Is nonlinear distortion of (small) speakers unimportant??

[pma]

Is nonlinear distortion of (small) speakers unimportant?

We can see plenty of measurements of very small monitors here at ASR, in last weeks. And we often read that nonlinear distortion of speakers is unimportant and the only important parameter is frequency response, on axis of axis at different angles etc., in other words power response. Distortion is said to be unimportant. Really?? As this forum is not interested at listening impressions at all, only in measured facts, I am posting some facts.

Two speakers were compared in distortion in higher bass region, CNO-T25 2-way small floor-standing project by Troels Gravesen,

http://www.troelsgravesen.dk/CNO-T25.htm

and JBL Control 1 Pro small monitor

https://eu.jbl.com/CONTROL+ONE.html

CNO-T25 has 6.5” Seas W18NX001 woofer/midrange driver, though JBL has a 4” driver. Does it make a difference? Yes, a huge one. 6.5” is still a very small woofer, but let's see the measurements.

JBL specs say:
Frequency Range (-10 dB): 80 Hz - 20 kHz
Frequency Response (+/- 3 dB): 100 Hz - 18 kHz
Power Capacity: 150W

So I made a test at 2Vrms output amplifier voltage approx. (1W for JBL), at 100Hz sine and 200+300Hz twin tone to see IMD at lower frequencies. Both speakers were measured from the same distance. 1W should be fine for a unit that has “Power Capacity” of 150W, isn't it? No, it is not fine. The speaker distorts as a hell at lower frequencies and it is audible even in the second room where is my PC and measuring SW. But who cares about listening, we want data.

JBL Control 1 Pro distortion spectrum at 100Hz

CNO-T25 distortion at 100Hz

JBL Control 1 Pro 200+300Hz IMD. The twin tone from this speaker does not sound like a twin tone, but like some cacophony.

CNO-T25 200+300Hz IMD

We can see that at 100Hz the distortion of JBL Control 1 Pro is simply unacceptable (and it is only at 1W!!!) and for 200+300Hz twin tone there is a plenty of high order harmonics that are much more audible than the low order harmonics.

My conclusion
Small speakers are toys that are not worth serious listening. I am looking forward the day when ASR starts to test something that could be used for normal listening conditions. So far we could see the small plastic things sometimes at enormous price like $6000 in the current review, with just few exceptions.

 

[miero]

Distortion measurements are already available in reviews:
- https://www.audiosciencereview.com/forum/index.php?attachments/genelec-8341a-sam™-studio-monitor-powered-speaker-spl-and-distortion-percentage-audio-measurem-png.51409/
- https://www.audiosciencereview.com/...aker-review.11566/#lg=attachment50850&slide=0
- https://www.audiosciencereview.com/...tion-percentage-audio-measurements-png.49063/
- . . .

the small plastic things . . .

Sometimes a die-cast aluminium looks like a plastic too ;-)

 

[maxxevv]

You have conveniently cited one of the least desirable examples to have come through to date.

So whats the point of doing that when there are others that measure (far) better ?

No big speakers?? Best you look through the list again.

 

[pma]

Distortion measurements are already available in reviews:
- https://www.audiosciencereview.com/forum/index.php?attachments/genelec-8341a-sam™-studio-monitor-powered-speaker-spl-and-distortion-percentage-audio-measurem-png.51409/
- https://www.audiosciencereview.com/...aker-review.11566/#lg=attachment50850&slide=0
- https://www.audiosciencereview.com/...tion-percentage-audio-measurements-png.49063/
- . . .


Sometimes a die-cast aluminium looks like a plastic too ;-)

With Y-axis distortion % in linear scale? It gives no view, distortion in % should be shown in Y-axis log scale. And spectra are much more illustrating of high order harmonics. This is what you hear, the ear would tolerate quite high level of low order harmonics. Numbers itself or linear Y-axis scale are not much helpful.

 

[MZKM]

It’s not not a factor, but it seems that in most cases, it’s not such a huge factor that you should choose a worse measuring speaker because it has lower THD.

 

[tuga]

Neumann talks about intermodulation distortion in loudspeakers in this piece.

 

[Jon AA]

Is nonlinear distortion of (small) speakers unimportant?.....

My conclusion
Small speakers are toys that are not worth serious listening. I am looking forward the day when ASR starts to test something that could be used for normal listening conditions. So far we could see the small plastic things sometimes at enormous price like $6000 in the current review, with just few exceptions.

