Since I have spent thousands of dollars over the past year of two to get my endgame solutions… What would you say to someone who isn’t going to make this a hobby? While leaving the door open for it to be a hobby?
I often see those in a tech forum I frequent looking for a solution rather than a hobby. Maybe they are in a country with a limited selection or inflated prices, it’s difficult to tell them that the Focal Elex “WAS” the budget king (can’t get it anymore) even though ~$500USD without periphs isn’t budget in any way.
I find that unless I am changing my hardware frequently… it just normalizes.
I have $3500 on my head right now and it’s glorious, but it’s an entire thing. I have recently been realizing that if I stick to one solution for long enough, it’s just the music not the hardware. There is that golden era with new hardware, but what is the tipping point where a $35 headset isn’t good and a $1k chain is?
“Here. Try this.” Hands them a pair of $3k headphones.
Each person who asks me for advice is a unique case and I don’t have a standard recommendation. Budget and whether the headphones will be used for gaming are the two biggest influences.
But seriously, I do have most people try my equipment before making a recommendation. If they aren’t the least bit impressed then I know they probably aren’t going to become an audiophile.
Yep. This is the case with anything else too.
It’s all part of your individual journey. I got tired of a $2k headphone and moved on. Then came back and with $50 worth of pads, it’s like a whole new headphone.
I can’t find it right now but DMS has a video on why you should keep old headphones around and revisit them once in awhile. This is a way to keep things fresh.
The way I see it, it’s easy to advocate getting a nice sounding headphone. Hearing your music with good headphones is an experience worth having. Spending $$$$ chasing tiny improvements to ASMR characteristics of a particular track is where the juice ends up not being worth the squeeze. For some people, that’s literally the way the approach this, and for them it becomes a hobby. We might say they’re the enthusiasts. But I think a person can and should take a more reasonable approach to buying a nice headphone. In other words, there’s a realm of sanity to all of this that is worth it IMO, even though perhaps the more visible parts of this space are very far removed from that sanity.
The main challenge we have is a lack of objectivity. We buy so many things based on accurate specs. At least I do. Things like DACs and speakers, and computers, just based on benchmarks and measured performance. I do not have to try them out 1st, cos the measurements that are important are well known.
Why are the measurements that cause a headphone NOT as well understood, widely?. Note I use the caveat “widely”. Some may know, but the info is apparently limited to a few wizards. Most public discussion on headworn listening devices, revolves around one thing. Frequency response.
Contrast that with the world of speakers where for decades, they have used so many other measurements to assess speaker performance. Impulse response, step response, waterfall plots, distortion. Why are manufacturers of headphones, NOT publishing these measurements, and also publishing the FR.(frequency responses), and the same question should be same for reviewers.
I am a bit tired of subjective reviews. They say nothing. Just show the facts, the measurements and then subjective opinions become a nice to have, a bonus, cos we cannot argue with facts.
But there is hope, and I encourage headphones.com to lead in this direction. I salute their efforts on measuring frequency response. Well done, but there is a lot more than FR, to describe the behaviour of a headphone or IEM.
Where I find the measurements I need, for a headphone or IEM, and can compare with others on my shortlist, far beyond just FR, I can take an informed decision, and at the very least narrow the effort to narrow down options into a shortlist. An objective approach to reviewing headphone and IEM expected performance based on data the include, but goes beyond just FR, would reduce the decision window, and increase sales. We would waste far less time checking subjective opinions, and be able to buy based more on measurements which are factual, and objective.
Please see the following thread, it explains my thoughts a whole lot more and spares me repeating them here.
I wish headphones.com would expand their reviews, to go in this direction. Happy to join their team, to contribute to reviews and the process of measurement, and willing to learn.
This is because headphones are minimum phase and many of those things don’t show you anything frequency response doesn’t already tell you. We’ve done a number of videos on this, but recently Mark Ryan also did a video on it. There are rare instances where time-based views will tell you something, but you can just plot the excess group delay to see if there’s a reason to care. And Harmonic distortion really doesn’t matter unless it’s awful.
