Hello, I do not have much experience with monster headphones, to assess if the volume is indicative, in my experience, I listen at low volume in conditions without distraction and it is where I visualize tones, timbre and details with their derivatives with greater clarity, on the contrary with the high volume I find everything more direct and present, but that does not give me any indication of recovery of nuances, that is my humble opinion,
Thanks for commenting friend … Greetings !!!
1 Like
Hello … And what you mention would only enter in over-ear or also in-ear, since the latter are limited by the wave field to reproduce the same density and naturalness of the sound, apart from its speed?
Hello!!! And for in-ears, how is this concept applied?
Detail is different than micro and macro dynamics, so I’m not quite sure what you’re inferring. Sorry.
BTW, I love your mega man avatar.
3 Likes
An article on the subject of headphone speed:
I agree with the observation that the dip in the mids may allow details in the bass and treble regions to be “retrieved” relatively with ease. It also allows the listener to raise the volume a bit helping further particularly the bass and fundamental frequencies.
1 Like
Detail retrieval is as much about the FR as well as the HP’s time-domain response (TDR). A pefectly transparent (and therefore capable of 100% accurate reproduction of sound, i.e., “perfect” detail retrieval) is a HP with 0ms impulse response decay. But this can only happen with a zero-mass transducer or infinite energy transfer, which is theoretically impossible. The take-away here is that lighter diaphragms and/or stronger magnets should theoretically make a more transparent HP. Arya, for example, would be relatively better than most dynamic driver HPs. LCD-X too. It is rather sad that FR has taken the center-stage for most HP talks. A lot of cans from mid-fi to truly hi-fi have low THDs, which means that EQ is generally possible to get each one to achieve a preferred FR. But it’s the HP’s performance in the time-domain that really separates relatively affordable HPs from the end-game ones. Deficiencies in time-domain performance cannot be cured satisfactorily by EQ bec these are very much hardware-dependent. LCD-X has a very fast transient response. No wonder that it is a detail monster. I really hope to see more reviews incorporate Log Impulse Response (IR) measurements.
Another factor that affects ‘detail retrieval’ is resonance. Resonance dictates which frequencies in the spectrum decay slower relative to others in time. This can be measured using a CSD plot (cumulative spectrum decay or so called “waterfall” plot). Frequencies from a “slice of sound” at time = 0ms that remain audible, say, 5 ms after would mask/colour the “slice of sound” at t = 5 ms. Resonant frequencies generally reduce transparency and overall resolution of the HP. Avoiding resonances is a tough R&D objective because each person has a different external ear structure (even between the left and right ears). An idealised ear structure (represented by HRTF) is used by each company to optimise their designs against resonance. Now, if you luck in and your ears come close to that, then happy days! You’ll hear music crystal clear (almost). Btw, HD800s has a very good CSD plot. Incidentally, it is often described as “bright” despite its FR…well, it is just that it’s very transparent with good IR. It reveals a lot of details. It’s also expensive. Anyway, I wish CSD plots are more common than FR.
Last factor is leakage response. Sealing adversely affects both frequency and time-domain responses. The effects of loose sealing vary across headphones. Good headphones consider leakages in their design to achieve more uniform TDR and FR…these manifest in earpad design for example. Not surprisingly, these are traits common to more expensive cans.
I guess my point here is that “detail retrieval” is not a made up thing. It’s real. I see it as a “sexier” term to express the transparency and resolution of a headphone, which are both dependent on FR and TDR. Log IR and CSD plots are good measurements that could provide insights on a headphone’s TDR. Reviewers should pay more attention to these measurements because arguably, TDR is more correlated to the HP’s hardware and hence, generally reflected in its price. These measurements may also help a certain person’s understanding of detail retrieval and be less dismissive towards its use in reviews.
Oh boy…
Okay so first of all, in order to determine what detail is, we have to solve the private language problem - that is to say, you can’t identify the thing if we can’t collectively agree on what the the term means even as far as our subjective experiences are concerned. Without first establishing that (and it could in theory be studied), we can’t move on to the next step.
