To crossfeed or not to crossfeed? That is the question...
Nov 7, 2019 at 5:57 PM Post #1,441 of 2,146
A blank slate! A unfettered mind, open to any and all possibilities! I salute you for your great discovery... You, Columbus and Neil Armstrong! George Washington and the Cherry Tree, ripe for picking! Let loose with the fireworks!

Yes, I had to discover crossfeed because they are not advertised anywhere. I had thought headphones are totally trouble free things until one day I realized the problem of excessive spatiality and started to study headphone sound. Crossfeed wasn't the only thing I discovered. I also learned out, that headphones often have so small mechanical resistance, that in order to have adequate damping, low output impedance is needed for the headphone amp. To my horror I realized headphones are not even close trouble free. In fact headphone sound is a mess! Too high output impedance here and not enough power there and recordings are mixed for speakers… …it's a mess and I bild myself a headphone adapter to deal with this mess and it helped a lot, so much so that a speaker guy became a headphone guy. I had listened to subpar headphone sound thinking it just is how headphones sound not realizing my output impedance was ridiculously high (no adequate damping + huge FR errors) and I suffered from excessive spatiality. When I got rid of those problems I found out that headphone sound can be awesome!

:)

HA2.jpg
 
Nov 7, 2019 at 6:09 PM Post #1,442 of 2,146
You prefer crossfeed. I do not. Picking one of two solutions, neither of which addresses a multitude of inaccuracies, doesn't make either "better".

Those are the options I have. Those are also the options implied in the name of this topic. Both my understanding of spatial hearing and my ears say crossfeed is often better out of these two options, given there is excessive spatiality in the recording (almost always is because recordings are mixed for speakers). Some recordings do not have excessive spatiality and I listen to them without crossfeed. The level of needed crossfeed varies from recording to recording and for some recordings it happens to be "minus infinity dB" meaning no crossfeed.
 
Nov 7, 2019 at 6:14 PM Post #1,443 of 2,146
I choose 3) speakers in a good room

Yes you do, but this topic is "to crossfeed or not to crossfeed?" Your option 3) is outside this topic and so are HRTF convolutions etc. I feel I am the only one here honouring the name of this thread and talking about choosing between the two options of this topic.
 
Nov 7, 2019 at 7:07 PM Post #1,444 of 2,146
OK. I'll get into the spirit of your pointless ranting...

Crossed sucks. It's for old ladies and wimps.
 
Nov 8, 2019 at 12:41 AM Post #1,445 of 2,146
Being a singular person in a group is nothing new for me. I am different. I am weird. I am excentric. That makes anyone a singular person in the group.
we're all singular people. you think gregorio is an example of typical dude? or bigshot? or me and my quasi-modo costume? almost everybody here is noticeably singular in some ways. that's a very bad excuse and it really has nothing to do with your outstanding tunnel vision on this topic.

I don't know what the problem here is. Are you saying given " the larger set of objective information " and two options:

1) no crossfeed
2) crossfeed

the larger set of objective information says option 1) is better? How is that not cherry picking and ignoring the fact that headphones give larger ILD than speakers to which recordings are mixed?
back to my previous analogy with masturbation as a sex simulation:

1) my hand
2) my hand while holding a bottle of water in the other
the larger set of objective information says option 1) is better? How is that not cherry picking and ignoring the fact that people are mostly made of water?



I believe those arguments are fairly equivalent. almost anybody can see the flaws in the reasoning and will pick 1) as preferred solution under most circumstances(or even better, real sex and actual speakers). but we can always stubbornly hang on to our pseudo scientific rational and pretend that when something is beneficial in complete isolation, then it's obviously going to be beneficial in the actual system. like how a larger tank in a car let us drive longer distance, so large tanks are always an improvement. why don't we have cars with much much larger tanks?

