[FlavioWolff]

I have read almost half of this thread and used the search as well, but I haven’t found any opinions on Roon’s dsp crossfeed. Weird considering it’s popularity. It has a “default” setting based on something about Bauer (yeah, I’m that ignorant), cmoy and Meier presets and also custom. Works wonderfully for me.

what do you most knowledgeable folks think about it?

 

[bigshot]

I would think that to judge a good crossfeed from a bad one, you would be looking at the flexibility of the settings. There is no one-size-fits-all setting. You would need to be able to have it be as flexible as possible to be useful.

 

[FlavioWolff]

I would think that to judge a good crossfeed from a bad one, you would be looking at the flexibility of the settings. There is no one-size-fits-all setting. You would need to be able to have it be as flexible as possible to be useful.

This is the implementation on roon: http://bs2b.sourceforge.net/
The custom settings lets you select Cut Frequency from 300 Hz to 2 KHz and "feed level" from 1 to 15 dB.
Seems very flexible for a "mainstream" product!

 

[castleofargh]

This is the implementation on roon: http://bs2b.sourceforge.net/
The custom settings lets you select Cut Frequency from 300 Hz to 2 KHz and "feed level" from 1 to 15 dB.
Seems very flexible for a "mainstream" product!

I was going to point to this. It's quite popular, and has been available as a free VST for years(so used well outside of Roon). In foobar I've used that a lot along with Xnor's Xfeed VST. As far as standalone crossfeed is concerned, they do the job.
The more recent and often non free offers tend to try and go beyond crossfeed, adding extra variables like room reverb or some more specific HRTF models. Ultimately you try and settle on what happens to feel nice to you.

 

[71 dB]

The custom settings lets you select Cut Frequency from 300 Hz to 2 KHz and "feed level" from 1 to 15 dB.

At 300 Hz the delay for the crossfed signal is about the same as for sounds that come from sides, 90° angle. This is "wide-crossfeed" and in my opinion some material may work well with it, but the miniature soundstage I hear is shaped very left to right and there is no feel of depth. Crossfeed level of -3 dB works nicely (crossfeed level is always negative because the crossfed signal is always quieter than the direct sound) because human head shadows low frequencies about that much when the sound comes from sides. The sound doesn't feel mono-like to me, because the delay of crossfed signal is large (about 700 µs) and crossfeeding stops at pretty low, in fact the lack of crossfeeding around 500-1000 kHz can be a small problem for me.

At 800 Hz we have the "standard crossfeed." The delay (about 250 µs) of crossfed signal simulates the delay with speakers in a room, 30° angle. At least for me this gives a miniature soundstage that is most similar to the soundstage with speakers and it feels quite natural in my opinion. There are depth cues for my spatial hearing. Crossfeed level is determined by how large stereo separation the material has: In my opinion some material don't need crossfeeding at all, some need just a little "finetuning" (say crossfeed level -12 dB) and some recordings with very poor spatiality requires crossfeed level -1 dB.

At 2 kHz the simulated angle of the sound has dropped to about 10° and use of higher crossfeed level can make the sound mono-like in my opinion. The shadow effect of human head at 1 kHz is hardly 10 dB, so I'd recommend using only the lower values of crossfeed level, -15 … -8 dB. This means that the ILD at low frequencies of the recording should be limited.

 

[gregorio]

... This gives a miniature soundstage that is most similar to the soundstage with speakers and it feels quite natural. There are depth ques. ...

Oh no, not AGAIN!!!!

You really mean that "it feels quite natural" TO YOU and that YOU perceive depth cues, not that it is actually "quite natural" and there are actually "depth ques" (or better depth cues)! Are you really going to go round and round this same circle yet again?

G

 

[71 dB]

Oh no, not AGAIN!!!!

You really mean that "it feels quite natural" TO YOU and that YOU perceive depth cues, not that it is actually "quite natural" and there are actually "depth ques" (or better depth cues)! Are you really going to go round and round this same circle yet again?

G

I don't care about your remarks anymore, but I did "subjectify" my post a little bit so maybe you are happier now.

 

[FlavioWolff]

Could someone please indicate me a track that supposedly doesn’t “need” crossfeed, so I can see if, for my tastes, the crossfeed good or bad on it?
According to Meier’s old article, crossfeed should be desirable even when there isn’t much stereo panning, because of fatigue.

 

[gimmeheadroom]

Most tracks don't need it. It's a correction for bad recordings.

