Amp Distortion vs Speaker Distortion

[watchnerd]

I was reading some reviews of some Klipsch "Heritage" series horn speakers, all of which have high sensitivities in the high 90s to low 100s and thinking about:

a) The alleged low distortion of horns and what that would mean at normal listening levels for total speaker distortion
b) That my amp has more distortion at low wattage than higher wattage (until the power limit is hit)
c) That the super high sensitivity of horn speakers would mean listening to my amp in the more distorted part of its range, at normal volumes

My amp is a Devialet Expert 440. These are the charts for the lower-wattage Expert 130, but I think the same pattern would still hold true with the higher output model:

Chart 2 - Distortion as a function of power output and output loading

(Line up at 20W to determine lines)
Top line = 4-ohm SMPTE IM distortion
Second line = 8-ohm SMPTE IM distortion
Third line = 4-ohm THD+N
Bottom line = 8-ohm THD+N

Chart 3 - Distortion as a function of power output and frequency

(8-ohm loading)
Red line = 1W
Magenta line = 10W
Blue line = 30W
Cyan line = 60W
Green line = 80W

[Charts courtesy of: https://www.soundstagenetwork.com/i...er&catid=97:amplifier-measurements&Itemid=154]

I usually listen in the 75-85 dB range, with peaks up to about 90 dB, at the listening seat.

Taking a hypothetical use case of a Klipsch Cornwall (102 dB), and plugging it into:

https://myhometheater.homestead.com/splcalculator.html

I get <0.3 Watts for my typical listening level and distance.

But 0.3 Watts / 300mW, from the chart above is about .01% THD + N, which is not awesome from an electronics point of view.

So am I better off with less sensitive speakers so that the amp can run in its "preferred power range"?

Or does the reduction in speaker distortion due to high efficiency trump anything happening on the electronics side?

 

[Kvalsvoll]

But 0.3 Watts / 300mW, from the chart above is about .01% THD + N, which is not awesome from an electronics point of view.

Your interpretation is incorrect. At lower power levels, the distortion is masked by noise, so the measurement graphs do not really show the true distortion level. Also, as level is reduced, the absolute distortion level is also lowered, which means the relative threshold for audibility actually is lower with reduced volume.

Well designed amplifiers have distortion levels below threshold of hearing. From looking at those graphs, it is not likely that this amplifier produces any audible distortion, at any volume, as long as it does not clip.

 

[watchnerd]

Your interpretation is incorrect. At lower power levels, the distortion is masked by noise, so the measurement graphs do not really show the true distortion level. Also, as level is reduced, the absolute distortion level is also lowered, which means the relative threshold for audibility actually is lower with reduced volume.

Well designed amplifiers have distortion levels below threshold of hearing. From looking at those graphs, it is not likely that this amplifier produces any audible distortion, at any volume, as long as it does not clip.

Thanks.

I always forget about the noise floor aspect of THD.

So what distortion level is audible? 1%?

 

[georgeT]

Thanks.

I always forget about the noise floor aspect of THD.

So what distortion level is audible? 1%?

If you are to believe Benchmark: "
If we want to listen at 80 dB, peaks will reach about 100 dB. At this playback level, the distortion must be lower than -100 dB (0.001%) to absolutely guarantee that it is inaudible.

If we want to listen at 90 dB, peaks will reach about 110 dB. At this playback level, the distortion must be lower than -110 dB (0.0003%) to absolutely guarantee that it is inaudible."

Then again, they are trying to sell their ultra low distortion amps.

 

[Julf]

If you are to believe Benchmark: "
If we want to listen at 80 dB, peaks will reach about 100 dB. At this playback level, the distortion must be lower than -100 dB (0.001%) to absolutely guarantee that it is inaudible.

If we want to listen at 90 dB, peaks will reach about 110 dB. At this playback level, the distortion must be lower than -110 dB (0.0003%) to absolutely guarantee that it is inaudible."

Then again, they are trying to sell their ultra low distortion amps.

Yes, I call BS on that. Your listening room will have a noise floor of 30 dBSPL or more, so for random noise (the worst possible), you need it to be at -70 dB (0.03%), but for harmonic distortion, where the distortion products are harmonics of the actual signal, even 1% is hard to hear.

 

[watchnerd]

If you are to believe Benchmark: "
If we want to listen at 80 dB, peaks will reach about 100 dB. At this playback level, the distortion must be lower than -100 dB (0.001%) to absolutely guarantee that it is inaudible.

