• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Complex Load for Power Amplifier torture testing

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
That "industry standard testing protocol" was made to draw 0,00001% THD and 5-100000 khz. In other words that protocol was implemented to sell crap.



Yes, it is a way to compare different amps. But it's a compete ignorance to judge how those amp will work on real load.



Sure it can. Take any 2-3 way speaker with more or less complex crossover and such load can simulate every situation. You all forget also that any speaker is also a generator (press on your woofer by your hand while it is connected to multimeter and you'll see it generates something). And when such generator is powered by amp with deep negative feedback it all cause very high distortion (sorry, my english is limited. I don't know all terms).
And when we're measuring even on such load as topicstarter made we don't see that effect.
1, Can you hear 20000hz? If answer is no then nothing is wrong with measuring up to 10khz for THD.
2, Negative feedback only improves immunity of back EMF not the other way around. Again, need proof.
 

JohnYang1997

Master Contributor
Technical Expert
Audio Company
Joined
Dec 28, 2018
Messages
7,175
Likes
18,292
Location
China
Answer on all questions is yes. But I'm pretty sure that when I'll show you what is happening on same frequency on resistor and on speaker on a screen of my oscilloscope you'll have 200 more questions and will try to deny my measurements cause they're oppoisite to what religious sect named ASR is used here.
Where is your proof? I don't need a YES that's actually false. Also I have designed amplifiers with extreme low distortion. I don't think most things you said is accurate or remotely correct.
 
OP
pma

pma

Major Contributor
Joined
Feb 23, 2019
Messages
4,591
Likes
10,727
Location
Prague
I would like to show an effect of inadequate (underrated) power supply transformer to amplifier parameters in case of heavy complex load.

Power amp module with 4 pairs of MJL3281/1302 output transistors is supplied by the power supply with 100VA transformer. This power supply gives 2 x 45Vdc at power amp supply terminals at idle (no input signal). The amplifier is loaded with R//(R+C) load, 4ohm//(0.44ohm+6.8uF). This load would not be unrealistic if we speak about ESL panels.

Now, in case of 1kHz 14Vrms output voltage, into this load, power supply DC voltage at amp supply terminals drops from 2x45V idle to 2x35V loaded. This determines maximum output power at low frequencies where the load is almost resistive and close to 4ohm.
If we increase frequency to 10kHz and keep same input signal voltage as in the 1kHz test, the power supply voltage drops to 2x25V, because of the inadequate, underrated power transformer, which is not able to feed current demands of the amplifier operating to 1.9ohm/-50° at 10kHz.

We can see what the inadequate transformer makes with power frequency response and distortion, in the plots below. The top curve is always for 17.6Vrms output voltage, curves are stepped by -1dB.

R-RC_load.png


trafo100VA_14V_step1dB.jpg

Frequency responses into complex load, with inadequate 100VA PSU transformer. From 17.6Vrms output voltage step -1dB.

trafo100VA_inadequacy.png

Distortion vs. frequency with inadequate PSU transformer, due to DC voltage drop. Fs=48kHz, so there is a jump at 7.5kHz in higher Vo plots, as H3 measurement finishes there.
 
Last edited:

eboleyn

Member
Forum Donor
Joined
Sep 28, 2019
Messages
92
Likes
98
Location
Portland/OR/USA
An interesting thread here, though I see it's gone quiet for a while. Plenty of, err, "spirited" debate. I'll preface this by saying I'm not up to speed on back-EMF from speaker drivers and reactive loads, but my interest in reactive load measurements are more oriented in low-wattage performance cases.

At one point a few years ago, I was comparing some amps at low levels, that is outputting less than 1 watt in my experiments for 50W+ rated amps to mostly 4 ohm but a few 8 ohm loads on a few different multi-way speakers (from 2 to 4 ways). I'd definitely noticed that "difficult" vs "simple" speaker loads can make a huge difference in the ability of some amplifiers to get said speaker to accurately reproduce the signal. The speaker models I won't comment on as I'm not sure it's relevant to the point here.