People THINK they’re hearing things that are outside of frequency response, because they hear things the graph doesn’t depict. But what they don’t know is that this is just frequency response variation that occurs as a result of their head and ears being different from the one used to take the measurement.
Think of this like… how speakers behave differently in different rooms. It’s not exactly like this because again for headphones, time domain is proportional to frequency response. But at least for response variation, there’s an in-use variable that matters significantly. And those other metrics create erroneous judgments about products because people will fixate on them, thinking they depict things they demonstrably don’t.
So while yes, there are other things we could show measurements of, this is more likely to mislead people than reveal anything novel about their performance. Really, what people need, is to see the frequency response measurements done on additional heads and ears, ideally ones closest to the heads and ears of the people considering buying the products. It’s the reason why we’ve gone into EEP testing with in-ear mics for on-head human measurements.
For most people, it’s a balance between dollars and their perceived value. As Hamlet says, “ah that’s the rub”
First value WHEN? Use cases!!! If this is FOR THE GYM, then there will be clear limits. If it’s MOBILE, there are others. If it’s to SHUT OUT THE WORLD another.
Assuming no special use cases, then as I’ve pointed out before, the 80/20 and 95/5 rules apply. So if you think that a $500 headphone is super awesome, $100 will get you 80% of the way there.
This old thread is instructive on how to think about things - and is strongly reccomended when giving advice to someone who is not yet an audiophile but dosn’t have tin ears .
P.S. 20% of the guys got to 80% of the girls in high school.
I have never purchased anything in my headphone hobby based on measurements or technical specifications. In fact, I have never looked at frequency response graphs or other measurement data when deciding what to buy.
My approach has been to seek which headphones, amplifiers, or DACs are generating the most interest and discussion, then try them for myself. Purchasing decisions narrowed by the influence of subjective impressions and subjective reviews shared by forum members and YouTube reviewers.
Over time I’ve learned which reviewers tend to align most closely with my own preferences. When those individuals recommend a product, I am most inclined purchase it. Ultimately, my buying decisions are driven by the subjective experiences and subjective observations of others and NEVER on measured performance. So far, this approach has served me well.
With all due respect, I beg to differ. Sincerely I think some reviewers are missing a trick on the impact of time on what we hear.
Not measuring the time related aspects of a transducer (drivers, shell, eartips, pads) and their impact on the audio, leaves a lot of information off the table.
There are no perfectly instantaneous impulse response transducers, with zero weight and infinite speed. All of them have inertia, a delay, and an overshoot, however small. I.e nothing has a perfect dirac impulse response. What is the variance from a perfect dirac impulse response? Both time and frequency. Easily observed in a waterfall plot, which shows both frequency response at start of the impulse, and how frequency varies with time.
Please think through this. For speakers one of the things that a waterfall plot easily shows us are things like the advantage of a non ported speaker over a ported speaker, with respect to resonances and roll offs in the bass region. And we can observe easily any other resonances, visually as lovely ridges, assessing how many and how quickly these fall in level over time. And compare the waterfall of two speakers.
It would be most dismissive to assert that such comparisons of headphones via impulse responses, and waterfall plots, and step responses, have no value. That would be an oversight, and I say this with all due respect. I have a lot of appreciation for the excellent work you chaps have done on the frequency response side of measurements, it has moved the dial forward.
You state that these resonances are negligible, but I beg to differ. As long as they are measurable, and we can see some devices, which are clearly better in avoiding these resonances, and distortions, they have a value. For example I have observed an amazing correlation between planar magnetic headworn devices, IEMs and Headphones, and in many cases their lack of distortion (really low - actually exceptionally low distortion). I would be most interested in also studying how such devices vary their frequency over time via a waterfall plot.
In my opinion, waterfall plots describe a behaviour of a headphone or IEM, which is relevant to whatever the listener hears, and can explain even more about how they perceptually sound, beyond just the frequency response. In the same way that a room’s acoustics has an impact on a speakers sound, and the damping inside the speaker contributes to this also, with headphones and IEMs, the eartips, the shells, closed vs open vs vented, and the filters used on IEMs, all have an impact on reflections which have to be audible, and add to time based variances between different devices., beyond just a change in frequency response.
