Hacker News new | past | comments | ask | show | jobs | submit login
Ear Speakers – Research, Design, and Evolution (valvesoftware.com)
96 points by mxfh 41 days ago | hide | past | web | favorite | 44 comments

If you are interested in software-driven headphones, the Smyth Realiser series has always fascinated me.




These things are awesome! They used to ship the A8 with STAX Ear Speakers... which when I discovered it almost 10 years ago was the first time I had ever heard the term "ear speakers" or had ever heard of electrostatic headphones, or the brand STAX (from Japan). All in all, The Realiser and the STAX were unbelievable! I got bored of the A8 after awhile but I never got bored of the STAX... I wish electrostatic headphones were more widely available in general, they're a true gem. Everyone that hears them is in awe.

And they are "easy" to build yourself. I wrote a guide about doing so here [0] which is based on this enormous head-fi thread [1].

0: https://github.com/Azrael3000/headphone-guide 1: https://www.head-fi.org/threads/my-diy-electrostatic-headpho...

Wow, very impressive to DIY! It is one thing to do speakers DIY, quite another to make headphones, let alone e-stats.

For those who are interested in entry-level, non-DIY route

Drop (formerly Massdrop) has their variant of the classic Koss ESP/950 electrostatic headphones for around $500. They sold out earlier this year.


> Humans in general are very sensitive to sounds within 2kHZ-5kHz range. If the frequency of a virtual sound doesn’t match up to what we expect it to be in reality, then we are more likely to identify the sound as “not real”. This is particularly true if you compare how easy it is to tell if someone’s voice is broadcast through a speaker vs. someone is talking beside you.

I never thought about that in regards to gaming or even conference room voice chat, it is extremely easy to tell the difference. Must be a challenging problem set to solve, sounds fun!

These look really interesting! I own an original AKG K1000 from 1989 and they definitely have a better sense of "solidity" and "presence" than ordinary headphones with drivers that seal against your head, due to the natural crossfeed.

Other than the K1000 successor that got released a year or two ago (mysphere 3) there has been very little development since then in the field of open-air driver headphones. Hopefully these can sound good without breaking the bank! Also this is the first headphone that uses a driver based on the BMR principle that I am aware of

HiFiMan has a series of open air planar magnetic drivers, and they're absolutely awesome to listen to. I've yet to find other headphones that can stack up. Their only downside is the weight.

Which Hifimans are you referring to? Just to clarify, I'm talking not about open-back headphones, but headphones that don't have pads and consequently don't form a seal between the ear and the driver.

There is no seal. A seal implies a lack of air movement and sound isolation associated with closed back headphones. Open back headphones have no sound isolation in or out. The parts that contact around your ears with HiFiMan's simply set the distance from the driver to your ear, in a comfortable way.

That is incorrect. When I say "seal", I am referring to the sealed cavity between ear and driver, not to anything behind the driver. This is actually quite important for planars in particular, as a broken seal between the ear and driver leads to rapid bass rolloff, due to the lack of excursion found in most planar drivers (ime this is most noticeable with audezes since their excursion is especially low but I notice it with my HE-6 when I don't have the pads on my head quite right and there's a tiny gap between my head and the pads somewhere)

What would be the sound difference between such open-air headphones (no seal between ear and driver), and placing my head in between regular speakers (few inches distance, ignoring comfort concerns)?

