speaker distortion

[fdegrove]

Hi,

who said they could hear 0.00001% thd?

Did anyone?

IMHO, THD speaker measurements, just like amplifier THD measurements, don't say didley squad about how they sound compared to the sound of a live, unamplified acoustic instrument.

If high-end companies stuck with accepted tools to measure performance of the gear they put on the market we'd probably still be stuck with all those "fabulous" Phase Linear amps and their ilk.
IOW, we would not even be talking about "high-end" anyway...

It must be rather disconcerting to discover that your trusted measuring gear doesn't even begin to correlate with what it's thought to do...

Cheers,

 

[tschrama]

Hi Till,

THanks for your reply. I think I'll make it my next project to make acoustic distortion measurements. We could do the same measurement and compare? Then it might be interesting to agree on some test-standard.. like 80dB, 600Hz, 1mtr etc?

Mean time I need to adjust my pure-sine-oscilator (which has abouyt -105dB THD) or make a test CD, make a good mic-preamp, make a microphone (got several spieces of panasonic capsules) etc.... I got a proffesional callibrated 16bit 200KHz DAQ ADC card (National Instruments) and software to do the THD measurements with ...

Greetings,
Thijs

 

[phase_accurate]

Just a few thoughts:

1.) Crossover distortion seems to be very nasty. Does a speaker produce it and do single-ended class-A amps produce it ? I think everyone knows the answer, and it is therefore not that hard to guess why (and I belive that it IS true) a class A amp with 0.01% THD @ o,1 Watt sounds better than an AB amp with 0.01%
THD @ 1 Watt, when driving a high-efficiency speaker.

2.) THD of speakers is heavily SPL dependant. So the figures given at the beginning of this thread are in fact meaningless. If this figures are from large P.A. speakers, driven a max power then I wouldn't care about 10% THD that much. Keep in mind how much THD the ear would produce at such insane SPLs ! Many P.A. speakers produce very low THD at domestic listening levels. Thats one advantge, amongst others (and there are NOT only advantages) of generously dimensioned P.A. components used at home.

3.) The same nonlinearities that generate THD, generate IMD as well. While the harmonics generated by THD are often very close to an instrument's natural harmonics (don't forget that not all harmonics of instruments are exact integer numbers of the keynote) this isn't true for IMD. There will be spectral content where there wasn't any before. And there will be content in the stereo image where there wasn't any either. Or in simple words: IMD is muddying up the sound (and every component in the chain makes the mud deeper). Since speaker drivers usually reproduce a small portion of the frequency range and since they are the last stage of the chain (disregarding room and listener) they might be the least culprits regarding IMD. IMO it wouldn't hurt to reduce an amp's good THD value of 0.0001 % down to 0.0000001 % even when driving a speaker with 1 % THD at 100 dB SPL - if it can be done at 1.) reasonable cost and 2.) without degrading other parameters.

Regards

Charles

 

[5th element]

Isnt speaker distortion far more benign then amp distortion anyway?

 

[till]

Isnt speaker distortion far more benign then amp distortion anyway?

this is what phase_accurate tells in the posting above your: yes, in case you look at normal push pull / AB amps. No in case you look at SE class a amps. The speaker will not make crossoverdistortion, the push pull amp will.

 

[tschrama]

Hi Phase_accurate,

1] Could be correct, but this was not the issue.

BTW the gainclone at 1W is more lineair than a ZEN at 0.1W. My measured spectra show little or no higher harmonis at a gainclone; crossover distortion seems to be pretty low then. (and I'm NOT a gainclone fan!)

2] Yes, THD for speakers is EXCURSION (is related to, but not identical to SPL, freq depended!) dependent, that's why I want to measure my own speakers at say 80dB spl.

3] THD and IMD are so much related that we could stick to THD for now. Personally I suspect the speakers to be the LARGEST culprits. Various publications show relative large Nonlineairity of speakers, check the web, the brochures os various speaker manufactures etc.

I think it's time for some hard numbers guys... going to the soldering iron now.. het thos measurements done!


