[gregorio]

My reading of this is that you're mastering for speaker playback - rather than headphone listening - is that right?

Not necessarily, what I was trying to explain is that it depends on what product I’m mixing/mastering and what reproduction environment/equipment the target consumers of that product is likely to be and of course the instructions of the people/entity who is employing me. An extreme example, say I’m working on mixing a theatrical feature film then the target consumers’ reproduction equipment/environment will be a cinema sound system, they’ll never be using some sort of HPs/IEMs and therefore I’ll never check the mix on HPs or make any concessions/compromises for HP use. On the other hand, if I’m mixing or mastering a product destined for distribution through say Apple Music, then I will make concessions/compromises for HP use (and I’ll probably at least check the mix on IEMs as well), unless specifically instructed not to.

G

 

[MrHaelscheir]

@jamesjames I'm with gregorio on modern HRTF rendering software superseding analog or digital implementations of basic crossfeed. I remember the initial pains of figuring out how to use VB-Cable and foo_record to get all my PC audio running into foobar2000. When I tried foo_dsp_crossfeed and foo_dsp_meiercf, I was rather underwhelmed per its sounding like a mere slider between stereo and mono, the latter being a minimum of clarity. It was with bs2b (https://bs2b.sourceforge.net/) that I was first truly impressed by the immediate, pleasant, and clear moving of the image forward with meagre perception of degradation. That implementation seems to be a generalization with presets emulating both Jan and C. Moy crossfeed filters. Any sense of lost clarity or detail seemed to merely be due to the sound no longer being unnaturally spread out. Nonetheless, it wasn't perfect, there still being some higher-frequency content and others that would image from the headphone drivers, evincing this filtering approach being much too simplistic.

I forget when I first learned of https://www.earfish.eu/, and though with their being university staff they can have a long turnaround, the payoff was immense. I also forget when I first had the idea of acquiring in-ear microphones so I could EQ my headphones to match a speaker response, but it was with Earfish that I learned I could also emulate speakers without dishing out on a Smyth Realiser A16, Impulcifier lacking support for head-tracking and requiring me to source my own in-ear mics (I'm still struggling to shop for an improvement upon my current Earfish pair for comfort, linearity, and noise floor). Anyways, in https://www.head-fi.org/threads/in-...frs-are-indeed-identical.970202/post-17801518 (post #61), you can see some of the effort that went into measuring my HRTF and EQing my Meze Elite to match such when running the the SPARTA HRTF rendering VST plugins. The end result is my best approximation of perfectly flat speakers in an anechoic chamber (no comb filtering potentially detracting clarity). Exquisitely coherent imaging, virtually no panned sounds erroneously imaging next to my ears or above my head, and I with SPARTA AmbiRoomSim can reposition the speakers for any distance and width (more convincing imaging of distance is still in the works for me). I can literally "look" at the sound sources mind with head-tracking turn my head toward them. It is simply a fantastic experience for someone without the means to set up a highly treated listening room for speakers. Tonality also remains "coherent" except perhaps expectedly at the extreme angles. And with my free-field EQ (though there are still flaws with how AmbiBIN and Binauraliser NF interact with my EQ for angles other than the reference 30-degree direction mind their differing phase response implementations causing AmbiBIN to for in-phase centered sounds technically be more relaxed than it should while Binauraliser NF gets too bright (I won't share the measurements here), excellent classical recordings sound wonderful.

 

[jamesjames]

@jamesjames I'm with gregorio on modern HRTF rendering software superseding analog or digital implementations of basic crossfeed. I remember the initial pains of figuring out how to use VB-Cable and foo_record to get all my PC audio running into foobar2000. When I tried foo_dsp_crossfeed and foo_dsp_meiercf, I was rather underwhelmed per its sounding like a mere slider between stereo and mono, the latter being a minimum of clarity. It was with bs2b (https://bs2b.sourceforge.net/) that I was first truly impressed by the immediate, pleasant, and clear moving of the image forward with meagre perception of degradation. That implementation seems to be a generalization with presets emulating both Jan and C. Moy crossfeed filters. Any sense of lost clarity or detail seemed to merely be due to the sound no longer being unnaturally spread out. Nonetheless, it wasn't perfect, there still being some higher-frequency content and others that would image from the headphone drivers, evincing this filtering approach being much too simplistic.

I forget when I first learned of https://www.earfish.eu/, and though with their being university staff they can have a long turnaround, the payoff was immense. I also forget when I first had the idea of acquiring in-ear microphones so I could EQ my headphones to match a speaker response, but it was with Earfish that I learned I could also emulate speakers without dishing out on a Smyth Realiser A16, Impulcifier lacking support for head-tracking and requiring me to source my own in-ear mics (I'm still struggling to shop for an improvement upon my current Earfish pair for comfort, linearity, and noise floor). Anyways, in https://www.head-fi.org/threads/in-...frs-are-indeed-identical.970202/post-17801518 (post #61), you can see some of the effort that went into measuring my HRTF and EQing my Meze Elite to match such when running the the SPARTA HRTF rendering VST plugins. The end result is my best approximation of perfectly flat speakers in an anechoic chamber (no comb filtering potentially detracting clarity). Exquisitely coherent imaging, virtually no panned sounds erroneously imaging next to my ears or above my head, and I with SPARTA AmbiRoomSim can reposition the speakers for any distance and width (more convincing imaging of distance is still in the works for me). I can literally "look" at the sound sources mind with head-tracking turn my head toward them. It is simply a fantastic experience for someone without the means to set up a highly treated listening room for speakers. Tonality also remains "coherent" except perhaps expectedly at the extreme angles. And with my free-field EQ (though there are still flaws with how AmbiBIN and Binauraliser NF interact with my EQ for angles other than the reference 30-degree direction mind their differing phase response implementations causing AmbiBIN to for in-phase centered sounds technically be more relaxed than it should while Binauraliser NF gets too bright (I won't share the measurements here), excellent classical recordings sound wonderful.

@MrHaelscheir this is very detailed indeed, and interesting, thanks. I'm not sure who gregorio is, but I don't doubt DSP and HRTF software have much to offer - perhaps even more in the future. I'm put in mind also of Keith Howard's writing some years ago about a 'seat-shaker' to emulate the physical effects of live or loudspeaker presentation. Re foobar etc, my suggestions above are directed to those new to the area and interested in the idea - and almost certainly without the resources needed to conduct the kind of investigations you have undertaken. My own current preference for analogue implementations is based to some extent on pragmatic considerations - I've not had the opportunity to look into all the current options - but of those I've used those I've mentioned seem to me to involve the most satisfactory trade-offs. I find a good amp with a good analogue crossfeed circuit to give excellent results - certainly better than no crossfeed! And readily available even to those with no technical background. Perhaps the most important point for me here, however, isn't to champion any particular solution, but to spread the word that it's an idea worth exploring.