Well for one, most speaker drivers aren't designed to cover the full range of 20hz-20khz, something headphone drivers are explicitly designed to do (historically there have been a couple of multi-driver headphones, but they were horrible without exception). Even a high quality midrange driver for a speaker will probably only have a good response between 300hz-5000hz and past the upper frequency limit you will get major treble spikes and rolloff due to the cone entering its breakup mode resonances, so it will sound predictably awful. Now there are fullrange speaker drivers, but these do not generally have flat response characteristics as they tend to roll off early on both ends of the spectrum, and fullrange drivers with multiple cones (aka "whizzer" cones) tend to have a breakup spike in the upper mids at around 3-5khz. I am a little less sure about the physics of this part, but I believe that even if you had a theoretically optimum speaker driver, you would run into Q-factor related issues because speaker drivers are meant to be listened to in the far field rather than being placed right up against your head, so if you had the driver right next to your head, your eardrums would see a huge FR spike in the midrange, due to a resonance/standing wave forming in the cavity between your ear and the surface of the driver. Headphone drivers, since they are designed to be placed up against the ear, are usually damped, mechanically or electrically, specifically to avoid this. The K1000 in particular is damped electrically (there's a small notch filter (?) on a tiny pcb that feeds the driver) since its sensitivity is already very low with the little amount of mechanical damping it has. One common issue between something like a K1000 and a headphone made with open air speaker drivers (no baffle or box behind the driver) is that they both experience rear wave phase cancellation at very low frequencies, which causes a 12db/octave rolloff below the primary resonance of the driver. This is due to the negative wave from behind the driver wrapping around to the front of the driver and decreasing the amplitude of the positive wave.

TLDR speaker drivers, compared to purpose-built headphone drivers, are highly sub-optimal due to limited FR range, breakup modes, and Q factor / improper damping for the application. Directivity and near-field effects (both of which would further distort FR) would probably also be an issue with speaker drivers but I don't know of anyone really trying to build a high quality headphone from speaker drivers so I can't comment on that.

Thank you, that was very informative! How did you learn so much about audio?

I use Sennheiser HD600 at home (mostly for classical music), and Bose QC-15 at work (mostly for pop/rock/electronic). I also have Grado SR125, but I rarely listen with them, because they are not very comfortable, even though the sound is pretty good - they feel more dynamic and lively than HD600, but that could be due to the difference in impedance (the source is Macbook Pro, which might technically be to weak to drive HD600, I'm not sure).

Based on this information, would you recommend any other headphones for me to try?

I've just been an audiophile for a long enough time I guess. The HD600 is great, it is endgame for many people and routinely beats headphones costing several times as much. Grados I quite dislike, they have really high distortion in the bass and have audible ringing at around 2-4khz. But if you like them, you do you I guess. The HD600 is in a unique spot where there are really no headphones out there that are a direct upgrade across all areas. Its main weaknesses are bass extention and dynamics, something like an hd800 or focal utopia (both of which are pretty different from one another) will beat it squarely in both of those areas, but those two headphones don't have the highly flat diffuse-field tuning of the hd600. If you like the hd600 and grados, you probably prefer a drier tuning, so a non-exhaustive list of some headphones you should try might be:

Focal utopia - These are somewhat v-shaped, extremely resolving, possibly the most "dynamic" sounding headphone, very narrow like the hd600, makes hd600 sound veiled and compressed in comparison.

Sennheiser HD800 - Definitely a polarizing headphone, bright in the lower treble but lacks some presence due to a scooped upper midrange @ around 2khz, so the opposite of grado in that regard, super enormous stage (second only to K1000) but imaging is not quite as precise as utopia or K1000. Great bass with minor rolloff, makes any hd6x0 sound wooly and imprecise. Extremely resolving.

AKG K1000 - Their presentation is a lot more like speakers than headphones. Very unique and incredibly "live" sounding. They are forward at 2khz but not as much as grados. They can sound alive and dynamic, or washed out, dead and veiled, all depending on the angle of the drivers, so you really have to adjust them perfectly to see what these can do. More resolving than hd600 but not as much as modern flagships like the two above. They have the widest stage of any headphone, it really has to be experienced. It's as if there are actually palpable instruments floating in the space around your head. Amazing realism, but highly uncomfortable and not ergonomic in the slightest. Ones with serial #s below 5000 are the most desirable as these are the "bass heavy" versions which were made before a factory revision that increased the primary resonance of the driver from around 40hz to around 60hz if i recall correctly)

Mysphere 3 - Modern K1000 successor, designed by the same guy who designed the K1000 in the late 80's. I have not heard them but they are said to improve on comfort, resolution, bass extention, and driver sensitivity.

