu]

My head hurts...

[Ski Bum]

Typical specs provided by speaker manufactures do not provide an indication of how well a speaker performs as SPL level increases.  

 

That's true.  Those makers of puny, direct radiator type speakers don't want you to realize their limitations.

 

Klipsch used to publish compression specs for the old Heritage speakers, and still do for the Cinema line...

 

Isn’t this a critical area of performance, especially if one has a large room and/or likes to listen at fairly loud levels?

 

 

Indeed it is, at least if realism is the goal.  You'll see this disclosed for the Cinema line, as well as similar offerings from the likes of JBL, QSC, Danley Sound Labs, etc, but those are all cases of product intended for big venues, where headroom is a critical consideration.  In a home environment, getting to reference levels is much, much easier, even with less sensitive speakers.  It's just not a huge concern in the broader market.

 

PWK said something to the effect that 99% of the folks don't care about the quality of reproduced music, and that his speakers are for the more discriminating 1%.  Wide dynamic range (no compression) is one ingredient to the recipe.  

[Chris A]

Isn’t this a critical area of performance, especially if one has a large room and/or likes to listen at fairly loud levels?

 

See https://community.klipsch.com/index.php?/topic/151087-why-horn-loaded-sounds-better-than-direct-radiating-faq/

 

If you listen below concert level for acoustic instrumentation, you'll still see SPL transients that extend above 100 dB at one metre...easily. Just listen to a jazz string bass player like Brian Bromberg. His energetic performance transients are significantly above 100 dB (at one metre).  Acoustic guitar performance, like Michael Hedges' Aerial Boundaries, can see very high transients with an average-to-peak (crest factor) of 23 dB. Percussion transients are even higher, even though most recordings are compressed in dynamics.  A simple snare drum or kick drum will produce crest factors in excess of 30 dB.   Symphonic orchestras playing 19th and 20th century compositions will typically see crest factors of 20 or more. 

 

All direct radiating woofers suffer from much higher levels of modulation distortion and compression distortion than if those same woofers were effectively horn loaded.  Paul Klipsch was the proponent of horn loading and low modulation distortion loudspeakers all of his professional career, from before building the original Klipschorn folded corner horn.  

 

It is the ability of horn-loaded loudspeakers to accurately reproduce music transients without audible modulation or compression distortion that sets them apart from all other loudspeakers.

 

Chris 

Edited May 26, 2016 by Chris A

[DrWho]

Seems like manufactures that are superior performers as SPL increases would generate and publish the data to tout this benefit.     Thoughts?

How would you quantify it?

 

Do you look at distortion versus SPL? Gain linearity versus SPL? What frequencies, and what amplitudes do you measure? What kinds of distortion? Do you use perceptual weightings on the results? Are there minimum noise/distortion targets too?

[Ski Bum]

How would you quantify it?

 

I think it typically shows up as "maximum continuous output", with some sort of qualifier, i.e. max continuous output with 1db of thermal compression.  That's what the old Heritage specs specified, if my memory serves, those old spec sheets seem to have disappeared down the memory hole.  At any rate, at some point the acoustical output fails to keep up with the electrical input, which is measurable.

Edited May 26, 2016 by Ski Bum

[adam2434]

 

Seems like manufactures that are superior performers as SPL increases would generate and publish the data to tout this benefit.     Thoughts?

How would you quantify it?

 

Do you look at distortion versus SPL? Gain linearity versus SPL? What frequencies, and what amplitudes do you measure? What kinds of distortion? Do you use perceptual weightings on the results? Are there minimum noise/distortion targets too?

 

I would have to defer to the speaker engineers, but I'll throw something out there.

 

Plot THD/IMD across a wide frequency range at multiple SPL levels.

 

Would this cover multiple sources of distortion, including thermal compression?

Edited May 27, 2016 by adam2434

[Deang]

How would you quantify it?

I think it typically shows up as "maximum continuous output", with some sort of qualifier, i.e. max continuous output with 1db of thermal compression. That's what the old Heritage specs specified, if my memory serves, those old spec sheets seem to have disappeared down the memory hole. At any rate, at some point the acoustical output fails to keep up with the electrical input, which is measurable.

Also known as Power Compression ...

"Power compression is the sensitivity loss at the specified power, measured from 50 to 500 Hz, after a 5 minute AES standard (50-500 Hz) pink noise preconditioning test at the specified power."

https://www.jblpro.com/pages/general_faq.htm

The majority of loudspeaker systems these days are quite small in comparison to what most of us here are using.