 

[castleofargh]

It depends on the amount of money and/or effort you wish to put into this. The minimum being some crossfeed that either pleases you or doesn't, and the max being full speaker simulation using as many of your own measurements as possible. I started with crossfeed, tried just about everything I could demo or get for free, and still own an old portable DAC/amp with 2 crossfeed levels, but I don't use it anymore. I slowly but surely moved toward more advanced and more customized solutions. I still tried just about anything along the way(OOYH(RIP), The 3D thingy from Waves with the head tracker(same thing I think Audeze used with the mobius), I demoed one Super XFI 3D product for a few minutes, but ended up recording impulses from speakers in my ears, because that's what finally gave me coherent placement and out of my head audio, whihc happens to be the most enjoyable to me and also the least fatiguing(beside real speakers). For a long time I wasted time and efforts doing so myself, now I have a Realiser A16(overly expensive piece of equipment that's quite hard to use properly, but what can I say, I love it). There are several in-between alternatives, DSPs using a generic HRTF(so not your own), some add a few settings for customizations that increase slightly the odds of feeling somewhat like you should. There are also head tracking solutions, that need some HRTF model(ideally your own) to work.
Some people have found satisfaction with just about all the mentioned solution types. It comes down to how much imagination, how close your HRTF is to what is used(be it oversimplified crossfeed delay and EQ or generic HRTF from a dummy head), and to some extent to you and your brain. I can't ever feel fully immersed in anything without head tracking, but I've discussed with people who have it and don't use it because they don't see the point. Different people...

If you wish to try recording real speakers in a room with binaural mics and find out how that's what you've been looking for all your headfier life, without investing a fortune into some expensive system, I recommend asking for information and advice here https://www.head-fi.org/threads/recording-impulse-responses-for-speaker-virtualization.890719/
You need binaural mics(the ones you stick in your own ears to record sounds around you), an ADC depending on the mics' plugs, impulcifer which is free https://github.com/jaakkopasanen/Impulcifer, an actual speaker system, and the patience to figure everything out and practice recording you impulses in the best and most stable conditions you can manage.
Be it the thread above or the main Realiser A16 Thread, people in it tend to be like me and have already tried a bunch of solutions(usually starting with crossfeed), so it's a good idea to just ask some questions there, even about alternative tools. Again, I don't want to be that guy, answering a question on crossfeed with "get the 4000+$ device". I'm saying to ask over there because people are likely to be in a unique position to tell you the pros and cons of the existing solutions, and all the systems that keep coming out, because they're passionate about this specific subject and know a lot.

 

[jamesjames]

In earlier posts I've outlined conventional crossfeed, implemented in DACs or amps by means of analogue circuits or digital signal processing. As noted, I find the analogue implementation in amps generally improves my experience of classical music over headphones. Two variations on the theme seem interesting to me and are sometimes discussed in connection with crossfeed (although I haven't heard them).

The first variation involves greater use of digital signal processing and is perhaps best known as implemented in the 'Smyth Research Realiser'. It involves a high degree of personalisation with a view to achieving a more 'spatially' realistic listening experience. Very broadly, to achieve the best result, the listener must first attend a studio set up with multiple speakers and sit at the point where the speakers focus. Microphones are placed in the listener's ears and music played through the speakers. Software plots the way the information is received by the microphones in the listener's ears to record a personalised 'spatialisation' model reflecting the physical attributes of the listener's head and ears. This model forms the basis of playback software in a special DAC/amp, designed to allow the listener to listen to music with any headphones and achieve the same spatial effect. A 'tracker' can be fitted which, via further DSP, adjusts for movement of the head during playback.

A second variation is probably the most passive, involving special headphones (eg, 'Crosszone' phones) but no special DAC or amp. Very broadly, each ear cup contains several drivers which deliver sound from both left and right channels. The earcups are designed to blend the different sounds fed to them by the various drivers - including by recreating timing and level differences like those occurring naturally at each ear in relation to sound from the 'other' side of the head.

 

[MrHaelscheir]

In earlier posts I've outlined conventional crossfeed, implemented in DACs or amps by means of analogue circuits or digital signal processing. As noted, I find the analogue implementation in amps generally improves my experience of classical music over headphones.

Two variations on the theme seem interesting to me and are sometimes discussed in connection with crossfeed (although I haven't heard them). The first involves greater use of digital signal processing and is perhaps best known in its implementation in the 'Smyth Research Realiser'. It involves a high degree of personalisation with a view to achieving a more realistic listening experience. Very broadly, to achieve the optimal result, the listener must first attend a special studio set up with multiple speakers at various places and sit at the point where these speakers focus. Microphones are then placed in the listener's ears and material played through the speakers. The microphones allow software, in effect, to plot a graph of the listener's experience of music in that 'real' environment. DSP then allows this personalised hearing effect to be reproduced with music played back through any headphones using a special processor/DAC/amp. A 'tracker' is also fitted to the headphones, referrable to the processor/DAC/amp, which adjusts for movement of the head. Clearly, this departs significantly from crossfeed in that it involves a personalised, DSP-driven system. It's also not clear to what extent it may adjust for the basic crossfeed between recorded stereo channels. It would be interesting to hear from someone who has heard this system.

A second variation is possibly the most passive - in involves only modified headphones (I'm aware of only Japanese 'Crosszone' phones). Once again, very broadly, each ear cup contains several drivers, to deliver sound from both left and right channels. The drivers are located and the earcups are designed more generally to blend the resultant sound - including 'delay' to address differences in arrival of information at each ear from the 'opposite' speaker in stereo loudspeaker situations. Once again, it would be great to hear from someone who has experience of these phones.

The first DSP category can be split into a variety of things that all involve in-ear measurements. I believe the Smyth Realiser A16, and Impulcifier fall under capturing the response and sound of an entire speaker system, the latter as of yet lacking the head-tracking component which I believe the Smyth Realiser A16 implements using special interpolation from less measurements. The "material" generally comprises sine sweeps (measuring how your ears modify the loudness of each frequency), and the "graph of the listener's experience of music" is the frequency response of the room, or when using convolution filters, you first convert that into a time domain impulse response. I don't know if these measurement methods actually go as far as to capture all the reflection properties of a given room, though with a sufficiently high-resolution sine sweep, it is in fact possible to see the delayed spikes from room reflections within the calculated impulse response. TIL that FIR and "convolution" EQ are synonymous: https://www.reddit.com/r/oratory1990/comments/go2grd/convolution_eqs_what_are_they_what_problems_do/. I guess the idea is that compared to the IIR filters used for parametric EQ, being able to use a .WAV file (recorded else generated from the sine sweep through the Fourier transform) instead of having to configure a bunch of parametric filters enables capturing all the miniscule details (the more reflective the room, the more jagged the plotted frequency response looks due to all the comb filtering) of a speaker's in-room response and perhaps the reflection properties.

Then you have full HRTF capture that ideally eliminates the sound of the room and I suppose approximates either your free-field (anechoic) or diffuse-field response (reflective room) in all direction (at least interpolated between a whole lot more measured directions). https://www.earfish.eu/ (which I use) and Genelec Aural ID mind those folks who actually walk into a specialized HRTF measurement lab falls under this. Basically, any solution that produces a SOFA or other HRTF file that you load into special binaural rendering software. I think BACCH is just one such example of software that accepts SOFA files for personalized HRTF. This rendering software generally runs on your own computer and through your DAC and amp of choice, as does Impulcifier.

I've already shared my experience with the latter system for the aim of simulating anechoic sound (removing reflections from the recorded impulses; I suppose I could use my original unsmoothed measurements to generate convolution filters for simulating listening to speakers in my backyard...). https://www.head-fi.org/threads/smyth-research-realiser-a16.807459/ and https://www.head-fi.org/threads/recording-impulse-responses-for-speaker-virtualization.890719/ should have the impressions of many users of the Smyth Realiser A16 and Impulcifier for simulating actual speaker sound.