Cya,
Thijs

 

[till]

BTW the gainclone at 1W is more lineair than a ZEN at 0.1W. My measured spectra show little or no higher harmonis at a gainclone; crossover distortion seems to be pretty low then. (and I'm NOT a gainclone fan!)

did you measure the chip also at very low Watt, and did you measure at some higher frequencies? like 5 or 10 kHz. And also the gain you set on the chip may influence it.

 

[ThorstenL]

Konnichiwa,

here is a particular speaker (by a well known high-end brand name) and its thd figures:

2nd harmonic:
100hz: about 10%;
1000hz: about 0.5%;
10Khz: about 5%

3rd harmonic:
100hz: about 10%;
1khz: about 0.2%;
10khz: about 0.5%.

Sounds quite reasonable for 96db/1m if that's where it is taken at and barely average for 90db/1W .

who said they could hear 0.00001% thd?

Hmmm. Depeneds entierly upon the actual spectrum of the 0.00001% THD and the presence or absence of other "non-harmonic", "non correlated" or "noisefloor modulation" (I am still looking for the best term to imply a signal dependent noisefloor whose content appears not correlated or harmonically/IMD related to the signal).

Some interesting work on distortion has been done by Earl Geddes:

http://www.gedlee.com/distortion_perception.htm

It is quite interesting to compare the sound samples given.

Sayonara

 

[phase_accurate]

BTW the gainclone at 1W is more lineair than a ZEN at 0.1W. My measured spectra show little or no higher harmonis at a gainclone; crossover distortion seems to be pretty low then. (and I'm NOT a gainclone fan!)

O.K. When I was younger I would have listened to a 109 dB efficient Speaker at 0.1 Watt or even 1 Watt.
Nowadays it would be closer to 0.01 Watt (even less on average). The less power, the better it looks for speaker and single-ended, linearity-wise.

Personally I suspect the speakers to be the LARGEST culprits. Various publications show relative large Nonlineairity of speakers, check the web, the brochures os various speaker manufactures etc.

If you look at the figures alone you are definitely right. But be aware that things get worse when they are already fed with garbage (and every single stage adds more garbage to the one that it was already fed with) thats why I think it is still worth to have low THD amps even when the speaker's THD is higher.
IMD generated by a nonlinear device is larger the broader the part of the audio range it has to "process". If you split the signal into several spectral parts before feeding it to a nonlinear device, you will have less IMD (or maybe less disturbing IMD) in the end. That could be one reason for nonlinearity within speakers being less severe than amp related nonlinearity. If you include the power amps into such a topology (i.e. active x-over) you will reduce amp-related IMD as well.

Regards

Charles

 

[ThorstenL]

Konnichiwa,

also typically you can almost eliminate speaker distortion by never exceeding 60% or so of rated RMS power.

Hmmm. You think so? From experience most speakers start to exceed 0.5% THD at around 1 Watt applied. The above is for cone/dome speakers operated within the linear throw and is a direct result of the magnet system construction and based on a cubic function of the current (hence the 1 Watt reference).

Measures can be taken to reduce this level of distortion, which are rarely found in "High Fidelity" and "High End" speaker drivers due to cost issues - even in supposedly "cost no object" design. Yet even with all distortion mechanisms minimised I find "no distortion until 60% RMS Power" to be a rather adventerous assertation, to be kind about it.

Also, I find that in many "HiFi" speaker drivers compression starts to set in around the 1 Watt Level. Again, there are measures that can be taken, but rarely are for "High Fidelity" Drivers.

Sayonara

 

[ThorstenL]

Konnichiwa,

Isnt speaker distortion far more benign then amp distortion anyway?

That will depend ENTIERLY on the Amplifier and Speaker you use in the example....


Sayonara

 

[tschrama]

Hi Till,

The LM3886 has about 50mA Iq in it's output stage, so it allmost works in class A up to 0.1 Watt in 8 Ohm. I only measured at 1Watt in 8 Ohm at 1 KHz. Offcourse we all know that THD increases with freq and closed-loop gain due to less feedback availeble.

