Amplifier Distortion Measurements at Low Power

[paulg1]

Hello,

A recent thread pointed out the relationship between output power requirements, speaker sensitivity, and the range of power preferred for different listening.
https://www.audiosciencereview.com/...s-decibels-and-sensitivity.62146/post-2278881. This suggests that 17-20bB is desired above the average listening level (50X to 100X times the average power) in a typical application for music/movies.

This got me thinking about typical listening and softer sounds. For me, in a domestic environment, typical average listening level is 60-85 dB. During listening, sometimes the level drops lower, and seldom, only for brief moments it will rise above 90 dB. So let's say we have a speaker with a sensitivity rating of 88 dB like in the example. So at the typical listening level of 78 dB the speakers are using 0.1W. In my case, with typical orchestral music, a lot of the musical details, are at a level considerably lower than this, 20-30 dB down from this level, maybe even lower in rare instances, such as harmonics, decays, hall reverberations, etc. That's 48 dB, so the amplifier power in those instances is 1mW-0.1mW and even perhaps lower in rare instances.

ASR is one of the few sites(only?) that have SINAD measurements of amplifiers down this low, and in some headphone amplifier measurements we see SINAD measured as low as 1uW. In reviewing these measurements, the typical SINAD graphs show it increasing with decreasing power beyond a certain point. Typically, we see SINAD vs. power level as a general V-shape overall, with the low point at 0.1-1W and usually increasing linearly with decreasing power level.
That behavior is curious and a bit counterintuitive, we would expect lower power to show ever decreasing SINAD. I suspect that this is an artifact of the measurement itself; SINAD is combined distortion and noise, and these low power measurements show noise in the test instrument itself as a percentage of the total measured SINAD. That would explain the linear behavior with decreasing power level. Question#1: Is that correct?

Question#2: is it possible to measure distortion (i.e., harmonic) independent of noise? I realize that this is a difficult proposal.

I had read that some studies had done exactly this by using a modified spectrum analysis software program. It uses which time-averaging to effectively cancel out noise products, leaving an identifiable signal and its related harmonics. This time-averaging approach is to identify extremely weak signals from spacecraft, amid a very high noise ambient background.

 

[pma]

Question#2: is it possible to measure distortion (i.e., harmonic) independent of noise? I realize that this is a difficult proposal.

Of course it is possible. And it is not difficult at all. REW, Arta have had such possibility always.

 

[Sokel]

Of course you can break SINAD and much more.
Here's an example breaking it to Noise and THD (along with THD+N=SINAD) :

..and here's breaking THD to its harmonics till H4 (you can go further too if you want) :

Furthermore, the above is not only one frequency SINAD but across the spectrum (valid up to 6kHz or so with these settings) .
You can also produce the same charts vs Level the same way.

 

[DVDdoug]

usually increasing linearly with decreasing power level.
That behavior is curious and a bit counterintuitive, we would expect lower power to show ever decreasing SINAD.

It's the "N" (noise) which is generally constant. Reducing the signal reduces the signal-to-noise ratio.

With class A/B amplifiers there is something called "crossover distortion" where the voltage changes between positive and negative. So as a ratio it's also worse with small signals. But in a well-designed amplifier it's below audibility and below the noise.

 

[paulg1]

I am very interested in how audio equipment measures at low signal levels. Here is an excerpt from another site that I referred to above. His comparison is between a very expensive solid-state amplifier and an absurdly priced single-ended triode vacuum tube amplifier:

"I have developed a measurement system which is able to quantify distortion levels at micropower quantities down to 1/1000th of a watt. These measurement techniques are unprecedented, and have revealed a number of revelations of amplifier distortion characteristics, at micropower levels, which are in direct opposition with traditional and scientific assumptions up unto this point.

The emphasis of my work is to definitively quantify low level signal linearity measurements of power amplifiers, and attempt to correlate these measurements with subjective listening results, as well as establishing the significance of low level distortion. Conventional test equipment generally does not resolve meaningful distortion measurements below 100mw, since the measurements become predominated by noise.

