[Somafunk]

Interesting……….I may apply peq on the upper treble on the Sony from 3k onwards to see what it sounds like and if it makes a noticeable difference in my impression of the bass response, I wonder if I can get to sound just like the meteor?. No way of comparing other than my utterly fallible audible memory and the time it takes to swap iem’s, so it’d be a moot point and a wasted exercise but I may try if it’s a damp n’ dreich weekend - not as if I’ve anything better to do other than sit on my arse

Also interesting and whoop-whoop at the greater resolution/accuracy of the 5128 on sub bass/treble frequency capture, this is good to see.

 

[Ghoostknight]

In its most basic form, frequencies are what we hear, and distortion is simply inaccuracy with frequencies. So I guess everything, aside from timing, comes down to frequencies.

you cant say this to distortion overall, what about different distortion profiles?
distortion is actually a complex thing IMO, what about different amplitude of distortion compared to the main tone?
since we are in the sound science forum i guess everything below -60db is inaudible anyway?

 

[KinGensai]

The causal link exists; if you change the FR, the CSD graph also changes. It is indeed true that CSD graphs show the information that we want more directly and straightforwardly. But that doesn't mean that the information is not already contained in the FR.

You wouldn't be able to quantify the duration of a Dirac pulse IR by looking at an FR graph. The Fourier transform would indicate the presence and amplitude of decay, but not the duration because there isn't a time axis recorded. It's only frequency and amplitude. A CSD does introduce the time axis thus is appropriate for that information.

For IEMs, the time domain is determined by the FR, because they are basically minimum-phase systems, unless there are horrendous phase-mismatch between the drivers in a multi-driver setup or if baroque housing structures cause a lot of internal reflections before reaching your eardrums, among other things. These generally don't happen.

Could you provide an example of this? I'm having trouble understanding how you could derive IR from just the FR graph, so an example would help.

The discussion around BA bass and DD bass is interesting as I have the Sony XBA-N3 with 1dd/1ba and Symphonium Meteors with 4ba as my two favourite iems and the bass response of each iem as measured by Precogvision below is practically identical from 300hz down, listening to them both I get a more impactful sound from the Sony with the DD, I’ll be keen to to view/read the research from Resolve with the BK5128 as to why this is, I think it must be the decay time/volume of air/pressurisation of the ear canal with the Sony’s DD.

https://precog.squig.link/?share=Precog_Target,Sony_XBA-N3,Symphonium_Meteor

DDs tend to have different harmonic profiles and lower infrasonic extension vs BAs, so I think that's what is causing the difference in conjunction with the longer decay (V14 measures ~1.7ms at 300Hz vs ~2.5ms MSE, both are essentially the same driver config outside of 4 bass BAs vs 1 DD and have distinct bass feel differences between them that I haven't been able to replicate with EQ matching).

you cant say this to distortion overall, what about different distortion profiles?
distortion is actually a complex thing IMO, what about different amplitude of distortion compared to the main tone?
since we are in the sound science forum i guess everything below -60db is inaudible anyway?

Audibility depends on your system and your listening health & skills. I'm at -39dB, you can test for your threshold using a blind test if you want to know.

 

[Ghoostknight]

Audibility depends on your system and your listening health & skills. I'm at -39dB, you can test for your threshold using a blind test if you want to know.

Which test you got -39dB at?

this is my result with two-tone 70Hz 800Hz (all guesses were right till the end...silly they just go to -69dB ) ..... with a music sample i got silly -21dB

 

[KinGensai]

Which test you got -39dB at?

this is my result with two-tone 70Hz 800Hz ..... with a music sample i got silly -21dB

Same site, used the J stone music track.

 

[Doltonius]

You wouldn't be able to quantify the duration of a Dirac pulse IR by looking at an FR graph. The Fourier transform would indicate the presence and amplitude of decay, but not the duration because there isn't a time axis recorded. It's only frequency and amplitude. A CSD does introduce the time axis thus is appropriate for that information.