However, if there is something there that’s collectively identified, it’s unlikely to be found in time domain information with headphones, and that’s because headphones are in the vast majority of cases minimum phase. This means time domain information is proportional to frequency response. In other words, it’s just an older, worse way of looking at the same thing in minimum phase devices.
And you can demonstrate this quite easily. A headphone with various peaks in its FR will have corresponding resonances visible in time domain, EQ out those peaks and the resonances will disappear accordingly. If they don’t, then there’s something seriously wrong going on, and you’d be able to see this in excess group delay.
It’s understandable that people would come to this conclusion based on what this stuff looks like, and it may even be preferable to have a ‘clean’ response there - but that doesn’t mean this isn’t an FR thing. It’s worth watching the section on this in this video.
As to this topic - descriptions of ‘detail’, including the ones that I’ve given at various times, are only ever descriptions of the subjective experience. They do not refer to any physical or acoustic property going on in the headphone. Presumably they are caused by an acoustic property, but we run the risk of misattribution of metrics and data (which we see all the time) when we conflate these descriptions.
3 Likes
Hmm, not quite sure about what you mean by “minimum phase”. Sure, I can convert a time-domain response in a linear system to frequency domain…this is what FFT is about…anyway, I don’t want to start a debate on this. True of course that you can infer resonances from a FR (and even from THD; the usual suspects as you say are the peaks) but CSD plot and log IR present these information directly with a lot of visual clarity. Also, injecting the group delay seems to concede that time domain response info is, in the end, useful…like without this, how would you know which peak to EQ out because you suspect it is resonant? Esp if it’s within your supposed target FR band? Also, it is not always true that you can EQ out resonant frequencies without seriously altering the headphone’s performance. Resonance is hardware related. You can’t make a Prius run like a F1 car by altering its engine tuning. Yes, we’re talking milliseconds decay in headphones but then sound is in that time scale.
It’s your opinion that detail retrieval is ‘subjective’. I concede that it’s a loose term. Perhaps I should stick to calling it a headphone’s transparency and resolution.
Also, there is such a thing called Fourier uncertainty principle, which states that generally humans cannot discern a difference in frequencies (pitch difference) and the precise timing when the change occurs…there is a theoretical limit for “detail retrieval” in time. Now, granted that golden ears have “breached” this limit, this is more the exception than the rule. This is somehow due to the nonlinearities in our ear system compared to the simplifying assumptions concerning Fourier transforms (linearisation). Even then, the uncertainty remains non-zero. So in conclusion, science differentiates FR from TDR. And it would help that we also get somehow the measurements in the time domain.
By the way, just to clarify…I’m not referring to you at all in my previous post. I’m referring to someone not from here who thinks SINAD is the most important parameter in audio.
Excess group delay, not group delay. This is basically the indication if something is wrong, not minimum phase, where we’d need to start worrying about time based views.
So, you’d see GD features for headphones with highly modal FR features (like many planars for example), but that doesn’t mean they’re not minimum phase.
There are other issues with trying to quantify things like detail and resolution, but if all things are equal there’s no reason to assume that assigned subjective goodness/badness isn’t down to just FR at the ear drum. The issue is we’re not measuring at the ear drums of individual people, nor do we have their corresponding HRTFs to test against.
And with regards to EQ, you absolutely can fix resonances, the issue is once again that measurement rigs have their own unique HRTFs. Deep phase cancelation issues are more of an issue there.
Thanks @Resolve, ok I get the point abt excess group delays…but who provides these measurements? I have not seen these. By the way, I checked GD measurements and they do somehow show general indications which ones have longer transient delays at which frequencies. Sundara for example has a very messy GD and its CSD is also quite messy although a lot of the frequencies decay very fast.