 

[jamesjames]

The first DSP category can be split into a variety of things that all involve in-ear measurements. I believe the Smyth Realiser A16, and Impulcifier fall under capturing the response and sound of an entire speaker system, the latter as of yet lacking the head-tracking component which I believe the Smyth Realiser A16 implements using special interpolation from less measurements. The "material" generally comprises sine sweeps (measuring how your ears modify the loudness of each frequency), and the "graph of the listener's experience of music" is the frequency response of the room, or when using convolution filters, you first convert that into a time domain impulse response. I don't know if these measurement methods actually go as far as to capture all the reflection properties of a given room, though with a sufficiently high-resolution sine sweep, it is in fact possible to see the delayed spikes from room reflections within the calculated impulse response. TIL that FIR and "convolution" EQ are synonymous: https://www.reddit.com/r/oratory1990/comments/go2grd/convolution_eqs_what_are_they_what_problems_do/. I guess the idea is that compared to the IIR filters used for parametric EQ, being able to use a .WAV file (recorded else generated from the sine sweep through the Fourier transform) instead of having to configure a bunch of parametric filters enables capturing all the miniscule details (the more reflective the room, the more jagged the plotted frequency response looks due to all the comb filtering) of a speaker's in-room response and perhaps the reflection properties.

Then you have full HRTF capture that ideally eliminates the sound of the room and I suppose approximates either your free-field (anechoic) or diffuse-field response (reflective room) in all direction (at least interpolated between a whole lot more measured directions). https://www.earfish.eu/ (which I use) and Genelec Aural ID mind those folks who actually walk into a specialized HRTF measurement lab falls under this. Basically, any solution that produces a SOFA or other HRTF file that you load into special binaural rendering software. I think BACCH is just one such example of software that accepts SOFA files for personalized HRTF. This rendering software generally runs on your own computer and through your DAC and amp of choice, as does Impulcifier.

I've already shared my experience with the latter system for the aim of simulating anechoic sound (removing reflections from the recorded impulses; I suppose I could use my original unsmoothed measurements to generate convolution filters for simulating listening to speakers in my backyard...). https://www.head-fi.org/threads/smyth-research-realiser-a16.807459/ and https://www.head-fi.org/threads/recording-impulse-responses-for-speaker-virtualization.890719/ should have the impressions of many users of the Smyth Realiser A16 and Impulcifier for simulating actual speaker sound.

Thanks for that. I was conscious of your post of course when I wrote the previous post - an attempt to introduce what is potentially a very complex subject to readers likely to have no background (except perhaps an interest in off-the-shelf crossfeed products). The Realiser and Crosszone products are the only two 'variant' off-the-shelf products I know. And I would be interested to hear from anyone who actually has them up and running. Your response above obviously touches on the comparison with crossfeed - and was of particular interest to me. While we're on that, and apologies if I've missed your discussion of this, but does your arrangement address 'crossfeed' as opposed to spatialisation? As you know, strictly, this isn't simply a slider between stereo and mono - it involves parameters to deal with interaural level and time differences.

I should add that my scan of the Realiser thread didn't yield up much on the comparison with crossfeed. It seemed also to focus on the detail of adjustment rather than 'big picture' impressions. I suppose I'm particularly interested in why a listener might prefer one over the other (which your earlier post addresses to some extent).

 

[MrHaelscheir]

Thanks for that. I was conscious of your post of course when I wrote the previous post - an attempt to introduce what is potentially a very complex subject to readers likely to have no background (except perhaps an interest in off-the-shelf crossfeed products). The Realiser and Crosszone products are the only two 'variant' off-the-shelf products I know. And I would be interested to hear from anyone that actually has them up and running. Your response above obviously touches on the comparison with crossfeed - and was of particular interest to me. While we're on that, and apologies if I've missed your discussion of this, but does your arrangement address 'crossfeed' as opposed to spatialisation? As you know, strictly, this isn't simply a slider between stereo and mono - it involves parameters to deal with interaural level and time differences.

I should add that my scan of the Realiser thread didn't yield up much on the comparison with crossfeed. It seemed also to focus on the detail of adjustment rather than 'big picture' impressions. I suppose I'm particularly interested in why a listener might prefer one over the other (which your earlier post addresses to some extent).

Ah. I suppose to clarify things, with the Realiser and HRTF rendering software, the "spatialisation" necessarily implements some form of "crossfeed" (they are not in "opposition"), just that it uses more sophisticated filters than traditional crossfeed. From my understanding, "crossfeed" is simply simulation of how a single sound source from a particular direction presents sound differently to both ears. https://bs2b.sourceforge.net/ gives a decent background on the "filters" used by traditional crossfeed; that is, traditional crossfeed effectively performs a shelf EQ and level change on a channel before sending it to the other ear along with adding a delay. This is a substantial simplification compared to the actual "crossfeed" naturally encountered by human ears.

Compare the below filter (green trace):

Figure 1: https://bs2b.sourceforge.net/ crossfeed filter construction.

To my actual left ear response ("blocked canal" as opposed to at-eardrum measurement) for the left and right channels in an equilateral stereo triangle:

Figure 2: Left ear receiving a sine sweep from a speaker positioned 30 degrees to the left. This measurement includes all the flaws due to reflections within my backyard.

Figure 3: Left ear receiving a sine sweep from a speaker positioned 30 degrees to the right. Here, you can see lots a lot more comb filtering perhaps accentuated by the diffraction of the sound around my head. This would ideally be smoothed out, but if I were to generated "crossfeed" convolution filters with this much detail, I would probably effectively end up simulating "backyard reverb". The right ear receiving a sine sweep from the left speaker should have a similar response as this other than differences due to measurement imperfections or actual physiological asymmetries.

Perhaps I should have clarified what a measurement and rendering implies. When you measure your HRTF or personal in-ear room impulse response, the software uses the sine sweep to capture how the sound source for that particular speaker direction relative to your head presents sound to both of your ears. the two previous graphs are higher-resolution versions of such. If you measured the speaker with it 30 degrees to the left, then the measurement software will capture the delays or phase differences in the sound received by the ears as well as how your own physical ears (auricles) directionally alter the levels of each frequency (frequency response as well as general) in your ears. That data is effectively the filter parameters for a much more detailed "crossfeed" implementation. The software then uses that measurement to apply those same delays and in-ear frequency response and level alterations to a given audio channel, aiming to replicate the same way a single speaker's sound would reach both of your ears. And when you have two or more virtual speakers at once, you necessarily experience having multiple virtual channels and their directional specifics being mixed into each ear. That replication causes you to hear that virtual sound source from the exact same position as where you measured that speaker from.

I don't know much about the Realiser's subjective impressions (maybe looking up "reviews" of such could reveal these) other than the claims that it convincingly reproduces the sound and I suppose imaging of any given speaker system (minus subwoofer effects and whatnot), which to me implies the capacity to also capture all of the flaws of the room such as comb filtering or asymmetries that can affect clarity and imaging coherency (e.g. if you play a centered sine sweep and you hear the tone wildly shifting left and right, that would imply poor imaging coherency where the frequency content of individual sound sources is being smeared into different directions). HRTF capture aims to remove the room reflections (more so for versatility between software) be it through taking the measurements in an anechoic chamber or "windowing" (truncating) the measured impulse responses at the cost of measurement resolution. From my understanding, the hallmarks of anechoic stereo listening is perfect pinpoint imaging between the two (virtual) speakers, and from my experience, sound sources properly in front of you that you can actually pinpoint "look at" or believe you are looking toward. If you play a given sound source or instrument, everything will coherently sound like it is coming from that location, provided that your HRTF measurement and simulation is accurate enough.

