@oneplustwo
I have played with that, and was done with it within the same day. Except for the surround field option.
The room modes were a bit too obvious for me. At least compared to the sound without it. It didn't last long.
I can't compare it to my ambient experiences in any way.
I don't need any virtual sub though.. not with my house curve.
I have played with that, and was done with it within the same day. Except for the surround field option.
The room modes were a bit too obvious for me. At least compared to the sound without it. It didn't last long.
I can't compare it to my ambient experiences in any way.
I don't need any virtual sub though.. not with my house curve.
Last edited:
@Pano, Sure,
I don't use any JRiver feature or enhancement effect anymore, haven't used it for a long time. I found the mid/side EQ got me better or more precise results. The 'surround field medium enhancement' boosts the side EQ about 3 dB above the mid level if one splits Stereo into mid and side.
Here we go:
The mains get pré EQ, measured at the listening spot and FIR corrected. After that I optimise the result between left and right channel with some EQ tweaks, based on the waterfall graphs I get with FIR correction applied.
** the room has damping panels to absorb the early reflections, essential for me to be able to measure at a single spot and have it translate to measurements around that microphone position **
Next the mains get the mid-side EQ tweaks where the FR looks like this:
The light blue trace is phantom center or mid (from mid/side). The dark blue is panned hard left or panned hard right content, single channel. The green trace is for L and inverted R.
This is the first trick that helps imaging, tonality between L,R and phantom C and works wonders for intelligibility. It makes the sides have a bit more bottom end (less bright if tonality was optimized for the phantom center) while the phantom center sounds similar in tonality to the sides.
Then we move on to ambient channels. No direct sound to the listening spot, everything is reflected to arrive at lateral angles from about ~120 degree at the LP if 0 degree is straight forward.
They get the mix of (L-R) and (R-L) with a bit of phantom center mixed in. Let me show how I've set that up in JRiver:
After this step we get 6 channels, 2x main, 2x phantom center for the back channels, 2x ambient (L-R and R-L). The low pass is set at 12 dB/octave. The bottom end cut-off of the band-pass is set in the FIR correction. It is a little steeper starting at ~150 Hz. SL and SR represent a band passed, delayed and attenuated (L-R) and (R-L), just as in the links I previously showed from diymobileaudio by werewolf. The bit of phantom mix in is my own tweak based on my listening tests. Hiding the phantom stereo cross talk dips in combination with the mid/side EQ.
They enter Metaplugin and each get their own EQ throughput:
For the mains, that's where the actual mid/side EQ takes place, with a linear phase EQ.
For the center or SL and SR, I can enter separate tweaks here, to deviate from the standard curve that mimics the dark blue FR curve as seen above (except for the band pass behaviour between ~150 to respectively 5 and 4 K). This is also where the definitive levels are set for the phantom ambient part (L+R) [gain is set to 0.7 dB] and the sides ambient parts (L-R) and (R-L) [gain is set to 7.5 dB, about 6 dB louder than the phantom part]. {remember I have a 14 dB reserve (head)room to deal with the mains EQ}
The lexicon plugin is used with about a 45% mix of Random Hall, where there are way to many variables to mention which all can influence the output once again. Roughly speaking the early reverb is set at a level of -4.5 dB, the late at -7.5 dB. No predelay is set within the lexicon plugin.
After this step each of them goes trough their respective FIR filter (still having 6 channels). (the FIR filter of the center and sub is the same as SL and SR).
In the next parametric EQ tab of JRiver the Center is added to SL (left ambience) and Sub (= the other half of the phantom center) is added to SR (right ambience).
I guess this is it in a nutshell. Does that help?
Easy to follow?
What I've played with are the delay numbers you see in that JRiver filters window. The ambient speakers are a bit closer to the LP so they arrive ~3ms in front of the mains if no delay is set. The slightly differing delay figures for left or right are due to small differences in distances from the reflective surfaces to the LP out in my real room.