While I'm not sure your methodology was the most appropriate, and I would note not only Amir's measurements but often his listening tests mention when speakers can't get loud and/or start sounding like crap when they do, I do sort of agree with your premise. Not with Amir specifically (the spin data is so much more valuable and difficult to get, I'd be happy if that was all he spent his time focusing on) but more for the industry in general.

Nobody in the industry seems to do reviews of speakers which include distortion and compression testing at high volumes to determine the maximum "usable" SPL of which a particular speaker is capable. This really leaves the consumer blind to a pretty important aspect of a speaker's performance. People are left to look at the sensitivity and max power handling, do a little math and make a guess. Sometimes a max SPL is given in the specs, but I'd venture to guess the vast majority of consumer speakers start really sounding like crap long before they reach the number listed in their specs. The Pro Audio side of the industry does a better job of providing this data, but most consumers never give such speakers much consideration.

I can understand why reviewers don't do it. If a certain reviewer gets a reputation for sending back speakers with blown tweeters, manufacturers will stop sending speakers to him. So that's a good reason for it, but not a solution. And I don't expect manufacturers to be helpful in this regard either, as the current trend of trying to get the lowest bass extension out of the smallest box possible runs contrary to that design goal, putting them at a fundamental disadvantage, so they're probably happy no reviewers actually push the capability (or lack thereof) of the speakers they're reviewing.

 

[QMuse]

Just a few measurment hints at the beginning:

- as speaker's sensitivity varies you should adjust power so that average SPL is say 90dB at 2m distance instead of using fixed power
- y log scale is ok but it would be better to have it in SPL dB

Now to the point.. hearing distortion of sine wave is very much different then hearing distorion with music. While we don't have much trouble detecting THD>0.5% even with low freqs it is not that simple with music because of masking. For that reason it turns out that THD is actually not a good metric for our perception of distortion with music. Earl Geddes and Lidia Lee have done very nice research on that topic and they proposed new metrics, more info can be found in their AES papers, Part I and Part II.

On this link you will find discussion and download link for a XLS calculator based on their metric.

 

[pma]

Now to the point.. hearing distortion of sine wave is very much different then hearing distorion with music. While we don't have much trouble detecting THD>0.5% even with low freqs it is not that simple with music because of masking. For that reason it turns out that THD is actually not a good metric for our perception of distortion with music. Earl Geddes and Lidia Lee have done very nice research on that topic and they proposed new metrics, more info can be found in their AES papers, Part I and Part II.

On this link you will find discussion and download link for a XLS calculator based on their metric.

I know Earl Geddes work and papers very well and I have never spoken about THD as a single number as being much important. All I have shown are measured spectra on 2 speakers under same conditions, only 4 significant plots shown from a huge number collected. There are 2 points -

1) distortion of the smaller speaker at lower frequencies is unacceptable in the frequency band where many acoustic instruments have enough energy.
2) the smaller speaker not only has higher distortion, but, it has tons of high order distortion components in the spectrum. So it is not about THD number, but about distortion profile (as is also mentioned by Earl Geddes). The problem I have with most published speaker measurements is that they mostly do not have enough resolution to show high order harmonics. And in case that linear scale on Y axis is used, then it is totally useless, as you would read nothing below some -40dBr. Yes H2 of -40dBr is not important, but H5 or H7 down to -60dBr is important. Conventional measurements are fine, but they should be accompanied by high resolution distortion measurements. Tell me, why do we inspect -120dBr distortion components on DACs and amplifiers and do not care about proper view to speaker distortion spectrum? Yes it is more complicated, reflections tend to devalue the result, however though it is difficult, it should not be omitted.
The forest of high order harmonics and intermodulations is audible not only on sines, but also on music signal.
------------------------------------
P.S.: I am ignoring and not replying to the posts that have no technical content, like #3 in this thread.
https://www.audiosciencereview.com/...-small-speakers-unimportant.11659/post-335633

 

[QMuse]

I know Earl Geddes work and papers very well and I have never spoken about THD as a single number as being much important. All I have shown are measured spectra on 2 speakers under same conditions, only 4 significant plots shown from a huge number collected. There are 2 points -

1) distortion of the smaller speaker at lower frequencies is unacceptable in the frequency band where many acoustic instruments have enough energy.
2) the smaller speaker not only has higher distortion, but, it has tons of high order distortion components in the spectrum. So it is not about THD number, but about distortion profile (as is also mentioned by Earl Geddes). The problem I have with most published speaker measurements is that they mostly do not have enough resolution to show high order harmonics. And in case that linear scale on Y axis is used, then it is totally useless, as you would read nothing below some -40dBr. Yes H2 of -40dBr is not important, but H5 or H7 down to -60dBr is important. Conventional measurements are fine, but they should be accompanied by high resolution distortion measurements. Tell me, why do we inspect -120dBr distortion components on DACs and amplifiers and do not care about proper view to speaker distortion spectrum? Yes it is more complicated, reflections tend to devalue the result, however though it is difficult, it should not be omitted.