I have modified a spectrum analysis software program, which uses time-averaging to effectively cancel out noise products, leaving an identifiable signal and its related harmonics. This time-averaging approach is to identify extremely weak signals from spacecraft, amid a very high noise ambient background. Using this method, I am able to resolve a standardized total harmonic distortion measurement down to 1/1000 watt and an approximate measurement down to 1/500,000th of a watt.

My measurement results oppose common engineering supposition, in that it is commonly believed that very low signal linearity is essentially 'virtually perfect', and that only high-level signal linearity is a relevant parameter. My measurements indicate exactly the opposite is true of this commonly held assumption, particularly for amplifiers employing solid state devices.

To give you an example: the Halcro DM58 amplifier measures .007%THD at 2 watts, whereas at 1/1000th watt, THD measures 8.9%! By contrast a Wavac SH833 measures .57%THD at 2 watts and 1.6%THD at 1/1000th watt. The Wavac exhibits significantly lower THD at low signal levels by orders of magnitude than the Halcro. I have measured numerous amplifiers, both solid state and tube..."

This doesn't seem real to me.

 

[NTK]

I am very interested in how audio equipment measures at low signal levels. Here is an excerpt from another site that I referred to above. His comparison is between a very expensive solid-state amplifier and an absurdly priced single-ended triode vacuum tube amplifier:

"I have developed a measurement system which is able to quantify distortion levels at micropower quantities down to 1/1000th of a watt. These measurement techniques are unprecedented, and have revealed a number of revelations of amplifier distortion characteristics, at micropower levels, which are in direct opposition with traditional and scientific assumptions up unto this point.

The emphasis of my work is to definitively quantify low level signal linearity measurements of power amplifiers, and attempt to correlate these measurements with subjective listening results, as well as establishing the significance of low level distortion. Conventional test equipment generally does not resolve meaningful distortion measurements below 100mw, since the measurements become predominated by noise.

I have modified a spectrum analysis software program, which uses time-averaging to effectively cancel out noise products, leaving an identifiable signal and its related harmonics. This time-averaging approach is to identify extremely weak signals from spacecraft, amid a very high noise ambient background. Using this method, I am able to resolve a standardized total harmonic distortion measurement down to 1/1000 watt and an approximate measurement down to 1/500,000th of a watt.

My measurement results oppose common engineering supposition, in that it is commonly believed that very low signal linearity is essentially 'virtually perfect', and that only high-level signal linearity is a relevant parameter. My measurements indicate exactly the opposite is true of this commonly held assumption, particularly for amplifiers employing solid state devices.

To give you an example: the Halcro DM58 amplifier measures .007%THD at 2 watts, whereas at 1/1000th watt, THD measures 8.9%! By contrast a Wavac SH833 measures .57%THD at 2 watts and 1.6%THD at 1/1000th watt. The Wavac exhibits significantly lower THD at low signal levels by orders of magnitude than the Halcro. I have measured numerous amplifiers, both solid state and tube..."

This doesn't seem real to me.

Here is the 5 W @ 4 Ω spectrum plot of the Benchmark AHB2. The 3 kHz (3rd harmonic) spike, clearly measurable with current standard methods, is -115 dB below the signal fundamental, which means it is 1.6e-11 W, or 16 picowatt (= 16 millionth millionth of a watt).

Or look at it a different way, speakers sensitivity are rated at 2 W into 4 Ω (= 2.83 Vrms), and 5 W is 4 dB above 2 W. If a speaker has a sensitivity of 100 dB @ 2.83 Vrms, 1 m, then the sound pressure level produced by this 3rd harmonic spike will give 100 + 4 - 115 = -11 dB SPL at 1 m distance.

Totally inconsequential, and no problem with current measurements to measure that.

 

[Blumlein 88]

Amir measures SINAD at various power levels on amps which usually start somewhere below 50 milliwatts. The better amps have SINAD below -80 db at 20-30 milliwatts. It probably is mostly noise, but we know at worst it is distortion at -80 db (.01% THD). Here is one example. Looks like it starts at 25 milliwatts. The shape of the curve tells me it is mostly if not almost completely noise. So likely any THD is another 20 db or more lower ( -100 db or .001%).