I don't use any JRiver feature or enhancement effect anymore, haven't used it for a long time. I found the mid/side EQ got me better or more precise results. The 'surround field medium enhancement' boosts the side EQ about 3 dB above the mid level if one splits Stereo into mid and side.
Here we go:
The mains get pré EQ, measured at the listening spot and FIR corrected. After that I optimise the result between left and right channel with some EQ tweaks, based on the waterfall graphs I get with FIR correction applied.
** the room has damping panels to absorb the early reflections, essential for me to be able to measure at a single spot and have it translate to measurements around that microphone position **
Next the mains get the mid-side EQ tweaks where the FR looks like this:
The light blue trace is phantom center or mid (from mid/side). The dark blue is panned hard left or panned hard right content, single channel. The green trace is for L and inverted R.
This is the first trick that helps imaging, tonality between L,R and phantom C and works wonders for intelligibility. It makes the sides have a bit more bottom end (less bright if tonality was optimized for the phantom center) while the phantom center sounds similar in tonality to the sides.
Then we move on to ambient channels. No direct sound to the listening spot, everything is reflected to arrive at lateral angles from about ~120 degree at the LP if 0 degree is straight forward.
They get the mix of (L-R) and (R-L) with a bit of phantom center mixed in. Let me show how I've set that up in JRiver:
After this step we get 6 channels, 2x main, 2x phantom center for the back channels, 2x ambient (L-R and R-L). The low pass is set at 12 dB/octave. The bottom end cut-off of the band-pass is set in the FIR correction. It is a little steeper starting at ~150 Hz. SL and SR represent a band passed, delayed and attenuated (L-R) and (R-L), just as in the links I previously showed from diymobileaudio by werewolf. The bit of phantom mix in is my own tweak based on my listening tests. Hiding the phantom stereo cross talk dips in combination with the mid/side EQ.
They enter Metaplugin and each get their own EQ throughput:
For the mains, that's where the actual mid/side EQ takes place, with a linear phase EQ.
For the center or SL and SR, I can enter separate tweaks here, to deviate from the standard curve that mimics the dark blue FR curve as seen above (except for the band pass behaviour between ~150 to respectively 5 and 4 K). This is also where the definitive levels are set for the phantom ambient part (L+R) [gain is set to 0.7 dB] and the sides ambient parts (L-R) and (R-L) [gain is set to 7.5 dB, about 6 dB louder than the phantom part]. {remember I have a 14 dB reserve (head)room to deal with the mains EQ}
The lexicon plugin is used with about a 45% mix of Random Hall, where there are way to many variables to mention which all can influence the output once again. Roughly speaking the early reverb is set at a level of -4.5 dB, the late at -7.5 dB. No predelay is set within the lexicon plugin.
After this step each of them goes trough their respective FIR filter (still having 6 channels). (the FIR filter of the center and sub is the same as SL and SR).
In the next parametric EQ tab of JRiver the Center is added to SL (left ambience) and Sub (= the other half of the phantom center) is added to SR (right ambience).
I guess this is it in a nutshell. Does that help?
Easy to follow?What I've played with are the delay numbers you see in that JRiver filters window. The ambient speakers are a bit closer to the LP so they arrive ~3ms in front of the mains if no delay is set. The slightly differing delay figures for left or right are due to small differences in distances from the reflective surfaces to the LP out in my real room.
Last edited:
I should note: when I add center to SL and Sub to SR it gets another 3 dB reduction.
That's how far down the phantom part is in the ambient mix.
We can't split the Stereo signal in actual left, right and center. The ambience is positive L and inverted R in one channel and positive R and inverted L in the other. Adding in that bit of L+R (which still has a piece of +L and +R in it's mix) will add a Left bias in the left ambient channel and a right bias in the right. Mix it in at a higher level and you're back at a Stereo signal.
