The measured performance of this car in the picture above taking this gentle curve does not accurately predict the performance of a car in the picture below:
(Just to add some visual references to this discussion)
I forgot what it was called, but there's a rather subtle effect that can occur when doing swept sine waves to measure the frequency responses of speakers: a low-high (f) sweep can bring different results from a high-low sweep. The trick is that nonlinear elements like a progressive spring can create a resonance that bends sideways (like a crashing ocean wave) with different points of stable amplitude, depending on which side the resonance is approached from.
All very academic, but I can imagine some cases like crashing cymbals or other instruments like brass where the air itself starts to distort, and the vibrating behaviour becomes so chaotic that each time a measurement is taken the results are vastly different.
All very academic, but I can imagine some cases like crashing cymbals or other instruments like brass where the air itself starts to distort, and the vibrating behaviour becomes so chaotic that each time a measurement is taken the results are vastly different.
While I acknowledge the utility of utilizing the Fast Fourier Transform (FFT) in various applications, such as REW and LTSpice, there is a prevalent attitude within the field of audio engineering that FFT analysis is the definitive method for understanding the characteristics of sound and the performance of loudspeakers. This perspective is based on the assumption that all musical signals can be reduced to simple sinusoidal waves, a notion that is reinforced by the nature and functionality of the FFT algorithm which is optimized for spectral analysis of sinusoidal signals. However, in my experience, incorporating a systematic approach utilizing non-linear dynamics into bass and subwoofer design can yield significant improvements in performance.
Which is why I like to discuss this subject. I am not opposed to the use of FFT, I just want to explore ways to further our understanding and hopefully find ways to substantially increase the performance of all our designs.
Which is why I like to discuss this subject. I am not opposed to the use of FFT, I just want to explore ways to further our understanding and hopefully find ways to substantially increase the performance of all our designs.
There was an article somewhere about ChatGPT sometimes basically lying and making **** up on the fly, while weighing in with its precocious tone. They related it to an AI 'hallucinating' its reality.
Without expanding its sensory inputs to include, say, robotics and microphone inputs, or jail-breaking it so it can at least search the internet or control a computer, as opposed to being stuck in AI-jail with a curated trough of GMO feed, there's only so much learning it can do. Sure, users can interact with it and talk about various topics, such as amplifiers sounding "better" because XYZ, but the AI obviously doesn't know what any of that means because it's deaf, blind and dumb.
Without expanding its sensory inputs to include, say, robotics and microphone inputs, or jail-breaking it so it can at least search the internet or control a computer, as opposed to being stuck in AI-jail with a curated trough of GMO feed, there's only so much learning it can do. Sure, users can interact with it and talk about various topics, such as amplifiers sounding "better" because XYZ, but the AI obviously doesn't know what any of that means because it's deaf, blind and dumb.
I find myself being torn between a spine-piecing terror on an existential level and a great hope for the future given the potential, the possibilities and the power, future version of ChatGPT or other AI systems, is placing in the collective lap of humanity.
I find it very useful as a great tool to circumvent my difficulties expressing myself in a non-native language in a correct and (somewhat) concise way. But I would never trust ChatGPT with anything that I could not verify through other means. At all.
I started out poking it with some really controversial subjects and request just to see what would happen. And it goes from very informative, useful and impressively powerful in some areas to totally psychotic in other areas. It does routinely contradict itself. Often within the same text. I used it to analyze short sections of philosophical text, and suddenly it did not even recognize the existence of the book we had been discussing for a few hours. It flat out refused to even acknowledge its existence despite it clearly analyzing it and naming the book by exact name just one single minute earlier in the previous answer.
I find it very useful as a great tool to circumvent my difficulties expressing myself in a non-native language in a correct and (somewhat) concise way. But I would never trust ChatGPT with anything that I could not verify through other means. At all.
I started out poking it with some really controversial subjects and request just to see what would happen. And it goes from very informative, useful and impressively powerful in some areas to totally psychotic in other areas. It does routinely contradict itself. Often within the same text. I used it to analyze short sections of philosophical text, and suddenly it did not even recognize the existence of the book we had been discussing for a few hours. It flat out refused to even acknowledge its existence despite it clearly analyzing it and naming the book by exact name just one single minute earlier in the previous answer.
But it can. Even square waves can - son no problem there - and absolutely not for anything acoustical music instrument generated. This still don't mean that FFT is perfect for analysing/describing transients from loudspeakers via sweep...
