• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Crossfeed for headphones

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,490
Likes
4,117
Location
Pacific Northwest
Crossfeed implementations are highly variable in quality. The best I have heard is Jan Meier's, and DSP implementations of it in tools like Foobar and Pulse Effects. It is subtle, adjustable, and very natural.

The article I linked explains how it works and has a sample circuit diagram. If you want to make your own, that could be a good place to start.

Personally, I don't use crossfeed on headphones except for material that has extreme L-R stereo separation. Like hard-panned in the studio as was often done in vintage jazz recordings, or some of Mapleshade's recordings which use PZMs mounted on a triangle which gives very wide stereo.
 

kemmler3D

Major Contributor
Forum Donor
Joined
Aug 25, 2022
Messages
3,358
Likes
6,885
Location
San Francisco
Crossfeed implementations are highly variable in quality. The best I have heard is Jan Meier's, and DSP implementations of it in tools like Foobar and Pulse Effects. It is subtle, adjustable, and very natural.

The article I linked explains how it works and has a sample circuit diagram. If you want to make your own, that could be a good place to start.

Personally, I don't use crossfeed on headphones except for material that has extreme L-R stereo separation. Like hard-panned in the studio as was often done in vintage jazz recordings, or some of Mapleshade's recordings which use PZMs mounted on a triangle which gives very wide stereo.
I bet I could throw together a decent crossfeed 'algorithm' with M/S plugins and a little EQ and some workmanlike mixer routing. However, IMO these approaches are just approximations of what actually happens in nearfield listening due to HRTFs and possibly some room reflections.

Impulses can act as measurements of what algorithmic crossfeed only approximates.

My thought was that this might be nice for listening, and it would also be important if I ever intend to do mixing work with headphones, which I might.

Heck, maybe I should just rent some in-ear mics and an hour of time in a nice studio and record the sweeps myself. :D
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,301
Likes
2,774
Location
Salvador-Bahia-Brasil
Resurrecting this thread because I just got some LCD-XCs, which as far as closed cans go, I think are endgame for me. With EQ it is hard to imagine much improvement.

So now I am wondering how far I can take it. I have never experimented with crossfeed, but I am starting to come around to the idea. However, I think the normal M/S or mixer-based solutions are a little crude. After reading some rave reviews of the Slate studio-simulating software, I started to think that convolution-based "crossfeed" would be the way to go, or at least an interesting way to go.

Does anyone know of sources of good impulses that would be suitable for a headphone crossfeed setup?

Has anyone rolled their own convolution-only crossfeed setup before?

I think the ideal is a binaural impulse recorded with someone having an HRTF like mine, in a world-class recording studio. That would be nice, one and done.

Failing that, I think it might be possible to do something reasonable by using some of the IRCAM impulses, although I am not sure if the sound source was necessarily placed where you'd place monitors...?

I found the IRCAM HRTF files but I will need to set up FTP to actually get them (they're ancient) and I am also not sure if that's the necessarily the right approach. Anyone have those handy and available to upload?

try these IRs: https://www.audiosciencereview.com/...sing-this-crossfeed-setup.43152/#post-1527755
start with WDR control room 1, as it should be the cleanest
 

IXOYE

Member
Joined
Jan 22, 2023
Messages
63
Likes
76
I use the Jan Meier option on my RME--ADI 2 with my headphones, and I can now listen for hours without fatigue, this option has taken headphone listening to a whole new level and I hope more manufacturers will add this option. in the future.
 

kemmler3D

Major Contributor
Forum Donor
Joined
Aug 25, 2022
Messages
3,358
Likes
6,885
Location
San Francisco
Just wanted to chime in having tried the Shanon Pearce convolution files last night. https://github.com/ShanonPearce/ASH-Listening-Set . Big thanks to @dasdoing for pointing me in the right direction, good example of why this is a cool forum. Ask and ye shall receive knowledge. :D

I am not sure if I was doing it right (just use the true stereo file, set and forget?) but the effect was immediate and very convincing, actually IMO too strong. For me the indirect sound felt like it was mixed in too high, hard to describe. It sounded quite real, definitely fools my ear to an extent, but very strongly like "speakers in a room" and not quite like I was listening to good speakers close to me in real life. It also seemed to chew up the treble in a way that I wasn't totally comfortable with.

