Measurement of speaker distortion

[PMA]

Hello,

I would like to contribute with my short study on loudspeaker distortion

Measurement of speaker distortion

Cheers,
Pavel

 

[Taco]

First of all, thanks for your efforts!

Do you have an estimate spl level for 0.6 Vrms (my guess would be < 85 dB for your speaker, which is rather low for speaker testing). The 2 Vrms are more representative.

Can the surprising 19 kHz + 20 kHz measurement be hampered by the extension of your measurement mic?

 

[PMA]

Hi Taco,

thanks for your reply. You are right, 2Vrms is more realistic, but frankly speaking I do not listen at higher levels. The speaker has above 90dB/2.83Vrms/1m sensitivity.

I was surprised by the 19+20kHz result for the reason that I expected much higher distortion (we can see 1kHz 2nd harmonic product and 18kHz 3rd harmonic product about 60dB below basic level, which is 6dB above 19 and 20kHz spectral line).

The decaying shape of spectrum background is a 1/f noise of microphone and microphone preamp.

 

[planet10]

Even 85 dB is pretty loud.

dave

 

[gedlee]

Hello,

I would like to contribute with my short study on loudspeaker distortion

Measurement of speaker distortion

Cheers,
Pavel

Hi Pavel

You made a huge error in your asumptions. The form of the nonlinearity for an amplifier may be completely different than for a loudspeaker, resulting in a situation where .1 % THD in the amp is audible on a loudspeaker with 10% THD. This makes it very difficult to draw any conclusions from you tests.

 

[PMA]

Hi Earl,

there might be a misunderstanding, I agree with you. I wanted to say that distortion of a current speaker may be quite low (and low order, as shown). There are many amplifiers, especially 'exotic' designs, that distort much more, and it is audible (like SE triode power amp - very very audible with philharmonic orchestra, though appreciated by many).

The system I use for measurement (and listening) has all the distortion components below -100dB at 2Vrms. Power amplifier is no global NFB, 1A idle current design.

Regards,
Pavel

 

[gedlee]

Hi Earl,

there might be a misunderstanding, I agree with you. I wanted to say that distortion of a current speaker may be quite low (and low order, as shown). There are many amplifiers, especially 'exotic' designs, that distort much more, and it is audible (like SE triode power amp - very very audible with philharmonic orchestra, though appreciated by many).

The system I use for measurement (and listening) has all the distortion components below -100dB at 2Vrms. Power amplifier is no global NFB, 1A idle current design.

Regards,
Pavel

But isn't the point simply that the numbers don't mean anything?

 

[cuibono]

Thanks Pavel for sharing - I'm not surprised by your results - having non-linear distortion down 60dB is not too hard to do for modern drivers, particularly at the levels you've driven them at.

I agree with a general statement that nonlinear distortion is less audible in loudspeakers than many people assume. Toole's recent book has an interesting analysis of nonlinear distortion in loudspeakers. The upshot is that auditory masking limits our ability to hear harmonic tones. I would also add that any natural instrument has many harmonics that are just as loud as the fundamental, and would mask nonlinear distortion in loudspeakers, up to a point.

Stored energy and crossover distortion are areas that may matter more, but that is still a fairly gray area IMO.

 

[PMA]

But isn't the point simply that the numbers don't mean anything?

Partly, IMO. They should not be completely overlooked or ignored. And we are not speaking numbers, but spectra. They are just a part of a whole issue.

 

[rdf]

There are many amplifiers, especially 'exotic' designs, that distort much more, and it is audible (like SE triode power amp - very very audible with philharmonic orchestra, though appreciated by many).

Do you have a sample of an SE to try? Based on dim memories of the form of transfer functions which generate 2nd I never saw an obvious reason for discounting the possibility of second order cancellation between electronics and speakers. If you're set up it would make an interesting experiment. An option is to pre-distort the test signal with different percentages of second and graph the results.

 

[gedlee]

And we are not speaking numbers, but spectra. They are just a part of a whole issue.

