only good for some songs but its a rarity
Anyone into crossfeed?
jamesjames
500+ Head-Fier
I recently found time to compare crossfeed with standard stereo and binaural versions of the same recording. I found it very interesting and have made a few notes on my impressions.
The recording in question, some Haydn violin concertos played by Lisa Jacobs and The String Soliosts, was made by Cobra records in 2017. I downloaded it from NativeDSD. The notes on the site indicate it was originally recorded at DSD64. Higher rate versions are also available which NativeDSD claims sound better as they move noise into higher and less audible frequency ranges. I agree they sound better and downloaded DSD128 stereo and binaural versions. The stereo version of the nine-track album (folder of files) was about 5.1GB in size, the binaural version about 5.3GB. I have no explanation for the difference.
My headphone system comprises a Marantz SA10 digital player on balanced cables to a Moon 430HA amp, in turn on balanced cable to Sennheiser HD800S phones. The downloaded files were copied to a USB drive which was plugged directly into the Marantz player. The Marantz upconverts all inputs to DSD256. The Moon amp has a switchable crossfeed circuit. I played tracks one-at-a-time. In each case I played the stereo version w/o crossfeed, the stereo version with crossfeed and the binaural version. I adjusted gain by ear, to accommodate differences in level. Finally, I listened to the stereo version through near-field monitors – laptop on USB to Sennheiser HDV820 DAC/preamp, in turn on balanced cables to Neumann KH120 monitors.
The recording was made at the Cunera Church at Rhenen in the Netherlands. The first point I’ve noted is that it’s quite ‘wet’; that is, reasonably reverberant. It isn’t a ‘dry’, close-up recording. It isn’t a ‘swimmy’ recording – where the reverberation swamps detail – but there’s a palpable sense of a string orchestra playing in an old church.
The second point I’ve noted is that the three presentations sound quite different. The stereo version to my ear is slightly 'confused' - moving between spacious and unnaturally ‘flat’ and close in places - solo instruments can seem overly isolated from the ensemble - but then dynamic range can seem unnaturally compressed. Stereo affects are to my ear accentuated, without natural separation of instruments in complex passages through the central part of the image (which generally seems a bit woolly and 'thick'). Crossfeed immediately adds a better sense of integration and more natural separation of instruments. The image of the performance is to my ear significantly more coherent. I have a better sense of the performance space, which is apparently more nuanced - with attack and decay of instrumental contributions more naturally represented. Solo passages are less strident and apparently better integrated in a larger whole. Lower frequencies are better delineated. Interestingly, reverberation and decay seem better managed on the whole, more realistic. The sense of real instruments placed at an appropriate distance, weaving in and out of the ensemble, is more convincing.
Moving to binaural brings even greater improvement. A greater sense of integration here exposes limitations of the crossfeed presentation which I wouldn’t otherwise have noticed; and to my ear the standard headphone presentation now seems even more congested and unnatural by comparison. I find it difficult to describe what’s at work here, but I have no doubt the binaural version presents an integrated whole that’s more engaging than that achieved with crossfeed. The binaural version somehow combines some of the richness and weight of the standard headphone presentation with the more refined separation and space of crossfeed presentation. I’m really quite surprised by how much better this seems to me.
Listening to the stereo version on near-field monitors confirms for me the binaural headphone presentation is closest to the loudspeaker presentation. The stereo version with crossfeed seems to me a close second, and the stereo version without crossfeed a distant third. It’s interesting to consider the monitor presentation – which brings greater weight to the music than any of the headphone presentations – although binaural is close. Which sounds ‘better’ or more natural is a difficult question. I find the binaural presentation more pleasing than the monitors – but it’s possibly gilding the lily to some extent. It might be introducing a degree of refinement that’s not quite there with monitors (and perhaps wasn’t there at the performance). I’ll never know, but I’m convinced the binaural presentation is tremendously good, better even than stereo with crossfeed, and on par with the loudspeaker presentation.
EDIT
I've revised this slightly to describe my impressions a little more clearly (I hope).