I ended up mixing the "wet" convolved signal in at about -9.5dB and it subjectively felt more like a real-life nearfield listening situation at that point. I ended up listening to all sorts of music I don't even like that much because the sound was so cool. This track in particular was a bit mind-blowing:


I definitely recommend at least trying these out for anyone who has been curious about crossfeed and uses EQAPO. It's pretty dead simple to set up.

Also, it bears mentioning that it has a general purpose circumaural correction EQ already applied, so your normal headphone EQ needs to be deactivated if you go 100% wet.

Is the proper way to use these to set up an actual crossfeed with L / R channels separately, or is the stereo file actually doing the job? I feel like this should be obvious but maybe I'm not getting enough sleep these days...
 

kemmler3D

Major Contributor
Forum Donor
Joined
Aug 25, 2022
Messages
3,358
Likes
6,885
Location
San Francisco
Update: Eventually I just ended up using the VST that OP mentioned in the first post!

The Shannon-Pierce files are really interesting and at 100% wet sound pretty real. I kept those going for a good while, but they overwhelm the tonality of the sound anywhere above -12dB. They introduce a lot of FR variation that I don't really want in my headphone listening. The effect on bass is particularly interesting, but ultimately it either sounded unnoticeable or like a special effect to me.

Using a simple meier crossfeed, you get the nice perception of a smooth, wider stereo field, without mucking up FR much. I doubted crossfeed for many years, but having tried it, it just seems obvious in retrospect. Hoping for a realistic stereo image without it now seems like watching a 3D movie with one eye closed.

Combined with EQ, this brings headphone listening almost on par with listening to good speakers for me.

To all those with decent cans - if you're not using crossfeed, why not?
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,490
Likes
4,117
Location
Pacific Northwest
...
Using a simple meier crossfeed, you get the nice perception of a smooth, wider stereo field, without mucking up FR much. I doubted crossfeed for many years, but having tried it, it just seems obvious in retrospect. Hoping for a realistic stereo image without it now seems like watching a 3D movie with one eye closed.

Combined with EQ, this brings headphone listening almost on par with listening to good speakers for me.

To all those with decent cans - if you're not using crossfeed, why not?
I agree, if you can get a good one.
I think the key problem is getting a decent implementation of crossfeed. Meier's is one of the best. Many others are heavy handed and ineffective, sound artificial and squash the music.
 

kemmler3D

Major Contributor
Forum Donor
Joined
Aug 25, 2022
Messages
3,358
Likes
6,885
Location
San Francisco
I agree, if you can get a good one.
I think the key problem is getting a decent implementation of crossfeed. Meier's is one of the best. Many others are heavy handed and ineffective, sound artificial and squash the music.
This VST (from the OP) has a free implementation of Meier-style xfeed. (it's one of the presets.) https://resonic.at/tools/bs2br

Ironically, I actually landed on this one independently after a bout of googling for one myself.

Anyone using EQAPO or a similar DSP setup that accepts VSTs - check it out!

To me, subjectively it sounds rather neutral except for the stereo effects.
 

kemmler3D

Major Contributor
Forum Donor
Joined
Aug 25, 2022
Messages
3,358
Likes
6,885
Location
San Francisco
try "dearVR Monitor" for room sim
In the end I didn't really like the Shannon-Pierce IRs because they introduced too much room influence on FR even at a small mix %. I am also not really interested in adding fake reflections for playback...
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,301
Likes
2,774
Location
Salvador-Bahia-Brasil
In the end I didn't really like the Shannon-Pierce IRs because they introduced too much room influence on FR even at a small mix %. I am also not really interested in adding fake reflections for playback...