The picture you have atached seems to be a bad example (which chip is it? is it a chip at all?) with zero or very little bias current in it's output stage.

But lets not talk about amplifier distortion ok?

Question is : What is the speaker distortion, when excursion is such that it produces, say 90dB (1Watt typical for my speakers)?

My gues is for more than 0.01%, judging from manufactures brochures, and some lost references.

Further speculation is pointless: I'll do the measurement, and I am interested in other's measurements too!

 

[till]

I only measured at 1Watt in 8 Ohm at 1 KHz. Offcourse we all know that THD increases with freq and closed-loop gain due to less feedback availeble.

if you measre at much lower levels i bet you will find THD increases also!

The picture you have atached seems to be a bad example (which chip is it? is it a chip at all?) with zero or very little bias current in it's output stage.

It is an extreme example, its 20 kHz. Its was an OPA5?? don´t remember exactly which.

But lets not talk about amplifier distortion ok?

ok, but i think the important ponit is to see both interacting: amplifiers THD characteristic AND speakers characteristic.

Question is : What is the speaker distortion, when excursion is such that it produces, say 90dB (1Watt typical for my speakers)?

My gues is for more than 0.01%, judging from manufactures brochures, and some lost references.

yes more, the lower the fequency the higher the THD i suspect.

 

[cunningham]

Certainly power compression will be reduced...but the actual cone movements is the worrying thing if talking of low frequencies.

hundred or two watts applied to a shiva 12" can bottom it at 20hz,but whatabout at 90hz! its only thermally limited there.......

The impeadence of the coil in the speaker electrically has more resistance to 90Hz than 20Hz, but I was mostly refering to cone travel and streching of materials, bass reflex and spider, causing the travel in and out to not be linear. This is a majior consideration when dealing with sealed air sub-woofers vs free air ones...

then again some people just like it to be loud!!

 

[phase_accurate]

We can take it as a fact that speaker distortion is displacement-dependant.
Displacement is dependant on input voltage and FREQUENCY. So it is just normal that you will have more distortion at 20 Hz than at 90 Hz for given input power.

And I think it is O.K. if we talk about amp distortion as well, since someone thought that it would be useless to have low amp distortion when they are driving speakers with higher distortion. My opinion on that subject was already posted.

Regards

Charles

 

[ThorstenL]

Konnichiwa,

Question is : What is the speaker distortion, when excursion is such that it produces, say 90dB (1Watt typical for my speakers)?

My gues is for more than 0.01%, judging from manufactures brochures, and some lost references.

Well, 0.01% is -80db. Now I would love to see those lost references. I love reading Science Fiction.

Now here the ugly and nasty reality.

A Speaker from a reputable manufacturer, B&W of England, the (arguably low price) DM303. Here the full measurements:

http://www.soundstagemagazine.com/measurements/bw_dm303/

And here the distortion measured at 90db/2m (96db/1m):

An externally hosted image should be here but it was not working when we last tested it.

To get some idea of what is going on, the 70db line is equal to 10% Distortion, the 50db kine to 1% distortion and the baseline in this graph is equivalent to 0.56% THD, meaning in this measurement any distortion below 0.56% THD is simply hidden.

Distortion in the midrange is in the region of 1-3% (500Hz - 3KHz) and at 100Hz has risen to 5% with the 50Hz THD > 10%!!!

Okay, the DM303 is a "budget speaker" and others do perform somewhat better. A few examples of more or less "lowish" distortion speakers (remember, again distortion below 0.56% is not quantified, it is simply "off screen").

Wilson Audio Watt/Puppy:

http://www.soundstagemagazine.com/measurements/wilson_wattpuppy7/

An externally hosted image should be here but it was not working when we last tested it.

Midrange distortion has some peaks of up to around 3% but remains broadly below 0.56%, LF distortion at 100Hz is also near the graphs limit at around 0.6% and even at 50Hz we see only around 1.8%.