I know Earl Geddes work and papers very well and I have never spoken about THD as a single number as being much important. All I have shown are measured spectra on 2 speakers under same conditions, only 4 significant plots shown from a huge number collected. There are 2 points -

1) distortion of the smaller speaker at lower frequencies is unacceptable in the frequency band where many acoustic instruments have enough energy.
2) the smaller speaker not only has higher distortion, but, it has tons of high order distortion components in the spectrum. So it is not about THD number, but about distortion profile (as is also mentioned by Earl Geddes). The problem I have with most published speaker measurements is that they mostly do not have enough resolution to show high order harmonics. And in case that linear scale on Y axis is used, then it is totally useless, as you would read nothing below some -40dBr. Yes H2 of -40dBr is not important, but H5 or H7 down to -60dBr is important. Conventional measurements are fine, but they should be accompanied by high resolution distortion measurements. Tell me, why do we inspect -120dBr distortion components on DACs and amplifiers and do not care about proper view to speaker distortion spectrum? Yes it is more complicated, reflections tend to devalue the result, however though it is difficult, it should not be omitted.

Small monitor and decent floorstander from the same line of products usually don't differ in lower LF -3dB and -10dB limits that much so I have seen many folks asking themselves why pay more for the floostander. Where they do differ however is in their ability to play loud bass. Decent floorstander will deliver 40Hz tone at 4m at 105dB with less than 1% of THD while smaller speaker will usually deliver hugely distorted sound, assumming of course they can deliver a 40Hz base tone at 105dB at 4m.

The forest of high order harmonics and intermodulations is audible not only on sines, but also on music signal.

Hm.. I don't remember seeing much research supporting that statement. FToole actually claims distortion is not a problem with mdoern speakers and has dedicated only a 2 pages to that topic. Do you know of some other research supporting your statement?

 

[pma]

Hm.. I don't remember seeing much research supporting that statement. FToole actually claims distortion is not a problem with mdoern speakers and has dedicated only a 2 pages to that topic. Do you know of some other research supporting your statement?

Certainly, and you will find numerous papers just using Google search engine.

https://www.audioholics.com/loudspeaker-design/audibility-of-distortion-at-bass/nonlinear-distortion

https://www.researchgate.net/public..._nonlinear_distortion_on_personal_sound_zones

https://pdfs.semanticscholar.org/2298/f8d16cfd5140002da418511204358b87de45.pdf

 

[pozz]

There are different perceptual mechanisms at play for music vs. pure tones (see Juan Roederer's Physics and Psychophysics of Music). I'm not completely clear on why, but there a few things that contribute:

• Saturation of neurons and their response/recovery times (temporal/level-based masking).
• Each frequency sensation excites more than just one area in the basilar membrane (i.e., auditory filter shape).
• Perception of added distortion tones with music, which has broad spectral content, is best answered by the theory of pitch sensation. When tones are close together they aren't perceived as separate tones, but as a timbral modification known as roughness. If close enough together the perception will be that the main tone, the fundamental, is louder. Then there are thresholds involved: as tones move apart in frequency, there is a sudden transitional point where the sensation moves from one rough tone to two separate tones—this is why high-order distortion is more perceivable than low-order. This means that with contrived signals very low distortion can be perceived, particularly if you are trained and know what to listen for. But with music the thresholds become elevated.

It's also the case that accurate distortion measurements are not possible without an anechoic chamber due to the contribution of room noise.

Small speakers are toys that are not worth serious listening. I am looking forward the day when ASR starts to test something that could be used for normal listening conditions. So far we could see the small plastic things sometimes at enormous price like $6000 in the current review, with just few exceptions.

I find this very rude and not in at all in line with the technical, academic tone you want to set for the discussion. There's no reason not to test cheap speakers since they provide valuable information about the engineering quality in the market as a whole, regardless of your goals as a listener or designer.

 

[QMuse]

Certainly, and you will find numerous papers just using Google search engine.

https://www.audioholics.com/loudspeaker-design/audibility-of-distortion-at-bass/nonlinear-distortion

https://www.researchgate.net/public..._nonlinear_distortion_on_personal_sound_zones

https://pdfs.semanticscholar.org/2298/f8d16cfd5140002da418511204358b87de45.pdf

None of those 3 articles don't mention conducting blind tests to establish audible threshold of distortion nor do they propose a metric to measure it which correlates with blind test results.