I totally understand, but this stuff is next level IMO. But sure not for someone looking for a simple crossfeed. For crossfeed, CanOpener seems to be standard in pro circles.
 

MRC01

Major Contributor
Joined
Feb 5, 2019
Messages
3,490
Likes
4,117
Location
Pacific Northwest
Yeah, crossfeed and reverb are two entirely different things. Some might like one and not the other.
BTW, Meier also devised his own reverb in DSP, and some other interesting effects, in a new device and he has audio samples on its product page, for those who are curious.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,301
Likes
2,774
Location
Salvador-Bahia-Brasil
today I tried HeSuVi and liked the Dolby Headphone DH-1 studio profile. It is more colored than using DearVR, but depending on the taste that can be a good thing as it sounds more like a real room.
 

OK1

Member
Joined
Apr 1, 2019
Messages
71
Likes
22
Update – I made some upgrades. Using 3 of the configurations from the plugin (there is a fourth one that’s pretty much identical to another one so I ignored it) and my custom configuration, I made four presets of crossfeed with varying level of strength. The best part is that I added shelf filters to each configuration to flatten out the frequency response and volume changes, so now when I switch between the presets, the only thing that changes is the perceived stereo width. Along with an off state, I get 5 levels of crossfeed to choose from.

I also discovered that different headphones appear to do better with different levels of crossfeed. My Focal Clear are best with the level 1 (just a touch), while the Audeze LCD2C are suddenly improved by applying level 2 or 3 (depending on the material).
Hi Fluffy,

It has taken me a long time, I think about 3 years to date, reviewing various approaches to addressing the need for crossfeed in headphones. For convenience, quiet in the home, and the power available on computers, as well as the ecosystem of plugins, I have arrived at a similar conclusion to you.

BS2R the BS2B version implemented by Liqube (albeit, one can achieve the same results with the original BS2B), has also become my go-to crossfeed solution. It took so long because on many occasions, I abandoned the search. What triggered this was an interest in binaural representations of stereo, and trying out a fair number of these headphone speaker simulation products, and not being satisfied with them. Eventually I arrived at a conclusion that for me, what was more important was the crossfeed. The advantage of BS2R over BS2B plugin is BS2R is 64 bit dll, which I think is where most audio nerds have transitioned to.

Similar to you, using the Resonic preset, which I found to be the best for me, of the presets, I've created two even more aggressive presets. For casual listening, I'll use the Resonic preset, but for critical listening, I'll use one of mine.

I once attempted to also use filters to compensate for the frequency response changes, in a similar manner to what you have done, but that was not conclusive, cos I was still figuring things out. I'll give this another go. Just to see if it improves anything.

Please note there is an anomaly in BS2R (which probably was inherited from BS2B code), where it behaves like a brickwall limiter, which will compromise the audio by adding some distortion. I have described this here, further to my discovering it a few days ago, and the suggested workaround is described in this other thread.


Specific workaround described here :

 

OK1

Member
Joined
Apr 1, 2019
Messages
71
Likes
22
I totally understand, but this stuff is next level IMO. But sure not for someone looking for a simple crossfeed. For crossfeed, CanOpener seems to be standard in pro circles.
I have also tried out CanOpener and am still reviewing it, but I find it useful only as an alternative listening tool. I've been exhaustively comparing probably as many as 20+ headphone listening solutions.

Attached is a partial image of the current finalists. I'm able to route and switch instantaneously from one option to another, and also setup multiple configs of the same solution, to compare them.

I've found the Resonic preset in BS2R to be just about the right compromise between detail, frequency response, listening engagement and stereo width.