on this you clearly confuse science with cherry picking the models and the facts. the guys who keep claiming that 24 or even 32bit files are so much better than 16, are following the same reasoning you do. yes 32bit allows for higher resolution, that one specific and very objective part is totally true. does that make 32bit files an obvious improvement? :thinking: I'd sooner agree that 32bit is an improvement, than agree to conclude that crossfeed is an obvious improvement over default headphone playback. because while irrelevant, at least with 32bit there are no unpredictable consequences on the listener's subjective experience. we can actually draw conclusions because we know how that model works. for crossfeed, we know jack crap about the consequences of having a partial, inaccurate (by how much? we don't know it's listener dependent), change applied to the sound. well what we know is that only a minority of people happen to enjoy using that on a regular basis. so we do know at least that it's not the improvement you make it to be. you deny reality in favor of your toy model of acoustic and your own subjective impressions. but want to play pretend that you're interested in facts and objective reality. that's denial.

anyway you're going to find a way to misinterpret that and stick to your ILD crap until the day you die, so I don't know why I keep replying to you on this topic.
 
Nov 8, 2019 at 2:24 AM Post #1,446 of 2,146
Yes you do, but this topic is "to crossfeed or not to crossfeed?" Your option 3) is outside this topic and so are HRTF convolutions etc. I feel I am the only one here honouring the name of this thread and talking about choosing between the two options of this topic.

Hi 71 dB,

Why do you feel that HRTF convolutions are outside the topic of crossfeed? HRTF is an advanced form of crossfeed.

https://en.wikipedia.org/wiki/Crossfeed#Digital_(DSP) :

A digital, or DSP-type, crossfeed is typically more sophisticated, mixing an amount of signal from one channel to the other, delaying the signal to mimic interaural time differences and applying other characteristics of head-related transfer functions (HRTFs) to mimic the changes between the left and right ears. Some digital crossfeeds include controls for varying the realism of the crossfeed implementation and which HRTF characteristics are used.

To this end, I have a practical question. I downloaded impulses from http://recherche.ircam.fr/equipes/salles/listen/download.html

I also installed the true stereo convolver: https://www.liquidsonics.com/software/reverberate-2/
I switched this convolver to the true stereo mode, loaded 2 (raw) impulses - and finally, everything sounds the way it should ! (the sound comes from 30 degrees and 330 degrees, if you assume that 0 is North).

My question is: In order to use these HRTF true stereo convolutions as a crossfeed, do I need to process the sound any other way additionally? Does the use of these impulses take care of ITD, ILD differences - and thus no further equalizing in the form of low pass filtering or treble boosting or arranging a delay is required? Please advise.
 
Last edited:
Nov 8, 2019 at 2:55 AM Post #1,447 of 2,146
Hi 71 dB,

Why do you feel that HRTF convolutions are outside the topic of crossfeed? HRTF is an advanced form of crossfeed.

https://en.wikipedia.org/wiki/Crossfeed#Digital_(DSP) :

A digital, or DSP-type, crossfeed is typically more sophisticated, mixing an amount of signal from one channel to the other, delaying the signal to mimic interaural time differences and applying other characteristics of head-related transfer functions (HRTFs) to mimic the changes between the left and right ears. Some digital crossfeeds include controls for varying the realism of the crossfeed implementation and which HRTF characteristics are used.

To this end, I have a practical question. I downloaded impulses from http://recherche.ircam.fr/equipes/salles/listen/download.html

I also installed the true stereo convolver: https://www.liquidsonics.com/software/reverberate-2/
I switched this convolver to the true stereo mode, loaded 2 (raw) impulses - and finally, everything sounds the way it should ! (the sound comes from 30 degrees and 330 degrees, if you assume that 0 is North).

My question is: In order to use these HRTF true stereo convolutions are a crossfeed, do I need to process the sound any other way additionally? Does the use of these impulses take care of ITD, ILD differences - and thus no further equalizing in the form of low pass filtering or treble boosting or arranging a delay is required? Please advise.
the impulses are already delayed as they should relatively to one another(otherwise you'd have a bunch of useless data regarding delays, and not the HRTF).
the compensated version accounts for some stuff they know about the rig, so it's probably a better bet. but clearly the best option is just to try and pick what works for you, as you're just looking for what subjectively comes close, and not actually using your very own measurements. so objectivity can only go so far in that approach.
you're still left with the headphone's FR. but as you've used it to determine which HRTF was closest to your hearing for frontal image, maybe we can consider that you're already good that way. kind of up to you to maybe EQ for a more theoretical neutral, then retry all the impulses to find the one that work best under those condition? might improve something, or maybe just make you waste a lot of efforts to end up with the same impulses. hard to tell.

so did you find out what was your problem the other day with the convolution schemes you had tried?
 
