1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

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

Discussion in 'Sound Science' started by jasonb, Oct 21, 2010.
76 77 78 79 80 81 82 83 84 85
87 88 89 90 91 92 93 94 95 96
  1. 71 dB
    I wonder why this "focus control" is able to do for you what crossfeed vst plugins were not. The user guide is not very technical about how it's done, but based on the description it doesn't sound much more sophisticated than your typical crossfeed plugin.
  2. 71 dB
    Very little if anything can be gained. That's my problem. How to gain? This has been one of my efforts to "gain" and it failed royally.
    Experiences like this are demoralizing. Why even try when all you can achieve is this? I am really really bad at selling my opinions
    to other people as you have seen. I feel I am not able to express what I realized in 2012 and it has something to do with system
    thinking and how people assume default positions. People assume the default position of headphone listening without crossfeed.
    Is that called for knowing recordings are mixed for speakers? In 2012 I realized the scientific default position of headphone listening
    is binaural sound and since recordings very rarely are binaural, they have to be modified for headphones. Crossfeed doesn't turn
    speaker stereophony into binaural stereophony, of course not, but it "scales" the speaker stereophony into something that has got
    similar ILD range compared to binaural stereophony. To my ears this mean a significant improvement over not doing anything and
    perhaps the best bang for the buck rations of headphone audio. I don't know how to get an improvement that big with $50 in any
    other way…

    …I have been convinced that this improvement my ears hear is explained by the science of human spatial hearing. It was
    applying the science to headphone listening that made me have the 2012 realization in the first place (before 2012 I never
    thought about headphone sound, because I was into speakers and headphones seemed so trouble-free as the transducers
    are at your ears, but ironically that turned up being the problem!) Crossfeed does round corners and is not perfect in any way.
    It's coarse. However, I believe it is still a significant improvement as speaker stereophony on headphones is such a huge
    problem. If you can use something better than crossfeed good for you, but all the HRTF-convolutions are technically crossfeed,
    summing a modified version of the signal to the other channel, just much more detailed. So, when I speak generally about
    crossfeed, I mean any way to reduce ILD of speaker stereophony to get something that has ILD similar to binaural

    Since recordings are mixed for speakers using studio monitors, I think it's rational to assume the playback to have acoustic
    crossfeed for the direct sound with a 200-300 µs ITD corresponding the speaker angle. Crossfeed usually mimicks this using
    similar ITD so I don't see why this is a significant problem. Not doing anything should be a much bigger problem. As for the
    combination of ILD and ITD values, I believe those pairs make much more sense after crossfeed, because ILD values tend
    to be so much "off" so even if ITD is change a little bit, getting ILD to the ballpark is much more important. Crossfeed is
    kind of a "mapper" that takes the original spatial parameters and "maps" them into parameters that are more natural to
    human ear. This is what my reason says and it's what my ears say. That's why it is so difficult for me to believe I am wrong.
    If I was wrong then the implications would be I can't trust my reason and senses at all.

    If I can learn something from the experience here it is this: I am a person of low self-esteem so when someone questions
    my knowledge of a subject I myself have believed knowing well I get hurt badly and I attack and insult people. The solutions
    are not easy:

    1) get better self-esteem => easier said than done. I try, but it is so damn hard!!
    2) know better => again I am a human with limits of learning and I don't know what new to learn about spatiality, especially
    when my own ears tell me what I know seems to work. If I agree with Gregorio I disagree with my ears!
    3) stay offlne. => that works, but human beings need interactions. I want to be involved, part of something.

    The solutions are the there, but they are not easy. I am annoyed of these difficulties. I want easy happy life!
  3. bigshot
    I think you've gotten to the point where no one cares. You're typing into an echo chamber. No one is bothering to read your word salad any more.
  4. 71 dB
    Does my voice here make a dissonant chord in your echo chamber? Sorry. I am this way. My opinions about crossfeed shouldn't even concern you as you are not a headphone user. You are a speaker user or at least so you keep saying ten times a day. Crossfeed is irrelevant for speaker users, isn't it? The question in the topic of this thread "To crossfeed or not to crossfeed?" should be meaningless to you. Why do you care about whether crossfeed improves headphone sound or not? You don't use headphones, do you?

    Can you blame me for losing my patience and getting frustrated when so much of the responses to my posts are about trying to downplay my credibility (attacks toward me rather than yje claims I am making) rather careful analyse of what is possibly wrong in my claims? No one bothers to read my word salad any more? No one cares? Are you sure? Have you asked everyone? My posts being word salad is YOUR opinion (and you are of course entitled to have such opinion), but do other members here share your opinion? It's possible, of course, but how sure are you?
  5. bigshot
    I think you are the one that should be asking whether people appreciate your posts. Take a break.
  6. 71 dB
    I post what I want and people appreciate it or they don't.