The one higher-end class D amp I tested is a NAD C390DD (goes directly from digital to amplified signal driving speakers, sometimes called a "PowerDAC"), and was comparing it to a class A (Pass Labs XA30.8) vs. a high-current A/B amp (Pass Labs X250.8). I tried a few others in passing but these were the ones I focused on.

The long and short of it is that the simpler the speaker load I used, the more the NAD unit shined. It really didn't seem to like the complex crossovers. The Pass XA30.8 was in the middle, where it always sounded pleasant, but had the well-reported result of being "weak" in the bass on the more difficult-to-drive speakers and in one case had some other artifacts. The standout was the Pass X250.8, which drove everything extremely consistently and pleasantly.

Having said the above, when recently I have been performing digital crossover bi-amp experiments with no passive crossover at all between the amps and the individual drivers, the NAD unit sounded simply stunning, possibly the best of all. For this setup the NAD is somewhat inconvenient to use since it's not well-suited to being just a power amp, so I'm acquiring some Hypex NC400 monoblocks to see if they behave similarly or not.

Anyway, all of this is just fodder for the commentary where both complex crossovers and the back-EMF from multiple drivers I'm sure gets messy, so love to hear about measurements that try to account for this even somewhat.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Perhaps you could share measurements and methodology?
 

eboleyn

Member
Forum Donor
Joined
Sep 28, 2019
Messages
92
Likes
98
Location
Portland/OR/USA
Sadly the above results were using human ears (mine and some friends), so not nearly the level of what's being used in this forum. We tried to correct for levels to get matching around 0.5 db or less, but it wasn't super-rigorous.

I'm quite interested in better dynamic measurements of whole chain response (from digital source out through a speaker then measured via microphone) and have been tinkering in that area due to some personal subjective experiences of "better quality" audio from a RAAL 70-20xr ribbon tweeter which objectively by THD standards measured no better than another tweeter, but don't have anything worthwhile to say on it so far.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Sadly the above results were using human ears (mine and some friends), so not nearly the level of what's being used in this forum. We tried to correct for levels to get matching around 0.5 db or less, but it wasn't super-rigorous.

Clearly. :cool:

The two things that matter are source impedance low enough to not have audible frequency response errors with the target load, and the ability to stably drive current into that load. Both incredibly well-understood for decades and both routinely done except for the very cheapest and very most expensive amps.
 

eboleyn

Member
Forum Donor
Joined
Sep 28, 2019
Messages
92
Likes
98
Location
Portland/OR/USA
Clearly. :cool:

The two things that matter are source impedance low enough to not have audible frequency response errors with the target load, and the ability to stably drive current into that load. Both incredibly well-understood for decades and both routinely done except for the very cheapest and very most expensive amps.

Yes I'm aware of this issue. The amps I listed all are sub-0.1 ohm across the frequency spectrum (modulo the "impedance setting" on the NAD unit which is kind of strange) and even tested output level at several frequency points to determine if the response was flat. But that was just calibration. The actual results were using human ears and so would still fall well into the "subjective" category.

I'm indeed interested in performing these kind of measurements more rigorously using equipment, but don't yet know the right criteria I'm looking for.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
The actual results were using human ears and so would still fall well into the "subjective" category.

Unfortunately, no, this was not by ear if you didn't put in basic ears-only controls. So the differences you think you heard may or may not be real. Start there to confirm that there actually ARE audible differences before trying to figure out what you need to measure.
 

amirm

Founder/Admin
Staff Member
CFO (Chief Fun Officer)
Joined
Feb 13, 2016
Messages
44,368
Likes
234,386
Location
Seattle Area
I'd definitely noticed that "difficult" vs "simple" speaker loads can make a huge difference in the ability of some amplifiers to get said speaker to accurately reproduce the signal.
I built a complex load emulating a 2-way speaker and started to run it on a few class-D amp. Except for very cheap amps (less than $100), it didn't show anything revealing other than a fraction of a dB variation around 20 kHz so I stopped running it.
 

March Audio

Master Contributor
Audio Company
Joined
Mar 1, 2016
Messages
6,378
Likes
9,317
Location
Albany Western Australia
An interesting thread here, though I see it's gone quiet for a while. Plenty of, err, "spirited" debate. I'll preface this by saying I'm not up to speed on back-EMF from speaker drivers and reactive loads, but my interest in reactive load measurements are more oriented in low-wattage performance cases.

At one point a few years ago, I was comparing some amps at low levels, that is outputting less than 1 watt in my experiments for 50W+ rated amps to mostly 4 ohm but a few 8 ohm loads on a few different multi-way speakers (from 2 to 4 ways). I'd definitely noticed that "difficult" vs "simple" speaker loads can make a huge difference in the ability of some amplifiers to get said speaker to accurately reproduce the signal. The speaker models I won't comment on as I'm not sure it's relevant to the point here.

The one higher-end class D amp I tested is a NAD C390DD (goes directly from digital to amplified signal driving speakers, sometimes called a "PowerDAC"), and was comparing it to a class A (Pass Labs XA30.8) vs. a high-current A/B amp (Pass Labs X250.8). I tried a few others in passing but these were the ones I focused on.

The long and short of it is that the simpler the speaker load I used, the more the NAD unit shined. It really didn't seem to like the complex crossovers. The Pass XA30.8 was in the middle, where it always sounded pleasant, but had the well-reported result of being "weak" in the bass on the more difficult-to-drive speakers and in one case had some other artifacts. The standout was the Pass X250.8, which drove everything extremely consistently and pleasantly.

Having said the above, when recently I have been performing digital crossover bi-amp experiments with no passive crossover at all between the amps and the individual drivers, the NAD unit sounded simply stunning, possibly the best of all. For this setup the NAD is somewhat inconvenient to use since it's not well-suited to being just a power amp, so I'm acquiring some Hypex NC400 monoblocks to see if they behave similarly or not.

Anyway, all of this is just fodder for the commentary where both complex crossovers and the back-EMF from multiple drivers I'm sure gets messy, so love to hear about measurements that try to account for this even somewhat.
Well you need to cautious of some of the information posted in the thread. Creating pathalogical situations and loads that are totally unrepresentative of the real world is not useful and in fact potentially misleading.

PMAs post above showing a load impedance falling to 1 ohm at 20kHz is encountered with some speakers, but not many, and is erring toward the more extreme. However the big mistake he has made there is applying a 14 volt signal across the entire frequency range. Music simply just does not behave like this so its an unrealistic test providing no useful information. Music will typically have signal levels at 20kHz that are 60dB or more lower than at 1kHz. This means a 14 volt signal at 1khz will fall to just 0.014 volts ! So the amplifier (or power supply) is simply not taxed, and will never be, as is indicated in that post.

So what he considers an "inadequate" power supply is in fact [probably what i would consider an "intelligent" design. It hasnt been over engineered to account for situations that just dont exist in the real word. A larger transformer that is not needed would just add significant cost for no benefit.

You can see my point perfectly demonstrated in one of PMAs graphs.

When the signal level falls by just 6dB (not 60dB) the output remains consistent to 20kHz.

1593395477503.png


Second thing to note is that music is transient in nature. Its not a consistent load. Unregulated supplies, as is typical with amplifiers, will cope better with transients and maintain their output voltage better than with constant loads.
 
Last edited:

AnLaoJin

Member
Joined
Mar 18, 2020
Messages
69
Likes
9
Location
China-HeNan
An interesting thread here, though I see it's gone quiet for a while. Plenty of, err, "spirited" debate. I'll preface this by saying I'm not up to speed on back-EMF from speaker drivers and reactive loads, but my interest in reactive load measurements are more oriented in low-wattage performance cases.

At one point a few years ago, I was comparing some amps at low levels, that is outputting less than 1 watt in my experiments for 50W+ rated amps to mostly 4 ohm but a few 8 ohm loads on a few different multi-way speakers (from 2 to 4 ways). I'd definitely noticed that "difficult" vs "simple" speaker loads can make a huge difference in the ability of some amplifiers to get said speaker to accurately reproduce the signal. The speaker models I won't comment on as I'm not sure it's relevant to the point here.

The one higher-end class D amp I tested is a NAD C390DD (goes directly from digital to amplified signal driving speakers, sometimes called a "PowerDAC"), and was comparing it to a class A (Pass Labs XA30.8) vs. a high-current A/B amp (Pass Labs X250.8). I tried a few others in passing but these were the ones I focused on.

The long and short of it is that the simpler the speaker load I used, the more the NAD unit shined. It really didn't seem to like the complex crossovers. The Pass XA30.8 was in the middle, where it always sounded pleasant, but had the well-reported result of being "weak" in the bass on the more difficult-to-drive speakers and in one case had some other artifacts. The standout was the Pass X250.8, which drove everything extremely consistently and pleasantly.

Having said the above, when recently I have been performing digital crossover bi-amp experiments with no passive crossover at all between the amps and the individual drivers, the NAD unit sounded simply stunning, possibly the best of all. For this setup the NAD is somewhat inconvenient to use since it's not well-suited to being just a power amp, so I'm acquiring some Hypex NC400 monoblocks to see if they behave similarly or not.

Anyway, all of this is just fodder for the commentary where both complex crossovers and the back-EMF from multiple drivers I'm sure gets messy, so love to hear about measurements that try to account for this even somewhat.


My point of view is the same as yours.
Class D amplifier itself is an excellent amplifier, but it is only suitable for speakers with active crossover, for narrow frequency amplification and simple impedance load. Passive speakers are limited by complex dividers with three or even four-way frequencies. The final result is complex back-EMF and intermodulation, as well as mixed resistance, capacitance and reactance. Using Class D to load the entire speaker will cause a mess of bad sound.
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
Class D amplifier itself is an excellent amplifier, but it is only suitable for speakers with active crossover, for narrow frequency amplification and simple impedance load. Passive speakers are limited by complex dividers with three or even four-way frequencies. The final result is complex back-EMF and intermodulation, as well as mixed resistance, capacitance and reactance. Using Class D to load the entire speaker will cause a mess of bad sound.

Evidence? That just seems like a buzzword salad.
 

AnLaoJin

Member
Joined
Mar 18, 2020
Messages
69
Likes
9
Location
China-HeNan
Evidence? That just seems like a buzzword salad.
Class D amplifiers are clean and efficient as simple loads and resistive loads. But the class D amplifier under complex load is far inferior to the class AB with large current and low internal resistance. . .
 

tmtomh

Major Contributor
Forum Donor
Joined
Aug 14, 2018
Messages
2,636
Likes
7,490
Class D amplifiers are clean and efficient as simple loads and resistive loads. But the class D amplifier under complex load is far inferior to the class AB with large current and low internal resistance. . .

Repeating the same assertion does not make it any truer. As @SIY asked before, what evidence do you have to support this claim?
 

eboleyn

Member
Forum Donor
Joined
Sep 28, 2019
Messages
92
Likes
98
Location
Portland/OR/USA
My point of view is the same as yours.
Class D amplifier itself is an excellent amplifier, but it is only suitable for speakers with active crossover, for narrow frequency amplification and simple impedance load. Passive speakers are limited by complex dividers with three or even four-way frequencies. The final result is complex back-EMF and intermodulation, as well as mixed resistance, capacitance and reactance. Using Class D to load the entire speaker will cause a mess of bad sound.

Well, to be clear, I was only speaking about that specific class D amp on hand, the NAD C390DD, because it was all I had at the time. The other ones were also specific class A (Pass Labs XA30.8) and class AB (Pass Labs X250.8) amps which have been famous for the characteristics mentioned. My comments were not meant to be a declaration that class D or A or AB amps necessarily all had those particular characteristics. FWIW, I did indeed get both Hypex NC400 and Purifi 1ET400A amps and have in my own listening found them to be better than the NAD C390DD's for driving said passive speakers from my earlier post, which doesn't seem surprising given the amazing measurements of the NC400 and 1ET400A, but that is fodder for a separate reply.

I love audiosciencereview's (and many of the users') dedication to real measurements and the science of audio as opposed to being satisfied by anecdotal descriptions. But science is a dualistic process in which personal/anecdotal experience often helps in our understanding of data. We develop theories and make measurements, then if they don't quite correlate with experiences, we search for a better explanation and better ways to measure.

