No, I still am planning on doing a video, but maybe just on my own channel to start, or a future live thing, because I don’t really have my thoughts fully organized yet on the subject and that seems like a good way to do so. Mainly the issue I’m struggling with is how the community treats some of this stuff, because at the end of the day, if people have better experiences because of their believies in driver stories… Is there much sense in correcting that? There may be good reasons to but there needs to be a point beyond just “I must tell the people that they’re wrong!”.
2 Likes
I own a few headphones that I prefer their factory tuning to EQing them to a target. They are all what I would consider expensive. Having said that I always end up doing some tweaking with EQ, at least in the bass region.
I myself would have a tough time EQing a cheap headphone to sound as good. I think this applies to most people as most of us are not very skilled at EQing by ear, and that is what is required.
But now that it’s been brought up, I’m actually very intrigued by the proposition of EQing a Focal Clear to sound like a Utopia. Since the designs are similar, it may be that the non-HRTF related HpTF variation between them is minimal. All I’d need is an accurate measurement of both headphones on the same rig with the Utopia pads. It may even be possible to blind test the results since the ergos will be similar. Anyone know of a source for such measurements?
1 Like
Some skeptical science/engineering people believe that the differences obvious to audiophiles between cables, amplifiers, use of “tweak” accessories, etc. are hype and illusion. If you want to believe that, go ahead.
But I’m an electronics engineer myself, and I have found through long experience as an audiophile that these effects are definitively real and pervasive in audio, and at least partially can be corrected by various measures.
This is a complicated topic. I agree that well and “scientifically”-conducted objective measurements, up to a point, are totally valid for what they can show. But what they can show is sadly limited. The issue is, there is a vast amount of subtle audio information important to audiophiles that is real and physical but at the same time is essentially not practically measurable because of the limitations of normally available instrumentation equipment. But this information is still perceivable by audiophiles and is very important, because the resolving power of the experienced audiophile’s ear-brain system is greatly beyond the capabilities of electronic and acoustic measuring instruments.
Caveat: For high quality/high cost components, cables and “tweaks” to be of benefit, your system already has to be good enough in terms of resolution, imaging, etc. to reveal their effects.
2 Likes
By all means if you have evidence of this, I’m sure we’d all love to see it. I still think you may need to consider again what something like non-HRTF related HpTF variation entails for some of the subjective experiences you and other audiophiles are getting from these products. In my opinion we still don’t have the clearest picture of it, but what it entails - what’s at stake here - that part is a lot more clear to me, that being of course a far better relationship between what we can measure and what we hear, including all the effects… the technicalities people like to speak about.
1 Like
I agree with the general principle that everything we hear, including technicalities, is contained in the FR at the eardrum. However I think that when we substitute the word “perceive” for “hear” in that statement, there may be exceptions. Two that come to my mind are punch/slam and openness/spaciousness. Both of them are likely to have some degree of influence from things which are not heard. Touch sensation from the air movement in a sufficiently sealed earcup might increase the sense of punch. Or say a sufficiently large earcup which doesn’t touch the pinnae might increase the sense of openness.
2 Likes
This one may actually be more interesting since yeah the designs are a lot more similar. Or say an Elex to a Clear for example. But yeah re EQing to a target, I just want to reiterate for folks who may not have read the stuff above, we’re not suggesting EQing to a target is necessarily going to get you better ‘technicalities’. We’re saying in theory you could EQ a headphone such that its in-situ response at your eardrum imparts this. The line on the graph doesn’t give you precisely what to shoot for given the head and ears is not your own.
1 Like
Yup, exactly. There are all kinds of additional factors, like the openness of a headphone allowing you to hear room tone outside of the headphone, and this contributes to those kinds of ‘technicalities’ perceptions. I’m less sure about the air movement and dynamics thing, but it could be tested. The core problem to all of this is - to get philosophical - the private language problem. What one person means when they say ‘dynamics’ is not necessarily what another person means.
