Crossfeed for headphones

[maverickronin]

This endeavor made me also wonder about the necessity of lower crosstalk. Basically, what these filters are doing is adding the equivalent of huge amounts of crosstalk - although more sophisticated and intentional. The proponents of balanced headphone connectors usually speak highly of how the lower crosstalk between channels widens the soundstage. If the difference is between -30 db and -60 db of crosstalk, that does sounds reasonable. But realistically the initial crosstalk is much lower than even -80 db, so the benefits to the soundstage are questionable. And even if they exist – creating more separation between the stereo channels is only widening the stage artificially. Music that is mixed with speakers is not really intended to be listened to in such a way that each channel is completely isolated to one ear. The widening effect is artificial and can only decrease the realism. In other words, those who claim that lower crosstalk in headphones gives you more realistic soundstage are basically wrong.

"Accidental" crosstalk like that is usually caused by capacitive coupling in the circuit and increases with frequency which is the opposite of what's needed.

Challenge for you all- pro or anti crossfeed. Find the track "Let it Flow" by Spiritualized (on pure phase album)- There is a sound that is panned in a 360 loop around your head (or at least I perceive that and believe that is the intention). I get it without crossfeed enabled. See what you think.

I have some similar tracks which can't decide if they want to be stereo or binaural. Tracks that are already binaural certainly don't need crossfeed and usually suffer from it.

 

[solderdude]

And then there is also floor and ceiling bounces and of objects in the room and HRTF differences as well as driver-ear distances and angles.
Acoustical mixing simply is not the same as electrical manipulation certainly not from the front.
Emulation from things like the Smyth realizer etc. can hardly be called crossfeed anymore.

The effects/preference will indeed differ from person to person.
I too suffer from the reverse problem having gotten used to headphones more so than speakers.

 

[Fluffy]

After a bit more adjustment, the custom filter looks like this:

Copy: L2=L R2=R

Channel: L2 R2

Filter: ON HSC 6 dB Fc 1000 Hz Gain -18 dB

Copy: L=0dB*L+-13.0dB*R2 R=0dB*R+-13.0dB*L2

Channel: L R

Filter: ON LSC 6 dB Fc 400 Hz Gain -1.5 dB

I raised the crossfeed in the bass by 3 db so it's a tiny bit more centered, though still not as much as Meier's filter.
Basically I have 6 levels now. I could probably make a lot more levels in between but the differences are already pretty minor.

 

[Jimbob54]

After a bit more adjustment, the custom filter looks like this:


I raised the crossfeed in the bass by 3 db so it's a tiny bit more centered, though still not as much as Meier's filter.
Basically I have 6 levels now. I could probably make a lot more levels in between but the differences are already pretty minor.

You able to quickly A/B between any filters or toggle on /off? Any noticeable difference in FR when off?

 

[dwkdnvr]

Everyone interested in 'more realistic' headphone reproduction should check out Impulcifier here https://github.com/jaakkopasanen/Impulcifer . It's basically a DIY version of the Smyth Realizer concept - measures the binaural HRIR of a set of speakers in-room, and then creates a set of convolution filters to recreate that over headphones. It requires a binaural mic such as these which I have https://www.soundprofessionals.com/cgi-bin/gold/item/SP-TFB-2-BLK-HS-MOG-XLR .
I'm still in the 'experimental' stage, but it's the real deal - works remarkably well. I've only done stereo, but the approach scales to 5 or 7 channels if that's your bent. You can of course emulate a 5/7 channel setup by repeatedly measuring a single channel at different orientations, too. I run the convolution filters in JRiver, but there is also a package called 'HESUVI' that works with EqualizerAPO that will run them.

 

[maverickronin]

Everyone interested in 'more realistic' headphone reproduction should check out Impulcifier here https://github.com/jaakkopasanen/Impulcifer . It's basically a DIY version of the Smyth Realizer concept - measures the binaural HRIR of a set of speakers in-room, and then creates a set of convolution filters to recreate that over headphones. It requires a binaural mic such as these which I have https://www.soundprofessionals.com/cgi-bin/gold/item/SP-TFB-2-BLK-HS-MOG-XLR .

