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.
90 91 92 93 94 95 96 97 98 99
101 102 103 104 105
  1. SoundAndMotion

    I think this response to you is better placed in @castleofargh 's thread here, but I didn't want to bring down the level there. Should this continue, perhaps a new thread is better, since it also doesn't belong here.

    I have neither the time nor desire to engage in one of your standard bullying, long-winded, pedantic, point-by-point, repetitive battles. Perhaps later, when I have more time, although my previous exchanges with you prove unsatisfying. You never, or at best rarely, acknowledge your mistakes or give credit to valid counterpoints to your points.

    I respect your expertise in the methods and standard practices of music recording and production. I can and have referred to your knowledge of that space. I also support your, or anyone's, calls for evidence to back up claims. I'm an evidence-based person and do evidence-based work. There are limits about when and how to call for evidence, but it is a valid and useful endeavor. But using your own words as evidence ("The facts and science back me up.... so disagreeing with me is anti-fact/anti-science") is such a blatant appeal to self-authority that DK problems are an obvious area to explore. You place people in 2 groups: those who show you proper obeisance or the audiophiles (using the negative definition, not one from a dictionary).

    Although you may know more than many lay people in the areas of digital and analog audio, electronics/physics, and perception, you are by no means expert in these. But you act as though you are and appeal to your own authority: classic DK. I can dig up your posts and PMs for evidence, but off the top of my head, among other things, you: misapply Shannon-Nyquist to sampling theory, interchange voltage and current incorrectly, confuse frequency and time, confuse measurement and recording, and don't understand the limits of perception. There are valid relationships among these things you use to justify your mistakes, but they do show your lack of understanding, let alone expertise.

    71dB has useful knowledge about auditory spatial perception and acoustics. Better than a lay person, but not an expert. He also likes crossfeed and wants to point out that there is a perceptual basis for some people liking it. Sometimes he leaves out the "some" from "some people", but he has gotten better. You want to dismiss the preference as a "vanilla ice cream tastes better than chocolate"-type preference. He repeats his justification for there being a true perceptual reason that the illusion is pleasing to some. You engage in pedantic, bullying, condescending harassment, which just provokes continued, flustered responses. In your exchanges, there is not a correct vs. incorrect side. You both make valid points, often talking past each other, but he struggles with communication (for language and social-interaction reasons) and that is your favorite exploit. The only winners in your exchanges are the popcorn-munching, rubber-necking voyeurs who used to shout "fight, fight, fight" on the playground, and you, because this achieves your goal. You provoke 71dB's worst behavior, and the only way to stop the train-wreck is to just stop.

    Most posters on audio forums are not experts in the related fields. So any exchange here must be taken with a grain of salt. The reason most exchanges are posted, rather than PM-ed is that the exchange may benefit others. 71dB's and ironmine's exchange, if they mutually benefit from it, should continue. Not because they write flawless posts, but because it helps both, they have fun, and it could help others. You were not elected hall monitor, and your classic "this is not the *** forum, this is sound science" really shows you seek validation here. We all know it's socially unacceptable to admit that... all except 71dB, he just admits it. And I'm willing to cut him some slack for his language and social awkwardness. You, as a bully, haven't earned that slack.
    Last edited: Nov 11, 2019
    71 dB and Hifiearspeakers like this.
  2. ironmine
    I almost gave up on using HRTF impulses from Listen HRTF Database website.
    They just don't provide crossfeed effect to a sufficient degree. And there's too much bass (as if music is playing in a basement).

    So, I will continue with my other "project": trying to build a better alternative to my currently favorite 112dB Redline Monitor using individual VST plugins.

    71dB, you mentioned earlier that the low pass filtered channels, before they are crossfed back to the main signal, must be attenuated from -14dB to -2dB - but you advised me, if I have to use one value, to apply the average -8dB. I just wonder is there any objective method (in the form of a VST plugin) to measure an audio track and see how much "hard stereo" there is, especially in in low frequencies? Can such phenomenon as "hard-panning" be objectively measured?
  3. 71 dB
    Actually I think the range of needed crossfeed is something like -12 dB (almost "binaural" recordings) to -1 dB ("ping pong" and often multichannel tracks downsampled to stereo).
    I think if you need to crossfeed at level lower than -12 dB, it's better to do nothing and just bypass crossfeed. That way you at least avoid all the negative things Gregorio accuses me of ignoring.

    I use my ears to determine the proper level of crossfeed and I think once you undertand what crossfeed should be doing it's quite easy. However, there is of course possible to analyse signals more objectively. I use S/(S+M) method:

    M = abs(L + R) (the mid signal, absolute value)
    S = abs(L - R) (the side signal, absolute value)

    We don't need to scale these in any way to match matrix rotations, because it won't affect the value of

    D = S / (S + M).