Nov 8, 2019 at 4:06 AM Post #1,448 of 2,146
[1] Did the inventors of crossfeed cherry pick? Did headphone amp manufacturers who incorporated croddfeed into their amps cherry pick?
[2] No, because the limits of crossfeed are understood. [2a] I don't cherry pick, because I know the limits.

1. I have no idea, possibly, but if they did, it was certainly nowhere near the extent to which you cherry-pick! In the 1950's when they invented crossfeed, stereo as a consumer product had only been released a few years earlier, it's adoption was slow and there was relatively little reliable evidence/science regarding the perception stereophony. Much/Most of the science/evidence was developed more than a decade later and continues to this day, so how could the inventors of crossfeed deliberately omit (and thereby cherry-pick) science/evidence which didn't exist at the time? If as you state, you have studied crossfeed, how is it possible you don't know even the basic time-line of it's history? Either you know far less than you think you do or your question is a ridiculous attempt to support your position, which is it?
2. Your statement is false! How any particular listener perceives crossfeed with commercial music recordings is NOT well/fully understood. Therefore ...
2a. This statement is completely backwards! You "know the limits" ONLY because you cherry-pick! Conversely, if you didn't cherry-pick then you could NOT "know the limits".

[1] I discovered crossfeed, because I thought about headphone sound and how large ILD is a problem.
[2] I used the science I have learned in university, my understanding of human spatial hearing.
[2a] You say this was cherry picking, but to me it was using science.

1. Clearly you didn't discover crossfeed, you simply learned that it had already been discovered about 60 years earlier.
2. You've ALREADY admitted that "the science you learned in university" didn't even mention music production and clearly demonstrated that you have little understanding of either the spatial information contained in commercial music recordings or how it is perceived! And OF COURSE, science is NOT defined by your personal understanding of it.
2a. You ARE using the science of spatial hearing, no one is disputing that. However, you're ONLY using the science of the spatial hearing of real/natural acoustics and even then, ONLY some parts of that science. This is pretty much the text book definition of cherry-picking!
1) no crossfeed
2) crossfeed
the larger set of objective information says option 1) is better? How is that not cherry picking and ignoring the fact that headphones give larger ILD than speakers to which recordings are mixed?
Why are you asking this question, how can you not already know, if as you state you have studied this? The "larger set of objective information" demonstrates that neither is intrinsically better, that it depends on individual perception, although it indicates that crossfeed is not better/preferred by most. Also, "the larger set of objective information" does NOT ignore the fact that headphones give larger ILD than speakers, in fact quite the opposite! If it wasn't for the reduced ILD with crossfeed then "the larger set of objective information" would (probably) demonstrate that not using crossfeed IS intrinsically better and indicate that even fewer people prefer it! So CLEARLY your question and statement are nonsense!
[1] Both my understanding of spatial hearing and my ears say crossfeed is often better out of these two options,
[2] given there is excessive spatiality in the recording (almost always is because recordings are mixed for speakers).
1. Science is NOT defined by YOUR understanding of it or by what your "ears say" and this isn't the "71dB's understanding of science and what his ears say" forum!

2. It is NOT "given" that there is excessive spatiality in the recording (when listening with headphones), you've simply made that up and it's contrary to the actual facts! Completely opposite to your FALSE assertion, recordings do NOT have too much spatiality, they have too little because if they are designed for speaker reproduction there will be the additional spatial information created by the listening environment. This is the actual fact and is simple common sense but you ignore both and repeatedly state the exact opposite of the actual facts, despite it being refuted almost as many times as you repeat it. Your assertion is not just false but clearly ludicrous, some of those who try to compensate for headphone listening do so by adding MORE spatial information (reverb) not by trying to remove/reduce what's on the recording!

Round and round you go. Hey @bfreedma, 71dB has discovered something, a perpetual motion carousel. The power of "agenda" is limitless, even Yoda would be overwhelmed! :)

G
 
Nov 8, 2019 at 5:03 AM Post #1,449 of 2,146
the impulses are already delayed as they should relatively to one another(otherwise you'd have a bunch of useless data regarding delays, and not the HRTF).
the compensated version accounts for some stuff they know about the rig, so it's probably a better bet.