I surely appreciate the FR of headphones. No doubt they convey a lot of info. However, I feel that the TDR has been forgotten and a there’s even a lot of confusion about how “fast” a headphone is. Anyway, maybe I’m just old-school. CSD has been a staple from my DSP days in uni and I quite miss the insights I gained from looking at them.
I do watch a lot of your reviews by the way. I hear what you say about the dynamics and detail retrieval although I interpret them quite differently.
We’re starting to do it now - you can see it in the Tungsten thread here.
To be clear, we’ve checked this stuff for a long time now, we just haven’t been posting it because there’s always a risk that people focus on the wrong things. But given how the conversation on GD has gone, where people will refer to “messy” performance there or indicating issues to do with resonances without noting the excess GD, it’s probably worth posting. And also to show that things are behaving normally in the vast majority of cases.
And this is sort of what I’m saying - you’ll find exactly this with more modal responses like with common planar magnetics. But that doesn’t mean what is heard isn’t just the FR and those modal FR features.
I think the common issue related to all of this is that people will go looking for things that explain their experiences, and they’ll note quite understandably that what they see on the FR graph doesn’t line up with what they hear, and so they start to dive into other metrics to see if that helps explain things. But when you study FR, HRTF and HpTF a bit more, and in particular try to understand how your own anatomy influences FR at the ear drum, it actually does a better job of explaining what you hear in terms of these things.
And, the reason I say this, is that the burden of proof would be strongly on the side of anyone indicating that time domain information is actually relevant in minimum phase devices (and thus not proportional to FR), or indeed that headphones shouldn’t be minimum phase, or that they’re not entirely minimum phase… etc.
I say this also in part because there’s apparently a talk on this very subject at CanJam NYC next weekend that I’m hopeful to attend to see if there’s any there there. I’m skeptical, but curious nonetheless.
1 Like
Thanks @Resolve. I read the review and I quite like it. Very informative and comprehensive. Still missing my CSD plot! Hahaha. But the addition of excess group delay, in conjunction with FR and THD helps shed light so I’ll happily take that.
I understand that one (or even few) metric/measurement(s) cannot reveal a HP’s performance, which is why I find a particular reviewer’s “scientific approach” amusing. Not to mention the very strong gospel-like conclusions from a sample of one.
Hmm, I’d say that what we “hear” is the sound propagating in time, which is physically the changes in air pressure over time, which is modeled in frequency domain as series of sine waves of varying frequencies with different amplitudes in discrete tiny windows of time, moving in time. This linear model (of a non-linear physical phenomena) is just a scientific construct, not the reality…it’s the FR snapshots moving in time.
We agree on the fact that one’s HRTF affects one’s perception of sound. But the cool thing is that 2 persons could have relatively similar HRTF, or that a person could match well with whatever HRTF the manufacturer has used as its basis of design. This makes someone’s “subjective” review and also the vendor’s (reviewer’s) objective measurements relevant to a person. The key thing, as you imply, is the understanding of how close (or far) your HRTF to the “standard” used by the OEM/reviewer. This is why I find listening to as many headphone reviews vital to allow me to see who’s the one likely to match with my perception of sound…as that could mean that his/her HRTF could be closer to mine.
Now back to TDR, I’m in the camp that the time-domain related measurements are valuable to “complete” the picture. Duality is inherent in all natural physical phenomena. FR and TDR are in theory inseparable. I’m skeptical that all relevant time-domain properties can be had from a frequency-domain snapshot to the point that TDR measurements aren’t required. I’d even argue that as we get closer to the theoretical Fourier limit (for the general population) and hence the highest resolution and transparency, the TDR measurements would matter more. Going back to my F1 analogy, the engine’s kinetic energy storage/withdrawal during braking and turning (which is reflected in lap time) matters more in an F1 car than in a Prius.
This is where there may be a benefit to the ‘view’ of time domain stuff, but we shouldn’t be assuming there’s anything new given there in minimum phase systems, which headphones typically are. So I’d say if you want to run with the idea that there’s something new or different from FR, that entails a commitment to the idea that headphones are either not minimum phase in practice, or not fully minimum phase at some level. At which point far be it from me to discourage your research on that front, but I would also say that… this is a road I’ve been down. I used to go looking for these things when it seemed like that’s what was going on, and I’ve since come around to what @Mad_Economist and others have been saying.
