Skip to main content
Topic: Crossfeed - The next generation (Read 12437 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Crossfeed - The next generation

Reply #25
Right so its just a matter of how much width to apply. I agree that at 1.65 tonality changes but it usually doesn't detract from my music (not test signals). From 1.05 ~ 1.20 the spatial effect can be boosted transparently more less.  Also the tonality WILL change to some extent because you cannot have it both ways. Same with EQ tonality will change but that doesn't have to be bad in real world listening.

The mono-center is when vocals / bass  sound too mono or muddy . Sometimes crossfeed makes it worse at least when using the higher settings. Sometimes the recording itself is to blame. Speakers make it more natural. Isn't the phantom center like we listen in real life ?
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #26
May I ask why you don't just increase the delay and crossfeed level in bs2b?

Vocals are supposed to be located in the center or slightly off center. Large stereo width just smears things all over the place.
"I hear it when I see it."

Crossfeed - The next generation

Reply #27
May I ask why you don't just increase the delay and crossfeed level in bs2b?

Vocals are supposed to be located in the center or slightly off center. Large stereo width just smears things all over the place.


Yes around the center but not over-dominant or mono sounding. With some added width they are still in the center but more like speakers 6ft apart - more 3D like.  Is Width=1.00 really correct ?

Now to add more crossfeed makes things deeper and more 3D - but makes the sides too narrow and some recordings too mono.
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #28
I find some tracks are unlistenable. Danzig- Brand new god. Its recorded mono like but in the car and on speakers sounds okay. On headphones no DSP (width=1.00) its really tough . Going up to just 1.04 makes a difference.
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #29
For a  'transparent' setting I arrived at bs2b -  J M + width = 1.15 **

Unfortunately it seems I need a big stereo width to get the sound out of my head.  My original setting of bs2b - CM + width = 1.65 does it.  It needs to be at least 1.20


** updated from 1.15 to 1.20 and I think it sounds really good and not in the head as much.
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #30
I have used bs2b as you suggested followed by the channel mixer to boost the side channel (stereo width 1.65) and the result is a mess imo.
There stereo image is all over the place and the end result is more out of phase than even the original stereo track.



Thanks for testing. There could be several things:

- I still use foobar 1.03. Maybe the DSP behave differently like bs2b isn't working with channel mixer. This can be checked by removing bs2b from the chain. On my setup it has to be the first one in the chain. If its not enabled then the stereo field is totally off as you describe.

- Stereo-width value is too much for the given setup - Try a lower one like 1.05 ~ 1.30

- Headphone needs EQ in the mid treble. This is the worst and most complex issue I faced thus far. If treble has spikes and not smooth then the sound will have a rough bombastic character.  You have to lose a bit of detail to  gain a better stereo image.


The only real reason to be concerned with crossfeed issues would be for headphone listening.

Crossfeed - The next generation

Reply #31
Thank you for pointing to bs2b. It was new to me, and it is quite good at compressing the often too wide stereo panorama for headphones. Can be used on a regular basis though sometimes the sound changes a bit to the less favorable side IMO. Depends on the music, and not always is bs2b needed for reasons of unnatural stereo panorama.

TUsing bs2b the sound feeling is still coming from inside the head (also when advancing/lowering the cans a bit). So I understand your usage of the channel_mixer. When upmixing to 4 channels a 3d effect drops in. However in order to keep a more or less natural sound I had to stay with stereo image width around 1.10, and use upmix mode 'copy'. Stereo image gets narrower by the way by this procedure. (It gets wider when not upmixing, but I can't see an advantage in using the channel_mixer this way).
lame3995o -Q1.7
opus --bitrate 140

Crossfeed - The next generation

Reply #32
Thanks for the feedback Halb27. I wonder if you could test the delay option @ 2M for front and rear channels. In my experience it enhances the 3D effect even more at the slight cost of detail.

I have tested a smaller widths like 1.10 but the sound is in my head. Adding delay + 2M rectifies this to a good degree. I still  needed width @ 1.20 to get the desired effect.
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #33
I found that the C Moy preset works very well with the mixer. Its already pretty good by itself but its slightly veiled character adds to the 3D effect IMO.

I tested an 'intermediate' setting: bs2b CM + Width= 1.20 ~ 1.25 + delay = 2m + CH=4 + Upmix=copy

The center is still pretty tight IMO. Using only bs2b CM and the sound is smaller and is back inside my head.
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #34
I also found that setting differing delay values for front / back adds a layer mystique and veiling that makes it more realistic.

Front L/R = 2.00
Rear L/R = 0.75
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #35
I tried your delay settings, but this is not for me. While I can see the effect I even heard artifacts and the sound is very muffled (when having bs2b as the first in the DSP chain). Lower values improved for me while still having the effect to a certain degree. But anyway: not my taste.
So I stick with width=1.10 upmix 4 copy no delay though you're right: sound is still 'in the head'.
Out of interest I put away with bs2b, and things are still good for me.
I reloaded bs2b, this time using it as the second tool. I was surprised that to me this sounds better. I use bs2b's default setting because otherwise stereo width is too narrow for my taste.
With this DSP order when using a delay sound isn't so muffled any more, but it's still not my taste - not mentioning the artifacts.