 

[paulg1]

Thanks. Everyone. Got it now. The shape of the curve shows constant noise as the signal level drops. None of the typical amplifiers measured here show any anomalies in harmonic distortion at low signal levels. Amir's measurements go low enough to satisfy me. The excerpt from the web is not real.

 

[pma]

None of the typical amplifiers measured here show any anomalies in harmonic distortion at low signal levels. Amir's measurements go low enough to satisfy me. The excerpt from the web is not real.

Are you sure you (and other members) understand completely? Amir's measurement show THD+N vs. power, that is distortion+noise. The descending plot slope is due to noise that dominates above distortion at lower power. But, If you split THD and N components from the THD+N plot, only then you can see contribution of distortion to distortion+noise plot. See the graph below.

With class AB amplifiers, you can see rise of THD at low power with decreasing idle current of the output stage. This may or may not be corrected by sophisticated feedback. With class D, it depends on circuit design. As always, there is no simple answer to a simple question. Or, there always exists a simple and wrong answer to every question.

 

[gwing]

Thanks. Everyone. Got it now. The shape of the curve shows constant noise as the signal level drops. None of the typical amplifiers measured here show any anomalies in harmonic distortion at low signal levels. Amir's measurements go low enough to satisfy me. The excerpt from the web is not real.

Why do you say the web excerpt is not real?

I can only read the information you have posted but although it could all be invented and spurious it seems entirely possible that two different amplifiers would have different THD to each other and that the THD would be very different at micro power levels to that at the power levels usually measured. A sufficiently low micro power level wouldn't be audible of course so maybe not too relevant although that doesn't stop us chasing performance numbers, and listening to different amplifiers without measurement does sometimes (however unreliably) seem to show that some amplifiers sound better at low volume than others (which may well have nothing to do with micro power level distortion either).

 

[Gorgonzola]

Are you sure you (and other members) understand completely? Amir's measurement show THD+N vs. power, that is distortion+noise. The descending plot slope is due to noise that dominates above distortion at lower power. But, If you split THD and N components from the THD+N plot, only then you can see contribution of distortion to distortion+noise plot. See the graph below.

View attachment 444226

With class AB amplifiers, you can see rise of THD at low power with decreasing idle current of the output stage. This may or may not be corrected by sophisticated feedback. With class D, it depends on circuit design. As always, there is no simple answer to a simple question. Or, there always exists a simple and wrong answer to every question.

Thank you, and it illustrates clearly how THD and noise may be separated.

My question is this (to @pma and other): To what extent does the noise floor establish a level below which no useful detail can be heard from the music being played?

 

[pma]

My question is this (to @pma and other): To what extent does the noise floor establish a level below which no useful detail can be heard from the music being played?

It depends, so again there is no simple answer. To my knowledge, there have been experiments when a signal 20dB below noise level was audible in a DBT. I assume that @j_j would be able to tell us more.

 

[gwing]

Thank you, and it illustrates clearly how THD and noise may be separated.

My question is this (to @pma and other): To what extent does the noise floor establish a level below which no useful detail can be heard from the music being played?

Does it matter? That seems far too low a target for anyone interested in music to be concerned about.

I would be much more interested in the level at which noise can be heard at all rather than the level at which noise drowns out all the music. Ditto for THD. And especially for low listening volumes.

 

[Gorgonzola]

It depends, so again there is no simple answer. To my knowledge, there have been experiments when a signal 20dB below noise level was audible in a DBT. I assume that @j_j would be able to tell us more.

Personally I have an additional noise problem: I have tinnitus, an affliction of over 60 years, (tells you something about my age). In my case the tinnitus isn't to bad; it's really only obvious at very low ambient sound and when I'm listening carefully which includes when I'm listening too music.