The recording in question, some Haydn violin concertos played by Lisa Jacobs and The String Soliosts, was made by Cobra records in 2017. I downloaded it from NativeDSD. The notes on the site indicate it was originally recorded at DSD64. Higher rate versions are also available which NativeDSD claims sound better as they move noise into higher and less audible frequency ranges. I agree they sound better and downloaded DSD128 stereo and binaural versions. The stereo version of the nine-track album (folder of files) was about 5.1GB in size, the binaural version about 5.3GB. I have no explanation for the difference.
My headphone system comprises a Marantz SA10 digital player on balanced cables to a Moon 430HA amp, in turn on balanced cable to Sennheiser HD800S phones. The downloaded files were copied to a USB drive which was plugged directly into the Marantz player. The Marantz upconverts all inputs to DSD256. The Moon amp has a switchable crossfeed circuit. I played tracks one-at-a-time. In each case I played the stereo version w/o crossfeed, the stereo version with crossfeed and the binaural version. I adjusted gain by ear, to accommodate differences in level. Finally, I listened to the stereo version through near-field monitors – laptop on USB to Sennheiser HDV820 DAC/preamp, in turn on balanced cables to Neumann KH120 monitors.
The recording was made at the Cunera Church at Rhenen in the Netherlands. The first point I’ve noted is that it’s quite ‘wet’; that is, reasonably reverberant. It isn’t a ‘dry’, close-up recording. It isn’t a ‘swimmy’ recording – where the reverberation swamps detail – but there’s a palpable sense of a string orchestra playing in an old church.
The second point I’ve noted is that the three presentations sound quite different. The stereo version to my ear is slightly 'confused' - moving between spacious and unnaturally ‘flat’ and close in places - solo instruments can seem overly isolated from the ensemble - but then dynamic range can seem unnaturally compressed. Stereo affects are to my ear accentuated, without natural separation of instruments in complex passages through the central part of the image (which generally seems a bit woolly and 'thick'). Crossfeed immediately adds a better sense of integration and more natural separation of instruments. The image of the performance is to my ear significantly more coherent. I have a better sense of the performance space, which is apparently more nuanced - with attack and decay of instrumental contributions more naturally represented. Solo passages are less strident and apparently better integrated in a larger whole. Lower frequencies are better delineated. Interestingly, reverberation and decay seem better managed on the whole, more realistic. The sense of real instruments placed at an appropriate distance, weaving in and out of the ensemble, is more convincing.
Moving to binaural brings even greater improvement. A greater sense of integration here exposes limitations of the crossfeed presentation which I wouldn’t otherwise have noticed; and to my ear the standard headphone presentation now seems even more congested and unnatural by comparison. I find it difficult to describe what’s at work here, but I have no doubt the binaural version presents an integrated whole that’s more engaging than that achieved with crossfeed. The binaural version somehow combines some of the richness and weight of the standard headphone presentation with the more refined separation and space of crossfeed presentation. I’m really quite surprised by how much better this seems to me.
Listening to the stereo version on near-field monitors confirms for me the binaural headphone presentation is closest to the loudspeaker presentation. The stereo version with crossfeed seems to me a close second, and the stereo version without crossfeed a distant third. It’s interesting to consider the monitor presentation – which brings greater weight to the music than any of the headphone presentations – although binaural is close. Which sounds ‘better’ or more natural is a difficult question. I find the binaural presentation more pleasing than the monitors – but it’s possibly gilding the lily to some extent. It might be introducing a degree of refinement that’s not quite there with monitors (and perhaps wasn’t there at the performance). I’ll never know, but I’m convinced the binaural presentation is tremendously good, better even than stereo with crossfeed, and on par with the loudspeaker presentation.
EDIT
I've revised this slightly to describe my impressions a little more clearly (I hope).
Last edited:
castleofargh
Sound Science Forum Moderator
- Joined
- Jul 2, 2011
- Posts
- 10,429
- Likes
- 6,041
And you have to consider that binaural recordings aren't using your HRTF, usually it's a dummy head, and sometimes just 2 mics in parallel. So even between binaural recordings, some will work better than others for you. And of course, the more degrees of customization toward your own HRTF you get, the more convincing a sound can become.