[Chris A]

Seems like manufactures that are superior performers as SPL increases would generate and publish the data to tout this benefit.     Thoughts?

 

How would you quantify it?

 

Do you look at distortion versus SPL? Gain linearity versus SPL? What frequencies, and what amplitudes do you measure? What kinds of distortion? Do you use perceptual weightings on the results? Are there minimum noise/distortion targets too?

 

I would have to defer to the speaker engineers, but I'll throw something out there.

 

Plot THD/IMD across a wide frequency range at multiple SPL levels.

 

Would this cover multiple sources of distortion, including thermal compression?

 

I don't believe that you have to do that. 

 

For amplitude compression measurement, see https://www.klippel.de/fileadmin/klippel/Files/Know_How/Application_Notes/AN_12_Amplitude_Compression.pdf

 

The generalized nonlinear distortion measurement procedures by type of distortion: https://www.klippel.de/know-how/measurements/nonlinear-distortion.html

 

The above is incorporated by reference into the international standard for testing loudspeakers: IEC 60268-5 "Sound System Equipment, Part 5: Loudspeakers".  The standard test procedures are defined.

 

Chris

Edited May 27, 2016 by Chris A

[adam2434]

 

Seems like manufactures that are superior performers as SPL increases would generate and publish the data to tout this benefit.     Thoughts?

 

How would you quantify it?

 

Do you look at distortion versus SPL? Gain linearity versus SPL? What frequencies, and what amplitudes do you measure? What kinds of distortion? Do you use perceptual weightings on the results? Are there minimum noise/distortion targets too?

 

I would have to defer to the speaker engineers, but I'll throw something out there.

 

Plot THD/IMD across a wide frequency range at multiple SPL levels.

 

Would this cover multiple sources of distortion, including thermal compression?

 

I don't believe that you have to do that. 

 

For amplitude compression measurement, see https://www.klippel.de/fileadmin/klippel/Files/Know_How/Application_Notes/AN_12_Amplitude_Compression.pdf

 

The generalized nonlinear distortion measurement procedures by type of distortion: https://www.klippel.de/know-how/measurements/nonlinear-distortion.html

 

The above is incorporated by reference into the international standard for testing loudspeakers: IEC 60268-5 "Sound System Equipment, Part 5: Loudspeakers".  The standard test procedures are defined.

 

Chris

 

I’ve heard of Klippel analysis – some speaker manufacturers use it for design/R&D.  It looks like a very comprehensive suite of tests.

 

Perhaps the numbers it generates are too complicated for the average consumer.  Plus, without comparative numbers from competitors’ speakers, the numbers would be difficult to interpret.

 

I noticed that JTR publishes a “Usable Output” spec for their residential speakers (for example “130db calculated peak 133 – 3db compression"), but it’s not clear how they generate that number.

 

My gut tells me that distortion/compression vs. SPL would be very telling and relevant data when comparing speaker performance.

 

Ever notice that some speakers make you want to listen at higher and higher volumes, but other speakers make you want to reduce volume once you hit a certain point?

Edited May 31, 2016 by adam2434

[adam2434]

In this PSA speaker review, the reviewer measured THD and IMD at 75/85/95 dB.

 

http://www.hometheatershack.com/forums/speaker-subwoofer-reviews/113313-power-sound-audio-mtm-210-speaker-review.html#post1111641

 

I wonder how other compression driver/horn speakers and dome tweeter speakers would compare.

[Chris A]

If you look at the K-402 multiple-entry horn thread, pages 13 and 22, you'll see a number of REW measurement plots that you can compare. 

 

However, all of these plots were taken at 100 dB@1 metre, which is significantly higher than the highest SPL that the loudspeaker that you referenced above was tested.  Most measurements of the K-402 multiple-entry horn design were taken outside in nominally "anechoic" conditions, i.e., no boundary gain was present other than the ground plane at about one metre away (half space).

 

Chris

[DrWho]

For amplitude compression measurement, see https://www.klippel....Compression.pdf

 

That kind of analysis works for really large artifacts, but for smaller artifacts (like what we get in electronics world) we should use gain linearity to show the minor differences.

 

Page 22 in the attachment starts talking about non-linearity, and Page 26 has a graph of what I'm referring to.

 

Of course this can be used in speaker world too.