If we want to listen at 90 dB, peaks will reach about 110 dB. At this playback level, the distortion must be lower than -110 dB (0.0003%) to absolutely guarantee that it is inaudible."

Then again, they are trying to sell their ultra low distortion amps.

I'm also going to call BS on that, but for a different reason:

What speaker is good enough to have .001% distortion at 100dB?

I'm pretty sure the distortion in the speakers will mask this.

 

[Julf]

I'm pretty sure the distortion in the speakers will mask this.

Of course.

 

[Blumlein 88]

Yes, I call BS on that. Your listening room will have a noise floor of 30 dBSPL or more, so for random noise (the worst possible), you need it to be at -70 dB (0.03%), but for harmonic distortion, where the distortion products are harmonics of the actual signal, even 1% is hard to hear.

Yes, and no. Most room noise is very low in frequency. Even domestic listening rooms may be very close to 0 db SPL in that 3-5 khz range (quite often 10 db spl in ERB bands. Meaning noise will not mask harmonics of frequencies in the meat of the midrange which can be fairly high in level. Of course this is with test tones.

I do agree however, that with actual music you'll mostly not hear it unless it exceeds 1 %.

 

[watchnerd]

Here are comments and graphs on speaker distortion from a JBL 4367 review:

"Distortion measurements for speakers are typically done at a 90 dB reference point but the distortion numbers for the JBL 4367s were low enough that I also measured them at 100 dB to really see what these drivers could do. This is considered loud by any stretch and some lesser drivers would balk at these levels. This particular graph shows a distortion level at 10 kHz of only 0.389017% at 100 db"

"At 1 kHz the distortion level at 100 dB drops even further to 0.096540%!"

etc.

There is much more at: https://hometheaterhifi.com/reviews...g/jbl-4367-studio-monitor-loudspeaker-review/

 

[Kvalsvoll]

Thanks.

I always forget about the noise floor aspect of THD.

So what distortion level is audible? 1%?

Depends on the signal, noise level in the room, type of distortion, listening volume.

For continuous sine wave signal, it is not so difficult to find the answer, for transient signals, a bit more complex.

For sine wave, these numbers were found for detection levels of harmonic distortion by listening at different spl:

440hz:

60dB
2h -34dB 2%
3h -50dB 0.32%
4h -50dB 0.32%
5h -53dB 0.22%
6h -50dB 0.32%
8h -58dB 0.13%

70dB
2h -34dB 2%
3h -50dB 0.32%
4h -54dB 0.2%
5h -65dB 0.056%
6h -64dB 0.063%
8h -72dB 0.025%

80dB
2h -34dB 2%
3h -34dB 2%
4h -43dB 0.71%
5h -65dB 0.056%
6h -72dB 0.025%
8h -76dB 0.016%

Plotted in a graph, this is how it looks:

We see the low-order is masked, and then the higher approach threshold of hearing. This means higher order components are most critical, and louder listening volume makes it easier to hear.

For bass frequencies, the situation is the same.

There are a couple more aspects that come into play here, but generally, we see that ordinary, well designed equipment can manage to keep distortion below threshold.

For music, it will be much more difficult to hear, because the signal is far more complex and contains harmonics on its own.

I believe there is at least one thread on this particular subject already, so for all I know everything I write here is redundant information.

 

[watchnerd]

Depends on the signal, noise level in the room, type of distortion, listening volume.

For continuous sine wave signal, it is not so difficult to find the answer, for transient signals, a bit more complex.

For sine wave, these numbers were found for detection levels of harmonic distortion by listening at different spl:

440hz:

60dB
2h -34dB 2%
3h -50dB 0.32%
4h -50dB 0.32%
5h -53dB 0.22%
6h -50dB 0.32%
8h -58dB 0.13%

70dB
2h -34dB 2%
3h -50dB 0.32%
4h -54dB 0.2%
5h -65dB 0.056%
6h -64dB 0.063%
8h -72dB 0.025%

80dB
2h -34dB 2%
3h -34dB 2%
4h -43dB 0.71%
5h -65dB 0.056%
6h -72dB 0.025%
8h -76dB 0.016%

Plotted in a graph, this is how it looks:
View attachment 32409
We see the low-order is masked, and then the higher approach threshold of hearing. This means higher order components are most critical, and louder listening volume makes it easier to hear.

For bass frequencies, the situation is the same.

There are a couple more aspects that come into play here, but generally, we see that ordinary, well designed equipment can manage to keep distortion below threshold.