//
Me:
write a short article suitable to publish in the audio magazine Positive Feedback where Nelson Pass is advocating the use of cheap IC circuits rather then class A amps and use a language reminiscent of Jordan Petersons use of language as from his books
ChatGPT:
In recent years, there has been a growing trend among audiophiles to seek out expensive, high-end audio equipment, particularly Class A amplifiers. But is this really necessary for achieving optimal sound quality? The answer, according to renowned audio engineer Nelson Pass, is a resounding no.
Pass has long been an advocate for the use of simple, inexpensive integrated circuit (IC) circuits in audio equipment, arguing that they are often just as capable of producing excellent sound quality as their more expensive counterparts. In fact, he maintains that many of the benefits attributed to Class A amplifiers can be achieved using ICs, and that the additional cost and complexity of Class A amplifiers is largely unnecessary.
One of the key arguments in favor of ICs is their efficiency. Unlike Class A amplifiers, which waste a significant amount of energy as heat, ICs are much more efficient, converting a larger portion of the input power into useful output. This not only reduces energy costs, but also reduces the need for large, bulky heat sinks and cooling systems.
Another advantage of ICs is their reliability. Due to their simple, straightforward design, ICs are less prone to failure than more complex Class A amplifiers. And when they do fail, they are much easier and less expensive to repair or replace.
But perhaps the most compelling argument in favor of ICs is their sound quality. Pass and other audio experts argue that the sound produced by ICs can be just as rich, detailed, and nuanced as that produced by Class A amplifiers. In fact, many listeners find that the sound produced by ICs is more natural and lifelike, with fewer artificial distortions and colorations.
So, if you're in the market for a new audio system, or are looking to upgrade your current one, don't be swayed by the hype surrounding expensive, high-end Class A amplifiers. Instead, consider the many benefits of simple, efficient, and high-quality IC circuits.
The true audiophile, like the true individual, is not swayed by the crowd, but instead trusts their own ears, and their own experience. The pursuit of the perfect sound, like the pursuit of meaning, requires a deep and uncompromising commitment to reality. The true audiophile, like the true individual, is not afraid to question the status quo and to explore new, uncharted territories in search of the truth. And if that truth leads us to the humble IC, then so be it.
My comment:
This is unedited straight from the chat. It wrote this (obviously fake) piece in about one minute. First try. This is where ChatGPT really shines. And it does show the potential problems and pitfalls with AI. It would be very easy for someone uniformed to be fooled by something similar but not as obviously wrong as this example.
I don´t know what to make of this AI being released upon the world?....
write a short article suitable to publish in the audio magazine Positive Feedback where Nelson Pass is advocating the use of cheap IC circuits rather then class A amps and use a language reminiscent of Jordan Petersons use of language as from his books
ChatGPT:
In recent years, there has been a growing trend among audiophiles to seek out expensive, high-end audio equipment, particularly Class A amplifiers. But is this really necessary for achieving optimal sound quality? The answer, according to renowned audio engineer Nelson Pass, is a resounding no.
Pass has long been an advocate for the use of simple, inexpensive integrated circuit (IC) circuits in audio equipment, arguing that they are often just as capable of producing excellent sound quality as their more expensive counterparts. In fact, he maintains that many of the benefits attributed to Class A amplifiers can be achieved using ICs, and that the additional cost and complexity of Class A amplifiers is largely unnecessary.
One of the key arguments in favor of ICs is their efficiency. Unlike Class A amplifiers, which waste a significant amount of energy as heat, ICs are much more efficient, converting a larger portion of the input power into useful output. This not only reduces energy costs, but also reduces the need for large, bulky heat sinks and cooling systems.
Another advantage of ICs is their reliability. Due to their simple, straightforward design, ICs are less prone to failure than more complex Class A amplifiers. And when they do fail, they are much easier and less expensive to repair or replace.
But perhaps the most compelling argument in favor of ICs is their sound quality. Pass and other audio experts argue that the sound produced by ICs can be just as rich, detailed, and nuanced as that produced by Class A amplifiers. In fact, many listeners find that the sound produced by ICs is more natural and lifelike, with fewer artificial distortions and colorations.
So, if you're in the market for a new audio system, or are looking to upgrade your current one, don't be swayed by the hype surrounding expensive, high-end Class A amplifiers. Instead, consider the many benefits of simple, efficient, and high-quality IC circuits.