Compared to traditional crossfeed, I find that the simplistic filters can cause individual sound sources or instruments' frequencies to be smeared. Usually, it is the higher-frequency content that fails to be pushed sufficiently forward within the image. Maybe some other crossfeed implementations are better at this. One test would be to play a sine sweep into just one channel and listen for whether its panned position remains consistent, else to play other test signals through just one channel and listen closely for whether you are actually hearing sounds only coming from one direction. Regardless, the forward imaging allowed by basic crossfeed can't be called "accurate" or a true simulation of speaker presentation. The advantage I find from proper HRTF rendering and "crossfeed" is the exquisite imaging coherence and the sheer enhanced ease with which you can hear and "see" all of the sound sources clearly laid out before you on a line between the virtual channels. It was a huge improvement compared to the incremental one from switching from not using crossfeed to using bs2b. And when you add head-tracking, those sound sources sit stationary in space as you rotate your head (unless they are imminently being panned within the track) other than some lag from tracking latency, your being able to rotate your head to be centered with individual sounds/instruments in the mix. I suppose if you think traditional crossfeed creates a nice forward image, unless you do already have experience with speakers, wait until you hear proper forward imaging.

 

[jamesjames]

Ah. I suppose to clarify things, with the Realiser and HRTF rendering software, the "spatialisation" necessarily implements some form of "crossfeed" (they are not in "opposition"), just that it uses more sophisticated filters than traditional crossfeed. From my understanding, "crossfeed" is simply simulation of how a single sound source from a particular direction presents sound differently to both ears. https://bs2b.sourceforge.net/ gives a decent background on the "filters" used by traditional crossfeed; that is, traditional crossfeed effectively performs a shelf EQ and level change on a channel before sending it to the other ear along with adding a delay. This is a substantial simplification compared to the actual "crossfeed" naturally encountered by human ears.

Compare the below filter (green trace):


Figure 1: https://bs2b.sourceforge.net/ crossfeed filter construction.

To my actual left ear response ("blocked canal" as opposed to at-eardrum measurement) for the left and right channels in an equilateral stereo triangle:


Figure 2: Left ear receiving a sine sweep from a speaker positioned 30 degrees to the left. This measurement includes all the flaws due to reflections within my backyard.


Figure 3: Left ear receiving a sine sweep from a speaker positioned 30 degrees to the right. Here, you can see lots a lot more comb filtering perhaps accentuated by the diffraction of the sound around my head. This would ideally be smoothed out, but if I were to generated "crossfeed" convolution filters with this much detail, I would probably effectively end up simulating "backyard reverb". The right ear receiving a sine sweep from the left speaker should have a similar response as this other than differences due to measurement imperfections or actual physiological asymmetries.

Perhaps I should have clarified what a measurement and rendering implies. When you measure your HRTF or personal in-ear room impulse response, the software uses the sine sweep to capture how the sound source for that particular speaker direction relative to your head presents sound to both of your ears. the two previous graphs are higher-resolution versions of such. If you measured the speaker with it 30 degrees to the left, then the measurement software will capture the delays or phase differences in the sound received by the ears as well as how your own physical ears (auricles) directionally alter the levels of each frequency (frequency response as well as general) in your ears. That data is effectively the filter parameters for a much more detailed "crossfeed" implementation. The software then uses that measurement to apply those same delays and in-ear frequency response and level alterations to a given audio channel, aiming to replicate the same way a single speaker's sound would reach both of your ears. And when you have two or more virtual speakers at once, you necessarily experience having multiple virtual channels and their directional specifics being mixed into each ear. That replication causes you to hear that virtual sound source from the exact same position as where you measured that speaker from.

I don't know much about the Realiser's subjective impressions (maybe looking up "reviews" of such could reveal these) other than the claims that it convincingly reproduces the sound and I suppose imaging of any given speaker system (minus subwoofer effects and whatnot), which to me implies the capacity to also capture all of the flaws of the room such as comb filtering or asymmetries that can affect clarity and imaging coherency (e.g. if you play a centered sine sweep and you hear the tone wildly shifting left and right, that would imply poor imaging coherency where the frequency content of individual sound sources is being smeared into different directions). HRTF capture aims to remove the room reflections (more so for versatility between software) be it through taking the measurements in an anechoic chamber or "windowing" (truncating) the measured impulse responses at the cost of measurement resolution. From my understanding, the hallmarks of anechoic stereo listening is perfect pinpoint imaging between the two (virtual) speakers, and from my experience, sound sources properly in front of you that you can actually pinpoint "look at" or believe you are looking toward. If you play a given sound source or instrument, everything will coherently sound like it is coming from that location, provided that your HRTF measurement and simulation is accurate enough.

Compared to traditional crossfeed, I find that the simplistic filters can cause individual sound sources or instruments' frequencies to be smeared. Usually, it is the higher-frequency content that fails to be pushed sufficiently forward within the image. Maybe some other crossfeed implementations are better at this. One test would be to play a sine sweep into just one channel and listen for whether its panned position remains consistent, else to play other test signals through just one channel and listen closely for whether you are actually hearing sounds only coming from one direction. Regardless, the forward imaging allowed by basic crossfeed can't be called "accurate" or a true simulation of speaker presentation. The advantage I find from proper HRTF rendering and "crossfeed" is the exquisite imaging coherence and the sheer enhanced ease with which you can hear and "see" all of the sound sources clearly laid out before you on a line between the virtual channels. It was a huge improvement compared to the incremental one from switching from not using crossfeed to using bs2b. And when you add head-tracking, those sound sources sit stationary in space as you rotate your head (unless they are imminently being panned within the track) other than some lag from tracking latency, your being able to rotate your head to be centered with individual sounds/instruments in the mix. I suppose if you think traditional crossfeed creates a nice forward image, unless you do already have experience with speakers, wait until you hear proper forward imaging.

That's very helpful indeed. For what it's worth, I've always assumed that standard crossfeed is a substantial simplification compared to the 'crosfeed' actually experienced by our ears. Thanks for clarifying the crossfeed component of the HTRF rendering and Realiser software.

As to my own experience, I've played for many years (classical music) and been involved in recording. I've many years' experience with speaker systems - domestic and professional - but for the last decade or so have preferred phones. I find phones avoid the problem of (enough) power and loudspeakers (particularly in the domestic setting). It's possible with phones to achieve fidelity I can't in practice achieve with loudspeakers. I find crossfeed - done well - makes listening better. The issue for me has always been the SQ trade-off with additional digital processing. In my experience processors often pay an audible price for processing load. This is one aspect of the HTRF rendering software and implementation that I'm particularly interested in - which is behind my interest in the impressions of others. It's one thing to achieve structure in the projected performance space - but do the violins still sound like violins?

 

[MrHaelscheir]

but do the violins still sound like violins?