Listening to this ambient mix on headphones can be very revealing. The more center is mixed in the wider it will sound. I determined each separate level in listening tests (not on headphones but out in the room).
For Home Theatre (I have no real center channel) I don't use the ambient mix (they just get the SL and SR info of the HT mix, delayed a bit) but change the settings in the mid/side EQ. Like the dip at 600 Hz in the main phantom (mid) part, this gets slightly more gain (thus creating a deeper dip).
It's actually easier to set it up for HT, as these mixes are more alike in tonal balance.
That's how far down the phantom part is in the ambient mix.
We can't split the Stereo signal in actual left, right and center. The ambience is positive L and inverted R in one channel and positive R and inverted L in the other. Adding in that bit of L+R (which still has a piece of +L and +R in it's mix) will add a Left bias in the left ambient channel and a right bias in the right. Mix it in at a higher level and you're back at a Stereo signal.
Listening to this ambient mix on headphones can be very revealing. The more center is mixed in the wider it will sound. I determined each separate level in listening tests (not on headphones but out in the room).
For Home Theatre (I have no real center channel) I don't use the ambient mix (they just get the SL and SR info of the HT mix, delayed a bit) but change the settings in the mid/side EQ. Like the dip at 600 Hz in the main phantom (mid) part, this gets slightly more gain (thus creating a deeper dip).
It's actually easier to set it up for HT, as these mixes are more alike in tonal balance.
A bit more info to think about here. Stereo cross talk is going to create a comb pattern at the ears. This creates dips at certain frequencies (most notable) in the absence of early reflections. In my case I could point down the biggest dip at the ear, in the sweet spot, at ~1750 Hz. This is a problem for the phantom part only. A sound that is panned hard left won't have a difference to real life perception. As it does not create the same combing pattern caused by the same material coming from the right speaker. It is mimicking the real sound more easily.
No matter what would be mixed in to the mains, does not change the perception of that (phantom) dip. Except when we use cross talk cancelation. That could actually alleviate the dip if done well. (think of the work of Prof. Edgar Choueiri)
Mixing in the room queues as JRiver does, adds those queues to the main signal. But the geometry between the speakers and yourself does not change. It colours the sound and will change tonality. It's an added effect in a way. It got boring fast.
The queues from my ambient channels arrive from a different angle, and as a result have a slightly different comb pattern at the ears. They are more like scattered reflections, coming from more than one angle in real life. They can actually fill in some of those cross talk dips, perceived at the ear. Especially when arriving within the Haas limit.
What to listen for... the things I listen for when determining the success of this combination of mid/side EQ and phantom ambient addition is in the central voices. I listen to many tracks with clear phantom center voices, both male and female and also concentrate on the words, can I follow their sentences more easily? What happens when I move, side to side a bit etc. Does the perception change? Is the voice more separated from other instruments? It also has an effect on the perception of depth of the stage
Sound coming from the lateral angles will be perceived differently than coming from the stereo triangle. The rest of the ambience I create is hiding later room queues in my room. I could see the RT60 graphs become more even in content with the addition of the ambient channels. I'm kind of 'creating' my own room's power response to mimic the direct sound, in a way. This adds to the sense of spaciousness, but not the same spaciousness as early wall reflections would create. A sense of space, the room or space we listen in "feels" bigger than it really is. This works best with eyes closed.
The eyes do detract and are powerful. However with all these tricks (cause that's what they are) even with eyes open the perception is altered (in a positive way).
Though all my tests I have become way more sensitive to room queues. Actually listening for them wherever I go. Not a good hobby
.But it can help you to try. I do not recommend this though. Just enjoy the sounds wherever you go. Sometimes this analysing mode of mine plainly s##ks.
Listening in my relatively normal room I cannot (kid myself to be able to) rid myself of the actual space completely. However these tricks I use help to work with it.
Read some David Griesinger papers, those were most helpful for me to get a grip on it. His work in this field is awesome.