2 Likes
Yes and even though I understood that before reading this thread, I still feel that part of what most will take away from it is that expensive headphones are not worth the money, just EQ your cheap headphones instead. And while that is true in theory for both tonality and technicalities, it is not practical in practice for the vast majority of headphone listeners. Which is probably why people posting from their own experiences are disagreeing with what the science is telling us on this topic.
2 Likes
Okay. Thanks for clarifying that. Fwiw, I understand some of your hesitation on this.
I watched about half of the livestream. So still haven’t finished it. But didn’t really catch the references to “driver story” the OP refers to here. Maybe that comes a bit later in the stream. I do want to listen to the section on distortion though.
Also watched all of Blaine’s recent CanJam presentation, which was referenced in the livestream. While it doesn’t really deal explicitly with the question of FR matching, it certainly touches on some of the related issues discussed here. Especially perspectives of different types of enthusiasts (the so-called subjectivists and objectivists.) So thought it might be of interest to some here, including possibly the OP. Overall, I thought it was fairly informative, and a pretty good watch. Though Blaine was clearly rushing through parts of his presentation.
I wonder if there’s a similar video for Doc Olive’s most recent CanJam talk, since I wasn’t able to attend any of these.
I was also recently watching some of Tyll’s older videos. And thought it might be fun to hop in our time machines to see where we were in this game of headphones, sound quality, and reviews more than a decade ago. This was a little before I got seriously into the field, and became aware of Tyll’s work on HeadRoom and graphing. Maybe these will also help to give you a little more perspective on where we are today, and where we were back then, not so very long ago. Sound is a bit loud…
Tyll was mostly a DIYer of amps, and maybe some other related headphone gear, but not the actual headphones as I recall. (Anyone feel free to correct me on this, if it’s wrong.) He was, and I presume still is, also a motorcycle buff.
This was one of the early Beats products Tyll reviewed. And they did get better afterwards… 
I owned a pair of the wired Solo2’s for awhile, but the clamp, microphonic effects/lack of damping(?) were just too much. Very warm and smooth FR though.
Howdy again, MJ. Just wanted to respond to a couple things in the other posts above.
I can’t really respond to the time domain questions, because I still don’t understand why impulse response is no different than FR. Or at least contained within it. Alot of the minimum phase stuff still goes over my head, even though it’s been explained to me several times.
The time domain is clearly an important factor with loudspeakers though, because of the way sound interacts with a room. And there are some headphone designers who use the analogy of a small room for their designs. The time domain in a headphone is generally much smaller though than for speakers in a room. And the acoustic transmission of sound is also different (coupling vs. propogation of waves). So I’m not sure how apt this analogy is. Some enthusiasts feel that the results have been pretty good though.
There are going to be strong opinions on both sides of the FR matching question. And it’s not something that should really be taken personally. Some folks invest alot of their time, energy, and money trying out and listening to different headphones and other audio gear. And there are also those who have invested small fortunes and much of their time and energy on measurement gear. And there are bound to be tensions arising between these two groups. I think the experiences and info they both have to offer has value though. And should not be summarily discounted.
I think there are also a variety of factors, besides just the basic FR, that can be quantified and measured in headphones that have the potential to give us a better understanding of how a product might perform in real world user conditions, with different kinds of gear. Blaine discusses some ideas along these lines in the CanJam video mentioned in my last post. But there are also other simpler parameters that I think might have value, if they were studied a bit more closely. I’m thinking of things like distortion, phase, noise, impedance, driver symmetry/matching, isolation, crossfeed/crosstalk, damping (both physical and electrical), clamping pressure, driver size/shape/stroke, leakage, and so forth.
There is alot that can be discussed (and measured!) beyond just FR that could potentially help us to understand our sonic experiences with different headphones and other audio gear. And it’s not all subjective.
Subjective impressions also matter though imo. Because even though we understand a lot about the science of audio, we still don’t understand it all.
While I agree that frequency response at the eardrum would be the best approximation of measured sound characteristics currently available (despite the fact that we can’t measure it at each and everybody’s eardrum apparently), I doubt that it is the only and sufficient characteristic.