Maybe not the best way to phrase it unless you are hearing truly drastic changes. Any loss of clarity I might hear I personally attribute to limitations of the mixing implementation or the fact that the sound the headphones had "stretched out" and "unmixed" by playing each channel into one ear at a time is now being reintegrated and also heard from less wide of a sonic image. Of course, a violin or a whole strings section can sound pretty different between different venues and different seats within each, and likewise between each recording. There are some live performances where I feel like the specific harmonic distribution I heard probably hasn't been captured in any recording and could not be achieved with EQ or speaker room setup unless you were present at the venue with neutral mics at ideal positions and the means for neutral mixing. Otherwise, unless one's subjective response to the knowledge of the use of DSP permanently inhibits one's perception of the DSP's transparency, the only "SQ trade-off" I could imagine would be errors in the HRTF measurements themselves and the calibration of the headphones to a given response. Even with SPARTA AmbiBIN's measurable shortcomings, Boulez' Mahler 5 (https://app.idagio.com/recordings/18124541) still sounds absolutely wonderful and tonally the closest to what I've experienced on two occasions at Roy Thomson Hall. The ability to image the aggressive and textured hugeness in front of you, and the vividness of the strings and brass. From my perspective, I'm coming very close to having that "headphones as portals for your ears into the concert hall" experience, there just being a few kinks to resolve for further perfecting head-tracked tonality and imaging. It at least certainly sounds better and more vivid than applying bs2b crossfeed to my previous neutral PEQ.

 

[jamesjames]

Maybe not the best way to phrase it unless you are hearing truly drastic changes. Any loss of clarity I might hear I personally attribute to limitations of the mixing implementation or the fact that the sound the headphones had "stretched out" and "unmixed" by playing each channel into one ear at a time is now being reintegrated and also heard from less wide of a sonic image. Of course, a violin or a whole strings section can sound pretty different between different venues and different seats within each, and likewise between each recording. There are some live performances where I feel like the specific harmonic distribution I heard probably hasn't been captured in any recording and could not be achieved with EQ or speaker room setup unless you were present at the venue with neutral mics at ideal positions and the means for neutral mixing. Otherwise, unless one's subjective response to the knowledge of the use of DSP permanently inhibits one's perception of the DSP's transparency, the only "SQ trade-off" I could imagine would be errors in the HRTF measurements themselves and the calibration of the headphones to a given response. Even with SPARTA AmbiBIN's measurable shortcomings, Boulez' Mahler 5 (https://app.idagio.com/recordings/18124541) still sounds absolutely wonderful and tonally the closest to what I've experienced on two occasions at Roy Thomson Hall. The ability to image the aggressive and textured hugeness in front of you, and the vividness of the strings and brass. From my perspective, I'm coming very close to having that "headphones as portals for your ears into the concert hall" experience, there just being a few kinks to resolve for further perfecting head-tracked tonality and imaging. It at least certainly sounds better and more vivid than applying bs2b crossfeed to my previous neutral PEQ.

I accept there could be expectation bias in personal impressions - re the awareness of DSP use. On the other hand, I'm in no doubt even the 'best' crossfeed changes the presentation through my phones. I accept some may perceive this change to be detrimental (although I don't hear it that way). There's a question here I think whether current designs actually compensate for absence of crossfeed by adjusting headphone presentation generally in a way that's compromised when crossfeed's added. If so, this might be a factor in personal assessments of accuracy/fidelity, etc. A listener used to the standard approach might regard it as 'correct'. At the end of the day my own test is whether my system is convincing - by which I mean apparently placing me in the audience at a performance at a venue with a flattering acoustic. It must be a little conjectural for the reasons you've already mentioned. Inherent limitations of driver technology, amplification and signal processing will all affect the result. But I find crossfeed helps. And it's very encouraging to read your observations about the results of HTRF software - which seem to be bringing you closer to a 'convincing' experience.

 

[jamesjames]

Maybe not the best way to phrase it unless you are hearing truly drastic changes. Any loss of clarity I might hear I personally attribute to limitations of the mixing implementation or the fact that the sound the headphones had "stretched out" and "unmixed" by playing each channel into one ear at a time is now being reintegrated and also heard from less wide of a sonic image. Of course, a violin or a whole strings section can sound pretty different between different venues and different seats within each, and likewise between each recording. There are some live performances where I feel like the specific harmonic distribution I heard probably hasn't been captured in any recording and could not be achieved with EQ or speaker room setup unless you were present at the venue with neutral mics at ideal positions and the means for neutral mixing. Otherwise, unless one's subjective response to the knowledge of the use of DSP permanently inhibits one's perception of the DSP's transparency, the only "SQ trade-off" I could imagine would be errors in the HRTF measurements themselves and the calibration of the headphones to a given response. Even with SPARTA AmbiBIN's measurable shortcomings, Boulez' Mahler 5 (https://app.idagio.com/recordings/18124541) still sounds absolutely wonderful and tonally the closest to what I've experienced on two occasions at Roy Thomson Hall. The ability to image the aggressive and textured hugeness in front of you, and the vividness of the strings and brass. From my perspective, I'm coming very close to having that "headphones as portals for your ears into the concert hall" experience, there just being a few kinks to resolve for further perfecting head-tracked tonality and imaging. It at least certainly sounds better and more vivid than applying bs2b crossfeed to my previous neutral PEQ.

By the way (off-topic - apologies), any thoughts on the relative merits of the phones in your signature - Elite, HEKSE, Arya? I've owned and enjoyed them all myself.

 

[MrHaelscheir]

By the way (off-topic - apologies) any thoughts on the relative merits of the Elite, the HEKse and the Arya. I've owned them all myself.

:O This is going to be fun.

https://www.head-fi.org/threads/mez...official-thread.959445/page-303#post-17509345 #4,541 covers the details of the journey of auditioning the Arya Stealth and within three months auditioning the Meze Elite, mainly because the HE1000se which I was more enamored by from reviews wasn't available for demo at any local shop. https://www.head-fi.org/threads/hifiman-he1000-se.886228/page-320 covers my initial findings of the HE1000se.

My "summary" from my present perspective:

So I started out with cheap desktop speakers that I had at least arbitrarily toed in since I thought that was how you were supposed to position them, then switched to earbuds through latter high school and early university, then cheap Sonies, and finally in latter university got the ATH-M50xBT which I found needed EQ in order to sound right, the lower midrange dip having caused it to sound dull, wrong, and effectively "worse" than my Sony at least for piano tone. Likewise, something special with the EQ profile I set up on my Kawaii ES8 still has that headphone sounding magical for the EX Concert Grand samples. Two years later for working from home during the pandemic, I got the Jabra Elite 85h as my first "neutral out of the box" headphone, and it indeed sounded excellent, though I had to get used to the pad seal. I had seen the Arya and HD 800 S on RTINGS.com and had always thought such exorbitant.

Arya Stealth:

Two years later, a bit past exactly a year ago, I came across Joshua Valour's video about the Edition XS and entered the rabbit hole of review videos. By then, I was "swimming" in cash and a bonus mainly thanks to my typewriter collecting hobby having died down (the remaining machines I was looking for becoming more sparse to find), so I within two weeks finally entered a headphone shop equipped with everything I had watched and read up on (including an assumption that cables most probably don't matter; I had finished a computer engineering degree). So I see this now $1.3k USD or $1.8 CAD headphone before me, no longer so mystical, just a venerated object (same deal up to my encountering the X9000 and HE-1). I was probably fascinated by the comfort. I then played my favourite YouTube recording of Brahms Piano Concerto No. 2. Lo and behold, I could not hear "soundstage". Everything was imaging on or between the drivers just as they did with my Jabra Elite, and when volume matched, I felt like my Jabra Elite 85h could sound just as big and detailed. Same deal with the HD 800 S they had. This was off of the Chord Hugo 2, then off of the Mojo 2 after that ran out of batteries. I think the Mojo 2 had rather discrete volume controls. It took two hours of back and forth amid the desire to not walk out of the store empty-handed, my eventually "fooling" myself into hearing more "definition" in piano tones through the Focal Clear and then Arya Stealth. So the first lesson was that I could not hear "soundstage", at least nothing "vastly wide", and that between multiple for the most part neutral headphones, the Arya Stealth only subtle sounded the best to me.