For music, it will be much more difficult to hear, because the signal is far more complex and contains harmonics on its own.

I believe there is at least one thread on this particular subject already, so for all I know everything I write here is redundant information.

Wow, a 40dB spread between 2nd and 8th harmonic.

But I've often heard that 'odd order' harmonic distortion is easier to hear / more disturbing.

Except for 3h, this graph doesn't seem to show that....

Is 'even order = benign, odd order = bad' a myth?

 

[Kvalsvoll]

Is 'even order = benign, odd order = bad' a myth?

I don't know, have to test it to actually find out. And to do that, you need to generate distortion above those limits - more than 2% 2h . If the distortion components are below threshold, it does not matter what distribution they have.

But I did note that the phase of the distortion matters, and a little 2h actually could make the tone sound like more pure and less distorted.

 

[Kvalsvoll]

As for amp vs speakers; those numbers were found listening to speakers. The distortion was monitored real-time, to ensure the added distortion components were not masked by distortion from other elements in the reproduction chain - speakers, amplifiers, room rattles.

 

[watchnerd]

How do I translate distortion dB to distortion %?

Below are the distortion measurements for the Dynaudio Special 40, close enough to my Dynaudio Contour 20 to be a good proxy (the Special 40 has an almost identical tweeter, 1 cm smaller woofer, smaller cabinet, shorter port, I'm also high-passing mine at 80 Hz).

Top curve: frequency response @ 90dB SPL
Bottom curve: THD+N @ 90dB (50Hz - 10kHz)

Data taken from:

https://www.soundstagenetwork.com/i...&catid=77:loudspeaker-measurements&Itemid=153

 

[Julf]

How do I translate distortion dB to distortion %?

Decibel calculator

Insert the dB value in the upper field (remember to enter it as a negative value - 0 dB is full volume), go "calculate", and multiply the result to get %.

 

[watchnerd]

Decibel calculator

Insert the dB value in the upper field (remember to enter it as a negative value - 0 dB is full volume), go "calculate", and multiply the result to get %.

If I'm doing it correctly, and eyeballing the graphs above, I get 2% at the 100 Hz spike, and 0.5% at the spike just above 1k Hz.

 

[Julf]

If I'm doing it correctly, and eyeballing the graphs above, I get 2% at the 100 Hz spike, and 0.5% at the spike just above 1k Hz.

Looks like it, yes.

 

[Frank Dernie]

Is 'even order = benign, odd order = bad' a myth?

Could be. John Atkinson has frequently noted that 3rd order sounds benign, but since I haven't ever tested this myself I don't know.
The harmonic giving the oboe its characteristic timbre is the 5th iirc and I love the sound of the oboe

 

[Blumlein 88]

How do I translate distortion dB to distortion %?

Below are the distortion measurements for the Dynaudio Special 40, close enough to my Dynaudio Contour 20 to be a good proxy (the Special 40 has an almost identical tweeter, 1 cm smaller woofer, smaller cabinet, shorter port, I'm also high-passing mine at 80 Hz).

Top curve: frequency response @ 90dB SPL
Bottom curve: THD+N @ 90dB (50Hz - 10kHz)

Data taken from:

https://www.soundstagenetwork.com/i...&catid=77:loudspeaker-measurements&Itemid=153

Easy shorthand for this. -20 db is 10%, -40 db is 1%, -60 db is .1% distortion. So 90-40 db means everything below 50 db on that graph is 1% or less.

 

[Kvalsvoll]

Could be. John Atkinson has frequently noted that 3rd order sounds benign, but since I haven't ever tested this myself I don't know.
The harmonic giving the oboe its characteristic timbre is the 5th iirc and I love the sound of the oboe

Excellent example to put things into perspective. If the 5h. is -6dB down (check spectrum graphs of oboe), how much distortion is necessary to change the level of this harmonic significantly? The answer is A LOT. 50% (at 5h.) will give something that can be easily heard, while anything below 10% will be very difficult to hear. No amplifier has 10% 5h distortion.

Fact is, non-linear distortion is not what determines sound character in any reasonably good sound reproduction system. This does not mean everything sounds the same, but the differences are largely caused by linear phenomenon, again caused by frequency response and radiation pattern of the loudspeakers and room acoustics.

In reproduction, non-linear distortion can be a problem when amplifiers clip, and for loudspeakers when pushed beyond limits. Small loudspeakers will distort and compress at surprisingly low spl at lower frequencies, and this is audible.