What makes me think so is the math of the physics. Frequency response in the essence is a spectral function, which is a derivative of an Amplitude(time) function. E.g. to get a spectrum of a waveform, you typically take an FFT2 of it. This is a lossy transformation, meaning that you cannot transform an FFT back to the exactly same waveform. With basic waveforms like sine waves or harmonics the spectrum is still a pretty complete approximation. With music consisting of a lot of complex waveforms like acoustic sounds, reflections, etc. the spectrum tells you how energy is distributed by frequency within a time window of measurement, but it doesn’t tell how the music would be perceived by the ear at every single moment in time.
For example, when talking about technicalities in headphones we often mention things such as reproducing the trailing ends of tones and fast transient sounds. How do you represent these on a FR plot? These are waveform characteristics where time is an important dimension, if you take a timeless derivative of it, this information is lost.
So, ideally we should be measuring an output waveform at the eardrum and comparing it to the input waveform, and doing it with music. Doing FR measurements is a by far easier and more standardized process.
1 Like
Thanks all for the discussion, it’s been great to read all your contributions.
Thank you both for your responses, I’m glad that I wasn’t completely off the mark. Also yes, definitely meant non-HRTF related HpTF variation, I’ll try to remember that mouthful.
I am super interested in this. I mentioned on the podcast topic suggestion page that I’m interested in the question of what tube sound is. Is it just the addition of harmonic distortion, or is there more to it? The only tube amp I’ve heard is my TA-26S, and I like how it sounds with my dynamic driver headphones (though it does pick up RF noise from my PC, which is very frustrating). I’d describe it as subtly richer and punchier than my solid state amps. In your videos you’ve mentioned tube amps aren’t really worth it until you get to the high end (e.g. kenzie, forge, euphoria etc.). One day I’d like to try those out. What are the qualities or differences that the high end tube amps impart that you like, and what’s your take on why they sound different?
This part of the discussion ties back to my earlier comment in the YT discussion. There, I’d said: I understand the point here that hypothetically, if we knew FR at the eardrum, then you could potentially EQ cheaper headphones (e.g. 400se) to sound as good as or better then expensive ones (e.g. susvara). That possibility means that, at least hypothetically, driver story (eg planar is faster, beryllium is faster than magnesium, etc) could be irrelevant in that scenario. But - I think there are still issues around this that shouldn’t be ignored or discounted, at least for now, and drivers remain a relevant part of the discussion in practice. For one, we don’t currently have widely available, cheap, reliable and repeatable means of measuring FR at the eardrum. I’m not sure how fine-grained any such measurements that can currently be made would be. Another issue is that while some headphones readily take EQ, others distort. Another issue is that readily available EQ tools like peace only allow for a limited number of adjustments, and don’t allow for fine-grained adjustments that might be necessary to directly mimic another headphone’s frequency response. In other words, even if you had an accurate picture of FR at the eardrum, and even if we can account for HRTF and HpTF, we still might not be able to make one headphone sound exactly like another in practice, or to make it so one person has the same FR at the eardrum as another person. Perhaps it might be possible to get close. Conversely, it seems to me that driver story remains relevant as one of a number of basic contributing factors that is relevant to predicting FR at the eardrum for the average user, at least where the headphones aren’t EQ’d (e.g. planars usually sound fast, have good instrument separation, have that plucked quality; dynamic drivers typically sound more punchy; heavily front damped headphones may sound more blunted, etc.). That’s not to discount your point at all, which I think is more directed at dispelling the idea of some special driver ingredient outside of FR, just that there are other elements to consider in terms of how the discourse engages with these concepts.
I think I broadly stand by that comment, save to clarify that I was referring to driver story there as shorthand for the contribution that the driver makes to FR, but not meaning that the driver itself has inherent or essential qualities outside of that contribution which could be changed with EQ.
I am of the view that components in the source chain can have audible effects on FR. To me it’s kind of a truism, since it is necessary to be true in order for electricity to be changed into sound waves. But further to that, my own experience has been to perceive subtle differences in the amplifiers that I have. One clear example is my Singxer SA-1, which has a set of 4 capacitors that can be bypassed with jumpers. A/Bing against my K5 Pro and Mojo 2, the Singxer in unmodded form very clearly resulted in transients being softened and bass sounding woolly. This was audible to me as well as to my dad, who I asked to see if he could hear any differences (but of course the tests were not blind). I wouldn’t say the differences were obvious, however. It’s possible I noticed these things because transients having a sense of edge or attack is something that I value and have been chasing a bit in my headphones. When I added the jumpers and the capacitors were bypassed, that quality was improved. I don’t doubt that the capacitors are capable of affecting FR in some, perhaps small ways. Perhaps it would be possible to test this with in-ear mics before and after doing the jumper mod.
To me, this is the real issue. Although I think different sources can be perceptibly different, and it is likely that the components in the signal chain of those sources make small differences, those differences are ultimately very marginal and unlikely to provide value for money, particularly compared to the differences that an equivalent amount of money spent on a new set of headphones would yield.
Having said all of this, I can’t speak to any experience with cables, interconnects or other kinds of ‘tweak’ accessories.
I think the openness of a headphone is a significant contributor to the spaciousness effect and probably other technicalities. Certainly it can make for a subjectively enjoyable experience, particularly for someone who hasn’t experienced it before. I remember when I tried my Sundara for the first time, coming from my K371 and HD560S, and I had my first ‘veil has been lifted’ experience. It was incredible.
Just on the ‘unveiling’ effect thing, which is discussed in the video as being associated with closeness to HRTF, I think I also experienced what that discussion was referring to with my Focal Clear. Everything seems more ‘real’ when I wear them, I wonder if perhaps they are a closer match to my HRTF than my other headphones. Alternatively, perhaps it also is coming down to how open they are, or maybe other qualities.
There is some discussion about this from about 40 minutes in, and then later again from GoldenSound at around 1:09.
Thanks for this, I haven’t watched it yet but will definitely get to it when I can!
I promise I am not taking anything that is said ITT personally, I just find it interesting. And as I said above, I don’t have a preconceived idea about what’s possible in terms of using EQ to improve a cheaper headphone or make it sound like a more expensive headphone, so I’ve been reading this discussion with interest.
I would be interested in what this looks like also. But I think there are two possible sides to this:
- First, there will be FR features that emphasise certain psycho-acoustic qualities, such as perceptions of transient speed, attack, punchiness etc. The video discussed some of these (I can’t recall the specific details, I think it was about having some presence at 1khz that isn’t overshadowed by too much bass or treble). Perhaps @listen_r could remind me.
- Second, there is presumably a way of visualising how a tone appears in terms of attack, decay, sustain and release - can a waterfall plot be used for this? I’m not sure this would be meaningful for looking at FR associated with music, since the musical information would just all be shown together without distinction between individual notes. Or at least I’m not sure how one could meaningfully analyse this. But I obviously have no specific knowledge about this, so please chime in if you know something.
We probably don’t talk about these things enough, but source components, such as players, DACs, amps, EQs, and also phono preamps, can absolutely effect what you hear in your headphones, in a wide variety of ways. Even though this is often downplayed by some of the experts in discussions here and elsewhere.
Higher impedance tube amps, for example, can effect the FR through their interaction with the transducer’s impedance curve. This is measurable. And it’s something many enthusiasts use to their advantage to improve the warmth and tonal balance of their open-back headphones, without having to resort to using EQ.
The output impedance of an amp can also effect the electrical damping of the headphone driver, which can result in more or less audible distortion. This will be more noticeable on lower impedance headphones with a lower damping factor than on higher impedance headphones though.
The gain of an amp also effects the amount of noise you can hear from the amp. This is dependent on the sensitivity of the headphone though, and is more noticeable on louder headphones than on those with lower sensitivity.
Clipping distortion can also occur in the opposite case, when using lower sensitivity headphones with a lower gain amp.
EQ’s can effect not only the FR, but also the phase, gain, clipping, noise, and effective dynamic range of the audio signal. This is why alot of enthusiasts prefer to use no EQ at all, and to make any changes to the FR or tonal balance on their headphones through other physical modifications instead. Clipping, gain, noise, and dynamic range effects are all fairly easy to grasp (at least conceptually). Changes to phase can be less obvious, but may also effect the imaging of your content in some subtle ways. So it’s probably one of the more insidious effects an EQ can have that might negatively effect the sound quality. Most of this is dependent on how large the FR amplitude changes are in your EQ curves/profiles. And how the EQ is implemented.
The gain/volume, sample rate, and bit depth settings on your other source gear (e.g. player, DAC) can also effect the frequency range, noise levels, dynamic range, and the clarity/detail of the content you’re listening to. Some of this could possibly be measurable at the DRP, and some may not, depending on how the measurements are done. But it’s something you can potentially also hear in your content.
The conventional wisdom is that you don’t really need a higher bit depth or sample rate than 16/44.1 CD spec for good sound quality. And as long as you’re not altering a digital audio signal from its original bit depth, sample rate, volume, or frequency response, that’s probably true. I recommend using higher bit depths and sample rates though depending on your audio content/source, and whether you use EQ. Because it can improve the resampling of the content. And provide potentially better matching to the source rates (most video content uses 24/48, for example, which is what I regularly use). The downside is that much higher depths/rates may also increase latency a small amount, with little to no perceptual benefits to SQ.
Much of this falls within the general rubric of proper gain-staging and impedance bridging (or what enthusiasts sometimes refer to as “synergy”), and just the correct set up of your gear… which is a non-trivial task in audio. And not as sexy as talking about HRTFs, diffuse fields, and target respone curves. It’s been one of my particular interests though since joining this hobby, because they can all potentially effect the performance and sound quality of the gear and content you listen to. And because I’m less able to afford much of the higher-end gear that can potentially be a little more flexible and forgiving in terms of performance on some of this… And because I have a tendency to dabble in both pro and consumer audio applications.
Authoring your own audio content can help you gain a better understanding of some of these things, because you’ll probably be required to use gear that’s of lesser quality, and have to figure out how to get the most out of it,… which can be educational.
Thanks! I’ll check it out.
So, the latter here is something you could reasonably correlate with FR, and again I say all of this with the recognition that this is just conjecture on our part. For all we know there could be additional metrics we’re not considering (like the psychoacoustic effects of acoustic Z for example). But I still think it’s likely to just be FR at the eardrum. The problem is it just doesn’t LOOK like “trailing ends of tones and fast transient sounds”. To understand how it relates you need to do a bit of analysis, and this is where the gaps typically are.
For example, something can sound more blunted if the FR at the eardrum is such that there’s a notable roll off, or dip, or withdrawn section in the treble. A reduction at 3khz can also cause things to sound muted or blunted. It’s just hard to make that judgment just by looking at the standard FR graph, because once again that’s not the FR at your eardrum that confers whatever these subjective experiences are. IMO that’s also why subjective reports, provided they’re genuine, are incredibly valuable.
With regards to the earlier point about music stimulus and its complexity, if I’m understanding your point correctly, this actually relates to Blaine’s previous presentation at CanJam, and I really need to do the video on this because it’s a common thing that comes up. But effectively you can use music as a stimulus, and when it sufficiently spans the full spectrum, it ends up generating the same result you get with a sweep or an FFT. It may look a bit messier, and not all music is the same, but you are able to measure the same sound pressure level at the eardrum with music as you do with test tones.
I think the key to much of this is that it can be very easy to confuse the cause with the effect. So a headphone experience may sound like X, and because that’s very compelling, we’re quick to look for X in data to explain it. But the problem is that X is a subjective phenomenon (like ‘trailing ends of tones’, bluntedness and so on). We probably shouldn’t be so quick to assume its cause is conceptually or visually a similar thing.
Very much no. And this horse is basically dust at this point, to use a Blaineism. You can actually do a simple example to demonstrate that in headphones, time domain stuff like waterfall plots and CSD is just a worse view of FR - and I really need to do a video on this as well.
