Amplifier distortions - what, and how much are audible

[PENG]

Given those two profiles for weighting our THD, I'd lean towards the speaker as being preferable. Imagine jacking up the Onkyo's 20V distortion to the point where the graph starts at 1% or more instead of 0.01% to better compare with the magnitude of loudspeaker distortion. Better to sh*t the bed on the low end where our hearing is less sensitive and be relatively clean through the rest of the range IMO.

I see you points Steve, but on the other hand, it is the reason why I added my cautionary notes on reference to our hearing being less sensitive to distortions at the low end, as it may lead people to think it is okay to have higher distortion at say below 200 Hz but not so at above 10,000 Hz when it really is more complicated, such that the opposite to what people think (again it depends on the terms of reference) could be true.

For better clarity, I think we have to state whether we are talking about the test signal's fundamental frequency or harmonic frequencies. In the linked listening tests using tones, the website explained that and it can be verified easily, as Pogre has done already.

Example:

Test signal frequency 200 Hz, harmonics would be 400, 600, 1,000, 1,400 Hz so the 3nd to 7th harmonics would begin to fall into the band we are sensitive to.

Test signal frequency 10,000 Hz, even the 2nd harmonic, would be at 20,000 Hz, and the more offending 5th and 7th harmonics would be in the ultra sonic range already.

So the Onkyo's 20 V distortions in the graph you attached really isn't much of a concern because if you look at the FFT, it will show that 0.2% THD are made up of harmonics in the ultrasonic range, when the fundamental of the test frequency is 20,000 Hz.

Let me know if I misunderstood something on this.

 

[highfigh]

Right, but keep in mind in a studio environment, using the best available electronics and loudspeakers you may be able to score higher than Steve81, but then it is still just -42 dB!!

So I think we can probably agree claims of night and day difference between a musical Arcam AVR/AVP and a clinical Yamaha AVR/AVP, or an ATI flagship amp vs the often trashed internal amp of even a flag ship AVR must be due to reasons other than those directly related to the said DUTs (devices under test).

As Peter Walker allegedly had said in the interview I linked earlier (still hoping @TLS Guy will confirm credibility of that interview):

An amplifier should, within its limits of voltage and rate of change of voltage,
(which is slew rate limiting) if you keep within those two it should be very much
better than any program material. These are the things that are measured at .01
per cent or .05 per cent. But what is listened to is usually a program with 2 or 3
per cent distortion in the first place. That's the least you can get on records,
tapes, and such things. Listening tests are usually not done in this region of .01
percent distortion. ...................................................................................................If
people test two amplifiers and say, "These sound different," there's no magic in
it. Spend two days, maybe a whole week in the lab, and you find out exactly why
they're different and you can write the whole thing down in purely practical,
physical terms. This is why these two sound different, and the cause is usually
peripheral effects. It is not really a case of good or bad amplifiers, it's that the
termination impedances are wrong, or something of that sort.

Coming from the founder of Quad Electroacoustics and the one attributed with the famous hifi quote "the perfect amplifier is a straight wire with gain", it is both surprising and unsurprising.

Back to your point, I supposed it is reasonable to aim for 0.05 to 0.01% even when 0.1 to 1 % may be all that matters. It is not totally unlike if you feel a couple of million is all you would ever spend post retirement based on your desired life style, yet you still won't consider it good enough unless you have saved up 5 millions.

I agree if we’re just testing our hearing acuity.

I thought we’re trying to see if we could hear THD+N in the real world environment.

For example, why should everyone care about THD+N of -100dB if we can’t even hear THD+N of -20dB in the real world environment with real speakers and amps and music, even in an extremely quiet room?

Our ears are the input reference and their performance needs to be known if we're going to develop a system that meets our needs.

It's the same circle that occurs when designing a microphone- how can someone know it's better if the playback and other circuits can't use what it's sensing and passing on and how can they know the playback equipment is as good as the device used to sense the sounds? We can measure the equipment objectively, but it's not as easy with our ears.

Audiologists should use the same kind of equipment or, since its application has been standardized, the specs need to be used to make sure the tests are similar enough that large variations don't occur if someone goes to more than one test site. OTOH, they don't generally concern themselves with anything above 8KHz because they work in speech intelligibility, but it's works the same.

A reference system is needed for these tests- without that, the variables are too common for people to think that they're hearing the same sounds.

 

[Steve81]

Let me know if I misunderstood something on this.