Unfortunately for critical listening, while CanOpener is an ok commercial alternative, and very well knows due to advertising, BS2R and the presets - are more accurate, allowing me to pinpoint exactly where every instrument is. CanOpener smears the clarity, in my opinion, and this is after about two weeks comparing it to the other top performers on my shortlist. Particularly the clarity of the image of anything in the center is so much more pin-sharp in BS2R. I think CanOpener adds some kind of reverberation, which unfortunately smears the end result.

Especially as the most important aspects such as vocals are usually centered, its important that that information is preserved. CanOpener unfortunately for me, does not do this very well.

At the end of the day we are talking about psychoacoustics, i.e. our biological system, and the theoretic explanations, are still subject to improvement. Hitherto Resonic preset in BS2R, is one of the best, if not the best. It's what I use all the time, for both casual as well as critical listening.

What I am not sure of is : Are there dependencies on the specific headphone. I'm using a cheap but as I have discovered and excellent pair of Sony earbuds. Could be CanOpener is developed with those using on ear or over ear headphones, which have a different distance from driver to the ears. Maybe.

Considering how many of us use headphones for listening today, I'm surprised that the only major player who has invested in this is Apple - with their spatial audio. Unfortunately the only Apple device I own is their Mains to USB-C charger, which I use for my Samsung phone, which was not supplied with a charger, so I cannot compare the options I have been reviewing with Apple Spatial.

1702123509444.png
 

OK1

Member
Joined
Apr 1, 2019
Messages
71
Likes
22
Update – I made some upgrades. Using 3 of the configurations from the plugin (there is a fourth one that’s pretty much identical to another one so I ignored it) and my custom configuration, I made four presets of crossfeed with varying level of strength. The best part is that I added shelf filters to each configuration to flatten out the frequency response and volume changes, so now when I switch between the presets, the only thing that changes is the perceived stereo width. Along with an off state, I get 5 levels of crossfeed to choose from.

I also discovered that different headphones appear to do better with different levels of crossfeed. My Focal Clear are best with the level 1 (just a touch), while the Audeze LCD2C are suddenly improved by applying level 2 or 3 (depending on the material).
I did try out the option to compensate for the change in frequency, using filters - i.e EQ. I did not like the end result., sounded fuzzy.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,301
Likes
2,774
Location
Salvador-Bahia-Brasil
I have also tried out CanOpener and am still reviewing it, but I find it useful only as an alternative listening tool. I've been exhaustively comparing probably as many as 20+ headphone listening solutions.

Attached is a partial image of the current finalists. I'm able to route and switch instantaneously from one option to another, and also setup multiple configs of the same solution, to compare them.

I've found the Resonic preset in BS2R to be just about the right compromise between detail, frequency response, listening engagement and stereo width.

Unfortunately for critical listening, while CanOpener is an ok commercial alternative, and very well knows due to advertising, BS2R and the presets - are more accurate, allowing me to pinpoint exactly where every instrument is. CanOpener smears the clarity, in my opinion, and this is after about two weeks comparing it to the other top performers on my shortlist. Particularly the clarity of the image of anything in the center is so much more pin-sharp in BS2R. I think CanOpener adds some kind of reverberation, which unfortunately smears the end result.

Especially as the most important aspects such as vocals are usually centered, its important that that information is preserved. CanOpener unfortunately for me, does not do this very well.

At the end of the day we are talking about psychoacoustics, i.e. our biological system, and the theoretic explanations, are still subject to improvement. Hitherto Resonic preset in BS2R, is one of the best, if not the best. It's what I use all the time, for both casual as well as critical listening.

What I am not sure of is : Are there dependencies on the specific headphone. I'm using a cheap but as I have discovered and excellent pair of Sony earbuds. Could be CanOpener is developed with those using on ear or over ear headphones, which have a different distance from driver to the ears. Maybe.

Considering how many of us use headphones for listening today, I'm surprised that the only major player who has invested in this is Apple - with their spatial audio. Unfortunately the only Apple device I own is their Mains to USB-C charger, which I use for my Samsung phone, which was not supplied with a charger, so I cannot compare the options I have been reviewing with Apple Spatial.