Take an undamped highly modal planar headphone, say like one of the HiFiMANs. It will show meaningful features in a waterfall plot. EQ out those modal features in the FR and the time domain feature will disappear with it. If it does not, then that means the headphone is extremely flawed, but I’ve literally only seen this once - but this is also why its worth showing excess group delay, because that will tell you if you have a reason to care.
Now, that’s all just to do with making sense of time based views and why they’re generally irrelevant in headphones. The question of what’s perceptually relevant in FR, or what contributes to subjective qualities like “attack and decay” and so on… that’s precisely what Listener, Cameron and I were discussing in the live stream. So like which measured characteristics do each of us associate with a given experience, and I actually think we come to some general consensus there.
But again I come back to the problem being that the psychoacoustic effects of sound don’t often visually resemble an FR graph, even if FR exhaustively does describe the acoustic cause… that being sound pressure level at the eardrum.
Forgive my ignorance about this stuff, particularly waterfall plots. What I’m getting at is there must be some way of looking at a frequency plot over time as a sound is played. Like the animated frequency spectrum plots on this website. But while it’s simple enough to understand a single tone and its harmonics looking at these, when you do this for music it seems difficult to understand much more than the relative spread of different parts of the spectrum, particularly because notes with overlapping frequencies just add up to a higher amplitude without anything to say “this bit is from the cymbal crash, this part is from the high hat”. But I’m sure there are experts with a better understanding who could extract more information than I can.
1 Like
My point was not about using music to measure FR. That to me is quite pointless, because when measuring FR you are still measuring energy distribution by frequency. Which tells you how the electrical-acoustic system transmits different frequency components of signal to the measuring sensor over a period of time.
My point is that we are not measuring whether this electrical-acoustic system transmits the same waveform as the original signal. Trailing ends of tones could be a good example of this: there could be 2 headphones, one of which cuts them down in a busy mix and the other one reproduces, because it reproduces the original waveform more closely. Both headphones can have similarly measuring frequency response though, because the frequency spectrum of those ends of tones has nothing special about it compared to other sounds.
This is my point exactly. Let’s say we hear X, and then we use FFT or other spectral analysis function to get FR(X)=Y. And then we are talking about certain features of Y, there is no causality, because Y is a derivative of X, not the other way around. There can be a correlation between a 3kHz dip and things sounding blunted. We may say that a 3kHz dip causes a headphone to sound blunted. But in reality it’s the other way around: some parts of the “bluntedess” effect measure as a 3kHz dip.
There’s a distinction to be made between visualizing the capabilities of playback equipment and sonic or musical content. But in playback equipment, FR still describes how all of that propagates at the eardrum.
I suppose you could phrase it that way. What we measure as frequency response is sound pressure level at the eardrum. What we perceive as ‘bluntedness’ is sound pressure level at the eardrum. Or put another way… the sound at the eardrum, which we can measure as FR causes the perception. But the perception itself is still unlike the FR, and I think that’s where people start looking at other metrics that don’t mean anything to explain their experiences. And this is made worse by people not understanding the consequences of non-HRTF related HpTF variation.
I’m not sure this scenario is possible though. Because you would see that difference in the FR. Like if that were the case, those two headphones would necessarily measure differently.
Don’t have time to get to deep into this discussion, but its become eminently clear that we need an article talking specifically about the myriad misconceptions when it comes to the “time domain” in headphone measurements, and all of the reasons why basically every visualization of this effect separate from simple FR plots is more likely to be misleading than anything else… so I’ll work on that. Keep an eye out, I spose. It’ll probably have to be two parts.
4 Likes
Isn’t this describing distortion?
Yes, very much. A step by step explainer about this would be really welcome and useful.
Btw as a sidenote, if my posts ever seem flippant or missing something that could be addressed with a google search (see ref to waterfall plots above), it’s probably because I have 2 young kids and only about an hour of time to myself on any given day. It’s part of why I really like the podcast format, because I can listen while doing chores or walking the dog.
4 Likes