    D tells how much of the total signal is in S, in other words it's a simple measure of how "wide" the stereo signal is.

    Mono sound => S = 0 => D = 0 / (0 + M) = 0
    Sound only on one channel (L) => S = L, M = L => D = L / (L + l) = 0.5
    Same sound on L and R, but out of phase (anti-mono, R = -L) => S = 2L, M = 0 => D = 2L / (2L + 0) = 1.

    So, "wideness" index D varies between 0 and 1. We have also target "wideness" Dt which is calculated:

    Dt = ( 1 - 10^(-target ILD / 20) ) / 2

    For example the target ILD at low frequences being 3 dB, Dt = (1-10^(-3/20))/2 = 0.15. If a recording has D = 0.4 for example, the needed crossfeed level to reach the target is

    beta = (D - Dt) / (D + Dt - 2*D*Dt) = (0.4-0.15) / (0.4+0.15-2*0.4*0.15) = 0.5814
    crossfeed level = 20*log10 (beta) = 20*log10 (0.5814) = -4.7 dB.

    This is very simple analyse and Gregorio will again say I "ignore" a million things as I do. However, to me this analyse does correlete pretty well with my ears and you can always finetune the target Dt to your perception. These calculations should of course be done to lowpass filtered signals (for example 800 Hz). One can do the analyse for example for octave bands as I do to see what's happening in my low end ILD-wise when mixing my own music (to create omnistereophonic sound). In prectise I calculate D this way:

    D = S / (S + M + epsilon),

    where epsilon is a very small number, like 0.00001 to avoid 0 / 0 calculation whenever S + M = 0. D itself is a signal telling the instantaneous "wideness" of the signal sample by sample so it's it should be averaged for a general D.
    ironmine likes this.
  4. gregorio
    1. I have never disputed that there is a perceptual basis for some people liking crossfeed, in fact quite the opposite. However, there are is also a perceptual basis for some/many people not liking it.

    2. But that is not the issue. 71dB effectively asserts/implies that those who don't like crossfeed don't like it because they are uneducated and/or lack the listening skills/experience. But I agree, he has "gotten better", these insulting assertions/implications are relatively "better" than previously, when he just said we were ignorant, idiots and various other overt insults.

    3. No I don't, it's more complex than that, which I've explained!

    4. And what's the point of repeating it over and over, when I and everyone else have already acknowledged this, and not just once but going way back nearer the beginning of this thread roughly two years ago?

    5. You have that backwards! This accusation (and the points above) indicate you haven't even read this thread before casting your aspersions! I won't be bullied and will push back strongly against anyone who tries, can't you tell the difference? If you can, why are you casting your aspersions against me, instead of 71dB?

    6. In this sub-forum there is a "correct side" vs "incorrect side"! The "incorrect side" being the promotion of personal preferences/perception as objective facts. Have you read any of this thread? Besides effectively stating you believe I'm incompetent, you did NOT answer a single question I asked, why is that?
    6a. That's not the only "but" or even a particularly relevant "but". The relevant "but" is that 71dB refuses to acknowledge ANY "valid point" (made by anyone) that contradicts his belief that his personal perception/preferences define the objective facts/science. This is my "favourite exploit", NOT his language skills as you FALSELY state!

    7. This too is a lie! I've made "my goal" perfectly clear; that 71dB stop posting his false assertions of objective fact. I would FAR rather he did that WITHOUT a "fight in the playground" but if he continues to make those false assertions, I will continue to refute them. Have you read ANY of this thread? And, your response appears to be an attempt to start a new "fight in the playground" yourself, is that "your goal"?

  5. 71 dB
    That is one fine post of insight of what has happened here! Totally fair to me in my opinion. :)
  6. castleofargh Contributor
    can't argue that gregorio has a hard time posting without attacking others. he would be so much cooler if he could stop that. something I've observed in this hobby, if you give it enough time, I believe that a guy with relevant knowledge(in a given domain) will at some point start keeping it to himself instead of trying to explain why some posts are wrong, or he will get banned for blowing a fuse. either one. people who bother addressing mistakes on a regular basis and remain zen anyway, they're like mythical creatures in this hobby.