MrHaelscheir
100+ Head-Fier
There is likewise the question of whether the binaural recording already/still contains the ear gain from the dummy head or if they already EQed that out, else it may actually be brighter than expected. Then you have the overlaying of your HD 800 S's frequency response on top of the binaural recording's HRTF. Another thing about the "refined" presentation could be the lack of room reflections compared to your studio monitors affording greater clarity.
jamesjames
500+ Head-Fier
Yes, I agree. There's also the difference between DACs. The headphone chain includes the Marantz upconverted processing whereas the monitor chain uses Sennheiser processing. The Marantz is clearly distinguishable from the Sennheiser when using headphones - I've used the Sennheiser as my headphone amp from time to time and compared the integrated DAC with the Marantz as an 'external' DAC.
Whatever ... I'm very impressed with binaural recording and will be acquiring more from NativeDSD!
Last edited:
gregorio
Headphoneus Supremus
- Joined
- Feb 14, 2008
- Posts
- 6,773
- Likes
- 4,046
A few points here: Firstly, the original recordings were DSD64, so the higher rate versions are effectively just DSD64 upsampled. Secondly (and incidentally), it’s not really “Native” DSD as it’s converted to PCM for editing/processing and then re-converted back into DSD. Thirdly, upsampling does not “move noise”, all the noise of the recording itself and the dither noise of from the original DSD64 conversion process is completely unaffected by upsampling. The only noise affected is the additional dither noise applied in the re-conversion (from PCM back into DSD) process, DSD64 applies this dither noise in the range around 25kHz - 1.4MHz, while DSD128 applies it from around 25kHz - 5.6MHz but as human hearing only extends to 20kHz (or ~16kHz in adults) this dither noise is equally inaudible with both rates. The take-away is: Don’t take audiophile marketing seriously or you could be biased into falsely perceiving that “they sound better” when in fact there’s no audible difference!
Adjusting gain by ear does not “accommodate differences in level”. This is a very common error which typically results in the perception of differences in quality, where in fact there isn’t any.
Crossfeed obviously feeds some of the signal from the left channel into the right channel (and vice versa) and therefore introduces less separation, less managed/realistic reverberation and a less coherent stereo image. However, for some people it may fool their perception sufficiently to result in a perception of the opposite (more separation, etc.).
It shouldn’t really be a surprise, binaural recordings take into account most/all the factors covered by HRTFs rather than only crossfeed and therefore should be sufficient to fool the perception of most people, and to a far greater extent. Although as @castleofargh mentioned, that will largely depend on how well the dummy head HRTF matches that of the individual consumer.
Monitor presentation of course includes/adds room acoustics, which is drastically going to affect the signal throughout the audible frequency range, especially in the lower freqs and also the sound waves will be impacting your whole body rather than just your ears. This will typically be perceived as having more “weight”, even with content such as the Haydn Violin Concertos that don’t have much bass. You *might* find this somewhat different with other content, say popular music genres or a more recent symphony, as your near-field montitors don’t have a good low bass response.
G
Last edited:
jamesjames
500+ Head-Fier
I'm conscious this thread contains a lot of detail that might not be interesting to all. So I'd like to mention again that my posts here have been intended to sketch out broadly why crossfeed appeals to me and how readers might try it for themselves if interested. I've mentioned earlier that, strangely, it's a topic that seems to prompt strong reactions - to the effect that crossfeed is a waste of time. I'm not sure why. Part of my intention here is simply to mention that it makes a big difference to me and to many other enthusiastic listeners. If it seems interesting, give it a go - don't be dissuaded by the nay-sayers.
castleofargh
Sound Science Forum Moderator
- Joined
- Jul 2, 2011
- Posts
- 10,429
- Likes
- 6,041
Crossfeed stands in the middle of the road. It doesn't exactly fix the stereo of headphones and IEMs(one channel only reaching one ear, which is objectively unnatural to a human). It's a fairly generic term to describe quite a range of products and settings, so those who tried maybe one or 2 are more likely to report that it sounds "meh". And of course, to those who wish to get the speaker experience on headphones, crossfeed just it. Even speaker simulator aren't enough, but can come impressively close with the right HRTF reference, head tracking, and maybe even a body shaker. Crossfeed is a few steps behind that.
There is also the simple fact that as headphones came to the market, records have taken them into account more and more and have overwhelmingly reduced the use of long one side panned instruments or voices, because they noticed how uncomfortable that could feel on headphones(30° on speakers becomes 90° on headphones). Crossfeed applied over a singer full left all the track, it might just become the reason to still listen to that song on headphones, but modern albums often feel like they now have too narrow and often just plain weird presentation with crossfeed turned on.
