• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Speaker Max SPL Criteria?

A Surfer

Major Contributor
Joined
Jul 1, 2019
Messages
1,125
Likes
1,230
Yikes, I would never intentionally test any of my speakers to see a max SPL. I might be being dramatic, but that sounds like a way to potentially damage and or shorten the operational lifespan. Can't theoretical models already tell us enough information that this type of thing can be estimated reliably enough?
 

Kvalsvoll

Addicted to Fun and Learning
Audio Company
Joined
Apr 25, 2019
Messages
878
Likes
1,643
Location
Norway
This is tricky.

For amplifiers, it is easy - set limit at 1% THD, and you will most likely hit the point where clipping just starts, and that is the limit. Or just look at the scope - easy to see where it clips.

Speakers usually do not have a sharp transition where it no longer generates more output, the sound just gradually gets more distorted, or something brakes.

Using limits for audibility for sine wave test signal will be wrong, as those limits are very hard to achieve even at modest spl. And distortion at lower frequencies is just as audible as for higher freqs.

For music, the limits are much higher, and very program dependent. Generally, distortion in the bass range can be quite high before it starts to sound horrible, and in the midrange and higher frequencies, limits can be much lower. I often observe the bass-system being overdriven far beyond its capacity, but you don't hear it. Also comparing bass-systems with very different distortion levels, generally does not reveal distortion from the lesser system - you can sense dynamic compression on transients, and perhaps low frequency roll-off, but it does not sound distorted.

What you want to find is the knee-point where distortion and compression starts to increase dramatically. The distortion level at this point can be different from different speakers - some may gradually increase distortion and have quite high distortion before giving up, while another may have low distortion until it suddenly rises. And some will break - mechanical failure, voice coil meltdown.

Frequency range is also important. A small bookshelf intended for modest spl and perhaps integrated with a separat bass-system should not be required to do low bass frequencies. And distortion limits should be frequency dependent.

Usually, the lower end of a speakers intended frequency range will be where capacity limits are first reached. If a satellite is to be used from 120hz up, it does not help if it can do 120dB at 1000Hz if it can do only 90dB at 120Hz - the limit will be 90dB.

For testing purposes, you need a defined and consistent and manageable set of rules, so all tested speakers are evaluated using the exact same limits, and the testing procedure is kept reasonably simple. A starting point could be to use the already established CEA limits, and just ignore the fact that limits should be more strict at higher frequencies.
 

TomJ

Active Member
Joined
Sep 16, 2019
Messages
129
Likes
177
Location
Palo Alto CA
Are simple (sinusoidal, steady state) test signals really sufficient to assess the performance of a loudspeaker? One way I evaluate the sound of a grand piano is to play a complex passage that requires the sustain pedal, and then listen to whether the individual voices remain clear or blur as the soundboard reverberates. It's surprising how much pianos can differ, eg when a piano has beautifully voiced hammers so each note sounds very good at its onset but quickly gets lost in the soundboard sustain. Inside the piano, the tone is produced by the harp (strings and brass plate) and reproduced by the soundboard with a complicated transfer function.

I've often wondered whether something analogous might be happening in speakers, who knows. Not trying to go OT, simply a comment on how complicated tone production and reproduction can get with a complex signal (like music) instead of a simple test signal.
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Seems to me the Klippel method with tone bursts is a good one.
 

MZKM

Major Contributor
Forum Donor
Joined
Dec 1, 2018
Messages
4,240
Likes
11,463
Location
Land O’ Lakes, FL
Whichever process you end up adopting, an additional “max SPL with sub” rating would be nice (perhaps defined as max SPL > 100Hz.
@hardisj does this (>80Hz).
Usually a decent difference unless it was simply compression limited and not distortion limited.

I would just do something simple. Wide band pink noise and stop when any threshold is met:
100Hz: -20dB (10%) THD
1kHz: -30dB (~3%) THD
10kHz: -40dB (~1%) THD

For any in-between frequency, just draw a straight line connecting all 3 points.
 