So I got home and straightway made an Equalizer APO variable-band graphic EQ based on the headphones.com GRAS measurements, resulting in an immediate improvement in clarity, no "burn-in" needed. Further improvements were obtained after learning how to adjust treble EQ by ear. I suppose there was a point where I felt content that the Arya Stealth was indeed "clearer" sounding than the Jabra, at least with EQ. I would later learn/measure that my Jabra actually had a whole lot more bass than I thought, probably from substantial pad compression after those two years, contributing to the perception of less clarity. I appreciated the larger presentation of the Arya Stealth, but with closer listening felt like imaging differences were meagre in terms of positioning and size, the only difference being the extent to which the headphone pads' size causes them to intersect with or make way for that image. It required special mixes like Yosi Horikawa or others playing with HRTF effects to get properly out-of-head sounds. Then I finally tried out software crossfeed, whose experience I've already described, such having been a revolution in itself. I was probably still quite skeptical in regard to perception of headphone "resolution" and other claimed "technicalities".

Meze Elite:

The Meze Elite audition was a lesson in how using other test rigs like headphones.com's measurements to produce EQ profiles is absolutely no guarantee that the headphones will actually have the same response on your own head and ears. In this case, the Meze Elite ended up having a 600 Hz elevation that made Rodrigo y Gabriela tracks sound "sweeter" in a wonderful and impressive way that yielded my first experience of being truly astonished. There was then this HRTF-effect track where I for the first time heard the distant panning sound source sound "ridiculously" far away; this only happened once, the differences in perceived distance growing smaller with later comparisons. The only "soundstaging" differences I heard were related to the sense of the slight differences in driver distance and the amount of space around my ears. I also did not hear any advantageous "impact" or "slam" out of the Focal Utopia. I also learned that the DCA Expanse despite being effectively Harman-tuned was at the time less preferable than the result from EQing the Meze Elite. I likewise learned that I preferred large planar presentation and tactility, the Meze Elite sharing a level of openness with the Arya Stealth though the latter seemed more incisive. Regardless, I have no regrets since the Meze Elite Tungsten version is freaking gorgeous and exquisitely comfortable, and I have in the past few months with my in-ear mics uncovered this headphone's absolutely exceptional distortion performance even when EQing up the wide upper midrange dip. I while A/Bing found that I could listen to pure tones in order to EQ the headphones to roughly the same response frequency by frequency, whereby I was able to transfer the sweetness of the Meze Elite's EQ to the Arya Stealth and the clarity of the Arya Stealth's EQ to the Meze Elite. By then, I was convinced that all matters of tonality are founded in frequency response and can be transparently adjusted with good digital EQ. I had also found that EQing down the 6 kHz to 8 kHz region in a slope of relatively equal loudness greatly improved clarity and the cleanness of my sound.

I didn't like the feel of the Angled Alcantara pads and didn't hear any soundstaging or imaging improvements. They do measurably and audibly drive things a bit toward neutral, but I much preferred EQing the sweet lambskin (I think) hybrid pads which could feel wonderfully cool on first wearing and likewise had better bass extension and hence distortion.

Then came my in-ear mics and the HRTF and free-field EQing results I had described. It so happens that the Meze Elite may just be the perfect headphone for binaural head-tracking since the headband is carbon fiber and makes it really easy to slot in the head-tracking unit, and the magnetic field happens to be very well contained within the drivers so as to not interfere with the head-tracker's magnetometer, at least my cellphone app showing that even my two previous dynamic driver headphones exhibit stronger magnetic fields at the headband. In the delay between receiving my in-ear mics and receiving my HRTF SOFA file, I discovered that my EQing by ear (and making the initial profile using GRAS measurements) was extremely inaccurate in the midrange which actually came to be very wavy for my in-ear measurements, my refining this into the "V3 PEQ" which wasn't far off from the smoothness of my measurement of the HE-1 and X9000 for which I was able to say that I preferred the sound, cleanness, and presentation of my EQed Meze Elite. My in-ear speaker measurements also revealed that neutral speakers actually have decently more ear gain than my Harman EQ, whereby some of the "vividness" I had been looking for was to be found in EQing in more ear gain consistent with my HRTF and of course getting used to such, whereby the V3 PEQ now sounds a bit relaxed by comparison. Now, at this moment, I am somewhat surprised that this free-field EQ allowed Zubin Mehta's 1977 recording of Mahler 5 whose second movement I particularly enjoyed to sound more competitively clean and vivid against my reference 1996 Boulez recording on Idagio whence I had previously felt that it didn't sound as "hi-fi" as the best recordings I had encountered.

HE1000se:
I only got this in December as a result of my finally deciding that the X9000 wasn't for me both in terms of comfort and presentation as well as its despite the estat reputation measuring to not have much better distortion performance than my Meze Elite. I was still ready to possibly be impressed, and was technically hoping that the reviews were true or that this different magnet array would feel more open. I was also at the time becoming discontent with the Meze Elite's comfort and wanted to explore pad-rolling the HE1000se partly to optimize its EQed distortion and CSD measurements.

And so it arrived, my as part of a burn-in debunking plan playing a Room EQ Wizard sine sweep as the first signal. Measurements showed very similar frequency responses, but at least better control of the 4 kHz peak and perhaps more consistently good driver matching where my Arya Stealth wasn't well matched in the ear gain region. I was happy to see better distortion performance through the HE1000se, though I later learned that this was only the case for the left drivers, its being switched around for the right drivers, and now lately per https://www.head-fi.org/threads/totl-disappointments.925164/post-17949397 (post #936), at least units from November exhibit clear distortion defects albeit not particularly audible in practical music listening, but still not so nice of a thing to come across for the supposed "best" out of the HE1000 line. I personally don't believe in burn-in or believe its effects to be psychological in nature. Regardless, I so have found the HE1000se to sound very similar to the Arya Stealth, not doing any magically soundstaging or imaging stuff, the Arya sounding just as detailed and textured. Now maybe without crossfeed, there were cases where something through the HE1000se would sound like it were positioned a bit higher, but this may have rather been a mismatch in the vertical positioning of the cups since the two headphones have different headbands. I personally prefer the wider and more cushioned headband of the Arya Stealth. They also both feel equally open and transparent to external noises, and probably similarly tactile for large bass transients. I measured similar impulse and step responses, and also hear similarly incisive transients when listening to isolated impulses, though I am now inclined to believe that such differences in the quality of extreme transients (playing http://pcfarina.eng.unipr.it/Acustica-samples/Dirac.wav really loud) are due to nonlinearities showing up, whereby in practical listening, differences in transient quality decrease. Otherwise, the main things to like are the looks and the lighter clamping force, though I can still feel pressure points when wearing them for long enough. Worn HiFiMan pads, even for just a few months, have a rather nice, soft, and cool initial feel compared to fresh pads.