I agree that THD for fundamentals in the high treble range isn't all that concerning, but that's not exactly what I'm talking about.

Looking at the speaker THD graph, distortion reaches a peak of -10dB from the fundamental around 60Hz, and from 150Hz-20kHz, it's better than -40dB.

Switching over to the Onkyo's 20V graph, lets start by looking at 20kHz, where THD is 0.3%. Distortion has only dropped by 10dB by the time we get to ~5kHz, 20dB by the time we get to ~1.5kHz, and 30dB by the time we get to ~350Hz. This is of course no big deal given the absolute levels of THD involved with the amp. However, if we artificially ramp those levels up so that THD @ 20kHz is more comparable to what the speaker is doing on the low end, that doesn't seem favorable at all.

 

[PENG]

I agree that THD for fundamentals in the high treble range isn't all that concerning, but that's not exactly what I'm talking about.

Looking at the speaker THD graph, distortion reaches a peak of -10dB from the fundamental around 60Hz, and from 150Hz-20kHz, it's better than -40dB.

Switching over to the Onkyo's 20V graph, lets start by looking at 20kHz, where THD is 0.3%. Distortion has only dropped by 10dB by the time we get to ~5kHz, 20dB by the time we get to ~1.5kHz, and 30dB by the time we get to ~350Hz. This is of course no big deal given the absolute levels of THD involved with the amp. However, if we artificially ramp those levels up so that THD @ 20kHz is more comparable to what the speaker is doing on the low end, that doesn't seem favorable at all.

Yeah the way you look at it, it favors the speaker, though in absolute sense the Onkyo amp is still better in terms of having lower THD.

Now obvious one could pick another speaker such as the popular Martin Logan Motion 40:

It doesn't look too good in the 200 to 4000 Hz right but I am sure a lot of people wouldn't mind the infusion of 2nd and 3rd harmonics for the fundamental frequencies in that range.

Regardless, thank you very much for showing me how to look at this way, very interesting, why didn't I think of it?

SoundStageNetwork.com | SoundStage.com - NRC Measurements: MartinLogan Motion 40 Loudspeakers

 

[Steve81]

Now obvious one could pick another speaker such as the popular Martin Logan Motion 40:

It doesn't look too good in the 200 to 4000 Hz right but I am sure a lot of people wouldn't mind the infusion of 2nd and 3rd harmonics for the fundamental frequencies in that range.

Yeah, I aimed to pick a reasonably well designed speaker to compare with reasonably well designed amps; the THD isn't what first catches my attention with that graph

Seriously though, when we look at loudspeaker distortion, it all comes down to driver choice, and how well the XO is designed to mask the flaws / limitations of said drivers. For example, it's not hard to imagine a cheap tweeter crossed at a relatively low frequency with a low order XO taking a massive dump in the midrange. Similarly at the low end, that's part of why we cross over to subwoofers.

Regardless, thank you very much for showing me how to look at this way, very interesting, why didn't I think of it?

Clearly, you're not as obsessive/compulsive about looking at graphs.

 

[PENG]

Didn't realize Amir has a thread on the Klippel test and held a context using the BIx (Bl(x): The variation of the force factor with displacement) test sample.

That must be a more difficult test because it only had the one kind of distortions, instead of all 3 from speakers:

You can see a lot of scores of the ASR members, some did really good but apparently only because they got experienced enough to find the clue/hint..

Distortion Listening Test | Audio Science Review (ASR) Forum

 

[Bonscott]

Maybe everyone should go to a doctor and have there hearing tested. Base how much you spend on an amplifier based on the results of the test

 

[PENG]

Maybe everyone should go to a doctor and have there hearing tested. Base how much you spend on an amplifier based on the results of the test

Funny but true, at least to some extents.

 

[PENG]

I think the following AP article is "must read" one for those interested in measuring distortions:

THD and THD+N – Similar, but not the same - Audio Precision (ap.com)

I wish I could add it to the first page as the first link included, and I guess a moderator may be able to do it?

 

[MasterApex]

Listening Test - How it works (klippel.de)
Blind listening tests you can try (no info on what kind of harmonics introduced to the test samples):
Listening Test (klippel.de)

Nice summary Peng.
I took the Klippel test with with my Dell Laptop since my main Stereo system is not connected to PC/Internet Browser.

It get hard after -24dB but I tried to focus on the signature of the distortion note , instead of the good note.
Here is my result with Dell Precision Laptop 5510 speaker

 

[AcuDefTechGuy]

So what’s the practical real-life takeaway message?

