To crossfeed or not to crossfeed? That is the question...

[castleofargh]

"Headphone panning" means "illegal/impossible" spatial information which doesn't make sense to our brain. It is no wonder such information annoys.



1Most people are not aware of the concept of spatial distortion so they don't know the full potential of headphone listening (I was one of these people until 6 years ago). I personally think investing money on expensive headphones and listening to them without cross feed while pretending the sound quality is good is silly. Those who don't like cross-feed (know about it and have tried it) do not in my opinion fully understand what cross-feed is about and they misunderstand in my opinion what being a purists means.



2"Basic" cross-feed is right in the sense that it reduces or removes spatial distortion making the signal "legal/possible" for the brain. In other words, a recording could have be done in real world that sounds exactly the same when played back perfectly. How do you know what the sound originally was? Isn't the most important thing that the recording sounds as if it is identical to the original sound? If cross-feed places the drums 5° too left, then what is the problem really? The drums could have been 5° more left in the studio. At least the drums won't sound "fake" and "all over the place" because of spatial distortion.



3How much can we improve sound quality by going from basic cross-feed to "customized solution?" Is it worth the increased complexity and price? I ask this because I am a "bang for the buck" guy.



Technically vinyl is abysmally bad compared to digital audio formats, but still many praise it being superior. This is an interesting issue and I think I have found some explanations for it.

1 yeah, lack of experience and in general simply poor choice of reference can most certainly lead to different ideas of what is right. but I know many people who simply don't seem to mind much. you and I simply aren't among them.
2 well that's the all Hi-Fi idea. getting a little closer to whatever reference. I'm like a very lazy audiophile in that aspect, but the all hifi bizz runs on people who find small improvements very relevant and "worth it".3 follows on that, to each his own idea of what is worth it. I can talk only for myself and say that I'm rather cheap about audio, my headphone rig right now is odac/o2 and a hd650. not exactly TOTL. but when I saw the kickstarter campaign for the Realiser A16, I took my credit card out of my pocket and started filling in the required information to get one. ^_^ didn't need to think, and didn't for a second reconsidered since, despite how we'll probably get it in 2037(well that's in the game with kickstarter).
and I'm not even after any idea of perfection or hifi, I just wish to get something close to the sound of my speakers when it's late at night and I can't use them. because ... neighbors. what I use now feels to me slightly better than using Xnor's crossfeed VST with settings that honestly worked pretty well for my head I sure was glad to have found it at the time. and that felt slightly better than any other crossfeed, and any crossfeed felt better than nothing at all. what's worth it is the most subjective thing there is.

 

[pinnahertz]

Better sit down, 71....

I've just done a bit of cross-feed experimenting. (Ok, you can get up now.) It had been years, and now I have software and a bit of free time today, so here's what I did.

First, I got VOX iOS, played a few tracks, tried the 3 different xfd presets. The results were so completely dependant on the original material it was unnerving. The effect on hard-panned material was quite obvious, but on more contemporary mixes the effect went from nothing at all to just discernable.

To my ears none of the presets were always an improvement, sometimes the were inaudible, sometimes I preferred the origina, sometimes I liked the crossfeed version. I didn't discern any significant differences between Chu Moi and Jan Meier, but the default, whatever that is is slightly different. There is no adjustment within a crossfeed method, and there really should be.

Being the tweak head that I am, I opened a couple of tracks into Audition where I have complete control over what's going on. I simulated the Meier crossfeed, the Chu Moi, one of my own, and simple variable separation. All sound different, and all work differently on each track. Grrr.

For each test track each method had to be adjusted for optimum, and they didn't seem to track each other at all. One that was fairly consistent was the one I whipped up, that was derived from L-R/R-L, delayed, filtered, then mixed back in. I then had control of crossfeed level, time delay, and filter response. Once I had it dialed in, when images became relocated they were solid and palpable. Other methods resulted in images that were somewhat smeared and vague. However, the whole thing still needed to be customized per track. Darn. Well, expected of course. I also tried inserting dynamic processing on the crossfeed signal. Yeah, pretty much how I remembered it working, not well. Relocated images begin to wander around, kind of swimmy. But the control sample was just being derived from L-R, which is technically wrong, so that will have to be revisited to work right, likely processing through a special side-chain.

The other method that was pretty consistent was simply separation reduction (mixing L+R into the stereo mix), though it was less satisfying, it of course cured the hard-panning problem.

Test material, in part, was:
Beatles Flying from Magical Mystery Tour (hard panned instruments, hard center vocal)
Bob Dylan: Man of Constant Sorrow (hard panned guitar and vocal)
Diana Krall: Besame Mucho (contemporary mix)
I tried a few more too, and will still play a bit, perhaps write a bit more about this. So far it's kind of what I expected...sorry 71!!!

If anybody cares I can list my processing settings to get the algorithms right. Otherwise this is just a brain dump. Just do with it whatever you usually do with a dump of any sort.

 

[bigshot]

I find that true of a lot of processing. I have about four DSPs that I juggle depending on the situation. And there are albums that have been mastered out of spec that I have to adjust EQ for. In multichannel, the volume of the rears to the fronts is sometimes wonky. No one size fits all solution.

Maybe it would be best to have a crossfeed where the parameters are all flexible and you can just adjust to each recording as you see fit.