Given this, the main reason for having the HE1000se is as a comfort and presentation alternative for the same tonality, and for it to feel like a more "worthy" complement to my Meze Elite than the Arya Stealth with its measurable driver matching issues for my unit. I've already devised a free-field EQ which I document in https://www.audiosciencereview.com/...n-susvara-headphone-review.50705/post-1853760 (post #1,137), whereby my HE1000se with these huge pads (though pads are exceedingly easier to swap on the Meze Elite, with the HE1000se, there are a lot more pad options, mind the ability for full customization through NTRAX Mod Design (NMD)) effectively sounds tonally identical to the Meze Elite, but with a whole lot more space around my ears. The problem is that it has a metal headband and a powerful magnetic field that currently makes it impossible to use with magnetometer-based head-trackers, a problem I am currently trying to solve. The custom pads feel excellently snug and comfortable, though even with the lighter clamping force of the HE1000se compared to the Arya Stealth, I can still after enough time feel pressure points or like circulation to my auricles is being constricted (same deal with the Meze Elite after enough time or on certain occasions; some days, the Meze Elite just feels perfect and I don't want to take them off or stop listening because they feel so darned snug). I suppose the HE1000se with NMD pads for the same EQ and very similar in-ear response can still feel a bit "airier" while switching back to the Meze Elite can have such feeling a bit smoother.

tl;dr: Meze Elite for exquisite distortion performance (also a pretty good CSD for a planar magnetic, but not as good as dynamic driver or closed-back CSDs) and sweet, soft, and initially cool-feeling lambskin pads. HE1000se for worse distortion, especially for multitone IMD, but with huge custom pads and maybe more incisive transients at the extremes. You can see https://www.head-fi.org/threads/rec...r-speaker-virtualization.890719/post-17955514 (post #1,801) for my latest multitone distortion comparison.

 

[jamesjames]

:O This is going to be fun.

https://www.head-fi.org/threads/mez...official-thread.959445/page-303#post-17509345 #4,541 covers the details of the journey of auditioning the Arya Stealth and within three months auditioning the Meze Elite, mainly because the HE1000se which I was more enamored by from reviews wasn't available for demo at any local shop. https://www.head-fi.org/threads/hifiman-he1000-se.886228/page-320 covers my initial findings of the HE1000se.

My "summary" from my present perspective:

So I started out with cheap desktop speakers that I had at least arbitrarily toed in since I thought that was how you were supposed to position them, then switched to earbuds through latter high school and early university, then cheap Sonies, and finally in latter university got the ATH-M50xBT which I found needed EQ in order to sound right, the lower midrange dip having caused it to sound dull, wrong, and effectively "worse" than my Sony at least for piano tone. Likewise, something special with the EQ profile I set up on my Kawaii ES8 still has that headphone sounding magical for the EX Concert Grand samples. Two years later for working from home during the pandemic, I got the Jabra Elite 85h as my first "neutral out of the box" headphone, and it indeed sounded excellent, though I had to get used to the pad seal. I had seen the Arya and HD 800 S on RTINGS.com and had always thought such exorbitant.

Arya Stealth:

Two years later, a bit past exactly a year ago, I came across Joshua Valour's video about the Edition XS and entered the rabbit hole of review videos. By then, I was "swimming" in cash and a bonus mainly thanks to my typewriter collecting hobby having died down (the remaining machines I was looking for becoming more sparse to find), so I within two weeks finally entered a headphone shop equipped with everything I had watched and read up on (including an assumption that cables most probably don't matter; I had finished a computer engineering degree). So I see this now $1.3k USD or $1.8 CAD headphone before me, no longer so mystical, just a venerated object (same deal up to my encountering the X9000 and HE-1). I was probably fascinated by the comfort. I then played my favourite YouTube recording of Brahms Piano Concerto No. 2. Lo and behold, I could not hear "soundstage". Everything was imaging on or between the drivers just as they did with my Jabra Elite, and when volume matched, I felt like my Jabra Elite 85h could sound just as big and detailed. Same deal with the HD 800 S they had. This was off of the Chord Hugo 2, then off of the Mojo 2 after that ran out of batteries. I think the Mojo 2 had rather discrete volume controls. It took two hours of back and forth amid the desire to not walk out of the store empty-handed, my eventually "fooling" myself into hearing more "definition" in piano tones through the Focal Clear and then Arya Stealth. So the first lesson was that I could not hear "soundstage", at least nothing "vastly wide", and that between multiple for the most part neutral headphones, the Arya Stealth only subtle sounded the best to me.

So I got home and straightway made an Equalizer APO variable-band graphic EQ based on the headphones.com GRAS measurements, resulting in an immediate improvement in clarity, no "burn-in" needed. Further improvements were obtained after learning how to adjust treble EQ by ear. I suppose there was a point where I felt content that the Arya Stealth was indeed "clearer" sounding than the Jabra, at least with EQ. I would later learn/measure that my Jabra actually had a whole lot more bass than I thought, probably from substantial pad compression after those two years, contributing to the perception of less clarity. I appreciated the larger presentation of the Arya Stealth, but with closer listening felt like imaging differences were meagre in terms of positioning and size, the only difference being the extent to which the headphone pads' size causes them to intersect with or make way for that image. It required special mixes like Yosi Horikawa or others playing with HRTF effects to get properly out-of-head sounds. Then I finally tried out software crossfeed, whose experience I've already described, such having been a revolution in itself. I was probably still quite skeptical in regard to perception of headphone "resolution" and other claimed "technicalities".

Meze Elite:

The Meze Elite audition was a lesson in how using other test rigs like headphones.com's measurements to produce EQ profiles is absolutely no guarantee that the headphones will actually have the same response on your own head and ears. In this case, the Meze Elite ended up having a 600 Hz elevation that made Rodrigo y Gabriela tracks sound "sweeter" in a wonderful and impressive way that yielded my first experience of being truly astonished. There was then this HRTF-effect track where I for the first time heard the distant panning sound source sound "ridiculously" far away; this only happened once, the differences in perceived distance growing smaller with later comparisons. The only "soundstaging" differences I heard were related to the sense of the slight differences in driver distance and the amount of space around my ears. I also did not hear any advantageous "impact" or "slam" out of the Focal Utopia. I also learned that the DCA Expanse despite being effectively Harman-tuned was at the time less preferable than the result from EQing the Meze Elite. I likewise learned that I preferred large planar presentation and tactility, the Meze Elite sharing a level of openness with the Arya Stealth though the latter seemed more incisive. Regardless, I have no regrets since the Meze Elite Tungsten version is freaking gorgeous and exquisitely comfortable, and I have in the past few months with my in-ear mics uncovered this headphone's absolutely exceptional distortion performance even when EQing up the wide upper midrange dip. I while A/Bing found that I could listen to pure tones in order to EQ the headphones to roughly the same response frequency by frequency, whereby I was able to transfer the sweetness of the Meze Elite's EQ to the Arya Stealth and the clarity of the Arya Stealth's EQ to the Meze Elite. By then, I was convinced that all matters of tonality are founded in frequency response and can be transparently adjusted with good digital EQ. I had also found that EQing down the 6 kHz to 8 kHz region in a slope of relatively equal loudness greatly improved clarity and the cleanness of my sound.

I didn't like the feel of the Angled Alcantara pads and didn't hear any soundstaging or imaging improvements. They do measurably and audibly drive things a bit toward neutral, but I much preferred EQing the sweet lambskin (I think) hybrid pads which could feel wonderfully cool on first wearing and likewise had better bass extension and hence distortion.