Last edited:

q3cpma

Major Contributor
Joined
May 22, 2019
Messages
3,060
Likes
4,416
Location
France
At least in the bass, your measurements and impressions seem to indicate that the audible limit is when H4 and H5 skyrocket (i.e. cross a certain threshold or the derivative of their curve does); of course, this just means the product reaches the audible range, so it depends on the frequency this phenomenon happens at.
 

okok

Senior Member
Joined
Oct 31, 2020
Messages
377
Likes
161
kh310 datas from neumann got this, max SPL is only at 1k or 2k, the FR like a peak, not flat anymore
 

thewas

Master Contributor
Forum Donor
Joined
Jan 15, 2020
Messages
6,747
Likes
16,182
Is there an identical one for IMD?
Unfortunately not that I would know about as I guess it would be too complex for a 2D chart as it depends also on the distribution of the multitones.
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
@hardisj does this (>80Hz).
Usually a decent difference unless it was simply compression limited and not distortion limited.

I would just do something simple. Wide band pink noise and stop when any threshold is met:
100Hz: -20dB (10%) THD
1kHz: -30dB (~3%) THD
10kHz: -40dB (~1%) THD

For any in-between frequency, just draw a straight line connecting all 3 points.

+1

Makes sense to be a bit conservative.
 

detlev24

Senior Member
Joined
Dec 3, 2019
Messages
305
Likes
291
@hardisj does this (>80Hz).
Usually a decent difference unless it was simply compression limited and not distortion limited.

I would just do something simple. Wide band pink noise and stop when any threshold is met:
100Hz: -20dB (10%) THD
1kHz: -30dB (~3%) THD
10kHz: -40dB (~1%) THD

For any in-between frequency, just draw a straight line connecting all 3 points.
Why 1% THD at 10 kHz [@hardisj]? Our hearing is significantly less sensitive roughly >6 kHz, if we refer to this chart. IMO it would only make sense if THD tolerance would be higher, where ear sensitivity is lower.

Just a gedankenexperiment, without any proof for accuracy of the numbers - reference of the frequency bands taken from the aforementioned chart:

Sub Bass ≤ 60 Hz : -14 dB (~20%) THD
Bass ≤ 250 Hz : -20 dB (10%) THD
Midrange ≤ 2 kHz : -30 dB (~3%) THD
High Mids ≤ 6 kHz : -40 dB (1%) THD
High Freqs ≤ 20 kHz : -26 dB (~5%) THD

In an acoustically untreated listening environment, I guess, this would still be quite conservative especially 'Sub Bass', where I would expect ~30% THD to be acceptable at a given target SPL.
 
Last edited:

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
Why 1% THD at 10 kHz [@hardisj]? Our hearing is significantly less sensitive roughly >6 kHz, if we refer to this chart. IMO it would only make sense if THD tolerance would be higher, where ear sensitivity is lower.

Just a gedankenexperiment, without any proof for accuracy of the numbers - reference of the frequency bands taken from the aforementioned chart:

Sub Bass ≤ 60 Hz : -14 dB (~20%) THD
Bass ≤ 250 Hz : -20 dB (10%) THD
Midrange ≤ 2 kHz : -30 dB (~3%) THD
High Mids ≤ 6 kHz : -40 dB (1%) THD
High Freqs ≤ 20 kHz : -26 dB (~5%) THD

In an acoustically untreated listening environment, I guess, this would still be quite conservative especially 'Sub Bass', where I would expect ~30% THD to be acceptable at a given target SPL.

The thing is, the chart illustrates 2nd harmonic distortion audibility, but the problem with trying to quantify distortion audibility in terms of dB and percent is that the frequency of the distortion in relation to the signal is actually more important that the level of the distortion, then (as you pointed out) the sensitivity of our hearing is also frequency and level dependent, and finally, auditory masking effectiveness is massively dependent on level.