Can people actually hear THD+N of less than 0.1% 20Hz-20kHz with their speakers in their rooms?

 

[PENG]

So what’s the practical real-life takeaway message?

Can people actually hear THD+N of less than 0.1% 20Hz-20kHz with their speakers in their rooms?

Sorry I am going to have to give a long answer and my comments below refer to THD+N only, as that's your question. Though if THD is low, IMD would tend to be low too.

After reading all that I could find on the internet about THD I would say no one (except real golden ear types) could hear a difference unless the harmonics are all or almost all in the higher order harmonics such as the 5th, through 9th. or even higher orders. If you look at the FFTs that Gene and Amir ever published on any >$1000 popular brand amps, receivers and AVRs bench test reports you won't likely find anything that would show a lot of high order harmonics shown in the FFTs.

My reasoning is simple:

- Most trusted reviewers, again including Gene, Amir or even JA (Stereophile) have at one time or another, alluded to 0.1% being below the threshold of audibility. But..., see the next..

- Some argue, mostly for argument sake that it is not the % that matters but the profile, or orders of the harmonic contents, but then..

- Most so called "well designed.." amps don't really have most or all of the harmonics concentrated in the higher orders higher than the 3rd.

- If you consider the typically used 1,000 Hz sine wave, the 11th harmonics is then 11.000 Hz, now find me an amp that has the 11th harmonics higher than even 0.001%

Talk is cheap, so let's just look at the last AVR Gene bench tested, and a couple of cheap entry level AVRs Amir measured:

First, take a look of Gene's measurements on the SR8015, that's a nice AVR but still just an AVR with 13 channels jammed back into one box! He measured it at 1 W, so keep in mind things should get better at higher output levels.

Marantz SR8015 Measurements & Conclusion | Audioholics

And by the way, note that Gene said:

"These figures should give any skeptics cause to rethink their preconceptions that receivers can't have stout amplifiers."

Next we'll see some examples using much lower AVR models that Amir measured:

Amir, and also Gene typically don't measure FFT on AVR speaker outputs because they usually have pre outs for them to measure, that's why I have to pick on the Denon AVR-X2200 that does not have pre out.

Amir measured the entry level Denon at 5 W output, but since THD would typically improve somewhat at rated output, measuring at 5 W would seem quite sufficient for this example.

See that the 5th harmonics was only about -90 dB, that was 0.00316%, and any harmonics at >10,000 Hz would be at less than -110 dB = 0.000316%, almost at the limit of Amir's or Gene's AP instrument!!

Denon AVR-X2200W Audio/Video Receiver Review | Audio Science Review (ASR) Forum

Same thing with the Yamaha RX-A1080:
Review and Measurements of Yamaha RX-A1080 AVR | Audio Science Review (ASR) Forum

 

[sterling shoote]

It seems that almost every other week, this forum will get questions on something like the following:

- Will adding a power amp to a receive improve sound quality?
- Do amps have their own sound signature even if their THD+N are "low", with flat FR 20-20,000 Hz, DF>20, SNR/DR>100 dB A weighted etc., that is, will they sound the same if the specs are the same and equally good?

I have read tons of articles on such related topics and I believe if one reads most of the one I linked below, one will find the answer, at least the answer they want to hear/read.
Otherwise, all I can say is, it depends.

For example, an amp with THD+N <= 1% at rated output may sound better than one specified <=0.05% but if both are rated <=0.05%, 20 to 20,000 Hz at output level between 0.1 W and 200 W/300 W into 8 and 4 Ohms will most likely sound the same, all else (such as FR, DF, IMD, XT, SNR/DR) being equal, and both driven to say 30 W maximum level, into an 8 ohm nominal moving coil loudspeaker. Other than that, I guess you'll have to read the good stuff, ignore the bad ones, and you'll be the judge.

Articles (I read at least parts of all of them, found no apparent misconceptions in terms of the EE theories and math. that I am familiar with):

Edit: I highly recommend the ones highlighted bold, if any of the linked articles appear under more than one section, it is because they fit under both, so for convenience I put them under the sections they fit in.

Note: The THD listening test at audiocheck.net that someone linked on ASR is fun to do, I hope some will try it and give feedback. The klippel test is fun too, I didn't do well on that so I guess SINAD of 75 dB is more than good enough for me, but again, we don't know how accurate and scientific it is, and your score would depend on your room noise, using headphone vs speakers etc.