The decay time is determined by what frequencies are produced at what amplitudes. Yes, you heard that correctly, that frequencies and amplitudes are enough for the purpose of determining time-related effects, if the system is minimum-phase (usually the case with iems). Essentially you want to reproduce a transient that stops the vibrations, which is decomposable into pure sine waves at many frequencies of different amplitudes via Fourier transform (see illustration below, a quasi-square wave is decomposed into various pure sine waves). If you iem changes these amplitudes according to its frequency response, the resulting waveform's shape also changes, and that has time-dimension effects. For example, a sharper turn followed by a steeper slope in the waveform can become a rounder turn followed by a flatter slope, which essentially means, longer decay times.

Could you provide an example of this? I'm having trouble understanding how you could derive IR from just the FR graph, so an example would help.

The maths involved are complicated. I don't have the skills to demonstrate that. I do have a Chinese article explaining this and illustrating the effect of changing of FR resulting in changes in the CSD graph, if you want to read it after machine-translating it yourself. https://zhuanlan.zhihu.com/p/70275578. Found an English article using the same data (it is actually from the same author, just writing in English) https://www.linkedin.com/pulse/csd-cumulative-spectral-decay-really-important-jason-dai/, from Audio Precision. Basically, simply cutting down the amplitude at a give frequency can reduce its decay time. That is why, all else equal, less bass means tighter/faster/drier bass, and more bass means looser/slower/wetter bass.

 

[Ghoostknight]

Same site, used the J stone music track.

ah this one just got me -15dB ... the chapman song got me -21dB .... all tests were first try for this year
i rememember getting something like -30-40db a while back on the music samples

strange how much easier its to hear on the two tone test... even with no errors
EDIT: as i approached -40db or so the 800Hz tone was pretty much useless to detect distortion, i had to concentrate on the 70Hz "rumble"

imo this definitely shows how complex the whole topic is... and i guess it proofs the point of objectivists saying that music can "mask" lower level details, tho im pretty sure with the right track i would be able to get higher results than the two music samples provided, but thats more of a guess

 

[KinGensai]

The decay time is determined by what frequencies are produced at what amplitudes. Yes, you heard that correctly, that frequencies and amplitudes are enough for the purpose of determining time-related effects, if the system is minimum-phase (usually the case with iems). Essentially you want to reproduce a transient that stops the vibrations, which is decomposable into pure sine waves at many frequencies of different amplitudes via Fourier transform (see illustration below, a quasi-square wave is decomposed into various pure sine waves). If you iem changes these amplitudes according to its frequency response, the resulting waveform's shape also changes, and that has time-dimension effects. For example, a sharper turn followed by a steeper slope in the waveform can become a rounder turn followed by a flatter slope, which essentially means, longer decay times.

The maths involved are complicated. I don't have the skills to demonstrate that. I do have a Chinese article explaining this and illustrating the effect of changing of FR resulting in changes in the CSD graph, if you want to read it after machine-translating it yourself. https://zhuanlan.zhihu.com/p/70275578. Basically, simply cutting down the amplitude at a give frequency can reduce its decay time. That is why, all else equal, less bass means tighter/faster/drier bass, and more bass means looser/slower/wetter bass.

I took a read, chinese isn't my native language so I had to use translate. From what I read there, what I'm taking away is that resonance and nonlinear distortion artifacts generated by bass frequencies occur at essentially the same time given the minimum phase characteristics of IEMs whether they are single or multi driver, which I don't disagree with and can understand deriving from the fourier transform denoted by the FR. What I'm not understanding is how this information then allows you to derive the exact length of the IR without a discrete time axis, let alone how that fourier transform allows you to separate how much of that amplitude is due to resonance from bass vs other resonances and fundamental frequencies at that frequency.

 

[KinGensai]

ah this one just got me -15dB ... the chapman song got me -21dB .... all tests were first try for this year
i rememember getting something like -30-40db a while back on the music samples

strange how much easier its to hear on the two tone test... even with no errors
EDIT: as i approached -40db or so the 800Hz tone was pretty much useless to detect distortion, i had to concentrate on the 70Hz "rumble"


imo this definitely shows how complex the whole topic is... and i guess it proofs the point of objectivists saying that music can "mask" lower level details, tho im pretty sure with the right track i would be able to get higher results than the two music samples provided, but thats more of a guess

Yeah, music tests are the important one IMO because it's the closest approximation to how we are actually using our gear, thus gives us a good idea of what actually matters in terms of distortion.