I would be cautious with planars. When they aren't tuned to sound like listening to music through a sock (audeze) they often have a weird habit of sounding simultaneously bright and veiled, which isn't exactly pleasant. Not sure why that is from a physics standpoint. the diaphragms are usually held to high tension, so it might have to do with standing waves forming on their surface at high frequencies? Idk. What planars excel at is mostly their unmatched low frequency capability, with bass extention flat to well below 10hz with a perfect seal. Good planars tend to have a sense of huge weight and solidity but without the bloatedness and blurriness of bassy dynamic headphones. My favorite planar is the hifiman HE-6, but the driver has a lot of ringing in its stock form and needs to be modded to reach its full potential. Also the drivers have a nasty tendency of dying, and I don't think the manufacturer can replace them anymore. Like the K1000, they are also extremely insensitive. I unironically use a speaker amp to drive these. If what you want is an upgrade in terms of resolution, planars probably aren't where you want to look, as they usually have some degree of haziness to them that you don't find with dynamic drivers.

Any stax - These are electrostats and require their own high-voltage amps. There are warm stax (sr-007, sr-L700) and bright stax (sr-009) but generally speaking they tend to be highly resolving but dead sounding and lacking dynamics. This might be due to the limited excursion and low driver mass, again not 100% sure why. Also they tend to have an upper mid shelf (wider than the narrow 2khz-ish boost present in AKGs and grados) from around 1000hz onward that lends everything a kind of plasticy timbre. They are really cool from an engineering standpoint but in my opinion overpriced and not that musically engaging.

For your preferences I would stay away from: Audeze (veiled compressed dead sock), non-flagship hifimans such as he-500 and he-560 (veiled and hazy to the max), anything mrspeakers (grainy compressed dead sock, also veiled), ZMF (legitimately nice sounding boutique dynamic driver headphones, but definitely colored and significantly warm of neutral), the senn hd700 (truly an abortion of a headphone), the hifiman HE1000 (veiled, warm, wide, sounds like listening through two cellphone speakers placed 6ft away from your head), and any fostex dynamic drivers (highly dynamic and v-shaped, opposite of hd600)

Most of these are really expensive and/or unobtanium, but as I said you have to do quite a bit to beat the venerable HD600. It would be crazy to buy any of this stuff blind, I would recommend finding a local high-end audio shop where they will let you sit down for a few hours and listen without interruption. Head-fi meets are also a great opportunity to listen to a bunch of gear and meet new people, I'm not sure how often those occur anymore though. Cheers!

I appreciate the detailed response!

What kind of music would you recommend listening to to evaluate headphones in a shop? I plan to grab my HD600 and Grado and compare them to HD800. Aside from Sennheiser and Grado, my local shop lists the following: https://www.missionaudiovideo.com/catalog/headphones-and-wir...

I was told the high impedance can be an issue when listening from a desktop/laptop/phone (as I normally do), so I was actually considering HD660S or even HD598/599. What do you think?

What about in-ear models? I hate the fit of the stock iPhone earbuds (they just don't fit me at all), but provided I find those that fit well I would like to own high quality earbuds. What's the difference in terms of sound between in-ear and around the ear designs?

Listen to anything you like that's well mastered and in a lossless format. Acoustic music recorded in natural spaces tends to reveal the capabilities of a transducer better than synthetic or overly close-miced recordings do, in my opinion. Classical and jazz tend to be most useful here because they tend to escape the dynamic range compression that pervades all of pop and electronic music these days, but there is well-mastered music in any genre. Any music you are super familliar with will basically tell you all you need to know really. A few albums I like to listen to when evaluating gear are Pat Metheny Group - The Way Up, Aphex Twin - Selected Ambient Works 85-92, and Miles Davis - Bitches Brew (not the remaster).

I don't think any of those other cans listed online will challenge the HD600 at all, I know the nightowl and se-master1 are both notoriously bad. But if they have an HD800, definitely go have a listen. It is a polarizing headphone and whether you like it or not will tell you a lot about your own preferences.