It's been my impression for quite awhile that the tinnitus, while usually much more audible than equipment noise, doesn't prevent me hearing details, including "air", etc., that I wouldn't expect to hear if the tinnitus level imposed an absolute audibility floor. Then again, maybe there's something special about tinnitus as variety of noise.

 

[pma]

Personally I have an additional noise problem: I have tinnitus, an affliction of over 60 years, (tells you something about my age)

Me too. I am 70 in a few weeks, I have tinnitus, but I can hear 14kHz with 100% success in a DBT. It is a curse, you know. Especially during silence in the night. Maybe we suffer from similar problems.

 

[Old_School_Brad]

Personally I have an additional noise problem: I have tinnitus, an affliction of over 60 years, (tells you something about my age). In my case the tinnitus isn't to bad; it's really only obvious at very low ambient sound and when I'm listening carefully which includes when I'm listening to music.

It's been my impression for quite awhile that the tinnitus, while usually much more audible than equipment noise, doesn't prevent me hearing details, including "air", etc., that I wouldn't expect to hear if the tinnitus level imposed an absolute audibility floor. Then again, maybe there's something special about tinnitus as variety of noise.

I hear you -pun intended. The effect you're describing comes down to the interaction between what you're hearing and the specific characteristics of your tinnitus. Masking isn’t absolute; it depends on how the frequencies of the tinnitus, the distortion or noise, and the musical details overlap or differ.
Edit; I have tinnitus as well, though I’m not sure of the exact frequency. That said, I can still hear up to 16 kHz, so I don’t consider my hearing to be impaired.

 

[paulg1]

I hear what you are saying about background noise not necessarily masking low level sound details. My curiosity about low level distortion is along the lines that gwing described in post #10:

"A sufficiently low micro power level wouldn't be audible of course so maybe not too relevant although that doesn't stop us chasing performance numbers, and listening to different amplifiers without measurement does sometimes (however unreliably) seem to show that some amplifiers sound better at low volume than others (which may well have nothing to do with micro power level distortion either)."

This is the real mystery to me, why some systems sound correct at low levels and others don't, when both sound correct at higher volumes. Measurements alone don't seem to explain this phenomenon. I was exploring the possibility that distortion at micro power levels below those normally measured could be part of the answer. Now I'm not so sure this is the reason. What is this phenomenon? Is it real or imagined. How could it be measured? I don't think I will find the answer in this thread. This question most likely belongs in another topic area.

 

[Gorgonzola]

I hear you -pun intended. The effect you're describing comes down to the interaction between what you're hearing and the specific characteristics of your tinnitus. Masking isn’t absolute; it depends on how the frequencies of the tinnitus, the distortion or noise, and the musical details overlap or differ.
Edit; I have tinnitus as well, though I’m not sure of the exact frequency. That said, I can still hear up to 16 kHz, so I don’t consider my hearing to be impaired.

My tinnitus seems be range of frequencies perhaps centered around 7-8 kHz. It's a bit louder in my left ear.

Another problem for me has been that I've been stone deaf above 10 kHz for decades. My high frequency deficit and tinnitus might be related. I used to shoot at an indoor range for about 4 years without adequate ear protection; I don't recall any tinnitus before and my hi-freq deafness might go back that far. My hearing limit may have sagged to as little as 8 kHz in recent years -- I can't complain too much, I'm 80 y/o Neither problem has reduced my enjoyment of music

 

[Old_School_Brad]

My tinnitus seems be range of frequencies perhaps centered around 7-8 kHz. It's a bit louder in my left ear.

Another problem for me has been that I've been stone deaf above 10 kHz for decades. My high frequency deficit and tinnitus might be related. I used to shoot at an indoor range for about 4 years without adequate ear protection; I don't recall any tinnitus before and my hi-freq deafness might go back that far. My hearing limit may have sagged to as little as 8 kHz in recent years -- I can't complain too much, I'm 80 y/o Neither problem has reduced my enjoyment of music

80 years -what a milestone. I truly hope I’ll still be able to enjoy music at that age. The mind is a remarkable thing, and I like to believe it’ll help fill in the blanks if needed when the time comes.

Wishing you many more years of happy listening!