As for the compensated versions, the description says:"the propagation delay is removed". Is it a different kind of delay (from the speaker to the listener perhaps?) that was removed?

you're still left with the headphone's FR. but as you've used it to determine which HRTF was closest to your hearing for frontal image, maybe we can consider that you're already good that way. kind of up to you to maybe EQ for a more theoretical neutral, then retry all the impulses to find the one that work best under those condition?

Yes, you are right, I never thought about that. I need to apply to these demo sounds first the convolution that adjusts the frequency response of my earphones that will be used for music playback. That's how I gotta choose the head that suits me most.

so did you find out what was your problem the other day with the convolution schemes you had tried?

No, I didn't! I got a reply from the guy who runs http://recherche.ircam.fr website saying that everything is correct on his website, North is 0 and the angle begins counting as we move from North to the left (90), back (180) and right (270). So the error was on my side. Maybe the VST host does not like running four instances of the same plugin at the same time. Maybe it caches them in some weird way, I don't know. But LiquidSonics works fine, I will use it for now.
 
Nov 8, 2019 at 8:09 AM Post #1,450 of 2,146
Hi 71 dB,

Why do you feel that HRTF convolutions are outside the topic of crossfeed? HRTF is an advanced form of crossfeed.

https://en.wikipedia.org/wiki/Crossfeed#Digital_(DSP) :

A digital, or DSP-type, crossfeed is typically more sophisticated, mixing an amount of signal from one channel to the other, delaying the signal to mimic interaural time differences and applying other characteristics of head-related transfer functions (HRTFs) to mimic the changes between the left and right ears. Some digital crossfeeds include controls for varying the realism of the crossfeed implementation and which HRTF characteristics are used.

To this end, I have a practical question. I downloaded impulses from http://recherche.ircam.fr/equipes/salles/listen/download.html

I also installed the true stereo convolver: https://www.liquidsonics.com/software/reverberate-2/
I switched this convolver to the true stereo mode, loaded 2 (raw) impulses - and finally, everything sounds the way it should ! (the sound comes from 30 degrees and 330 degrees, if you assume that 0 is North).

My question is: In order to use these HRTF true stereo convolutions as a crossfeed, do I need to process the sound any other way additionally? Does the use of these impulses take care of ITD, ILD differences - and thus no further equalizing in the form of low pass filtering or treble boosting or arranging a delay is required? Please advise.

Yes, HRTF is an advanced form of crossfeed. At this point I need a break from this thread sorry. So tired of this. I just don't care anymore. Maybe I leave and enjoy my life with crossfeed and restore sanity to my life. Big mistake coming here...
 
Last edited:
Nov 8, 2019 at 8:29 AM Post #1,451 of 2,146
Speakers - lot of spatial information because of acoustic crossfeed, ER and reverberation, BUT small ILD at low freq => natural spatiality
Headphones - less spatial information, BUT large ILD at low freq. => excessive spatiality.

That's HOW I define this and I DON'T CARE HOW YOU DEFINE!!!
 
Nov 8, 2019 at 8:32 AM Post #1,452 of 2,146
How can you say anything "beyond" your understanding? Gregorio can't understand EVERYTHING, he is not God, is he? How can you do anything without cherry picking? How can we improve anything? We can never …. so c
 
Nov 8, 2019 at 8:34 AM Post #1,453 of 2,146
What is improving headphone sound without cherry pick? WHAT IS IT???
I improve headphone sound for myself: It's crossfeed. Cherry picking or not, It is a huge improvement for me!!!!!!
 
Nov 8, 2019 at 8:35 AM Post #1,454 of 2,146
I like crossfeed.
I stopped caring what you think becase I don't like you!
Keep this board to yourself!
 
Nov 8, 2019 at 8:49 AM Post #1,455 of 2,146
As for the compensated versions, the description says:"the propagation delay is removed". Is it a different kind of delay (from the speaker to the listener perhaps?) that was removed?

Delay from source to ears are removed, but the SAME among is removed left and right so ITD remains.
 

Users who are viewing this thread

Back
Top