 

Perhaps the numbers it generates are too complicated for the average consumer.

I think analyzing distortion numbers is too complicated for engineers too. These types of analysis are invaluable for verifying designs, but they don't tell you a lot about how they're going to sound. The design decision that creates the artifacts that shows up in the measurements are what dictate the audible result. The measurements are just tools to make sure the shortcomings of a particular topology are minimized.

 

In other words, the true art in audio design is in the architecture and selection of topologies. For a given topology, if two designs are implemented well, then they are going to sound very similar. By extension, the consumer really doesn't need to get bogged down in the technical detail. Find the topology you like, and choose a company that you trust to implement it well.

 

Trying to find a holistic unit of measure to compare every possible typology is kinda ignoring the original purpose of measuring in the first place.....we don't use measurements to understand how something sounds. We listen with our ears to understand how something sounds. We use measurements to fix (audible) problems...

AMH-1.pdf

[Chris A]

I think the point of bringing up the IEC 60268-5 standards and the Klippel measurement procedures and systems is that there are measurements that aren't being reported, and I would guess aren't being performed by loudspeaker manufacturers either, and that should be measured.  Those measurements ARE useful. 

 

The fact that you can't find the measurements tells you where the bulk of the loudspeaker manufacturers (especially the small hi-fi/boutique shops) are in terms of engineering: it's not engineering but rather...well...  To strong-arm consumers of extremely expensive hi-fi loudspeakers into, "well, those measurements really don't help..." is to say "well, you consumers are idiots anyway, so we aren't going to let you know how bad our loudspeakers measure, because the other manufacturers aren't reporting it, either."  That brings me to my concise appraisal of those that say it isn't important...

 

 

Chris

Edited June 7, 2016 by Chris A

[DrWho]

Right, because sales and marketing is a complete waste of time and money. It's usually only the engineering types that have the right combination of arrogance and ignorance to think that engineering is the only thing that matters in this world. These same people often like to parade around using numbers to extol their subjective preferences and convince the world that they're further along than everyone else.

 

The fact that most audio companies don't publish every single measurable parameter of a device is because it is a horrible marketing strategy! Never mind the fact that it would be confusing to even the most educated of customers. Heck, it's confusing to the most educated of engineers too! Why the heck burden a sales force with such complicated subjects?

 

The exploration into the subjective correlation to objective measure will forever be a topic of intense research - and the experts in the field certainly are nowhere near agreement (not surprising given the subjective nature of the field). I hope you're pinning that BS Button to your shirt Chris because you're way out of your league here.

 

The other funny thing here is the arrogant-ignorant-number-bashing-idiots talk so much about removing bias from the listening/purchasing equation. An open-minded person would realize that the "non-audible" factors when listening actually have a huge impact on the audible impressions of the person. The audiophile boutique craziness exists because that very craziness enhances the listening experience! We should be embracing those realities, and even engineering them.....not running around beating our chests about how we're free from the vices of aesthetics and a proper human purchasing experience. The sales / business types totally understand this - the engineers are too proud to admit it.

 

None of that changes what measurements do tell us, but to get on this kick that somehow we can quantify the human experience is simply childish.

 

Btw, I do absolutely conduct listening tests where we try our best to remove the bias factor. It's very important in the design world to isolate the objective impact of a single variable. However, it's equally important to conduct the same listening tests when bias is a factor.

 

Music is art, not documentation.

[Ski Bum]

The audiophile boutique craziness exists because that very craziness enhances the listening experience! We should be embracing those realities, and even engineering them...

 

 

Like snake oil salesmen and homeopaths?  That aspect encapsulates everything that's wrong with the hobby, and many of us cringe at the pseudoscientific marketing BS and out right lunacy that goes on.  But whatever.  If self delusion get's you off, more power to you.  

 

None of that changes what measurements do tell us, but to get on this kick that somehow we can quantify the human experience is simply childish.

 

 

This is an illogical deflection.  The question has nothing to do with quantifying the human experience, but rather specific aspects of speaker performance which do indeed matter (to some of us at least).    

[Chris A]

Perhaps the numbers it generates are too complicated for the average consumer. Plus, without comparative numbers from competitors’ speakers, the numbers would be difficult to interpret...My gut tells me that distortion/compression vs. SPL would be very telling and relevant data when comparing speaker performance.