It is precisely this issue that should be discussed. Spectra IS a good first step, but there is no standardized approach. Then as you proceed, you will find that the orders for loudspeakers are in general very low and that the orders for electronics can be very high. This means that it's more likely that the electronics will be the source of audible nonlinear distortion than the loudspeaker. Follow on with the fact that some perception issues are not linear in SPL and you have a major complication that makes the viewing of nonlinear distortion in a loudspeaker a relatively unimportant thing. Most people who have studied this problem have come to this conclusion (Toole for one, as noted above). But we all know that loudspeakers DO have audible problems. So what are they if not nonlinearities? This needs to be sorted out far more than anything else.

 

[PMA]

I would respectfully partly disagree. In some electronics, maybe. In my system I have, at the level shown, all the harmonics are well below -100dB, and only 2nd and 3rd visible. Please tell me, would this be audible? I am sure sometimes we are demonizing order and level of distortion of electronics, especially the well designed for audio purposes. I am sure there is another reason of sonic differences. But I cannot accept, at any level and order, when the amp has higher distortion components than a speaker.

 

[gedlee]

I would respectfully partly disagree. In some electronics, maybe. In my system I have, at the level shown, all the harmonics are well below -100dB, and only 2nd and 3rd visible. Please tell me, would this be audible? I am sure sometimes we are demonizing order and level of distortion of electronics, especially the well designed for audio purposes. I am sure there is another reason of sonic differences. But I cannot accept, at any level and order, when the amp has higher distortion components than a speaker.

It's not the harmonics at high levels that matter its the nonlinearity at very low levels - down into the noise that matters. These are what's audible. If ALL the harmonics all the way out to 10-20th order are 100 dB down when the signal is at the noise floor, then yes, nonlinear distortion in your electronics is not an issue. But then it's also not an issue in the speakers either, so its time to move on to something else.

 

[PMA]

We both speak low level and do not disagree

Anyway, there are still audible differences, even at low levels and electronics with almost no measurable harmonics. But we are off topic with this here.

My point is that amplifier distortion should not exceed the speaker one, again, at any level and any harmonics order. But, some well accepted amplifiers (by some) do exceed this, even at power level as low as 1W. It is audible.

It was believed that 3% of 2nd harmonic is inaudible. If I add 1% of 2H to 1kHz at about 80dB SPL, I can clearly hear the difference. If it is 3H 1%, I can distinguish it in a blind test at 100% rate. It is not so easy for 2H, but also distinguishable. Not with the amp or speaker with 2% 2H distortion - and this is where the old numbers come from, IMO.

 

[gedlee]

We both speak low level and do not disagree

Anyway, there are still audible differences, even at low levels and electronics with almost no measurable harmonics. But we are off topic with this here.

My point is that amplifier distortion should not exceed the speaker one, again, at any level and any harmonics order. But, some well accepted amplifiers (by some) do exceed this, even at power level as low as 1W. It is audible.

It was believed that 3% of 2nd harmonic is inaudible. If I add 1% of 2H to 1kHz at about 80dB SPL, I can clearly hear the difference. If it is 3H 1%, I can distinguish it in a blind test at 100% rate. It is not so easy for 2H, but also distinguishable. Not with the amp or speaker with 2% 2H distortion - and this is where the old numbers come from, IMO.

Yes, well your numbers certainly disagree with mine, but we are looking at different things. You are talking about thresholds for pure tones, a rather purest test that doesn't correlate with anything found in the real world. I am talking about the levels of THD that are found to be perceptable to annoying for a large group of listeners using music as a source. A far more real world scenario IMO.

 

[PMA]

Again I think we do not disagree. One singer + one saxophone tolerate high level of low order harmonic distortion, and still accepted well by listeners, even prefered by some. A philharmonic orchestra, well recorded at high dynamics (like Linn SACD Linn Records - Mozart Symphonies 38 - 41 e.g.) does not tolerate high level of distortion and an experienced, trained listener can easily find the difference.