Dynaudio Confidence C4:

http://www.soundstagemagazine.com/measurements/dynaudio_confidence_c4/

An externally hosted image should be here but it was not working when we last tested it.

All the THD in the midrange lies below 0.56% and only below around 80Hz does distortion rise to around 3% at 50Hz. Dynaudio is one of the few companies who have consistently manufacturerd low distortion motor systems, shame their cone/dome materials make the sound less then ideal, pretty lifeless and undynamic.

Finally another "bad" example, the North SM 6.9, interresting as it is build with Drivers known and available to DIY Enthusiasts, rather than the custor Drivers used by B&W, Wilson and Dynaudio.

http://www.soundstagemagazine.com/measurements/norh_sm69/

An externally hosted image should be here but it was not working when we last tested it.

Here we do see midrange distortion above 0.56% but usually below 1%, again at low frequencies the distortion rises to around 1% at 100Hz and around 10% at 50Hz (note that 50Hz is also > 10db down on the midrange sensitivity, hence the Distortion is only 20db below the signal).

At normal volume levels most commonly found dynamic speakers will average between 0.3 and 1% THD in the midrange (mostly odd order too) rising to as much 10% near full Power, with LF distortion usually between 2 - 10% at normal volume levels and approaching and exceeding 50% (it means more harmonic output than fundamental frequency Output!!!) at full power.

As remarked previously, many HiFi drivers will show around 1db compression at 1 Watt and > 6db at rated power.

I hope that illustrates the severe fallacy of the commonly held believe that speakers are "low distortion" devices.

Sayonara

 

[phase_accurate]

I hope that illustrates the severe fallacy of the commonly held believe that speakers are "low distortion" devices.

Having seen many measurement reports and manufacturer's data sheets, I never had any doubts about these figures.

But the interesting question remains: How do speaker generated- and amplifier generated- distortion compare ? It seems as if many speakers are "transparent" to distortion generated in amps etc, although their own contribution to distortion is much worse.

Regards

Charles

 

[ThorstenL]

Konnichiwa,

But the interesting question remains: How do speaker generated- and amplifier generated- distortion compare ? It seems as if many speakers are "transparent" to distortion generated in amps etc, although their own contribution to distortion is much worse.

Yes, Speakers (cone/dome) have usually two main distortion mechanisms.

One we shall call excursion non-linearity. This lumps together magnet field non-linearities, suspension non-linearities and air nonlinearities. The distrotion is almost always of a quadratic nature, related directly to the excursion of the driver and hence highly frequency dependant (more distortion at lower frequencies for the same SPL) and producing low order harmonics with 2nd harmonics dominant and a directly related reduction with excursion (1/10 excursion = 1/10 distortion) - sounds like SE Valve Amp? Damn right it does!

The other we shall call eddy current loss non-linearity. It comes from having a coil filled a solid steel piece (pole piece). This is a cubic function related to the current in teh voice coil. Being cubic in nature menas odd order harmonics are produced and that with rising level the distortion rises very fast, a 1/10 reduction in current resulting in around 1/100 reduction in distortion.

Other distortion sources exist, but the two "lumped" effects descrive can account for the majority. The simple solution to low distortion? Limit excursion and voice coil current, in other words large diameter drivers and high sensitivity.

Sayonara

 

[millwood]

i don´t need any mic to get this number.

we knew that very well,

If high-end companies stuck with accepted tools to measure performance of the gear they put on the market we'd probably still be stuck with all those "fabulous" Phase Linear amps and their ilk.
IOW, we would not even be talking about "high-end" anyway...

Cheers,

I would agree with that too: those high-end guys would then be really making something that is based on sound science, rather than voodoos and moon-rock cables, .

for those of us who want to rely on science, it is probably reasonable to say that speaker distortions are a few magnitude higher than amp distortion. and in our quest to lower distortion, amps aren't probably the bottleneck.

 

[till]

it is probably reasonable to say that speaker distortions are a few magnitude higher than amp distortion. and in our quest to lower distortion, amps aren't probably the bottleneck.

indeed