In theory that's the best, but in practice requires you have access to an excellent listening room.

 

[Fluffy]

You able to quickly A/B between any filters or toggle on /off? Any noticeable difference in FR when off?

Yes! Found a simple trick – I put each preset in its own txt file in the config folder, and named them "crossfeed 0-4". Then I use the "include" command to include one of the presets (for example: "include: crossfeed 0.txt"). And then all I have to do to instantly change to another preset is to change the number, say from 0 to 1. When I hit enter it immediately change to the new preset (the analysis panel say initialize time of 2.5 ms, so virtually instantly). You need to be in instant mode for this to work, of course. I can also then toggle it on/off by turning off the filter. What I can't do is do a quick A/B back and forth because each time I need to highlight the number, change it, and press enter, which takes several seconds. If there would be a simple a/b command with a button to click on to switch that would be easier. But my current method is enough to easily compare between one mode and another.

The main noticeable difference is the bass spread, and logically so because that's the main difference between the filters. If I'm listening to music with a lot of rumble or bass content, a stronger filter gives the impression of less bass. Note that I made sure that the actual FR remains perfectly the same (to an accuracy of 0.2 db across the spectrum). So this is not really a change in overall FR, just a perceived change of less bass because it's more concentrated in the center. Without my FR correction, the filters by themselves are measurably affecting the FR by elevating the bass – which makes sense, because each ear is receiving some of the bass from the other ear. So correcting this makes it actually less authentic. But it's better for comparing the net effect of the crossfeed on soundstage.

 

[MRC01]

If anyone's interested, I attached the measurements I did for the five crossfeed states (including off, and 1-4). The method was generating a logarithmic chirp of 20hz-20khz, hard panning it to the right, and loopback recording with each crossfeed active. It's easy to see (and hear) what is the exact effect of each filter across the spectrum. For example, the strongest one (level 4) have a mere 5 db difference between L and R at 20 hz, and slowly expands this difference until it reaches about 32 db at 20khz. ...

You made me curious about the crossfeed implemented in the Corda Soul. Perceptually, it simply narrows the stereo image without any other side effects in the sound. I can hear a slight difference in FR with the analog crossfeed on the Corda Jazz headphone amp. But I can't hear any difference in FR with the Corda Soul. I even tested using pink noise which usually accentuates differences in FR making them easier to hear. But all the crossfeed modes sounded the same.

So I wired up a couple of XLR to TRS cables, connected it to my sound card's balanced analog inputs, and took some measurements with REW. No measurable differences in FR at any crossfeed setting. A picture's worth 1,000 words, so here are all 5 crossfeed settings, plus all 5 "speaker cross feed" settings (DSP that does the opposite of crossfeed, widening stereo), plus the FR with all DSP disabled. The curves are all exactly on top of each other, with a Y scale of 0.1 dB per division. The overall curve shape is my sound card's analog inputs being slightly non-flat. This was at 96k sampling, but I also measured 44.1 and it was exactly the same.

So, Meier's implementation of crossfeed in the Corda Soul has no impact on FR at any of the settings, which is different from his analog version, and the software versions you're testing.

 

[Fluffy]

@MRC01
What signal was used to measure this?

I tried to verify that my presets don't change the frequency response using the FR display in EQ APO and with music. But I'm still getting a feeling that the bass responds differently. The pink noise method sounds interesting though, I should try it. is it pink noise in mono or stereo?

 

[MRC01]

To measure it, I used REW's frequency sweeps at 96 kHz, 10 Hz to 48 kHz. When REW is set up, just click the "Measure" button. BTW, just because the FR is the same doesn't mean it SOUNDS the same. Phase and inter-channel time differences can make frequency ranges SOUND louder or quieter, even if they're at the same level. However, with the Soul's crossfeed there are no perceptual differences in FR, at least not to me.

Pink noise is the best sound / test signal that I know of for hearing differences in FR. It doesn't seem to matter whether it's mono or stereo. Differences in FR that are hard to detect with music just leap out and become obvious with pink noise.