High headphone impedance is only an issue if the source you're listening out of has a high output impedance, if the output impedance of the source is high enough, the frequency response will be altered and there will be a FR bump around the primary resonance of the driver (usually around 100hz in most dynamic drivers) due to a lack of electrical damping. A dedicated headphone amp is always best, there are a lot of good ones out there these days that don't break the bank.

I'm not an IEM expert, I've never really felt the need for them and have never found a comfortable pair either. Over-ear designs should have more spatial realism as the sound field interacts with your whole ear (research HRTF and pinna transforms if you want to learn more) compared to IEMs which fire directly into the ear canal and so skip the HRTF entirely.

What do you think about Apple Homepods?

Haven't heard it. I know a lot of people were hyped up when it first came out but I don't really know if it delivered or not. It's supposed to be well DSP'd so I wouldnt be surprised if it beat most of the speakers in its price range though.

I'm thinking about getting them as desktop speakers, as an alternative to using headphones at work, because my ears get hot after a couple of hours, even in HD600.

I live articles that show the process of designing completely new hardware.

I do wonder if this article is missing any dead ends; only showing the iterative improvements?

Looks like it will be impossible to play with someone else in the same room, right?

Sure but you're already flailing around with a vision-blocking helmet, waving your hands. Noise pollution seems like less of a hazard for a bystander than getting whacked by the player.

How does sound affect us between receiving via headphones vs a speaker?

Oh, wow, this really is cool.

Elephant in the room is that it still has a wire sticking out.

That's not the subject. But do you really think they didn't even consider wireless?

Turns out it is really hard. You need a low latency, reliable ~10Gbps link. You also need the battery to be power all the that stuff. Putting everything inside the headset would be ideal but you also have to consider that weight affects comfort and immersion significantly.

The first semi successful attempt is TPCast but it is far from perfect. I don't know much about the Vive wireless but it seems to work better but still have its own set of issues. Both are also rather expensive (~$300).

So for now, we seem to be stuck with fully integrated low end or wired high end. I think we still need to wait a bit more for wireless high end.

I've been using a Vive with the official wireless adapter and it works flawlessly.

The subject is immersion.

My point is that with all that goes into making the audio immersive, the wire out back instantly defeats any gains there.

I wonder if the F/R graph has been pre-corrected against human perception?

It looks really flat -- and you'd ordinarily think that was good, but it turns out that humans don't have equally sensitive hearing at all frequencies. There are a range of curves from different studies, and most musical transducers pick one of those curves and try to emulate it in hardware, so that a smooth frequency sweep from low to high continues to sound equally loud.

Perhaps Valve is planning on adding that through digital equalization.

These folks are rather clearly a lot more sophisticated than "What's a Fletcher Munson Curve?"

In all likelihood they've had to tune things to a very fine degree, both in dsp and in the physical driver design. Because these are used in the near field, with the specific intent of leveraging acoustic cues from the outer ear, they won't be able to just reuse a reference curve developed for headphones or speakers.

They have a dummy head for binaural measurements, so this isn't a particular hard thing for them to calibrate.

> These folks are rather clearly a lot more sophisticated than "What's a Fletcher Munson Curve?"

And yet they're reinventing open headphones and calling it revolutionary? Only a little sarcastic.

If open-back headphones are not closed-back headphones, these off-ear headphones are not open-backed headphones — there seems to be as much differentiation to me.

And they didn’t say “revolutionary” or similar anywhere, as far as I can see.

You're, if I can be frank, wrong. The difference between the two different types of headphones is the amount of isolation the headphones provide. Open back headphones provide minimal, if any, isolation. Closed back provide significant sound isolation. This difference was outlined by a sibling comment which was sadly downvoted.

These are simply open backed headphones that use a different method of setting the distance between the driver and the ear. That's all.

Please see my other comment and dial back the hostility a touch. You are in fact mistaken.

Here's a whitepaper on the driver technology they're using: https://www.tectonicaudiolabs.com/wp-content/uploads/2019/02...

This is the first time I'm aware of anyone has used this class of driver for headphones, and they have novel properties that support their idea of a spaced nearfield monitor rather than an open or closed back cup.