For such test using music, try the Klippel one, that is however, more about loudspeaker distortions.

Definitions related:

Total harmonic distortion - Wikipedia
Confused About Amplifier Distortion Specs? | Analog Devices
Dyamic_Performance_Testing_of_Digital_Audio_D/A_Converters
Understanding, Calculating, and Measuring Total Harmonic Distortion (THD) - Technical Articles (allaboutcircuits.com)
Understanding instrumentation for measuring total harmonic distortion (testandmeasurementtips.com)
Total Harmonic Distortion | Universal Audio (uaudio.com)
Total Harmonic Distortion (THD): Is It a Good Indicator of Sound Quality? | Audioholics
10 Things about Audio Amplifiers You've Always Wanted to Know | Audioholics
THD and IMD Distortion—Sidebar | Audioholics Note: (I would be careful reading the first par by Steve Feistein, not saying he's wrong about anything but the part about people's sensitivity to low frequency distortions could be misleading imo).
Distortion In Power Amplifiers (douglas-self.com)
Intermodulation Distortion (sound-au.com)
Microsoft Word - AES06Gedlee_ll.doc
Microsoft PowerPoint - The Perception of Distortion.ppt [Compatibility Mode] (gedlee.azurewebsites.net)
Audio Distortion: What is It and Why Should You Care? | AV NIRVANA
Audibility of Phase Distortion (silcom.com)
Application notes - Audio Distortion Measurements (bo0385) (bksv.com)
Interpreting THD Measurements - Think dB not Percent! - Benchmark Media Systems
How Much Distortion Can We Hear With Music? - Blog | Axiom Audio | Axiom Audio (Be careful reading this one so you don't get mixed up noise with distortion, a couple times the author said distortion but he clearly meant noise).
The sound of Distortion (pmillett.com)

THD Measurements related:

Confused About Amplifier Distortion Specs? | Analog Devices
Dyamic_Performance_Testing_of_Digital_Audio_D/A_Converters
THD and THD+N – Similar, but not the same - Audio Precision (ap.com)
More about THD+N and THD - Audio Precision (ap.com)
AMH-1_2nd_Ed-4.vp (ap.com)
Mathew-AudioPrecisionSpecifications.pdf (sengpielaudio.com)
Total Harmonic Distortion | Universal Audio (uaudio.com)

Listening tests related articles/video:

Amplifier Distortion (Bob Carver, 1973) - Interviews/news - TheCarversite!
Audio Myths Workshop - YouTube
Listening Test - How it works (klippel.de)

Blind listening tests you can try (no info on what kind of harmonics introduced to the test samples):

Listening Test (klippel.de)
THD - The Extended Set (125Hz) (audiocheck.net)
Online Intermodulation Distortion (IMD) Sound Test (audiocheck.net)

Some DBT results in recent tests:

Archimago's Musings: BLIND TEST RESULTS Part II: "Is high Harmonic Distortion in music audible?" Respondent Results

Edit 1: Added one written by Doug Self, on "The distortion mechanisms of a generic power amplifier".
Since I am interested in DIY amps, about to start the 2nd one, I may buy his book:
The Audio Power Amplifier Design Handbook (douglas-self.com)

Edit 2: I think the following AP article is "must read" one for those interested in measuring distortions:
THD and THD+N – Similar, but not the same - Audio Precision (ap.com)

My greatest sense for a need to match amplification to speakers was back in 1986, at which time I paired a Sony TA-N55ES Power Amp operating at 300 watts in bridged mode to a 15 inch JBL B380 Subwoofer, which disregarded JBL's recommendation to power the speaker with a 600 watt amp. At any rate, 300 watts resulted in frequent clipping, which sounded like a sledge hammer hitting a steel pylon, so I connected a 560 watt amp to the sub and no more clipping. In my other application for power amps, powering my JBL L100t3's, I have experimented with amps having from 100 watts per channel to 560 watts per channel. Not having been able to discern one iota of divergence in sound at any volume from one amp to another, I am now powering my L100 mains, as well as center and rear channels by a 21 year old Sony TA-N9000ES 110 watt per channel 5 channel power amp. It gets the job done, so the take I have on all of this is just put your speakers on the amplification diet they need and don't be too persuaded by pro reviewers who wax eloquently about any amplifier.