On a related note of useful sounds / signals / recordings to keep around for testing, another one is any close-miced recording of jangling keys on a key chain made with a good microphone. It has tons of extreme HF content that highlights any differences in high frequency and transient response. Some audiophiles use castanets or other small percussion sounds for this, and they work OK, but jangling keys is even better.

 

[maverickronin]

So, Meier's implementation of crossfeed in the Corda Soul has no impact on FR at any of the settings, which is different from his analog version, and the software versions you're testing.

That kind of measurement won't usually reveal the real changes crossfeed filters make to FR since the changes usually come from the mostly mono bass in actual music summing with the opposite channel. Since pretty much all testing using identical signals in each channel the summing will remain the same across the frequency spectrum and no deviation will occur.

You'd probably need a very special test signal to actually see the difference in one quick measurement.

 

[MRC01]

That kind of measurement won't usually reveal the real changes crossfeed filters make to FR since the changes usually come from the mostly mono bass in actual music summing with the opposite channel. Since pretty much all testing using identical signals in each channel the summing will remain the same across the frequency spectrum and no deviation will occur. ...

That's true, but the Headroom amps did show a change in FR with the crossfeed enabled. I believe it was because their cross-channel time delay caused a comb filter effect that attenuated the high frequencies. This attenuation was significant and easily audible. Headroom had a gentle high pass "filter" switch to offset it. I don't know how Meier avoids this obvious FR change, but he does. His analog crossfeed implementation has a perceptible FR shift that is much more subtle than Headroom, and in the opposite direction - it seems to lift the mids & lower treble just a bit. Meier's DSP implementation of crossfeed is the only one I've heard that doesn't change the FR perceptually (at least for me) or in measurements.

 

[maverickronin]

I believe it was because their cross-channel time delay caused a comb filter effect that attenuated the high frequencies.

Interesting. I wouldn't have expected that.

Meier's DSP implementation of crossfeed is the only one I've heard that doesn't change the FR perceptually (at least for me) or in measurements.

I would guess it's counteracting the FR change with a corresponding EQ then. (Probably calculating it in real time for all the FR plots to line up so perectly...) TB Isone has a feature crosstalk spectrum compensation (CSC) which seems to do the same thing but I don't remember if I ever actually measured it to confirm through...

 

[solderdude]

Why not record the output of various settings using excerpts that really show the effect (MP3 VBR0 is already enough as it is quite audible) and or unleash PKanes software on it (not using MP3 )

 

[dasdoing]

I tryed all kinds of crossfeed/vitual room aplications and they all sound shit. Best I found is Abbey Road Studio 3, but it still colorizes af.
So I took some binaural IRs from WDR studios and tweeked it until I got most colorization out.
at the time I had a wide angle setup in my main speakers so I used a wide angle for this. I will create a regular stereo triangle version in the future since I went back to the standard:

what I found out is that people who like the "in the head sound" wont actualy like a speaker like soundstage on headphones. People who use manly speakers will apriciate it though when they use headphones

 

[gino1961]

... people who like the "in the head sound" ...

strange people indeed ...

I don't know how you guys do it. I flat out get headaches from any decent full size headphone without crossfeed. ... things really sound like they are coming from in front of you

Hi ! could you please upload a short sample of recording able to provide this feeling ? even a short one ... like 30 seconds.
I thought it were impossible. But hearing is believing.
And sorry but i missed what kind of crossfeed HW or SW are you using.
Thanks a lot, gino

 

[maverickronin]

Hi ! could you please upload a short sample of recording able to provide this feeling ? even a short one ... like 30 seconds.
I thought it were impossible. But hearing is believing.

The effect is also heavily dependent on the headphones so just uploading something run though a crossfeed algorithm probably won't have the dramatic results you're hoping for. What headphones do you have?

For local music files I can play through foobar I use TB Isone. For local video files I use the 5.1 upmix and headphone HRTF DSPs in ffdshow. For everything else (streaming media, games) I usually use the level 4 crossfeed setting on my RME ADI-2 DAC.

TB Isone is the most effective for me but it takes time to dial in the settings to find something that works well.