 

[napilopez]

I'd also contest the point that listening tests are irrelevant here as speaker measurements are only valuable insofar as we can correlate them to listening tests.

Put another way, I'd argue everyone here knows that distortion matters to some extent! The problem is that 'some' extent. The audibility of frequency response problems and how they relate to preference is very well documented. Meanwhile speaker designers and acousticians are still trying to figure out how much distortion actually matters in music and the best way to measure it in a relevant manner.

The onus is still on the distortion camp to prove that distortion is more of a problem than frequency domain issues.

Also no need for the hate on small speakers, especially when they're meant to be crossed with a sub. I'll still take a small speaker with a sub, or something like the Devialet phantom reactor which can actually hit 20hz at useful SPLs on its own, before a larger speaker with less distortion but less extension. Besides, not everyone is looking for speakers that can fill a living room.

If you don't need the SPL capabilities of a large speaker and have enough headroom on your small ones, you aren't going to gain much from going bigger. I've heard many small speakers that sound better than big speakers in my own home.

 

[QMuse]

Put another way, I'd argue everyone here knows that distortion matters to some extent! The problem is that 'some' extent.bigger.

Before defining an audible extent (threshold) you need to define a relevant metric, and as it seems THD is not it.

 

[pma]

There are different perceptual mechanisms at play for music vs. pure tones (see Juan Roederer's Physics and Psychophysics of Music). I'm not completely clear on why, but there a few things that contribute:

• Saturation of neurons and their response/recovery times (temporal/level-based masking).
• Each frequency sensation excites more than just one area in the basilar membrane (i.e., auditory filter shape).
• Perception of added distortion tones with music, which has broad spectral content, is best answered by the theory of pitch sensation. When tones are close together they aren't perceived as separate tones, but as a timbral modification known as roughness. If close enough together the perception will be that the main tone, the fundamental, is louder. Then there are thresholds involved: as tones move apart in frequency, there is a sudden transitional point where the sensation moves from one rough tone to two separate tones—this is why high-order distortion is more perceivable than low-order. This means that with contrived signals very low distortion can be perceived, particularly if you are trained and know what to listen for. But with music the thresholds become elevated.

It's also the case that accurate distortion measurements are not possible without an anechoic chamber due to the contribution of room noise.

I find this very rude and not in at all in line with the technical, academic tone you want to set for the discussion. There's no reason not to test cheap speakers since they provide valuable information about the engineering quality in the market as a whole, regardless of your goals as a listener or designer.

Thank you for the reply. Many studies do show importance of nonlinear distortion to music perception, the research is not restricted to Toole + Olive, which are the most cited names here. And the audibility is firmly influenced by music signal structure.
For sines, it is just masking, known well for 70 - 80 years. The distorted output may be easily recorded and compared each other or with a pure tone, in a DBT, as we like.

"It's also the case that accurate distortion measurements are not possible without an anechoic chamber due to the contribution of room noise." - and that's why I am measuring distortion in near field. I expect the objection of microphone distortion - then please take into account that at the same SPL from 2 speakers and same microphone position one had 3% distortion (+ plenty of high order harmonics) at 100Hz and the other one had 0.2% (+ few higher order harmonics). This would not be called as a result of microphone contribution, at least I would hope so.

Re what is rude - I should not comment but this atmosphere is not created by members.

 

[pma]

Before defining an audible extent (threshold) you need to define a relevant metric, and as it seems THD is not it.

Why do almost anyone speak about THD. THD is just a single number, and is quite pointless. Similar as SINAD, which is also a pointless single number. On the other hand, distortion spectrum is very revealing and giving a good picture about a tested object, if performed at many frequencies and signal levels.

 

[miero]

It would be interesting to determine a lower bound of a frequency where high order (3rd and above) distortion components fall under a "safe" threshold, for example 60dB.

 

[QMuse]

Why do almost anyone speak about THD. THD is just a single number, and is quite pointless. Similar as SINAD, which is also a pointless single number. On the other hand, distortion spectrum is very revealing and giving a good picture about a tested object, if performed at many frequencies and signal levels.

When I say THD I'm referring to a distortion spectrum measured at various SPLs and not to a single number, but based on the research done by Geddes and others I don't agree that distorion spectrum is a metric which correlates well with blind tests. You failed to provide a single research to support your statement.

 

[LTig]

Do you know how much of the single distortion products in your plots is created by the measurement rig (mic, preamp, ADC)?