Correlating the objective and the subjective is very difficult when it comes to distortion. We need a lot of measurements and a very complex and nonlinear correlation function, since to be effective it must take into account, at least in a rudimentary way:
  • the frequency relationship of the distortion to the signal
  • the highly nonlinear nature of auditory masking
  • the nonlinear relationship between frequency, level and audibility
I think we’re kidding ourselves here if we think we can come up with an adequate metric. Much better IMO to either:
  1. Continue with our basic measurements and not assume too much can be read into them;
  2. Expand our testing to include max. SPL and power compression, which at least gives us an indication of how loud the speaker can play (if not how nice or nasty it will sound on the way there); or
  3. Adopt a metric for which there is already a solid theoretical and scientific basis. In this case I’d nominate one of the de Santis metrics, which are the only ones I know of that have achieved high levels of correlation in controlled studies (albeit ones conducted by the author of the metric himself). (Geddes and Lee would be another possible contender that may be simpler to access/apply).

Sorry @detlev24, obviously very little of that post was actually directed at you :)
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
The thing is, the chart illustrates 2nd harmonic distortion audibility, but the problem with trying to quantify distortion audibility in terms of dB and percent is that the frequency of the distortion in relation to the signal is actually more important that the level of the distortion, then (as you pointed out) the sensitivity of our hearing is also frequency and level dependent, and finally, auditory masking effectiveness is massively dependent on level.

Correlating the objective and the subjective is very difficult when it comes to distortion. We need a lot of measurements and a very complex and nonlinear correlation function, since to be effective it must take into account, at least in a rudimentary way:
  • the frequency relationship of the distortion to the signal
  • the highly nonlinear nature of auditory masking
  • the nonlinear relationship between frequency, level and audibility
I think we’re kidding ourselves here if we think we can come up with an adequate metric. Much better IMO to either:
  1. Continue with our basic measurements and not assume too much can be read into them;
  2. Expand our testing to include max. SPL and power compression, which at least gives us an indication of how loud the speaker can play (if not how nice or nasty it will sound on the way there); or
  3. Adopt a metric for which there is already a solid theoretical and scientific basis. In this case I’d nominate one of the de Santis metrics, which are the only ones I know of that have achieved high levels of correlation in controlled studies (albeit ones conducted by the author of the metric himself). (Geddes and Lee would be another possible contender that may be simpler to access/apply).

Sorry @detlev24, obviously very little of that post was actually directed at you :)
http://hifisonix.com/wordpress/wp-content/uploads/2017/11/Perceptual-Levels-of-distortion.pdf

I assume this is the de Santis metric you are referring to?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,523
Likes
37,056
Yes, that one.
We just need to get @pkane to incorporate it into Deltawave. He can provide distortion upon request using his Distort software, and if Deltawave could compute the DS metric value it would make all this easier to do. Of course easy for me to say since Paul would do all the work. I think all the needed formulas for the DS are in that paper.
 

TomJ

Active Member
Joined
Sep 16, 2019
Messages
129
Likes
177
Location
Palo Alto CA
Continue with our basic measurements and not assume too much can be read into them...

Reminds me of a comment years ago by a physics professor who said “trust your measurements, but never trust that your measurements fully describe the system”.

Thanks for the de Santis & Henin 2007 ref, see also Tan et al 2004 (pdf download at www.aes.org/journal/online/JAES_V52/7_8/). Yes the DS metric looks promising, likewise Rnonlin if feasible (https://www.mathworks.com/matlabcentral/fileexchange/50230-rnonlin_calc).
 
Last edited:

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
We just need to get @pkane to incorporate it into Deltawave. He can provide distortion upon request using his Distort software, and if Deltawave could compute the DS metric value it would make all this easier to do. Of course easy for me to say since Paul would do all the work. I think all the needed formulas for the DS are in that paper.