For "off-ear" headphones there are only 3 different ways to build them: closed, open or half-open, solely based on amount of air flow given to a driver (and the type of driver doesn't matter).

These are quite different from other open-back headphones such as HiFiMan, etc.

There's no seal on or around the ears like traditional headphones. Additionally, they sit much farther away from the ears compared to existing headphone designs.

Open ear headphones have no seal, either. The parts that contact your head simply set the distance between your ear and the driver. This is simply open ear headphones being held a set distance from your ear in a slightly different way. Anything else is pure marketing.

No, these are quite different. They transition from axial motion to radial motion at around 3khz. The source in the near field is quite a bit different from an open back headphone driver.

The world around you doesn't adjust it's volume to match human hearing sensitivity at different frequencies - so why would you want to do that when trying to achieve VR immersion?

It is worse than that, the curve of human hearing is not only everything else than flat, it also changes with volume.

One thing to note is that it's not all doom and gloom. Unless aiming for a full VR positional audio experience like Valve is doing here, an audio playback system (headphones, speakers, etc) doesn't have to compensate for all of the various nuances of human hearing. After all, sounds "in real life" don't do that.

All an audio playback system needs to do is reproduce the original sound accurately, perhaps with some Fletcher-Munson compensation to enhance bass at low listening volumes.

(Interestingly, this sort of feature was present on home stereos for several decades. The poorly-named "loudness" knob increased bass so you could enjoy satisfying bass without increasing the overall system volume...)

The loudness button's functionality lives on, in an improved form. Any reasonably modern AV receiver with Audyssey or equivalent room correction/DSP will apply dynamic EQ based on the volume setting, to compensate for volume-dependent frequency sensitivity.

It's a great way to get speakers to sound "bigger" than they are, at volume settings.

Audyssey is absolutely wonderful. It makes a dramatic difference and is light years ahead of a simple loudness knob.

However, I have to wonder - for those with Audyssey-equipped receivers, what % of owners bother to complete the setup process with the calibration microphone? I suspect that number is unfortunately low. :-/

Also, I don't think too many people listen to music on AV receivers these days. The product category itself seems to be fading.

Home theaters have swung toward soundbars, and music is listened to on computers, mobile devices, and/or Bluetooth speakers.

I'm saying this as a pretty old-school stalwart myself, with a fairly capable AV receiver in the living room for video and a bunch of audiophile-ish two channel gear for music.

I don't think the AV receiver is going away anytime soon.

Just like what has happened to stereo amplifiers, what is happening is that the opportunistisk low end of the market is moving on. Cheap radios, turntables and amplifiers were replaced with cheap plastic mini/micro stereos, which were in turn replaced by all-in-ones, which were replaced by bluetooth speakers, and so on.

Cheap AV receivers and especially cheap home-theater-in-a-box type products have been replaced by sound bars, which I would say is actually an improvement in a lot of cases. Some of the low-end AV receivers that were on the market in the 90s and 2000s were unmitigated garbage and a waste of money.

Quality fifi separates live on, and you can still buy mini/micro stereos and all-in-one radio/media player devices. You can still buy AV receivers, but the worst cheap ones have disappeared. I'm not sure you can actually buy HTIB junk anymore, but good riddance to that rubbish.

The appreciation for good sound lives on, but the more complicated devices move to more of a niche role, sticking around for those of us who desire really good sound, and don't mind a bit of a complicated setup process. For everyone else, a sound bar with a wireless subwoofer gets them most of the way, with significantly less effort.

Listening on a computer is simply the next replacement for turntables, tape decks, CD players, DVD/Bluray players, standalone streaming devices and so on. My primary media source is an ultra small form factor PC streaming FLAC and DVD/BR rips from my NAS, hooked up to my receiver over HDMI. It's almost too easy to get amazing sound quality today.

Similarly a smartphone will produce great sound quality with decent source material. The playback device has become a footnote, they're all capable of great sound quality today, as long as you pipe it to something higher quality than a cheap soundbar/bluetooth speaker.

Lots of newer devices compensate for that. The HomePod and QC35 for example.

Applications are open for YC Winter 2020

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact