@crymeanaudioriver
Current limiting occurs in amps without specific current limiting circuits too. As an example, a heating up power transformer coil may in effect serve as a current limiter. Another example is insufficiently sized capacitive power bank.
The article, nor I differentiated what was current limiting, however, I believe the two examples you gave are not. This comes back to the math above. Some EE's could probably jump in on that.
Some amplifiers employ dedicated current limiting circuits. For instance, distribution amplifiers - the ones powering loudspeakers in restaurants and such - tend to have them. This is usually advertised as “soft clipping” feature.
It makes more business sense to sacrifice some audio quality if a restaurant employee accidentally turned up the volume too much, rather than to blow up, or to significantly degrade longevity, of the amp and speakers.
Other amplifiers may only have emergency current limiters, perhaps the simplest of which is a fuse in amp power rails. Yet others, while having a power cable fuse, rely upon “natural” current limiters, like the ones I originally described, to protect the amp and speakers.
Look at the curves of THD vs output power characteristics of amplifiers, and you’ll see that typically, there is a rise in THD (and by extension in IM) long before the amp clips. The degree of such deterioration is typically frequency-dependent too.
Even I know that is about how the amplifier is designed and feedback. The feedback goes down as the frequency goes up. Going back to the math I learned today, as the feedback goes down, the stability will improve.
It is not only about stability. It could be, for instance, about sheer non-linearity of solid-state amplification devices. The more they swing, the higher the level of distortions that then need to be kept in check by the feedback.
This is indeed one of the mechanisms explaining the phenomenon of some of the amps distorting significantly more while they are connected to a speaker compared to when they are connected to a dummy resistive load.
I think you made that up. That does not make sense.
Logic is simple here. Reactive loads at certain times may require significantly more current provided by the amp. To provide this current, the solid-state amplification devices need to swing wider, which increases the level of distortions.
However, just like with the discussion of THD and IM, we need to take into account that the amp-speaker system can “ring” for some time, instead of turning into a self-supporting generator. Some of the replies in the thread below describe precisely such occurrences.
Which brings us back to the 99.9% of the time it does not happen. "Ring for some time"? You mean unstable. Again, even I know that. Perhaps you should not be the person trying to lecture me on this.
I take “unstable”, under certain conditions, at its classic Control Theory definition: self-generating under certain conditions. Ringing is different: it is a process that a stable amplifier, excited by a change in input, goes through while arriving to equilibrium.
For purely linear time-independent systems, there are equivalence theorems: you can usually see discussions of “poles” of transfer function or of functions related to it, which would characterize both phenomena through one mathematical framework.
For realistic systems, with meaningful nonlinearities, these equivalence theorems do not hold. So, you can have a stable amplifier ringing significantly more than the linear theory would predict, or vice versa.
However, some of the replies highlight the fact that in some other market segments, including that of affluent audiophiles, larger speakers employing exotic transducers and much more sophisticated crossovers are more prevalent, and thus the events of ringing and self-generation are much more probable.
This is conjecture on your part.
I guess we could ask opinions of bona fide audiophiles on Audiogon. Are you and your fellow hobbyists more likely to own speakers employing exotic transducers and more sophisticated crossovers, compared to general public?
Going back to the math I learned today, if the designer knows the transfer function, they can estimate with high probability if the amp will be unstable. You may want to read this:
https://d1.amobbs.com/bbs_upload782111/files_28/ourdev_548669.pdf
I saw this article - “Simple Self-Oscillating Class D Amplifier with Full Output Filter Control” by Bruno Putzeys - some years back. What can I say? Bruno is a competent amps designer.
Still, look at the diagrams in Chapter 2.3. If you replace the purely resistive load, depicted as the rightmost element, with a realistic crossover of a three-way speaker, you’ll realize that the effective output filter may be quite different from the original filter depicted.
It can be indeed immaterial, as the change in amp characteristics would remain within inaudibility corridor. Or it may become material, changing the frequency response and distortion profile of the combined system enough to be noticeable.
I had to read it 3 times, but this is very interesting too.
https://linearaudio.net/sites/linearaudio.net/files/volume1bp.pdf That pokes holes in all the so called arguments about feedback.
Saw this before too. Once again, Bruno is a competent amps designer. The discussion covers suitably enough nonlinear systems theory too, even though time-dependent aspects are simplified away.
Personally, I like amps with balanced amount of feedback. Both on engineering merits, and on the way they sound. So, I’m with Bruno on this one.
The communication part is letting the patient think they had influence when they had none.
In my book, this is called “deception”.
When someone quotes another poster, it is normal, as one would also do here, to cut out what the replyer considers extraneous content w.r.t. their reply. That is not redacting, that is editing. I have not seen nor experienced ASR "redact" anything, though they will suggest less harsh language.
Removing the original post, like ASR sometimes does, may leave out the context in which a certain statement, quoted by another member, was made. Then, a reply to that statement may employ all kinds of rhetorical tricks, for instance moving the goalposts.
Nope, topic at hand is relevance of the testing Amir does on specifically power amplifiers to the subjective perception of audible distortions contributed by amp A vs amp B when connected to a specific speaker.
Except we come back to 99.9% or more of amplifier / speaker combinations will not have stability problems, ASR does not test tube amplifiers often, and based on my research, however, limited, that even audiophile speakers do not commonly have extreme characteristics, then there will be no change, at the amplifier level, with almost all speakers.
Here we’d need to ask our bona fide audiophiles. Do you believe the issue of an amp synergy with a speaker was only important in 0.1% of pairing cases you personally experienced?