 

[Ghoostknight]

and i guess it proofs the point of objectivists saying that music can "mask" lower level details

Tho, the objectivists need to explain now how i was able to hear -69db distortion with a volume of maybe 60-70db

 

[Ghoostknight]

Yeah, music tests are the important one IMO because it's the closest approximation to how we are actually using our gear, thus gives us a good idea of what actually matters in terms of distortion.

it might be a good generalization but i think its highly dependent on the source material, tho a whole study would be needed..

 

[Doltonius]

I took a read, chinese isn't my native language so I had to use translate. From what I read there, what I'm taking away is that resonance and nonlinear distortion artifacts generated by bass frequencies occur at essentially the same time given the minimum phase characteristics of IEMs whether they are single or multi driver, which I don't disagree with and can understand deriving from the fourier transform denoted by the FR. What I'm not understanding is how this information then allows you to derive the exact length of the IR without a discrete time axis, let alone how that fourier transform allows you to separate how much of that amplitude is due to resonance from bass vs other resonances and fundamental frequencies at that frequency.

Here is the English version by the same author. https://www.linkedin.com/pulse/csd-cumulative-spectral-decay-really-important-jason-dai/. I think you misunderstood. It really says that the CSD graph follows from the FR; changing the FR, i.e., just cutting the frequencies at which there is long decay time, will shorten the decay time, for example.

You see, Fourier transformation breaks whatever complex waveform into simple sine waves. All the information contained in the sine wave is described by its amplitude, phase (dislocation in time), and frequency. In a minimum-phase system, phase is not to be worried about, because it is does not change. That leaves us just with frequency and amplitude, which the iem will change according to its own FR. And by doing that, the waveform's shape is changed. This will have time-dimension effects, if you are looking at a single impulse. To learn exactly how the numbers are calculated, then you really need to learn the details of fourier transformation and the mathematical function that describes the impulse.
This above is inaccurate. The whole truth probably too complex to explain. The message is still that in minimum-phase systems, the FR uniquely determines the phase.

 

[KinGensai]

@Doltonius
It seems we are talking about different things. I get what you are saying here, that doesn't mean that you know what the actual duration of the IR is, just that you can shorten some harmonics of bass resonance by destructively interfering with that frequency via EQ.

 

[Doltonius]

@Doltonius
It seems we are talking about different things. I get what you are saying here, that doesn't mean that you know what the actual duration of the IR is, just that you can shorten some harmonics of bass resonance by destructively interfering with that frequency via EQ.

We are talking about the same thing, just that you don't quite understand my point or the article's point. The author manipulated the frequency response, i.e., just changed the amplitude of a narrow band of frequencies. He did nothing that is directly related to the time-domain. However, it made the decay time of the frequencies in that band shorter. This is a clear example of the FR affecting the time-domain. It follows from tenets of Fourier transform and the properties of a minimum-phase system.

In essence you can generate the CSD graph from the FR graph, given that your system is minimum-phase. It is really that deterministic. You don't need to separately measure IR, as long as you know the FR. All you need is the FR, the algorithms of Fourier transform, and a mathematical description of the impulse.

 

[Doltonius]

@Doltonius
It seems we are talking about different things. I get what you are saying here, that doesn't mean that you know what the actual duration of the IR is, just that you can shorten some harmonics of bass resonance by destructively interfering with that frequency via EQ.

Also try this paragraph from an ASR post:

1. Headphones/IEMs being minimum-phase systems means FR fully describes their output.​

​

Minimum-phase is a mathematical concept that applies not just to loudspeakers. In this case, it means that at any time, the loudspeaker's movement (and the resulting sound pressure at the ear drum) is within one phase-cycle of the input signal, and does not lag behind. This means there is no such thing as "attack", "decay", "driver speed", etc because the driver is tracking the original signal perfectly in the time domain. Now, you can totally get a subjective sense of "this headphone has bad decay compared to this other headphone", but what you're describing is a difference between their FRs.​

​

"Wait, you're saying the headphone tracks the input signal, how come different headphones sound different?" Well, how a headphone changes the input signal is called frequency response, but the point is the "not lagging behind" thing (the output is stable and time-invariant).​