 

[sterling shoote]

Here's a thought, if the measurements from two amps are grossly divergent yet the amps sound alike, which could be a typical scenario, then is it possible that the things we are measuring are less useful than things not measured or measurable?

 

[AcuDefTechGuy]

Here's a thought, if the measurements from two amps are grossly divergent yet the amps sound alike, which could be a typical scenario, then is it possible that the things we are measuring are less useful than things not measured or measurable?

One reason hardly anybody does measurements anymore. As many of us have been saying for years and years - all these specs are inaudible to non-golden-ears.

 

[PENG]

Here's a thought, if the measurements from two amps are grossly divergent yet the amps sound alike, which could be a typical scenario, then is it possible that the things we are measuring are less useful than things not measured or measurable?

Of course, but chances are that for something as simple as audio amps (relatively speaking) you would think that what needs to be measured have long been figured out so it is highly unlikely that we are not measuring things that are useful. So if by "grossly divergent" means something like 5% vs 0.1% THD+N then if they sound alike to someone, it could mean that "someone" simply could not, for whatever reasons, tell a difference between 5% and 0.1% THD and that is very possible because there are people who are not bothered by even 10% THD (yes it depends on the kinds of harmonics too).

Have you taken the Klippel test yet, to see at what THD level you can tell a difference? Even if you don't believe its validity and/or relevance it could still be fun..

Listening Test (klippel.de)

 

[MasterApex]

Here's a thought, if the measurements from two amps are grossly divergent yet the amps sound alike, which could be a typical scenario, then is it possible that the things we are measuring are less useful than things not measured or measurable?

Just like food tasting, we have to eat it , not just measuring the ingredients (% salt, sugar, protein, etc) to appreciate the full presentation.

I believe there are aspects of audio gears (speakers/amps) music reproductions that are beyond current measurement (%THD, etc) that is based on sinewave signal. That is why I look at spec but also do listening test.

 

[PENG]

My greatest sense for a need to match amplification to speakers was back in 1986, at which time I paired a Sony TA-N55ES Power Amp operating at 300 watts in bridged mode to a 15 inch JBL B380 Subwoofer, which disregarded JBL's recommendation to power the speaker with a 600 watt amp. At any rate, 300 watts resulted in frequent clipping, which sounded like a sledge hammer hitting a steel pylon, so I connected a 560 watt amp to the sub and no more clipping. In my other application for power amps, powering my JBL L100t3's, I have experimented with amps having from 100 watts per channel to 560 watts per channel. Not having been able to discern one iota of divergence in sound at any volume from one amp to another, I am now powering my L100 mains, as well as center and rear channels by a 21 year old Sony TA-N9000ES 110 watt per channel 5 channel power amp. It gets the job done, so the take I have on all of this is just put your speakers on the amplification diet they need and don't be too persuaded by pro reviewers who wax eloquently about any amplifier.

I was able to find the SM and OM of the TA-N55ES and it seemed like an excellent amp for its time. You simply pushed it too hard trying to get the best of a big passive subwoofer!! Every time you pushed it pass the 500 W (8 ohms) mark it's only to clip, no doubt about that.. The amp is really only rated 300 W into an 8 ohm load bridged to mono.

The 600 W rating could only be reached (or exceed, for short durations) without clipping if the impedance of the subwoofer happened to be at 4 ohms at that moment. It would be great if you have the impedance/phase angle curve of that JBL B380.

 

[cpp]

One reason hardly anybody does measurements anymore. As many of us have been saying for years and years - all these specs are inaudible to non-golden-ears.

Yep and for some as we get older, could careless if equipment measurements "are awesome", hell , I'll let my old ears tell me what I like. Hell I can't hear measurements anyway

 

[AcuDefTechGuy]

Yep and for some as we get older, could careless if equipment measurements "are awesome", hell , I'll let my old ears tell me what I like. Hell I can't hear measurements anyway

Yeah, I've said this before. But these electronic measurements only became a "issue" when publications (Stereophile, S&V) STOPPED doing them!

Back in the days when everyone was doing these electronic measurements, we were all talking about the measurements for fun. But it never seemed like a big deal. We still looked at SNR, Crosstalk, THD+N and even damping factors and slew rates. It wasn't about sound quality. More about bragging rights.

But since most publications have stopped doing these measurements, all of the sudden, it seems like measurement is the only thing that matters to some people.

There is nothing wrong with having all the measurements in the world. The problem is what some people do with the measurements.