Personally, I already said I tried anything that I would find, and used crossfeed for years, sometimes to include old tracks into my IEM playlists, and mostly just to reduce the fatigue I could feel from listening to unprocessed IEM stereo. But IMO, Stuff using HRTF models can do better than crossfeed(some crossfeed actually do even if it's only the EQ of one direction, and just decide to still call themselves croossfeed because of that limited use). Personal HRTF, something created by recording sine sweeps from speakers at your own ears, or by taking video or pictures of your ears(ideally the all head), to use a 3D model as reference and pick the closest HRTF from a catalog, are almost sure to feel more accurate and realistic than generic solutions or basic crossfeed.
So, crossfeed in stuck in the middle, as a solution of the past for a problem of the past. It's often too much change for those who got used to the headphone presentation(maybe they don't know some artists beside how they sound on headphones or IEM, so now it is their subjective reference of proper sound), and not enough/incorrect change for those who seek speaker sound.
I'm not saying this to spit on crossfeed, I'm saying it so that people who actually seek speaker simulation, don't miss out by stopping at crossfeed. Beyond that, if someone likes a certain crossfeed, that's cool. There is no point or need to try and justify personal preferences. Taste is taste.
There is also the simple fact that as headphones came to the market, records have taken them into account more and more and have overwhelmingly reduced the use of long one side panned instruments or voices, because they noticed how uncomfortable that could feel on headphones(30° on speakers becomes 90° on headphones). Crossfeed applied over a singer full left all the track, it might just become the reason to still listen to that song on headphones, but modern albums often feel like they now have too narrow and often just plain weird presentation with crossfeed turned on.
Personally, I already said I tried anything that I would find, and used crossfeed for years, sometimes to include old tracks into my IEM playlists, and mostly just to reduce the fatigue I could feel from listening to unprocessed IEM stereo. But IMO, Stuff using HRTF models can do better than crossfeed(some crossfeed actually do even if it's only the EQ of one direction, and just decide to still call themselves croossfeed because of that limited use). Personal HRTF, something created by recording sine sweeps from speakers at your own ears, or by taking video or pictures of your ears(ideally the all head), to use a 3D model as reference and pick the closest HRTF from a catalog, are almost sure to feel more accurate and realistic than generic solutions or basic crossfeed.
So, crossfeed in stuck in the middle, as a solution of the past for a problem of the past. It's often too much change for those who got used to the headphone presentation(maybe they don't know some artists beside how they sound on headphones or IEM, so now it is their subjective reference of proper sound), and not enough/incorrect change for those who seek speaker sound.
I'm not saying this to spit on crossfeed, I'm saying it so that people who actually seek speaker simulation, don't miss out by stopping at crossfeed. Beyond that, if someone likes a certain crossfeed, that's cool. There is no point or need to try and justify personal preferences. Taste is taste.
gregorio
Headphoneus Supremus
- Joined
- Feb 14, 2008
- Posts
- 6,773
- Likes
- 4,046
I’m not sure why you’re “not sure why”, as it’s been explained several times by several different posters. Actually, crossfeed is worse than just “a waste of time” because not only does it waste time/effort but also sounds worse to some/many.
Crossfeed has been around for many decades but never became common/mainstream because it did not make a positive difference to many/the majority of enthusiastic listeners. Clearly crossfeed makes a positive difference to some listeners though, otherwise it wouldn’t still exist 60 years or so after it was invented. However, it most likely will die out, maybe entirely in time, as more sophisticated and easy to use HRTF implementations come to market and more enthusiastic listeners discover HRTF solutions to be preferable, as indeed you yourself have.
G
My personal view is that crossfeed is a nice thing to have to 'fix' hard-panned tracks that were clearly mixed solely for speakers so that they don't sound jarring on headphones, but besides that it's usually not an actual benefit for most music, especially now that much of modern music is being mixed with headphones as more of a consideration. I find that it often forces stuff in front of you even if it shouldn't be, and tends to 'smear' centered content in many implementations
Obviously when it comes to spatial enhancement, the endgame is something either such as an atmos like setup with many speakers, though that's usually expensive and difficult to set up and you've got the considerations of room acoustics.
