Using Distortion Measurement to Compare Speakers

[redboat77]

Can anyone comment on how important harmonic distortion measurements are in comparing speakers?

For example, @amirm recently posted harmonic distortion measurements on ASR of the Revel F208, copied below, and these appear to compare favorably to measurements of the Philharmonic BMRs posted here and in-line below. For example, at 2000 Hz and 96 dB @ 1 meter, the F208s show a THD of less that 0.1%, while the Philharmonic BMRs are closer to 2% at similar sound levels. This is a 20x difference. Yet both speakers receive very strong positive reviews.

Is harmonic distortion even relevant when comparing two speakers? How?



Revel F208

Philharmonic BMR

 

[posvibes]

May I ask a more rookie question in regards to distortion?

As I have stated a few times I tend to listen a relatively modest levels and have done for many years for various reasons. I have had quite a few pairs of speakers come and go and some I have held onto and generally I have liked all of them for their various ways of presenting performance. I have in my memory never found a pair of speakers that I outright disliked and only in comparison with other speakers have made note of their weaknesses according to my preferences.

This is also the case for headphones.

I have had the fundamentally the same room for 20 odd years and never found the room to be a culprit for bad performance.

I understand that at certain volume levels room modes can be excited/activated etc.

Would I be correct to assume that most speakers @amirm has measured would measure acceptably better at lower volumes than the required 96db? I know that Speaker manufacturers usually state the parameters of their speakers in regards to minimum required amplifier power and minmum/maximum sound power levels etc, but if the listening measurement level is well below the test standard of 96db would speakers perform measurably much better?

I suppose I am trying to evaluate as to whether I have any critical faculties at all and that I have not a clue for what to listen for to evaluate a speaker, although I can hear differences in presentation I am unable to judge in the least as to what is good, bad or indifferent?

I may actually be the most undiscerning audiophile on the planet.

 

[amirm]

I understand that at certain volume levels room modes can be excited/activated etc.

No, room modes are activated at all listening levels. The physics only depends on the reflection path/dimension and what absorption you have. Perceptually though, our low frequency thresholds are different based on volume but still, you have definitely have room modes to fix regardless of your listening levels.

 

[amirm]

Would I be correct to assume that most speakers @amirm has measured would measure acceptably better at lower volumes than the required 96db?

I perform my speaker measurements at 86 dBSPL. This is loud enough to overcome ambient noise without stressing all but the tiniest speakers. I then run a secondary sweep at 96 dBSPL for distortion. So you have data points at both levels.

 

[posvibes]

@amirm, thanks for that it only goes to prove my underlying anxiety that I certainly am a cloth eared listener. I knew I wasn't a pair of golden ears and I never felt comfortable with the word audiophile, on the positive side I am easily pleased/placated.

Great site keep up the great work. despite everything I am chuffed with the LSR305p's, and my Stax Lambda's gave up the ghost after 30 odd years of service and I have ordered a pair of Hi-Fi Man 400x to replace them at a fraction of the original cost of the Stax's I bought when I was literally a young man.

I think that should see me out as far as Hi-Fi is concerned unless a lotto win comes along and then I will consider listening to a pair of Revel's.

Thanks again

 

[Sancus]

Is harmonic distortion even relevant when comparing two speakers? How?

BMR may or may not have audible midrange distortion issues, it's noted in the review: "I did hear some things that sounded different and now having seen the data, I wonder if the increased harmonic distortion between 2-4kHz might have been a contributor but I cannot say with any degree of certainty that it was."

You need to remember that 96db @ 2khz is extremely loud. Just about all music is down at least -10dB by that frequency from peak(which is typically around 100hz but always below 1khz).

In most cases, to hear 2khz @ 96dB you need to be playing at 110dB+ which is too loud for the BMR in general so it's rather a moot point.

So yes, distortion is relevant, but you need to put it in the context of how loud you'll actually be playing and what the speakers are capable of below 1khz, because that is where most SPL demands come from.

 

[tuga]

Can anyone comment on how important harmonic distortion measurements are in comparing speakers?

It would help a lot if the vertical scale on those graphs were standardised.

 

[redboat77]

"In the general population of consumer loudspeakers, it has been very rare for distortion to be identified in the overall subjective ratings. This is not because distortion is not there or it is not measurable, but it is low enough that it is not an obvious factor in judgements of sound quality at normal background listening levels". Floyd E Toole, Sound Reproduction, Loudspeakers and Rooms, Focal Press, 2008 (page 453).

In terms of nonlinear distortion, my guess is that it becomes important when the resulting harmonic and intermodulation distortion products are significant relative to the room background levels (typically 30 dB - 50 dB) and not masked by the original source. In particular, one concern would be the mapping of energy out of the band where the source material lies into other audible frequencies where masking does not occur. In this regard, nonlinearities that are discontinuous (or that just have large derivatives) would be of particular concern. Of course, this applies to the entire signal chain, and not just the speakers. Unfortunately, THD measurements, which are sums of harmonics generated by a single tone and are relatively easy to perform, are not a particularly good measurement of this since all harmonics are weighted equally, and smooth nonlinearities tend to generate most of the distortion in the immediate 2nd and 3rd harmonics, not in the higher harmonics that are less subject to masking, and which are therefore more likely to be perceived. Hence smooth nonlinearities could generate high THD yet still not be perceived, while a sharp nonlinearity could generate a relatively low THD and be much more objectionable to a listener. I assume this explains the preference some have for the smooth nonlinearities of tube amplifiers, despite their high THD ratings. THD does actually provide a good metric of the distortion we hear; but it's even worse than that since it tends to overemphasize the distortion we do not perceive. The paper by Lee and Geddes (L.W. Lee & E.R. Geddes, Auditory Perception of Nonlinear Distortion, AES 115th Convention Paper, 10-13 Oct 2003) looks at this. They propose a distortion measurement that penalizes for high second derivatives in the nonlinearity, and also penalizes more heavily when the input levels are closer to 0 (presumably because that is where the inputs are most of time, unless the system is driven into clipping), and show through testing that this measure correlates with human perception, while THD does not.

 

[pma]

Can anyone comment on how important harmonic distortion measurements are in comparing speakers?

For example, @amirm recently posted harmonic distortion measurements on ASR of the Revel F208, copied below, and these appear to compare favorably to measurements of the Philharmonic BMRs posted here and in-line below. For example, at 2000 Hz and 96 dB @ 1 meter, the F208s show a THD of less that 0.1%, while the Philharmonic BMRs are closer to 2% at similar sound levels. This is a 20x difference. Yet both speakers receive very strong positive reviews.

Is harmonic distortion even relevant when comparing two speakers? How?

Maybe you might like to measure another kind of distortion, like IMD, and use spectrum rather than a cumulative plot?