 

[FlavioWolff]

At 300 Hz the delay for the crossfed signal is about the same as for sounds that come from sides, 90° angle. This is "wide-crossfeed" and in my opinion some material may work well with it, but the miniature soundstage I hear is shaped very left to right and there is no feel of depth. Crossfeed level of -3 dB works nicely (crossfeed level is always negative because the crossfed signal is always quieter than the direct sound) because human head shadows low frequencies about that much when the sound comes from sides. The sound doesn't feel mono-like to me, because the delay of crossfed signal is large (about 700 µs) and crossfeeding stops at pretty low, in fact the lack of crossfeeding around 500-1000 kHz can be a small problem for me.

At 800 Hz we have the "standard crossfeed." The delay (about 250 µs) of crossfed signal simulates the delay with speakers in a room, 30° angle. At least for me this gives a miniature soundstage that is most similar to the soundstage with speakers and it feels quite natural in my opinion. There are depth cues for my spatial hearing. Crossfeed level is determined by how large stereo separation the material has: In my opinion some material don't need crossfeeding at all, some need just a little "finetuning" (say crossfeed level -12 dB) and some recordings with very poor spatiality requires crossfeed level -1 dB.

At 2 kHz the simulated angle of the sound has dropped to about 10° and use of higher crossfeed level can make the sound mono-like in my opinion. The shadow effect of human head at 1 kHz is hardly 10 dB, so I'd recommend using only the lower values of crossfeed level, -15 … -8 dB. This means that the ILD at low frequencies of the recording should be limited.

Thank you for the explanation.
Indeed, the "1-15db feed value" on roon seems to be negative. When I set to "15" db, there is much less crossfeed applied than "1db".
Black sabbath's first albums are so much better with xfeed! (i don't need to state that this is subjective evaluation, do I?)

 

[71 dB]

Could someone please indicate me a track that supposedly doesn’t “need” crossfeed, so I can see if, for my tastes, the crossfeed good or bad on it?
According to Meier’s old article, crossfeed should be desirable even when there isn’t much stereo panning, because of fatigue.

Electronic music example:


Baroque music example:


I don't use crossfeed with these. No need. Crossfeed can only make these worse in my opinion. The Graupner is in my opinion an awesome example of how to get headphone spatiality right. The ILD levels are so correct. The recording works very well on loudspeakers also, so it's what I call an omnistereophonic recording.

I use crossfeed on almost everything. In my opinion about 98 % of all stereophonic recordings benefit from some crossfeeding, but the rest 2 % should be left alone as crossfeed can only make them worse. So, I disagree with Meier 2 % of the time.

 

[71 dB]

Thank you for the explanation.
Indeed, the "1-15db feed value" on roon seems to be negative. When I set to "15" db, there is much less crossfeed applied than "1db".
Black sabbath's first albums are so much better with xfeed! (i don't need to state that this is subjective evaluation, do I?)

No problem!

It's about how you define it, but the logical and widely used way is to compare the crossfed signal to the direct signal and since it's always quieter, the dB value for crossfeed is negative. So, yes "15" means less crossfeeding than "1". I'm not into Black Sabbath myself, but I listened to "Black Sabbath" from 1970 and yes, it benefits a lot from strong crossfeeding at level -1 dB in my opinion. This is very typical for this kind of music from that era. The channel separation/spatiality is quite uncontrolled and harsh for headphone listening.

 

[71 dB]

Most tracks don't need it. It's a correction for bad recordings.

Well, something is by definition bad if correction is needed. It's about our experiences of whether the recording is bad or good and if bad how bad? My spatial hearing works so that most recordings benefit from crossfeed and there is no mystery in it in my opinion since recording are mixed primarily for speakers so that low frequency ILD is easily more than a few decibels.

 

[gregorio]

[1] The Graupner is in my opinion an awesome example of how to get headphone spatiality right. The ILD levels are so correct. The recording works very well on loudspeakers also, so it's what I call an omnistereophonic recording.
[2] I use crossfeed on almost everything. In my opinion about 98 % of all stereophonic recordings benefit from some crossfeeding, but the rest 2 % should be left alone as crossfeed can only make them worse.

1. On the other hand, that is a particular recording situation which is effectively mono (just the reverb/acoustics are stereo). However, this isn't the case with a lot of other baroque music, antiphonal renaissance/baroque being a particularly obvious example. Making such an antiphonal piece "omnistereophonic" as you call it, would be a serious error of judgement/musicality! Also, the vast majority of original club mixes are mono (or very close to mono) and with vinyl, which is the case in your example, the bass freqs have to be mono.

2. If we're talking pure perception/opinion, then my opinion is that about 2% of all stereophonic recordings benefit from crossfeed and the other 98% should be left alone as crossfeed can only make them worse!

[1] Well, something is by definition bad if correction is needed. It's about our experiences of whether the recording is bad or good and if bad how bad?
[2] My spatial hearing works so that most recordings benefit from crossfeed and there is no mystery in it in my opinion since recording are mixed primarily for speakers so that low frequency ILD is easily more than a few decibels.