Then came my in-ear mics and the HRTF and free-field EQing results I had described. It so happens that the Meze Elite may just be the perfect headphone for binaural head-tracking since the headband is carbon fiber and makes it really easy to slot in the head-tracking unit, and the magnetic field happens to be very well contained within the drivers so as to not interfere with the head-tracker's magnetometer, at least my cellphone app showing that even my two previous dynamic driver headphones exhibit stronger magnetic fields at the headband. In the delay between receiving my in-ear mics and receiving my HRTF SOFA file, I discovered that my EQing by ear (and making the initial profile using GRAS measurements) was extremely inaccurate in the midrange which actually came to be very wavy for my in-ear measurements, my refining this into the "V3 PEQ" which wasn't far off from the smoothness of my measurement of the HE-1 and X9000 for which I was able to say that I preferred the sound, cleanness, and presentation of my EQed Meze Elite. My in-ear speaker measurements also revealed that neutral speakers actually have decently more ear gain than my Harman EQ, whereby some of the "vividness" I had been looking for was to be found in EQing in more ear gain consistent with my HRTF and of course getting used to such, whereby the V3 PEQ now sounds a bit relaxed by comparison. Now, at this moment, I am somewhat surprised that this free-field EQ allowed Zubin Mehta's 1977 recording of Mahler 5 whose second movement I particularly enjoyed to sound more competitively clean and vivid against my reference 1996 Boulez recording on Idagio whence I had previously felt that it didn't sound as "hi-fi" as the best recordings I had encountered.

HE1000se:
I only got this in December as a result of my finally deciding that the X9000 wasn't for me both in terms of comfort and presentation as well as its despite the estat reputation measuring to not have much better distortion performance than my Meze Elite. I was still ready to possibly be impressed, and was technically hoping that the reviews were true or that this different magnet array would feel more open. I was also at the time becoming discontent with the Meze Elite's comfort and wanted to explore pad-rolling the HE1000se partly to optimize its EQed distortion and CSD measurements.

And so it arrived, my as part of a burn-in debunking plan playing a Room EQ Wizard sine sweep as the first signal. Measurements showed very similar frequency responses, but at least better control of the 4 kHz peak and perhaps more consistently good driver matching where my Arya Stealth wasn't well matched in the ear gain region. I was happy to see better distortion performance through the HE1000se, though I later learned that this was only the case for the left drivers, its being switched around for the right drivers, and now lately per https://www.head-fi.org/threads/totl-disappointments.925164/post-17949397 (post #936), at least units from November exhibit clear distortion defects albeit not particularly audible in practical music listening, but still not so nice of a thing to come across for the supposed "best" out of the HE1000 line. I personally don't believe in burn-in or believe its effects to be psychological in nature. Regardless, I so have found the HE1000se to sound very similar to the Arya Stealth, not doing any magically soundstaging or imaging stuff, the Arya sounding just as detailed and textured. Now maybe without crossfeed, there were cases where something through the HE1000se would sound like it were positioned a bit higher, but this may have rather been a mismatch in the vertical positioning of the cups since the two headphones have different headbands. I personally prefer the wider and more cushioned headband of the Arya Stealth. They also both feel equally open and transparent to external noises, and probably similarly tactile for large bass transients. I measured similar impulse and step responses, and also hear similarly incisive transients when listening to isolated impulses, though I am now inclined to believe that such differences in the quality of extreme transients (playing http://pcfarina.eng.unipr.it/Acustica-samples/Dirac.wav really loud) are due to nonlinearities showing up, whereby in practical listening, differences in transient quality decrease. Otherwise, the main things to like are the looks and the lighter clamping force, though I can still feel pressure points when wearing them for long enough. Worn HiFiMan pads, even for just a few months, have a rather nice, soft, and cool initial feel compared to fresh pads.

Given this, the main reason for having the HE1000se is as a comfort and presentation alternative for the same tonality, and for it to feel like a more "worthy" complement to my Meze Elite than the Arya Stealth with its measurable driver matching issues for my unit. I've already devised a free-field EQ which I document in https://www.audiosciencereview.com/...n-susvara-headphone-review.50705/post-1853760 (post #1,137), whereby my HE1000se with these huge pads (though pads are exceedingly easier to swap on the Meze Elite, with the HE1000se, there are a lot more pad options, mind the ability for full customization through NTRAX Mod Design (NMD)) effectively sounds tonally identical to the Meze Elite, but with a whole lot more space around my ears. The problem is that it has a metal headband and a powerful magnetic field that currently makes it impossible to use with magnetometer-based head-trackers, a problem I am currently trying to solve. The custom pads feel excellently snug and comfortable, though even with the lighter clamping force of the HE1000se compared to the Arya Stealth, I can still after enough time feel pressure points or like circulation to my auricles is being constricted (same deal with the Meze Elite after enough time or on certain occasions; some days, the Meze Elite just feels perfect and I don't want to take them off or stop listening because they feel so darned snug). I suppose the HE1000se with NMD pads for the same EQ and very similar in-ear response can still feel a bit "airier" while switching back to the Meze Elite can have such feeling a bit smoother.

tl;dr: Meze Elite for exquisite distortion performance (also a pretty good CSD for a planar magnetic, but not as good as dynamic driver or closed-back CSDs) and sweet, soft, and initially cool-feeling lambskin pads. HE1000se for worse distortion, especially for multitone IMD, but with huge custom pads and maybe more incisive transients at the extremes. You can see https://www.head-fi.org/threads/rec...r-speaker-virtualization.890719/post-17955514 (post #1,801) for my latest multitone distortion comparison.

I find this particularly interesting I suppose because - as mentioned - I've owned these phones myself and spent some time trying to decide what to think about them. Like you, it seems, I've also been interested in how they perform 'spatially' (for want of a better term).

I should say at the outset that, while I haven't kept these phones, I found all of them good in the sense that I would have been quite happy to keep them had I not encountered others that I preferred. I think my experience was quite similar in finding the Arya and HEKSE have much in common - perhaps with the HEKSE exhibiting a less polite top-end but more audibly resonant behaviour (I know this is controversial - but I feel Hifiman phones sound a little like they measure). The Meze seems to me more controlled than either through higher frequencies in particular but too warm for me (at least without EQ). I do feel the Hifiman phones naturally project a more spacious image - which is important to me.

Having owned many phones - including the Stax SR-009S and other electrostatics - I've learned I have a preference for the presentation of dynamic drivers - with or without crossfeed or EQ. They seem to me more refined in terms of dynamic 'shading' - which lends music much of its life I think. My preferred phones are also tuned to a diffuse-field reference - something which seems virtually absent in the world of planars at least. I now have only the HD800S and the ADX5000, both of which seem to me to work well with crossfeed. I think the HD800S seems particularly capable with recordings demanding a spacious image - although the AHX is no slouch. The HD800S is also the most 'transparent' headphone (transducer?) I've encountered - very transparent to sources I think - a real chameleon with very little personality of its own.

Focal phones have been a disappointment to me - and not greatly improved by crossfeed I think. So not all dynamic phones work in my terms - I would probably prefer the spacious feel of the HEKSE to the rather cramped world view of the Utopia. I would certainly prefer the SR-009S.

 

[jamesjames]

For anyone following this thread who may be interested, there is indeed a separate thread for Crosszone headphones - which has some interesting stuff.