I think it is a huge opportunity. That for whatever reason headphones.com, and other reviewers are leaving a lot of the table, by not paying attention to the time based behaviour of headphones and IEMS, and I speak not about phase, but actual time over the 1st few milliseconds, and this has to be audible, and explain even more about the differences between products., and what we perceive.
In my case, long before I ever knew I could find waterfall plots of a similar device to my ARTTI T10, I was hearing a definite “echo” especially with silicone eartips, similar to a bit of reverberation, compared to listening on single dynamic driver IEMs. 18 months later, I found the answer on this site (link below) and the waterfall plot fully explained the “echo” I had been hearing.
From this other website, here is the waterfall (aka CSD - Cumulative Spectral Decay) plot for a single DD IEM I own, and have had for well over a year, the KZ SAGA Balanced
Each of the colours represents 1 millisecond. Blue, Green, Light Brown, Dark Brown, Dark Grey.
Kind of rather telling that the SAGA waterfall shows a faster decay( i.e a larger area of blue), at the start of the decay, and after 5 milliseconds, when we get to the grey areas, we have less of the grey area, i.e more of the high frequencies are already below -50dB compared to the S12 Pro.
If one examines the highest point (i.e the loudest level after 4 milliseconds) i.e the start of any of the grey colours, in the upper mids and highs, from about 2 Khz and above, the SAGA is at -45dB approx, while the S12 Pro is at -36dB.
To completely dismiss the impact of time based measures, in assessing the performance of a headphone or IEM, is a missed opportunity, and I would urge a rethink.
Here is one more example
This next waterfall plot is for the Hifiman Sundara, which shows a definite departure in the speed of decay, i.e definitely slower than any of the aforementioned IEMs. Even after 5 milliseconds, the Sundara still has a lot going on, compared to the IEMs. How can we completely dismiss the import of a variation so glaring?. How audible these variances are, can be debatable, but that is a debate probably best left to those who study this for a living, in the walls of academia. We are predominantly hobbyists, hopefully looking for information, that is objective, and here is objective information. The big question is, should we be so dismissive of this information?
And here is the Sennheiser HD 600 below. Should we not be asking the question - why are these two products (the Sundara and the HD 600) are so different, in their waterfall/CSD? And what impact does this have on what we hear from these devices?
I’ve been working part time in audio engineering for almost 30 years and in music, as a singer, and part time musician(piano predominantly, and violin when I was much younger), for even longer, so it could be my ears and brain are more able to note these differences, not just in frequency, but in time, by reason of use.
Minimum phase device just means that time domain information is proportional to frequency response, and it’s also predictable. This means in headphones, waterfall plots or CSD visualizations just show a different view of FR (mostly). Can you plot the excess group delay for any of these examples? The S12 Pro in particular looks interesting. This will show if the device isn’t behaving conventionally. I’m not saying there’s never a reason to check, just that it rarely shows there to be any spot where it’s not just FR.
I do not have or am aware of how to get the raw data for the measurements on the Russian site. So have no idea how one would without the raw data, calculate the excess group delay. I may check on the website if that excess group delay is already published. It just might already be there.!
I’ll say this though. I’m now really interested in getting some gear so I can measure IEMs, and also any time related behaviours. Well over a decade ago, I’d be measuring and correcting speakers via EQ, but did not bother with correcting phase at that time, using REW. With speakers, its somewhat a lot more commonplace now for more of us to measure our speakers, now that measurement microphones with calibration data are no longer the cost of hen’s teeth.
Hopefully we’ll see more of this with headphones and IEMs over time. Inexpensive tools, to measure them., for the hobbyist.
At the very least, the benefit would be using the data to derive EQ corrections, for IEMs I already love their sound and taking them one step further, than just EQ by ear alone, which will still be part of the process, cos tools and my ear do not hear the same thing. All this bearing in mind the caveats of accuracy in the cheaper measuring tool clones, or incompatibility with measurements done by anyone else using much better gear…! Well we all have to start somewhere…
Measurement of IEMs, interesting, my next rabbit hole journey … That should be fun !!