1. Obviously that's not true! For example, there are more than a few bass-heads out there, for whom just about all recordings need correction (additional bass). Are just about all recordings therefore "bad" or is it just a case of their particular perception/preference? Probably no recordings exist that someone, somewhere doesn't think needs correction and therefore, according to your logic, all recordings must be "bad".

2. OK, if we're again going with personal perception/preference rather than the facts/science: My spatial hearing works so that most recordings do not benefit from crossfeed and there is no mystery in it, since crossfeed does not emulate my experience of listening to speakers and doesn't even claim to.

This is largely why we have a Sound Science subforum in the first place, so that we're not just arguing between different individuals' impressions/preferences/perceptions.

G

 

[71 dB]

1. On the other hand, that is a particular recording situation which is effectively mono (just the reverb/acoustics are stereo). However, this isn't the case with a lot of other baroque music, antiphonal renaissance/baroque being a particularly obvious example. Making such an antiphonal piece "omnistereophonic" as you call it, would be a serious error of judgement/musicality! Also, the vast majority of original club mixes are mono (or very close to mono) and with vinyl, which is the case in your example, the bass freqs have to be mono.

2. If we're talking pure perception/opinion, then my opinion is that about 2% of all stereophonic recordings benefit from crossfeed and the other 98% should be left alone as crossfeed can only make them worse!

3. Obviously that's not true! For example, there are more than a few bass-heads out there, for whom just about all recordings need correction (additional bass). Are just about all recordings therefore "bad" or is it just a case of their particular perception/preference? Probably no recordings exist that someone, somewhere doesn't think needs correction and therefore, according to your logic, all recordings must be "bad".

4. OK, if we're again going with personal perception/preference rather than the facts/science: My spatial hearing works so that most recordings do not benefit from crossfeed and there is no mystery in it, since crossfeed does not emulate my experience of listening to speakers and doesn't even claim to.

This is largely why we have a Sound Science subforum in the first place, so that we're not just arguing between different individuals' impressions/preferences/perceptions.

G

1. Yes, the harpsichord is a "mono-like" sound source unless the mic(s) is near the instrument compared to the dimension of the instrument in which case the mic(s) doesn't see a point sound source. Also, unless the instrument is exacly in the middle, there will be some ILD, ISD and ITD generated, but yes the result is from the instrument alone is pretty mono-like. The acoustics of course are an essential part of the sound: We don't want a recording done inside an anechoic chamber. We want to record the instrument in good acoustics. Of course omnistereophonic recordings become more challenging when the amount of instruments/singers increase. It doesn't mean it's an error of judgement/musicality. That's just something we are not able to do with our knowledge and recording technology (using a Jecklin/Schneider disk could work maybe), but the real reason in my opinion is that hardly nobody cares. As long as the spatiality is good on speakers only people like me worry about spatiality with headphones. It seems these rare omnistereophonic recordings are just "happy accidents."

The bass indeed has to be mono or at least near mono on vinyl, but the track has pretty much same spatiality on CD. It doesn't matter why the sound is mono-like at lower frequencies. Crossfeed is not needed when it is. To my experience this kind of club music was become less mono-like and newer stuff often needs crossfeed. If not because of bass (below 200 Hz) then because of 200-800 Hz range having too much ILD.

2. I know your opinion. It's pointless for me to argue about it. It's your opinion. You are the one who "controls/owns" it.

3. In my opinion the (allowed/natural) parameters of spatiality come from things like HRTF and are set. You have to modify your body/head to change HRTF. The level of bass is not like that. There is no "set" value for how much bass is correct. It's an artistic choice. Music is mixed for speakers. Headphone spatiality is by default wrong. Sometimes wrong in a lucky way so it doesn't matter. Most of the time just wrong and I use crossfeed to address that. Room REGULATES spatiality so it's ALWAYS within reasonable natural values no matter how the recording is done, but headphones don't regulate anything so it's wild, ILD can be anything. Crossfeed is a regulator limiting what ILD can be. If you crossfeed at level -6 dB, that' the largest ILD you can have no matter what recording you play.

4. You have told me this many times. Crossfeed is by nature about our individual perceptions and that's why opinions about it differ so much. There clearly isn't truth. That's what I have realized after all this time on this board. I have my opinion about crossfeed and I don't care anymore if your's is different. FlavioWolff is asking our opinions and I am telling him what I think. He can compare his experiences to our opinions. I know I will NEVER change your mind so I don't even try. Please don't try to change my mind. After using crossfeed for almost 8 years I know by now crossfeed is my thing and I have to live with how MY spatial hearing works. It would certainly make my life easier if 98 % of all stereo recordings didn't need crossfeed, but it is what it is.