That would be great... @pkane? ;)

Thanks for the de Santis & Henin 2007 ref, see also Tan et al 2004 (pdf download at www.aes.org/journal/online/JAES_V52/7_8/).

Yes, that's the one! Perhaps I was mistaken in thinking de Santis developed it. Had been a while since I'd read the studies.

Yes the DS metric looks promising, likewise Rnonlin if feasible (https://www.mathworks.com/matlabcentral/fileexchange/50230-rnonlin_calc).

Ah! Hadn't realised there was a calculator available :D
 

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
Since I was mentioned, I have been doing this for a while now. Here's the breakdown.
  1. I do perform IMD testing but I do that only on driver units because that doesn't provide me with a max SPL. It just provides me with a distortion profile.
  2. To get max SPL I use a dedicated test for that. I perform max SPL for both drive units and loudspeakers.


I perform the tests differently. With drive units, my max SPL thresholds are tighter than they are with speakers. Why? Because a loudspeaker has crossovers implemented and raw drive units are more "wide open" in my testing. Still, I do perform a "full" and "extended" set of measurements for both. All of this is documented in my reviews.


I'm going to split the rest of my reply up in to separate posts following the above.
 
Last edited:

hardisj

Major Contributor
Reviewer
Joined
Jul 18, 2019
Messages
2,907
Likes
13,908
Location
North Alabama
1) Raw Drive unit testing with both IMD and max SPL. Probably not of concern here since ASR doesn't test raw drivers, but it's a point of data and shows the difference in how I test drive units vs loudpseakers.




I'm going to quote myself. This is from the Purifi 6.5" midwoofer test I did a couple months back:
https://www.erinsaudiocorner.com/driveunits/purifi_ptt65w04/

Intermodulated Distortion (IMD):
Klippel’s 3D-DISTORTION MEASUREMENT (DIS) module is used to calculate the Intermodulated Distortion for this drive unit.

Measurements were completed in the nearfield. Multiple output levels were tested to provide the trend of distortion component profiles and to provide a comparison against other drive units I have tested. The SPL provided is relative to 1 meter distance, averaged in the noted bandpass region.

Unlike Harmonic Distortion - which is a measure of only harmonics from a single tone - IMD is tested with two tones at the same time: a low frequency “bass tone” near Fs and a higher frequency “voice tone” much greater than Fs. For example a speaker driver plays 30Hz at the same time it plays 200Hz. Any distortion artifacts created by the sum and/or difference of the speaker playing both tones at the same time is a result of IMD. In this example, the speaker is supposed to play only 30Hz & 200Hz at the same time. Thanks to IMD, it plays 200Hz ± 30Hz. Second order IMD would be 200 ± 30Hz = 170Hz & 230Hz. Third order IMD would be 200 ± 60Hz = 140Hz & 260Hz. If these ‘side bands’ are high enough in output then they are heard as distortion. If not, they are not.

With that in mind, I used Klippel’s DIS module to test this drive unit in two ways:

  1. “Bass tone” fixed at the driver’s Fs, with a “voice sweep” where the 200Hz - 6kHz region is played, one tone at a time (over 50 individual tones).
  2. The same as above, but the “bass tone” fixed at 80Hz.
The purpose of me testing with two methods (different “bass tones”) is to see the difference between what happens when the driver plays with high(er) excursion vs when a typical HPF is used. All similarly sized and similarly purposed speakers are tested in the same manner. For better or worse. This means a 6-inch midwoofer is tested the same way an 8-inch midwoofer is. Ultimately, this is for my sanity, because having numerous measurement methods for all sizes of speakers would muddy the waters quickly and wouldn’t give us an idea of when performance is great (say, a 6-inch midwoofer that has much less distortion than an 8-inch) or vice-versa.

The above is tested at 3 voltages each. The first voltage is always 2.83v. The other two voltages are increased to provide higher output (usually targeting at least 96dB for one). As is the case with the multiple HD tests, the multiple IMD levels provides an idea of what the speaker’s IMD profiles look like as the output of the speaker is increased.