This shows there’s no reason for us to care about time-based views in this device. It may be different for the S12 Pro though.
Listener has a useful explanation as to why EGD is helpful here:
"Headphones are “minimum phase systems” because the phase shift they produce is essentially identical to what a computer looking at a frequency response would calculate as the minimum-possible phase shift necessary to produce the same frequency response.
Now, that brings us to Andrew’s measurement, which is of Excess Group Delay, which is distinct from Group Delay.
Group Delay is the sum total delay between the signal generated by the source (in Andrew’s case, REW on his computer) and what is generated at the output. In simpler terms, its “how long does it take for the digital signal to become sound pressure produced from the headphone.” Pretty easy to understand.
Excess** Group Delay** however, is a measure of how much delay remains after you’ve derived the difference between the actual acoustic measurement of the headphone’s phase response to the computer-generated minimum-phase response."
Let’s take a step back. And start from 1st principles.
If all that is relevant to describe a headphone or IEM is the FR, why bother spending more on an expensive item, since we now have tools like the AutoEQ on various IEM measurement databases, like squig.link, to adjust any measured device in that database, to a target.
And if FR is all that matters, we can make any two devices sound exactly the same, if we correct their EQ to the same target.
Pretty certain, an Apple Earpod(or a 7Hz Zero or Zero 2), eq’d to the same FR target as a Crinacle Reference, will NOT sound the same..!
So what exactly distinguishes these devices, if we can via EQ get their frequency responses to be identical? i.e if we completely take FR out of the comparison, cos we can make this anything we want it to be, on any headworn listening device, via EQ correction, beyond FR(since this is now easy to emulate or change, and do this quite precisely), why spend more to get a better headphone or IEM? What difference do the better sounding devices bring to the table, beyond a better tuned “out of the box” FR?
You can make two devices sound exactly the same, but not by correcting them to the same target. This is the biggest misconception people have about FR. What you see on the graph is not the FR in situ, because you are not the measurement rig.
For the same reason stated above. If you could EQ it to the same FR in situ, then it would sound the same. But just because you EQ them to the same target using a measurement rig (any rig), that doesn’t mean the FR at the eardrum is the same. The FR is emergent depending on the interaction with the ear load and acoustic impedance. This is also why you can EQ two IEMs to match in the bass on 711 couplers that show meaningful differences when measured on the B&K 5128. The same is true of real humans as well, plus additional variables like length mode differences due to insertion depth and ear canal size/shape.
Any other perceptual qualities that aren’t only to do with the sound. These are things like openness, occlusion effects and so on. This can have a meaningful impact on the perception of these products.
So you don’t have to do the EQ. Like I’m not downplaying this, EQing properly by ear takes time and effort. 99.9% of people aren’t going to do this.
Now, these fine folks did something most people won’t and/cannot do by getting their HRTF measured, but the experiment paid off because they have proven you can make two different transducers sound the same, at least tonality wise. Comfort, occlusion and one headphone being more open than another by design likely can’t be mimicked, but the tonality ostensibly can. Unfortunately you and I may never be able to do this, but the headphone show gang has put in the time, effort and capital to back up their claims…which are now no longer claims but rather proven hypotheses. They essentially have an entire video series on this, which I highly recommend watching. Really cool stuff!
Your response focussed on differences between our hearing and how devices measure.
Let;'s take this difference out of the equation.
The chaps at headphones.com have recently measured their own HRTF’s, and are therefore able to alter any measurements to reflect how each of them is likely to hear any measured device. This would be a different result, for each of them.
Assume any one of us, also measured our own HRTFs, and therefore could arrive at an HRTF adjusted FR measurement of any headphone or IEM. Lets call this the HRTF Adjusted FR for device A.
It should be possible to take this HRTF Adjusted FR and modify it to a known target we call X, using EQ.
It should also be possible to take the HRTF Adjusted FR of another device B, and also modify it to the target X, using EQ.
This completely takes out the issue of any difference in hearing, between the listeners hearing and what the device measured.
Are we saying that an HRTF Adjusted Device A to Target X via EQ, will sound exactly the same as another HRTF Adjusted Device B to Target X via EQ, irrespective of the differences between the topology of drivers, shell, damping, nozzles, eartips on device A and B?