View attachment 332770

nice test. I can't really talk about pure cross-talk since that never was enough for me. I just mentioned CanOpener since I see it recommended between audio engineers all the time
 

OK1

Member
Joined
Apr 1, 2019
Messages
71
Likes
22
nice test. I can't really talk about pure cross-talk since that never was enough for me. I just mentioned CanOpener since I see it recommended between audio engineers all the time
Once upon a time the audio engineers were actually engineers, and would be able to take out a multimeter and perform tests on consoles, or preamps, etc etc, - who actually understood what was going on in their gear.

There is this lovely recent video (link below) where Amir discusses the highpoints of his presentation to the AES, E in this acronym being Engineering, a body of real audio engineers. One of the conclusions I got from this video, was the positive impact of quantitative scientific evaluation, on the quality of products delivered to consumers (including professional consumers such as audio engineers). But I must add here, on the professional side of things, companies like EchoAudio and Emu, published very detailed specs of their gear, which could be relied on, and have been doing so for many years. I have devices from Echo Audio and Emu, with spec that are still far more than good enough for music production today, unfortunately these businesses either moved on to other things, for whatever reason.

I find a similar major issue with software plugins such as CanOpener, or any other. No one checks using objective tools, how good these tools are. I am not a professional audio engineer(making money from audio engineering is not my regular income steam), yet on at least 5 occasions in the last few years, I have, based on both accidental discovery and deliberate testing, discovered serious bugs in software from companies as major as Waves, which led to their rewriting the published product spec - where they were not willing to make the ideal changes, or to their fixing the bugs.

If today's audio engineers were worth their salt, on receipt of a new plugin or plugin revision I'd expect them to subject the plugin to a barrage of objective tests, to ensure that all the basics are covered. I.e the plugin is doing what it is supposed to do., and also make sure there is nothing unexpected taking place within, however inaudible)

A while back there was an unimplemented oversampling feature, in the included compressor plugin that is provided with the Reaper DAW, and thousands of "Audio Engineers" had been mixing with placebo between their ears - until someone checked ! The unimplemented feature was deleted from the user interface, at the next release, once the cat was out of the bag. I was one of those who had been fooled, cos I trusted the Reaper developers implicitly.

Accidentally, and confirmed via proper tests, I found out that there is a bug in the BS2R plugin, which I've documented here, and sent an email to the developer about this. I only found out cos I use tools regularly to check levels throughout the signal path in the digital domain (i.e the DAW).


I am not an audio engineer, i.e a proper audio engineer who studied audio engineering, based on the maths, physics, psychoacoustics, etc @ University, otherwise I might have been able to put together some proper tools to understand what is really going on, in the many options available today, many with unsubstantiated claims.

I am confident, someone will soon be able to provide an objective analysis, that allows us to compare headphone crossfeed and other virtual listening simulation tools. I do have a clue based on some crude testing tools, but not anything that I can unashamedly publish, and ask others to independently verify. Until then we remain in the wild wild west of all kinds of snake oil - Reminds me of the days of Alchemy, where it was all guess work, and spooky ignorance, cos no one really had a clue about chemistry or biology.

Virtual listening on headphones, seems to be at that stage, still in the dark ages, with too many profiting from our ignorance. Have you seen how much Genelec charges for having your HRTF measured? Ridiculous. Even they who should know better, cos they are supposed to be proper engineers, are cashing in, on the abracadabra - the magic arena of binaural listening tools.

Hopefully the kind of work that Amir has done, on measuring other aspects of audio, will be applied by him or others, to headphone listening, so we can choose based on real knowledge, not some waffle by well paid industry endorsers, and their teeming army of placebo induced "audio engineers", who have never used a multimeter in their lives.

 
Last edited:
Top Bottom