In hindsight… it wasn’t the best use of my time hahah.
Yeah, I’m cognizant of that the info gained from time-domain measurements augment and not supplant the info from frequency response. It’s quite valuable when trying to make sense of the subjective audio performance differences between headphones. For example, I bought the more expensive Arya over Sundara based on most reviewers’ subjective assessments concerning improved resolution corroborated by a log IR result that shows Arya’s faster transient response than Sundara. This is even to standard pink noise, which is less complex than most musical composition. Arya’s CSD plot is also a lot cleaner than Sundara’s, which could mean than complex musical arrangements with substantive dynamics and non-periodic content (percussive) could sound a lot better in Arya than Sundara, which is indeed the case when I performed an A/B listening test. LCD-X is even better than Arya for this kind of music (e.g., EDM, hard rock with lots of distortion and heavy dose of fast quiet-loud-quiet dynamics). Not surprisingly, LCD-X time-domain measurements are better than Arya’s. I don’t think you can honestly see all these from FR charts.
Now I imagine you saying that we need to be careful in ascribing subjective experiences to measurements. As an engineer, I really understand cognitive biases and such so I’m conscious of this. But yeah, I really believe that CSD plots and Log IR do provide additional info that are valuable to resolution and transparency.
1 Like
Truly if you can find something demonstrable there, then by all means, I genuinely wish you luck! One always has to keep an open mind to new information.
Thanks! For now, I’d be happy to see excess GD plots that you’d be including in the reviews, moving forward. And yes, the decay time is for the most part irrelevant and the system is largely time-invariant (which is the basic assumption for FFT)…except when it’s not (which can be seen from excess GD, in absence of CSD plots).
Also, transient response is not very relevant when listening to relatively simple musical compositions. The “details” in this case are mostly timbral (overtones) in character and perhaps the general sense of “thumpiness” in the sub-bass extension than discernible pitch differences in time. Anything below 20 Hz registers as undifferentiated sonic pulses without a definite pitch and spectral content above 10kHz generally show pitch sensations that are quite inaccurate, registering as something we subjectively called “naturalness” (wrt to our individual sense of timbre). Fwiw, in this case FR plots are more than enough.
I appreciate reading this back and forth. I read your intro also @Jourdan. The key word is REVEAL.
I don’t disparage measurements. They certainly show something. In particular they show what a microphone shows on a particular rig, or what a scale shows on a particular planet. They can clearly be useful for headphone designers and audio engineers.
HRTF makes your own ears the only reliable guide for you. So does your specific amount of hearing loss, tinnitus, etc. If you look at @SenyorC’s reviews, you will see time and again that there is only a reasonable positive correlation between his measurements and what he hears and likes.
We do have tech experts here, and I hope you enjoy your experience here as contrasted with SINAD-fanboy-landia, the dimension not of time, not of space, but of measurements.
7 Likes
Microphones: The silent partner/contaminator of all audio measurements. I walk away from any review that suggests that I “listen to a recording of speakers to understand how they sound” for the obvious logical error. Take the speaker, add a mic, add a recording device, add a storage medium, add a playback amp and playback imperfections, and add a second set of speakers…okay…what system am I listening to now? 
Measurements can be internally consistent, but still very divorced from human-perceived reality.
Yep. High-range hearing imperfections can damage or destroy detail retrieval and sense of space.
4 Likes
Totally agree. I believe that people should only buy what their ears they can hear. If one’s tone deaf, then it makes no sense to buy a $5k set of cans. A pair of moondrop chus would be absolutely fine for listening purposes. Now of course, some buy expensive gear for reasons other than listening to music, like comparing them with their buddies online and using their gear’s SINAD performance to discredit other people’s experiences.
1 Like