    can't argue with you when it comes to making claims, whoever makes them. I also would like more supporting evidence/studies/ proper demonstrations for the sake of people who read the topics, and simply to maintain as you say an evidence based conversation. but in this particular case the facts about acoustic or psychoacoustics aren't that significant(strangely enough), because what we contest are mostly @71 dB flawed rational and empty claim about factual improvement. it's mostly about the basics of experimenting and reaching a conclusion at the end if we happen to get conclusive circumstances, instead of coming up with a conclusion we like, and then making up the argument that agrees with it by cherry picking everything.
    let's be clear, @71 dB is the only one who could settle this matter, he's the root of this particular problem as he's the one who stubbornly kept posting the same false stuff and empty claim almost word for word, despite our best and many efforts to point them out and make him understand. we gave him many opportunities, he didn't even have to admit to being wrong(as some people would rather die), we suggested to just agree that crossfeed was a subjective tool(which it is) having subjective and different results on different listeners(which it does). to no avail, he must claim improvement at more than his subjective level.
    so let's not make @gregorio the big bad wolf oppressing well informed good guy @71 dB. even if greg is no angel, all that mess isn't his fault. if @71 dB had stopped acting like he's crossfeed's mum, always finding excuses to its flaws, always idealizing its good side, fighting whoever attacked the kid like only a mother would, we could have avoided about 2 years of erroneous spam and the many corresponding and often pissed off replies.let's not mistake a consequence with the actual cause.

    just a few of @71 dB's best hits:
    - insisting on some idea that there is a proper or accurate spatiality when discussing playback of stereo albums. unless we first pick a bunch of more or less arbitrary references, that concept doesn't exist. nothing in the recording, mixing, mastering and playback process focuses on preserving accurate localization cues. plus he kept writing it in the context of crossfeed, which is obviously not accurate or proper anything.
    - the fallacy of considering a badly oversimplified EQ as being the listener's ILD. our brain usually knows better, so pretending otherwise only leads to using a faulty rational where we assume that this EQ has the psychoacoustic properties and role of the listener's own ILD. which obviously would need to be confirmed for many listeners before going further, and would most likely be proved not to be the case for all but the luckiest of listeners.
    - describing crossfeed as an acoustic model(not existing in reality!) to justify calling it an objective approach. and then make the claim of objective improvement. that made up model allows to guess stuff or make hypotheses. but drawing conclusions from that fake acoustic model and calling them objective...

    from a more empirical approach, the fact that different people get different impressions from using a given crossfeed setup, and that not many people like it or wish to keep using it, is IMO, evidence that crossfeed isn't the obvious and very factual improvement 'mother' claims it to be.
  7. 71 dB
    Maybe I should not, I am kind of forced to defend myself here clarifying things:

    1 - Proper spatiality in this context doesn't mean "accurate localization cues". It means spatiality with parameters that seem accurate/rational/natural. Whatever the accurate localization cues are, my ears expect ILD of about 3 dB at low frequencies and that is what my ears get when I listen to speakers. Headphones without crossfeed may give 10 dB, 30 dB… completely off, not even close to 3 dB. I think I have explained this 100 times…

    2 - Yes, crossfeed EQ is not listener HRTF, but the way I think the system, headphones without crossfeed = EQ that blocks all frequences. Crossfeed EQ is MUCH closer to HRTF than a filter that blocks everything. That's why crossfeed EQ, oversimplified as it is is superior to no crossfeed. All of this is because people can't see no crossfeed a process too, a process much worse than crossfeed because it even further of HRTF. I think I have explained this 100 times…

    3 - To measure objective improvement we need models to simulate hearing and I'm sure those models find crossfeed beneficial. The reduction of ILD to it's "natural" levels is so huge it must mean improvement objectively using any reasonable model for human hearing. Any objective model saying huge ILD at low frqequencies is no problem is completely useless.
  8. bigshot
    I got a set of AirPod Pros today and I've been playing with them. I've never had headphones that poke in my ear before. These are comfortable and I can almost forget I'm wearing them. The noise cancellation is amazing, but when it's turned on, I am stone deaf to anything happening outside of my own head. They seal my ears snugly, so it isn't much better with noise cancellation turned off. But it has an interesting passthrough feature, which uses the microphone in the AirPods to channel ambient sound in. It's a very odd effect because even though it is stereo and the mike is roughly in the position of my ears, it strips off all of the distance cues. The ambient sounds all are contained within my head. The spatiality is all boxed up. Even turning my head is different. It is very hard to determine direction and distance with the AirPods. Spatiality is a lot more than just cross feed between channels.
  9. ironmine
    Is there any VST plugin who can do this analysis automatically? Can one of Mastering The Mix plugins do it?