Results are provided in GIF form to help understand how the increased output levels impact the distortion profiles and levels. Also, I have provided the (calculated) excursion at the provided tests’ output level so you can see how much excursion the speaker is under at the bass tone. Naturally, with a lower frequency there is higher excursion than with a higher frequency bass tone. This also helps when you are comparing performance against another drive-unit because different speakers have different response profiles; one drive unit may have 2x the excursion from 30Hz to 80Hz while another may have 3x the excursion. Just something to keep in mind.

Test 1: Bass Tone at Fs, Voice Tones from 200-6000Hz

Purifi%20PTT6.5W04_IMD_Fs.gif






Test 2: Bass Tone at 80Hz, Voice Tones from 200-6000Hz

Purifi%20PTT6.5W04_IMD_HPF.gif







Maximum Long Term SPL (Multitone Distortion Testing) & Compression
Klippel’s Multi-Tone Measurement (MTON) module is used to calculate the maximum SPL for this drive unit.

Unlike IMD testing, Multitone testing incorporates a complex stimulus, used to emulate actual music. However, it is a bit more limited in evaluation purposes so I am using it here to characterize the maximum SPL of a drive unit at 1 meter distance as well as provide how much compression it exhibits with an increase in voice coil temperature.

The below data provides the metrics for how Maximum Long Term SPL is determined. This measurement follows the IEC 60268-21 Long Term SPL protocol, per Klippel’s template, as such:

  • Rated maximum sound pressure according IEC 60268-21 §18.4
  • Using broadband multi-tone stimulus according §8.4
  • Stimulus time = 60 s Excitation time + Preloops according §18.4.1
Each voltage test is 1 minute long (hence, the “Long Term” nomenclature).

The thresholds to determine the maximum SPL are:

  • -30dB Distortion relative to the fundamental
  • -2dB Compression relative to the reference (1V) measurement

When the speaker has reached either or both above thresholds, the test is terminated and the SPL of the last test is the maximum SPL. In the below results, the legend indicates the average SPL within the same passband as is provided for the Frequency Response measurements. The highest SPL in the legend is the maximum SPL referencd to 1 meter distance. I provide the data showing which threshold was exceeded. The graphic(s) with a red line indicate the threshold that failed and capped the maximum SPL.


This measurement is conducted twice (with a 30-minute break between to let the voice coil cool down):

  1. “Typical” - 80Hz to 1600Hz
  2. “Extended” - 40Hz to 3200Hz
The purpose of me testing with two methods is to see how a driver performs in a more “typical” passband vs when it plays an “extended” passband. All similarly sized and similarly purposed speakers are tested in the same manner. For better or worse. This means a 6-inch midwoofer is tested the same way an 8-inch midwoofer is. Ultimately, this is for my sanity, because having numerous measurement methods for all sizes of speakers would muddy the waters quickly and wouldn’t give us an idea of when performance is great (say, a 6-inch midwoofer that has much less distortion than an 8-inch) or vice-versa.

You can watch a demonstration of this testing via my YouTube channel:


Test 1: Typical

Multitone compression results.

Purifi%20PTT6.5W04_MTON_Compression%201.png




Multitone distortion results.

Purifi%20PTT6.5W04_MTON_Distortion%201.png






Test 2: Extended

Multitone compression results.

Purifi%20PTT6.5W04_MTON_Compression%202.png




Multitone distortion results.

Purifi%20PTT6.5W04_MTON_Distortion%202.png






The maximum SPL referenced to 1 meter for each test are:

  • Max SPL for Test 1 is 100.6dB. The compression threshold was exceeded above this SPL.
  • Max SPL for Test 2 is 100.6dB. The compression threshold was exceeded above this SPL.
Does this mean the speaker will not play above this level? No. It simply means, within a set of limits, the above values are the maximum the speaker output is. Above that SPL, the limits are further and further exceeded.
 
Last edited:
Top Bottom