So theoretically it is possible to EQ two different devices, to make them sound identical to any listener, from the perspective of an identical FR target. But having completely taken out the variation in FR, by making them the same, my point is, these different listening devices, will still sound different, cos FR is NOT and cannot be the only distinguishing factor in headworn listening devices.
Whatever happened to transient response, i.e how fast the drivers respond to an impulse, and how fast they also respond or rather STOP, when there is no signal. Different driver and shell and damping will create very different transient responses, and this has to be audible.
Why bother with things like electrostatic drivers and planar magnetic drivers, if all we need is EQ to make any dynamic driver sound similar, by adopting a similar FR.
Clearly FR cannot be the only criteria for assessing the sonics of a headworn device, it simply cannot. And my quest is - how are we measuring this transient response. Why do we not discuss this more, cos as I said earlier FR is a moot point, we can achieve identical FR between two different devices, with ease. We have the tools - parametric equalisers a plenty, that aspect is not a challenge.
How do we demonstrate with measurements the superior transient response of better drivers.
This way we can do away with all the waffle of our subjective comments on headphones and IEMs, by simply showing the measurements of transient response. Job done. And buyers can simply pick whatever transient response they prefer, cos EQ is a moot point, anyone with access to squig.link can easily generate a decent enough parametric EQ to fit a target, and accept the caveats, such as the need to go through a 2nd round of hand tuned EQ optimisation, using a tone generator, to fit their personal hearing and/or preferences.
This focus on FR, is so overdone, and we need to move past it. So many DACs now have EQ, that unless we have a deviant total outlier FR from the manufacturer of a headworn device, we have the tools to make sufficient FR changes, to improve the device. But that achievement alone, will NOT remove issues like resonances or audible distortion, based on poor drivers or poor design of the acoustic properties of a device.
Will such objective measuring of other things besides FR, make the reviewers’ subjective opinions, on these devices, no longer as relevant?
It happened with other products like DACs, and speakers, which one can now buy with a lot of trust simply by looking at the specs, cos all the relevant specs are measured and published, sometimes by the manufacturer. One day it will happen with Headphones and IEMs, soon enough, only a matter of time.
I rest my case. I’ll leave it there. No further responses from me.
How could you know this though? The in-situ FR for you is not something available to you, so you can’t know that what your hearing isn’t FR.
I hate to break it to you but this is all captured by FR.
Yes.
I will say, there are advantages and disadvantages to each driver type that are perceptually relevant and not down to FR. Like with estats they can be extremely acoustically open, more so than any DD. And that matters.
You’d need in-ear mics for this. But yes, we can. When you do that, they sound the same.
To be clear, transient response, as often used by audiophiles, is not any real acoustic property. There’s a subjective effect that many use this term to describe but the cause of this perception is a particular frequency response relationship.
You seem to be hung up on FR being a very limited and outmoded metric that doesn’t describe your experience sufficiently. But again, this is a characterization of FR based purely on the graphs you’re used to seeing, not what FR actually is.
If we are talking strictly about the sound, then it would be the imaging of the headphone (spatial representation of sound) for me. If the headphone images well, it’s a good headphone imo. If it doesn’t, then it’s not. Good imaging incorporates alot of what’s being discussed here though imo.
Some of the most important factors are the tonal or timbral balance, bandwidth or extension in both the high and low frequencies, symmetry (how well the drivers are matched), dynamic range, distortion, pinna interaction, speed, and openness. “Speed” is simply shorthand for a driver that plays high frequencies at a decent volume without distorting or breaking up modally.
There are other factors to consider though than just sound, including comfort, ergonomics, build-quality, serviceabililty, customer support, etc. EQ-ability is certainly also worth considering.
And the longer I think about this kind of thing, the more things I seem to want to add to the list. I think EGD graphs like the one Resolve posted also need to use a smaller scale so delays more on the order of a millisecond (or less) can be discerned.
I think EGD graphs like the one Resolve posted also need to use a smaller scale so delays more on the order of a millisecond (or less) can be discerned.