    Ideally, I want to end up with this scheme:
    1. I download an album that I want to listen via headphones.
    2. I run the album quickly through the monitoring plugin which gives me a value, or at least some sort of understanding, how much crossfeed effect to apply to the album.
    3. I convert the album from flac to flac while applying crossfeed with the recommended value.
  10. bigshot
    You might want to scan this thread, Ironmine. There's info on a lot of DSPs and how to get them working. https://www.head-fi.org/threads/the-dsp-rolling-how-to-thread.867258/ There's no reason you can't get a real time DSP to do it on a comp. And anything that can be done real time, can be bounced down if you have a sound editing app. There may be a pro plugin you can use. Gregorio would know about that.
    Last edited: Nov 12, 2019
  11. castleofargh Contributor
    yeah yeah. your ears, you preferences of what feels natural to you, you expectations, your cherry picked model, your half truth conclusions, therefore it's an objective improvement... is it just another semantic issue where when you write objective, you mean subjective? that would make so much sense TBH.
    I've given you the same fallacious argument you give but instead of cherry picking ILD, I picked some room reverb. which is also part of giving "spatial whatever impressions", except reverb is even more natural as it also occurs outside of speaker playback, we always have some in our lives. what could possibly be more natural? but then pretty much nobody goes on to add some fixed reverb and only reverb while listening to headphones. because of course it feels weird. and that's also how crossfeed feels to many listeners. the brain works with all the cues and all the expectations and all the preconceptions. giving it one variable from a completely different playback experience is more likely to feel weird than it is to feel more natural. it's pretty obvious and what people would tell you if you ever cared to ask instead of trying to push your world view onto them.

    suddenly the recording does matter again and you're not just pretending that crossfeed is the objective compensation to simulate ILD and ITD from virtual speakers at about 30° on each side. funny how you just shot yourself in the foot.
    silly me had the idea that once you've placed your speakers in a room, you play all albums on them the same way. guess I've been doing it wrong all this time.
  12. castleofargh Contributor
    nah, nothing there would be what he looks for. not convinced it exists at all, but then again there are so many guys out there coding some pretty crazy stuff, so who knows?
  13. 71 dB
    Sorry, I'm not aware of such VST.
    I recommend using your ears to decide the proper crossfeed level.
    That way you get what's best for your ears and I think this is easy when you get the hang of it.
  14. 71 dB
    1. To me this is clear. I believe there is objective justification for crossfeed. My ears just happen to agree. Speakers give small ILD at low freq, headphones give large ILD, so how surprising is it lowering ILD is an improvent? My claim is not that crossfeed does stellar job (it doesn't). My claim is that headphone spatiality is so totally off (ILD at low freq. not even close to what it should be) and doing just "something" can be an improvement. Crossfeed isn't just "something", it's coarse simulation of acoustic crossfeed and so to me it's not surprising it can improve headphone sound. It's about the rule of diminishing returns. At first it's easy to improve headphone spatiality, but it becomes harder and harder so that improving headphone spatiality further from crossfeed is hard.

    2. So, you say we can ONLY fix ILD if we also fix reverberation? Or we should fix nothing, because we are used to the sound as it is? This has nothing to do with objectivity. It's about your subjective preference not to learn alternative ILD-fixed headphone sound. To me this is easy. The recordings are mixed for speakers so it's not surprising the spatiality is off with headphones. Headphones lack ER and reverberation, but recordings tend to have some reverberation. When I listen to a CD of church organ music, the T60 time of the church is 10 times that of my listening room, so listening to it with speakers increase reverberation only by 10 % and the amount of reverberation with headphones is almost the same. Dry sounds dry on headphones, but I kind of like that "precision", but that's me. You may disagree.

    3. What are you talking about? Suddenly what? How is explaining objective methods making this LESS objective? First of all recordings are mixed differently so there are different levels of ILD from record to record. Some recordings have so low ILD they don't need crossfeed at all. Since crossfeed isn't perfect, one should use the lowest possible crossfeed level to get rid of excessive ILD. Listening room is ILD-regulator. No matter what recording you play (mono, ping pong…) you get about 0-3 dB of ILD at low freq. Selecting crossfeed level is same kind of regulation. You could change the acoustics of your room to optimaze it for each recordings, but you don't do it because it's too much work, it's totally impractical. So you listen everything using the same speaker set up and acoustics. Crossfeeders can have simple selection/adjustment of crossfeed level so you have the opportunity to easily optimaze it. So much for me shooting my foot!
  15. 71 dB
    Unfortunately I'm bad at coding. I don't know how to code VST plugins. I have my Nyquist plugins, that's it…..:robot:
90 91 92 93 94 95 96 97 98 99
101 102 103 104 105

Share This Page