 

The aim of continuing the development and use of measurement standards is to arrive at useful and relevant measures that are practical and easy to interpret...at least it's been my experience in all the engineering disciplines that I've worked in. 

 

In this discipline (loudspeaker design and development), that list or group of standard measures can be used in conjunction with listening to form insights and associations with the measurements used either singly or used in groups--as you have proposed, above.  But first, the measurements must be made.  "Art" is something that is necessary, but I've found that reducing the "art" down as far as practicable to measurements is the reason why loudspeakers are developed differently than the way that they were done in the 1950s.  Measurement is the difference, and better measurements helps more. 

 

What you propose above can be done using REW right now and all it costs is a laptop, a calibration microphone (connected via USB or a mixer), some time and a little effort.  After running two sweeps at lower and higher SPL, then REW can plot the difference between any two traces quite easily, and normalize them.  I've done it before on my K-402 multiple entry horn design.  Works great.

 

Chris

Edited June 8, 2016 by Chris A

[DrWho]

Like snake oil salesmen and homeopaths?  That aspect encapsulates everything that's wrong with the hobby, and many of us cringe at the pseudoscientific marketing BS and out right lunacy that goes on.....  

 

I used to cringe about it a lot more in the past, but them crazies are doing things of great value that the cold-hearted engineering approach doesn't address. People buy into that stuff because it's the experience of it all that they're enjoying. Ya, it's often insane lunacy - but audio is the only hobby I've ever seen where the "purists" are so averse to addressing the rest of the listening experience.

 

For example, car racing enthusiasts add stickers to their car because "it adds horsepower". If anything, the stickers make the car slower, but they don't care...the fake authenticity of looking like a real sponsored racecar makes them enjoy the racing event more. And that totally fake feeling of authenticity actually improves the mood of the drivers and then they actually drive faster. It's all part of the entire experience. The difference here is the car enthusiast crowd doesn't have such a vocal group of people crying foul about someone that chooses the fake authenticity.

 

So instead of complaining about the snake oil insanity, I think we should be asking the "authentic solutions" to also provide the rest of the experience. Part of that experience might actually be not getting so hell bent on trying to quantify minutia that takes dozens of measurements and a college degree to understand.

 

 

 

Btw, CEA2010 is another fun spec that is more focused on the amplitude linearity of subwoofers:

http://www.audiomatica.com/wp/wp-content/uploads/appnote_007.pdf

 

And an interesting article getting to my point about how hard it is to have absolute measures of anything:

http://www.soundandvision.com/content/cea-2010-good-baseline-bass#FdPPIXRu4WYb84pW.97

 

 

 

 

Gain non-linearities as small as 0.3dB on some measurements are easily audible in blind listening tests. In others, the exact same measurement is totally inaudible. The reason is related to the mechanism that caused that non-linearity. Now go from one measurement rig to another.....those 0.3dB variations can measure quite differently - say now 0.4dB and happening slightly earlier now. Even with the same gear there will be variation - and sometimes that 0.3dB variation doesn't show up. Now I'd argue we're using the wrong measurement to identify the shortcoming of the system, but this is a common occurrence when analyzing gain / amplitude linearity. Some things are indicative of problems and other are not.

 

When looking for an amplitude linearity spec - how then does Manufacturer A differentiate their inaudible irregularity from Manufacturer B with a much more audible problem? From a sales and marketing perspective, neither manufacturer wants to call attention to the irregularity in the first place. And if they do, then the customers get confused and start distrusting science when they realize that A sounds better than B even though they have this measurement telling them "A is worse". You seriously need about 30 incredibly accurate measurements of a system before you can really start getting a feel for how it sounds - and even then you're filling in the blanks based on your knowledge of the topology.

 

The only clean non-technical story is to simply talk about the major design elements....like the topologies used. And that's exactly what we see in the audio industry. We talk about speakers that are hornloaded or ported, or amplifiers that are Class A or AB, etc....big picture topologies. Honestly, it matters way more than trying to throw measurements at something to understand why it sounds the way it sounds.....assuming of course it was designed properly. Granted, things often aren't designed and built properly, but that will come out in the listening anyway. At which point, do we really need measurements to tell us what our ears have already heard?

 

 

The reason I'm asking these questions and pushing down this path is because it really matters why you're looking for an amplitude linearity spec.

[Ski Bum]

I think you're being far too accommodating to the forces of evil, specifically brainless consumerism, but otherwise a meaty post.