 

[pinnahertz]

The Parrot app lets you couple tunings and crossfeed settings to individual tracks with your settings maintained in a database you can even share. So you can play a tune, adjust, and save, then it recalls those settings next time you play the track. The problem is it doesn't work without an internet connection.

edit: And it's crossfeed sucks.

 

[Malfunkt]

My brain has pretty much adapted to listening without any kind of crossfeed or emulation. When I'm listening to classical or jazz, without any processing, I may have improper spatialization, but timbre is intimately enjoyable. I listen often for melodic input, and so only a certain level of resolution is necessary for my brain to fill in the blanks and to become engaged.

For movies and some gaming, I now use Dolby Atmos for Headphones, which can handle crossfeed for 2-channel as well as simulate 5.1 and 7.1 sound in headphones. It works remarkable well with the right material. This is on PC, and I would still say that a lot of my listening happens without crossfeed. I also listen to a lot of electronic music, where the spatial side effects of listening without crossfeed typically suffer less .

On iphone / android , I know @bigshot mentioned how many apps don't get ongoing support, and this is important. But I don't mind spending a bit of money now and then to get an app that has some DSP involved. I wish Apple built it right into their headphones, or even licensed Dolby Atmos so as we move forward with the future of VR and media (which will likely involve a lot of headphone listening) we can work around some sort of standard for full spatialization.

For iOS and Android, I'm using a well supported app called nPlayer https://nplayer.com/ They recently implemented DTS-X and using the 'normal' ' over-ear headphones' setting, it is worth listening. Much better than BS2B in my opinion. DTS-X though will also work very well on multi-channel content, and so its perfect for watching movies on the go.

(above post typed while immensely enjoying a lowly Denon AH-D2000 without any DSP processing. Vox Player on Mac listening to Entheogenic 'Anima Mundi')

 

[71 dB]

Better sit down, 71…. ...a lot of venting... ...sorry 71!!!

No need to apologize me. I'm sorry to hear cross-feed is so problematic for you.

 

[71 dB]

If you can use DSP to convolute HRTF-impulse responses with the music then of course do that. That's cross-feed too, just in more sophisticated form.

I don't have any hardware or software to do that, so normal cross-feed is the best I can do. This is not a money issue. I have money, but in my opinion it's more important to safe money for bad times/pension than use it all in fancy stuff while you are young. I enjoy my music on $200 cans with $50 cross-feed headphone adapter. I think I can sustain such level of consumerism and still be able to pay my bill when old. If I won millions in lottery and my life was completely secured financially, I would go for Smyth and whatever luxury products there are out there… …but I understand that's only for privileged people and it doesn't even make people happier.

My message with cross-feed is that since basic cross feed is quite affordable (for one guy I designed a simple DIY cross-feeder that cost $20 to build and he was happy with it) and help with spatial distortion, there is no reason to be without. Cross-feed as such is not to make your all headphone fantasies true.

 

[pinnahertz]

But would you spend $2 for DSP cross-feed done right?

 

[71 dB]

But would you spend $2 for DSP cross-feed done right?

$2, $20, even $200 if it really is superior to simple cross-feed.

 

[pinnahertz]

The typical app is around $2. If someone wanted to popularize cross-feed and get it into millions of hands/ears, that's probably the best way. The DSP load required is well within the capability of every smart device in the last few years.

 

[bigshot]

Add a good digital equalizer to it and I'd pay a ten spot

 

[pinnahertz]

What makes a "good digital equalizer" in your world?

 

[bigshot]

Either a graphic equalizer with at least 15 bands or a parametric with at least 5 bands. Perhaps auto normalizing to avoid clipping.

 

[71 dB]

The typical app is around $2. If someone wanted to popularize cross-feed and get it into millions of hands/ears, that's probably the best way. The DSP load required is well within the capability of every smart device in the last few years.

Probably yes, but it wouldn't hurt if these smart devices had it integrated as a default. DSP load is low for "normal" cross-feed.

However, more advanced methods are another story. Convolution of HRTF impulse responses is much more demanding and then there is of course the problem of using right HRTFs.

 

[pinnahertz]

Probably yes, but it wouldn't hurt if these smart devices had it integrated as a default. DSP load is low for "normal" cross-feed.

However, more advanced methods are another story. Convolution of HRTF impulse responses is much more demanding and then there is of course the problem of using right HRTFs.

You convolve once, then it's a low load fixed filter that is easy to run. It's been done, though not specifically an impulse response for HRTF. Audyssey's AMP app (now discontinued...nuts!) ran headphone EQ which started by essentially taking impulse response in their lab, developing the "tuning" filter, then users downloaded it and ran it on their iOS device. It worked fine even on devices several generations previous to current. In fact, the apps issues had nothing to do with running the DSP filter, that was actually the easy part.

I'm not sure why you're arguing about using the "right HRTF", when any generalized HRTF beats any 'normal' cross-feed. It's clearly impractical to measure every users HRTF, so it must be a generalization, perhaps a choice of 2 or 3, like small-medium-large, or whatever.

All I'm saying is the technology is in our hands now, and there's a market gap. The missing critical information is market perceived need, which is probably why it hasn't been done. None of the details of the DSP algorithm are road blocks at all. It sounds like "Mr. Cross-Feed" is pushing back on the idea of a $2 DSP cross-feed app. Huh.