My position, as is the position of majority of ASR members with practical experience in designing and repairing power amplifiers who cared to express their opinions, is that the testing Amir has been conducting is marketed as more definitive than it shall be based on scientific understanding of the limited nature of the tests.
I will only state that you have no provided any concrete examples of where this is the case, not even strong potential examples, though I have accepted tube amplifiers could be most at risk here.
Based on my understanding of issues involved, I decided to avoid speakers with passive crossovers, and only used active studio monitors in my living room and office systems since about 2008.
This year, I finally decided to extend to Atmos setup, and having so many quality active speakers in the living room became untenable. Some of them had to be passive.
So, I ordered and tried out some of the highly ranked by ASR D-class amplifiers. The results were disappointing. I had to either sell or return them.
That got me deeper interested in power amplifiers and passive speakers again, and I advanced my thinking on audibility of distortions of such setups.
As a result, I ordered some amplifiers that Amir deemed marginal. Yet in the context of my setup they worked noticeably better than the highly recommended ones.
You are pushing me out of my comfort zone, but I will respond with what I know, what I read, and my newfound knowledge of the math of stability and feedback. Looks like those math courses were not single minded! I read in one technical article that the electrical simulation models using resistors, inductors, and capacitors are both realistic and valid models of real speakers including the movement of the cones. As these are all linear elements, at least for the purposes of our discussion, then they can be simplified to magnitude and phase. Hence we are right back to our stability discussion and 99.9% it does not matter. Audiophiles may be interested. It does not mean their interest is relevant.
For small signals and to a degree, yes, this approximation works well enough for practical amps and speakers. Larger signals reveal nonlinearities, both in amps and speakers, and the divergence between the model-predicted behavior of the system and its real-life behavior may become large enough to be perceived as audible distortions, whereas theoretically there should be none.
Indeed, thermal drift of a transducer coil resistance value due, to ,say, a loud music passage, is a factor that a good amplifier must somehow compensate for.
This is not the purpose of the amplifier. How would it know it was the transducer coil, and not some other element. This would be the job of the speaker designer to compensate for.
This statement of mine requires qualification. Amplifiers embedded into some studio monitors, into some active subwoofers, as well as those found in some mobile phones and laptops, routinely compensate for thermal and other drifts of transducers.
I agree that compensating for such drifts in arbitrary pre-made speakers is far more challenging. To some degree, amps with current-based feedback attempt doing that, yet with mixed results.
I recall once seeing an amp with two sets of speaker outputs: one with voltage-based, and another with current-based feedback. As I understood, this approach wasn’t hugely successful.
One of the cases is simply running our of amp’s power supply current capacity. A music passage may be such that at some point all the bags will be moving towards the boxer, overwhelming him with the combined impulse.
Testing power output at different frequencies and 8,4,2 ohms would cover your argument. Again, even I know what. Using a reactive load that is not the same as your speaker is not going to provide easy guidance.
Well, theory of linear time-invariant systems predicts that such easy guidance would be valid. The amount of energy stored in the speaker system won’t exceed the amount of energy supplied by the amp during a characteristic system cycle.
Yet both amp and speaker may not obey that theory in certain regimes. Let’s say there is some hysteresis or ratcheting effect going on in the speaker, causing it to store energy over several characteristic cycles. Let’s say in a non-linearly-behaving transducer suspension elements.
Releasing such stored energy back into the amp output connectors may indeed overwhelm an amp perfectly tested on one-half and maybe even one-quarter of normal resistance loads. Will it actually ever happen? Maybe, or maybe not. Better be tested in my opinion.
This type of deficient behavior may be exhibited on some music passages by certain class A, A/B, and especially class D amplifiers, with their open loop bandwidth insufficient to deal with such combination of the speaker and music passage.
I feel this statement is made up. I don't think it based in theory or reality.
On the contrary. I have reasons to believe that this is exactly the behavior exhibited by the class D amplifiers I was disappointed with.
What is the characteristic reaction time of a classic class-D amplifier? It is one divided by half of its switching frequency, because it needs two cycles to produce required charge at the output.
Switching frequency tends to range between 400 and 800 KHz in modern class-D amplifiers. Thus the reaction time is between 2.5 and 5 microseconds.
A class A or A/B amp reaction time could be tens of nanoseconds. Let’s say 50 nanoseconds, which is 50 times shorter than the best class-D amplifier reaction time.
On a purely resistive load, this doesn’t make much difference. The amp input voltage may appreciably change only once in 5.2 microseconds, assuming the input is driven by 192 KHz sound source. Unpredictable back flow of current from the load is not present.
With a realistic load, which could include a complex network of resistive, inductive, capacitive, and non-linear components, situation is different. Now the amplifier needs to react not only to the changes in input voltage, but also to changes caused by the back flow of current.
A class A or A/B amp might do up to 50 adjustments through the feedback loop while the class D amplifier will only do one. Depending on the nature of music and characteristics of the speaker it may or may not matter, w.r.t. audible distortions.
An example of an undesirable process that class A or A/B amp may deal better with is ringing of tweeter, at a frequency significantly higher than 20 KHz. While not audible by itself, it may cause intermodulation distortions while contributing to the back flow of current.
A class A or A/B feedback loop may be fast enough to dampen such ringing quickly. A class D amplifier may not even “notice” it, or may otherwise react to such back flow of current in a way not conducing to the quick dampening of undesirable tweeter ringing.
Audibly, such deficiency may manifest itself as a lack of transparency, and timing errors, especially in music produced by dozens of instruments playing at once.
This is how I subjectively perceived those class-D amps driving my passive studio monitors. Highly accurate on slow-evolving parts of music, especially loud ones. Sounding strange and unconvincing on tightly placed lower volume transients.