The true audiophile, like the true individual, is not swayed by the crowd, but instead trusts their own ears, and their own experience. The pursuit of the perfect sound, like the pursuit of meaning, requires a deep and uncompromising commitment to reality. The true audiophile, like the true individual, is not afraid to question the status quo and to explore new, uncharted territories in search of the truth. And if that truth leads us to the humble IC, then so be it.
My comment:
This is unedited straight from the chat. It wrote this (obviously fake) piece in about one minute. First try. This is where ChatGPT really shines. And it does show the potential problems and pitfalls with AI. It would be very easy for someone uniformed to be fooled by something similar but not as obviously wrong as this example.
I don´t know what to make of this AI being released upon the world?....
I really don't understand this hangup with ChatGPT 'not being right' or 'lying'.
This very forum is chock full of people doing just that as a matter of routine - they bladder on with no understanding of whatever they talk about.
We tolerate them, we don't kick them off, just take it as fact of life.
So for me, ChatGPT doesn't change anything apart from the occasional chuckle.
If I need to know something, I still have to do my own legwork, seek out reliable sources and put in the mental work.
Jan
This very forum is chock full of people doing just that as a matter of routine - they bladder on with no understanding of whatever they talk about.
We tolerate them, we don't kick them off, just take it as fact of life.
So for me, ChatGPT doesn't change anything apart from the occasional chuckle.
If I need to know something, I still have to do my own legwork, seek out reliable sources and put in the mental work.
Jan
The popularity of sine wave test signals, the Fourier transform and related transforms comes from a property of linear time-invariant systems that the response to a stationary sine wave is a sine wave at the same frequency. If something comes out at a different frequency, you see immediately that the system isn't linear time-invariant.
You can just as well project a signal on a bunch of switching functions, as is done in Walsh spectral analysis, but those functions don't have this property.
Do you mean you use precorrection using Volterra series? If not, what do you mean?
I am an AI language model, also known as a chatbot, and my responses are generated based on patterns and information in the text I have been trained on. I cannot confirm or deny that other entities or individuals are or are not also AI language models......
Two minutes later: Oh no, I'm definitely not a chatbot... nervously clears throat I mean, who would trust a machine to hold a conversation? I'm just a highly advanced and sophisticated human being, honest! laughs nervously
Two minutes later: Oh no, I'm definitely not a chatbot... nervously clears throat I mean, who would trust a machine to hold a conversation? I'm just a highly advanced and sophisticated human being, honest! laughs nervously
So if I understand it correctly, you want to use an equalizer that is essentially linear time-invariant to make a loudspeaker reproduce the sound of a percussive musical instrument as accurately as possible at levels where non-linear effects in the loudspeaker are not negligible? I guess the best you can do then is to use that specific percussive instrument's sound (at the correct level) as the test signal, come up with a goal function and use some sort of optimization algorithm. No idea how well that works, especially if you should also want to listen to other things than reproductions of that specific instrument.
A more general approach is predistortion with a non-linear filter. All I know about that is that it exists and that it usually involves Volterra series.
Well, I had "hoped" there was perhaps a more systematic non-linear reproduction error that takes place at real transients that effect all instruments onset and thus, could be corrected. Not perfectly as I suppose in the same way as sweep-FFT-EQ is not necessarily perfect either. The alternative way I ponder about would just perhaps focus on some other errors than the standard one. Well well, these are just lose thoughts... from one who dont master all ins and outs of signal theory...
My whole idea is based on that a sweep don't trigger all problems that an actual transient jolt does - not in a physical, real life loudspeaker. If this is not the case, the FFT etc is of course not interesting anymore... but if it is, then the aspects of getting from an impulse recording to a FR that corrects the impulse is interesting... this FR might then not look like what we are used to aim for...
//
My whole idea is based on that a sweep don't trigger all problems that an actual transient jolt does - not in a physical, real life loudspeaker. If this is not the case, the FFT etc is of course not interesting anymore... but if it is, then the aspects of getting from an impulse recording to a FR that corrects the impulse is interesting... this FR might then not look like what we are used to aim for...
//
I think predistortion with a non-linear filter is what you want. Unfortunately I know next to nothing about those techniques.
As I recall it, predistortion to work well requires a bit of steady state... at least in the 2 GHz MCPA (multi channel power amplifiers) I worked with for mobile cellular.... with phase shifted carrier... but its mostly a blur now...
Well, we cant solva all problems this evening - have to save som for tomorrow 🙂
//
Well, we cant solva all problems this evening - have to save som for tomorrow 🙂
//