Static tone THD was considered the standard for many years as a basic "can it even drive a static frequency at all". Static tone IMD has been added by many because it's been shown that they only loosely correlate.

I'd meant my comments in my earlier reply mentioning complex back-EMF issues mainly to support the notion that we could probably use some more kinds of measurement-based audio analysis. Another example, human hearing is much more discerning to the relative difference in amplitude between 2 tones nearby in time or attack and decay than the absolute amplitude of a tone with no reference.

Having said all of that, I'll be the first to admit that I don't know what such "other kinds of measurements" should be, though I am extremely curious how to tease out those kind of issues!
 

SIY

Grand Contributor
Technical Expert
Joined
Apr 6, 2018
Messages
10,383
Likes
24,749
Location
Alfred, NY
But science is a dualistic process in which personal/anecdotal experience often helps in our understanding of data. We develop theories and make measurements, then if they don't quite correlate with experiences, we search for a better explanation and better ways to measure.

No, this is not true. An "experience" needs to be validated first to see if it's related to sound or vision or whatever we're talking about. If it can be actually demonstrated, then one can look at how to measure it. When it comes to electronics, no-one has yet demonstrated that anything beyond relatively simple measurements is necessary.
 

eboleyn

Member
Forum Donor
Joined
Sep 28, 2019
Messages
92
Likes
98
Location
Portland/OR/USA
No, this is not true. An "experience" needs to be validated first to see if it's related to sound or vision or whatever we're talking about. If it can be actually demonstrated, then one can look at how to measure it. When it comes to electronics, no-one has yet demonstrated that anything beyond relatively simple measurements is necessary.

I won't dispute your comment that experiences need to be validated in some sense to be related to the topic at hand, but ... to be more concrete:

Science is a process of creating models and testing of models through experiments, taking data, and data interpretation. The "model" you use is hopefully related to the reality around us, and that very much *also* requires validation. The models of gravity have been revised multiple times throughout history, to the point where we now have what the scientific community considers an extremely accurate model (Einstein's General Relativity) where singularities in the equations lead to situations that we don't understand. My point being that it is a human endeavor where finding models that match a wider array of data is valuable.

When you say that in electronics, no-one has yet demonstrated that anything beyond relatively simple measurements in necessary, I find that a difficult statement to swallow. It would be quite easy to build a machine that gets great static THD and IMD yet is terrible for reproducing real music. Static THD and IMD are just proxies for the quality of an audio channel which have been found to be pretty reasonable on audio chains otherwise designed to handle the full bandwidth of human-perceivable frequencies.

Human perception of sound is also a difficult one to work with, as we have multiple sensor networks in our ears that don't just measure sounds in one way. At the very least we know there is something like an FFT machine that perceives different frequencies directly, plus there are other "edge" detectors which help us with soundfield positioning/localization. We also know that many humans seem to *like* some kinds of distortion, commonly referred to as a "Tube-y sound" since analog audio chains built with vacuum tubes can often be tuned that way.

So my curiousity about non-THD/IMD measurements and how variations in them might affect human perception came about because I'd had some experiences in audio equipment of essentially equivalent THD/IMD producing perceptibly different soundfield results. I'm not claiming that this obviates the science of "better audio quality", simply that I'm curious about how that might come about, or was I just strongly confused by something. I've been slowly setting up to perform some experiments on audio drivers, non-linearity in sound reproduction, and seeing if I can find a way to generate violations of dymanic relationships that the human ear cares about which would not be detected with static THD or IMD. I'm not sure I'll ever get anywhere with it, but I can't help but feel curious about it.
 

restorer-john

Grand Contributor
Joined
Mar 1, 2018
Messages
12,579
Likes
38,280
Location
Gold Coast, Queensland, Australia
It would be quite easy to build a machine that gets great static THD and IMD yet is terrible for reproducing real music.

Tell me, what is remotely 'static' about a sine wave? Also, what is 'real music' comprised of?
 
Top Bottom