 

[gino1961]

The effect is also heavily dependent on the headphones so just uploading something run though a crossfeed algorithm probably won't have the dramatic results you're hoping for.

Hi ! this is very interesting. in your experience which are the best in this specific task of virtual soundstage also in the front ? this is for me by far the thing i miss most with HPs compared to natural listening. I do not say speaker listening but natural because in reality i can place precisely a person speaking in front of me ... not so with HPs. With speakers of course yes.

What headphones do you have?

i have many cheap ones ... akg 240, 242, k500, k701 beyer 880 old 600 ohm, 990 ... other from takstar ... and others cheap

For local music files I can play through foobar I use TB Isone. For local video files I use the 5.1 upmix and headphone HRTF DSPs in ffdshow. For everything else (streaming media, games) I usually use the level 4 crossfeed setting on my RME ADI-2 DAC.
TB Isone is the most effective for me but it takes time to dial in the settings to find something that works well.

Thanks a lot indeed. I will try TB Isone then. I am sure that the day when front sound can be perceive with HPs they will take over speakers in many many situations. I only heard once a distinctive sound clearly in front of me with HPs watching youtube ... silly me that i did not save the link. It was surprising.
But this experience gave me the feeling that it is possible. Tricky but possible.
Kind regards, gino

P.S. anyway if you have a track where this effect is pronuonced i could listen to it with all the HPs i can get hold of for sure. That track also would make for a very powerful tool to select also the best soundstaging HPs ... this is for me the best quality by far.

 

[maverickronin]

Hi ! this is very interesting. in your experience which are the best in this specific task of virtual soundstage also in the front ?

The best if you're using minimal processing the HD800. For me it can do it with just regular crossfeed. I don't own one, but they are always at shows and meets so I've heard them a lot. With most other headphones plain crossfeed mostly serves to keep me from getting headaches.

After dialing in TB Isone and EQ on my Stax Lambdas they also sound like a source several feet in front of me.

I've never been able to try the HD800's with TB Isone though.

k701 beyer 880 old 600 ohm, 990

You should be able to get pretty good results with those as long as you apply some EQ as well. I find both TB Isone and plain crossfeed work best when you EQ the headphone close to a diffuse field response first. Diffuse field is close enough to the midrange and treble portion of the Harman headphone curve so you can probably find some EQ presets for that and save some time.

 

[gino1961]

The best if you're using minimal processing the HD800. For me it can do it with just regular crossfeed. I don't own one, but they are always at shows and meets so I've heard them a lot. With most other headphones plain crossfeed mostly serves to keep me from getting headaches.

Hi, thanks for the valuable advice. I wonder why the concept of the HD800 is not more popular. I guess that their performance should depend on the placement of drivers. For sure the ones in the HD800 must be high end but if the concept is right also drivers more standard should provide a pretty similar effect.
I would relate the headache to an unnatural experience. The brain is confused and then stressed. It is like with the sound in the head. This is unnatural and leads to discomfort. This is a pity because HPs especially closed one are very interesting. I am always looking for good sounding and well isolating HPs to be used outside for some recordings i have in mind. I prefer closed HPs. The isolation factor for me is very important.

After dialing in TB Isone and EQ on my Stax Lambdas they also sound like a source several feet in front of me. I've never been able to try the HD800's with TB Isone though.

this must be amazing. Never experienced that. I have to study this SW indeed. For me just one meter in the front would be more than enough.
I cannot pretend a cavernous soundstage with HPs. But i am optimistic.
I have to listened to recordings done with the Neumann KU100 ... that is for me the concept to follow (i have even a diy project in mind ...

You should be able to get pretty good results with those as long as you apply some EQ as well. I find both TB Isone and plain crossfeed work best when you EQ the headphone close to a diffuse field response first. Diffuse field is close enough to the midrange and treble portion of the Harman headphone curve so you can probably find some EQ presets for that and save some time.

I have clearly to experiment. I am sure that some DSP in the end should work. Maybe it will not be perfect but useful to get some front sound.
The HPs are just to convenient to be left aside. Especially for recordings outside they are a must, of course.
Thank you very much for all your kind and very valuable advice.
Kind regards, gino