OR
Something that properly takes HRTF and localization cues into account, similar to how a lot of current VR audio solutions work. Though this requires head tracking.
There are some things you can do though with regular stereo tracks that provide a much better experience than crossfeed. I'm actually working on a product right now which I'm very excited to get in people's hands that has what I think is a much better 'spatial enhancement' than anything I've tried on gear previously (though ofc I'm biased there). Needs more compute power than what you can do in a typical DAC though.
Personally I found some of what companies like Valve have done to be really interesting, Valve open-sourced their steam-audio solution and there's some pretty fascinating stuff in there.
Some that is specifically related to games, such as physics-based sound propagation within the virtual environment, but some cool stuff with HRTF & Ambisonics that wouldn't be specific to a game necessarily. As well as there being some cool things you can do by calculating/extrapolating direction and/or location from a stereo track and modifying with approaches such as this....
Obviously when it comes to spatial enhancement, the endgame is something either such as an atmos like setup with many speakers, though that's usually expensive and difficult to set up and you've got the considerations of room acoustics.
OR
Something that properly takes HRTF and localization cues into account, similar to how a lot of current VR audio solutions work. Though this requires head tracking.
There are some things you can do though with regular stereo tracks that provide a much better experience than crossfeed. I'm actually working on a product right now which I'm very excited to get in people's hands that has what I think is a much better 'spatial enhancement' than anything I've tried on gear previously (though ofc I'm biased there). Needs more compute power than what you can do in a typical DAC though.
Personally I found some of what companies like Valve have done to be really interesting, Valve open-sourced their steam-audio solution and there's some pretty fascinating stuff in there.
Some that is specifically related to games, such as physics-based sound propagation within the virtual environment, but some cool stuff with HRTF & Ambisonics that wouldn't be specific to a game necessarily. As well as there being some cool things you can do by calculating/extrapolating direction and/or location from a stereo track and modifying with approaches such as this....
Last edited:
drummerman1
100+ Head-Fier
I find most music on HP's without applied Xfeed unnatural except perhaps electronica/rap.
I don't like instruments/voices to eminate from behind me, it's not how I hear music live or through my Hifi. That's ok if I'm in a cinema but I don't usually sit in the middle of a band or symphonie orchestra whilst they play.
Whilst Crossfeed does not eliminate this completely it at least helps to solidify a central image.
Furthermore, it can help adding meat/punch to what otherwise can be an overly spread out, thin stage.
As I said in another thread, the above could be a contributing reason many audiophiles do not like to listen to headphones.
So yes, I apply some Crossfeed to all my (humble) phones with most genres. Easy with Roon. I guess it's still a very rudimentary way of going about it but until someone creates more sophisticated DSP plugins or headphones with more 2channel speaker like presentation at affordable prices (and current weird looking, usually expensive Frankenstein implementations except ...), it will have to do.
I don't like instruments/voices to eminate from behind me, it's not how I hear music live or through my Hifi. That's ok if I'm in a cinema but I don't usually sit in the middle of a band or symphonie orchestra whilst they play.
Whilst Crossfeed does not eliminate this completely it at least helps to solidify a central image.
Furthermore, it can help adding meat/punch to what otherwise can be an overly spread out, thin stage.
As I said in another thread, the above could be a contributing reason many audiophiles do not like to listen to headphones.
So yes, I apply some Crossfeed to all my (humble) phones with most genres. Easy with Roon. I guess it's still a very rudimentary way of going about it but until someone creates more sophisticated DSP plugins or headphones with more 2channel speaker like presentation at affordable prices (and current weird looking, usually expensive Frankenstein implementations except ...), it will have to do.
Last edited:
MrHaelscheir
100+ Head-Fier
That would be the https://www.earfish.eu/ implementation I had described earlier in this thread (page 3), and if you don't need head-tracking and can supply your own in-ear mics, https://github.com/jaakkopasanen/Impulcifer.
Last edited:
TYATYA
1000+ Head-Fier
UAPP may be meet your need
Attachments
drummerman1
100+ Head-Fier
Thank you. That all reads way to much involved for me.
Users who are viewing this thread