I try to be conservative because I want to have one universal setting with which the headphone experience is improved while keeping general sound quality as close as possible to 100%. I don't like to bother whether things get significantly worse on occasion when using such a procedure.
lame3995o -Q1.7
opus --bitrate 140

Crossfeed - The next generation

Reply #36
After listening a lot I made up my mind not to use the channel-mixer. I prefer a sound closer to the original. I also tried several crossfeeders and found that the Meyer crossfeed plugin, crossfeed level 2, works best for me.
Keeps the spectrum closer to the original than bs2b, and sufficiently keeps from suffering headphone-inadequate stereo panorama. Changes the soundstage slightly, usually to the more favorable side for me and the kind of music I usally listen to.

After all it's a matter of taste what to use.
lame3995o -Q1.7
opus --bitrate 140

Crossfeed - The next generation

Reply #37
Thanks for testing . I tried a lot of setting and the hardest is to choose a universal-one-fits-all setting but that should be the goal. Interestingly everything I tried inc crossfeed only, no crossfeed + W 1.04 ~ 1.10 + 2ch / 4ch, no crossfeed + 4ch only , yielded better result than using NOTHING at all !. 

ATM i am using bs2b CM + 4ch + W=1.20..1.25

Updated the 'high' setting to W=1.60
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #38
Does anyone know an easier way to do "binaural" convolution than with the matrix mixer and gapless convolver plugins? If not, I can explain how it's done. That way you can use HRTFs.
"I hear it when I see it."


Crossfeed - The next generation

Reply #40
There was it seems.
"I hear it when I see it."

Crossfeed - The next generation

Reply #41
Robertina uploaded the binary here and I can upload its sources if anybody wants

Crossfeed - The next generation

Reply #42
Yeah: that's the best way to get the sound out of the head IMO. I haven't tested all of the files from the IRCAM page yet, but so far individual 1020's wav file does the job fairly well for me.
The subjective frequency response however also is pretty different from the original, so I will try some equalizing to bring back brilliance.
lame3995o -Q1.7
opus --bitrate 140

Crossfeed - The next generation

Reply #43
Also, which headphones are you using?  I find that crossfeed produces a more noticeable effect with closed headphones, but is very subtle with my HD800's.  If ambient noise and disturbing others isn't s concern, you might try some open-backed cans.

Crossfeed - The next generation

Reply #44
I've finished testing all the files from the IRCAM page and found individual 1031's data to fit my hearing best.
I've been doing quite intensive listening tests with my B&W P5 (the only closed one), Grado SR80e, Philips Fidelio X1 and STAX SR 202 cans.
From scratch the 3d experience is wonderful while being realistic. After listening for a while this way and then switching the convolver off flattens the soundstage making you immediately wish to reuse the convolver.
To me the closed P5 didn't have an advantage in this respect.
The bad new is as mentioned that the frequency balance changes. The drawback isn't huge with the Philips and the STAX, but for the B&W and the Grado things are getting bad here. So as shadowking mentioned already for his approach equalizing is necessary for at least some cans.
lame3995o -Q1.7
opus --bitrate 140

Crossfeed - The next generation

Reply #45
I think the bauer may work better with lower cut off HZ for some ears or heads. I used 360hz / 8db and its sounds more natural and still works on hard panned material. I combined it with width=1.25 upmix=4ch copy for a better spatial 3D effect. The tonality is much more natural than W=1.60 .I checked if the panning effect of bauer was still working and it seems that its fine when combined with more modest width values. So double win in a way. I am pleased with this setting -  it either enhances the sound or does almost nothing when the others were at times too wide or narrow.

New setting: bs2b 360hz / 8db , ch mixer W=1.25, upmix=4ch copy
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #46
Does anyone know an easier way to do "binaural" convolution than with the matrix mixer and gapless convolver plugins? If not, I can explain how it's done. That way you can use HRTFs.

Please explain.  Does it work as well as the test file near the end of this page?  I've heard a few binaural recordings, but none created the cognitive dissonance I get from this sample:

http://www.audiocheck.net/soundtests_headphones.php

Maybe it's just the extreme stereo separation and simplicity of the sound effect that makes it so convincing.

Crossfeed - The next generation

Reply #47
No, that's something different. The simple convolution I was talking about is using e.g. +/- 30° HRTFs, where speakers would be located. That's what the stereo convolver does.
To move the sound source around you'd have to change HRTFs.

Also, naturally there is quite some crosstalk. The head acts as a shadow only at high frequencies and even there the attenuation is not great (~20 dB) and it's as low as <1 dB at low frequencies. You only get high separation if e.g. someone is whispering into one of your ears.
"I hear it when I see it."

Crossfeed - The next generation

Reply #48
I would like some feedback on :  bs2b - 360hz / 8db + channel mixer -  stereo-Width=1.25, upmix=4ch copy

So far I cannot find any regressions on my music , switching it off always resulted in more fatigue / worst quality. The aim is near transparency in relation to the original "set it and forget its there" [effect is milder than most default crossfeed settings IMO] yet improved listening quality.
wavpack 4.8 -b3hx4c

Crossfeed - The next generation

Reply #49
I've also observed an interesting phenomenon that staring at speakers or a TV across the room while using headphone + keeping head still reinforces a positive 'illusion' and thus results in better impression of stereo image. Staring at the PC screen isn't as good.
wavpack 4.8 -b3hx4c

 
SimplePortal 1.0.0 RC1 © 2008-2019