    I don't tell you to take a break (not my business).
    All I know is you are appreciated here.
    I have seen very good posts by you.
    Sometimes you are perhaps a little bit condescending, but that's your style. I have my style.
    The subject we seem to agree the most about with each other is "16 bit / 44100 Hz is all you need".
  7. Hifiearspeakers
    Listen, you can all continue to bash @71 dB all you want, but he makes a good point in his post above. This thread is all about CROSSFEED and that really only applies to headphones/earphones. So when you all can’t stop chirping about speakers, then you all are really the ones off topic. So disagree with him all you want, but to call for him to be banned for pounding away on the actual topic (redundant as it may be) is just asinine.
    71 dB likes this.
  8. kukkurovaca
    what if they held a flamewar and nobody came
  9. bigshot
    I like your name kukkurovaca. It sounds like some sort of Australian marsupial.
  10. kukkurovaca
    Thanks, it's a joke about how bad I was at doing Sanskrit homework
  11. castleofargh Contributor
    it's hard not to discuss speakers, be it to explain crossfeed, or to debunk empty claims about it.
    if we were to treat crossfeed for the subjective process that it is, then of course we could focus on headphones and subjective impressions. but somehow, @71 dB NEEDS for crossfeed to be more than something you appreciate(or don't). so speaker talk, it is.

    once again wikipedia was right. :smile_cat:
  12. Dawnrazor
    No idea man. I just know that it works and vst crossfeed plugins never seems to make me happy.
  13. ironmine
    Just for the fun of it, I would like to try creating a crossfeed myself in the VST host chainer (ART Teknika Console), using individual VST plugins and connecting them to one another.

    71 dB, can you advise me how to do it in the right manner? I mean the general principle and scheme (diagram).

    As far as I understand, I need to mix a bit of the left channel with the right channel and vice versa. Also, simple mixing is not enough, the left channel needs to be delayed a bit before it is mixed with the right channel. Does this delaying need to be frequency-dependent? What other processing I should apply to the left channel before it's mixed with the right one?
  14. ironmine
    Here's the interesting post at GearSlutz from the guy who analyzed different crossfeed plugins and posted screenshots about them comparing the ways they process the sound.
    castleofargh likes this.
  15. 71 dB
    I am not good at coding and I have never wrote VST plugins, but I certainly know about the general principle and scheme.

    Crossfeed is about makiing both ears hear the content of both channels (as happens with all enviromental sounds) simulating coarsely the differences in sound at ears when the sound comes from an angle (typically something like 30° mimicking a normal speaker setup).

    So, you mix a little of left channel ( L ) to right channel ( R ) and vice versa as you said. Human head blocks high frequencies more than low frequencies, so that more bass is leaked to the other ear than higher frequencies. The size of human head dictates the frequency point at which the head starts to block sound "aggressively". This frequency is about 800 Hz. So, up to this frequency we should leak a lot of sound to the other ear and above that frequency less and less with frequency. That's why we low pass filter the leaked sound at 800 Hz (1st order butterworth is a common choice):

    Lf = L lowpass filtered at 800 Hz
    Rf = R lowpass filtered at 800 Hz

    Even at bass there is a small ILD so we don't "leak" these filtered versions without some scaling. To have gain x dB we need to scale them by a factor k of

    k = 10^(-x/20)

    which comes directly from the fact that x = 20*log10 (k). Note than k < 1 meaning log10 (k) is negative so the gain is negative decibels. To have gain -1 dB k becomes:

    k = 10^(-1/20) = 0.89125… …(0.9 does fine! The error is less than 0.1 dB)

    What about the delay? Lowpass filtering at 800 Hz using 1st order butterworth creates a delay of about 200 µs at bass dropping to about 140 µs at 800 Hz where the leaked signal has attenuated 3 dB more and then the delay dies out with frequency. It's possible to do an additional delay to have the angle of sound increased, but this works nicely as it is. ILD varies between 0 and about 700 µs and depends almost linearly of the angle between 0° and 90° for sound on horizontal plane so that the delay in microseconfs depends on angle roughly:

    delay in microseconds = 7 * angle in degrees for distant sounds
    delay in microseconds = 8 * angle in degrees for very near sounds

    The difference comes from geometric reasons and in my opinion the upper one is more reasonable approximation, because reducing ILD also means more distant sound. So, 200 µs corresponds about 29° and 140 µs equals 20°.

    Mixing L and Rf (R and Lf) means increasing low frequencies compared to high frequences. That's why we need to boost frequencies above 800 Hz to have balanced sound. If low frequencies are totally mono to begin with the increase at low frequencies (assuming zero phase differency) is (for k = 0.9) 20*log10 (1+0.9) = 5.58 dB. However if they are uncorrelated, the increase is smaller, 10*log (1*1+0.9*0.9) = 2.58 dB. Music is typically something in between, not totally mono, but not uncorreleted either. So the typical increase is between 2.58 dB and 5.58 dB, but in practise the increase is smaller, because of the phase delay caused by filtering. Typically crossfeeders make a high frequency boost of 2-3 dB.

    If someone at this point calls this "messy", let me remind you how "messy" speakers in a room are! That's messy! The acoustics combined with the radiation patterns of the speakers create a mess that make these minor things of a couple of dB look totally flat in comparison. If a crossfeeder sounds too bright or bassy to your ears they can be tweaked to give what you want. These calculations are made to give balanced sound having similar spectral balance to the original sound.

    Boosting the high frequences of L and R create an additional delay of about 50 µs so that the final delay is about 250 µs (~35°):

    L crossfed = L hf boosted + k * Rf
    R crossfed = R hf boosted + k * Lf

    However, this works well for "ping pong" recordings with huge stereo separation! If the recording has more sensible spatiality we want to crossfeed it much less! In practice the usefull range for crossfeed level is from -12 dB (almost headphone ready spatiality) to -1 dB (ping pong or "movie surround sound").

    So, the value of k has to be adjustable ranging from 0.25 to 0.9. Also, the need for treble boost is smaller when k is smaller. Below is the treble boost analyse of a my crossfeeder (which nicely scales the treble boost with the crossfeed levels):


    Hopefully this helps...
    Last edited: Oct 19, 2019
    ironmine likes this.
76 77 78 79 80 81 82 83 84 85
87 88 89 90 91 92 93 94 95 96

Share This Page