Last edited:
This thread points us to a similar discovery by Linkwitz.
Added realism with a dip at 3 kHz, 4 dB down. Hmm, if he had done the tweak to the panthom center only, the realism in the side sounds would have been kept. Throw in the change at ~600 Hz in mid/side and: boom!
A thing to notice: his proposed dip is actually deeper than what I use for stereo recordings. This works best for his "in ear" recordings.
That's similar to what I do for Home Theatre or other material mixed for a real center. Is this all starting to make sense yet?
Added realism with a dip at 3 kHz, 4 dB down. Hmm, if he had done the tweak to the panthom center only, the realism in the side sounds would have been kept. Throw in the change at ~600 Hz in mid/side and: boom!
A thing to notice: his proposed dip is actually deeper than what I use for stereo recordings. This works best for his "in ear" recordings.
That's similar to what I do for Home Theatre or other material mixed for a real center. Is this all starting to make sense yet?
Last edited:
OK, I've studied your layout and have 5 or 6 questions.
- I see what you are doing with the surrounds, L-R and R-L with a low pass and delay. Check. But what about the center and sub channels, which are mono sums? What are they used for? Are they used further along in the processing chain?
- You say that center gets added back into the surrounds, where does this happen?
- How many playback channels do you have (speakers)? L, R, LS, RS and.....? You show 6 in your flow chart.
- The EQ graph surprises me because your center EQ is about opposite from what I would expect. I think we've been over that before, I'll have to search for it in the thread.
- There is a FR trace that's opposite the center, which you call Left and inverted right. This are the surround EQ?
OK, I've studied your layout and have 5 or 6 questions.
[*]I see what you are doing with the surrounds, L-R and R-L with a low pass and delay. Check. But what about the center and sub channels, which are mono sums? What are they used for? Are they used further along in the processing chain?
The L-R and R-L is used as ambience. The center and sub channels (I used those channels for convenience, as I use a 5.1 container giving me 6 channels to play with) are used to fill the tonal holes created by the Stereo cross talk. Their angle towards the ears are different and that creates a difference where the perceived holes are from their respective cross talk.
After running trough a 6 channel FIR correction, where the center + sub get the same treatment as the SL and SR. After the convolution step I mix the center to SL and sub to SR (with a further 3 dB reduction in level of both center and sub.
I have 4 active channels, wired to amps.
Yes, you've mentioned it before
. It is all about balance. If we look how the stereo cross talk sums at the ear, we follow this graph:
As you can see we have sums at 3.7 kHz, 7.2 kHz and also at 500-600 Hz.
These peaks and dips are from the wrapping of sound around our heads from the opposing loudspeaker reaching that ear. But... our head is in the way of some of those dips too... so the will be less severe above 5 kHz and below about 4-500 Hz the waves get so long that it sums anyway.
By creating that dip you see, centered at 600 Hz in the phantom part of the mains (mid channel of the mains) it will sound brighter. The 3.7 kHz dip removes excess information (like mentioned in that Linkwitz description) that creates a haze if it's left there.
Meanwhile the sides don't get that dip at 600 Hz, usually almost anyone I know balances his sound for a pleasing phantom center experience. This makes the sides sound brighter. Almost kitchen table bright. By not having that same dip at 600 Hz it will actually sound much more balanced, very real actually. The sound from the left, does not get cancelled by the same sound from the right channel here, no cross talk effects. Whatever does wrap around the head to enter the right ear is the same as if a real sound source would create. Hence the straighter line for the side frequency curve (no peaks or dips to compensate for, it mimics the actual sound in its purest form, coming from a single speaker)
Nope, that is 3 graphs representing what happens with the EQ under the varying circumstances. If we feed the mains with the same signal in the left and right channel (which would best represent a phantom sound) you get the curve with the dips at 600, 3.7 k and 7.2 k.
If we feed it with a sound panned to one loudspeaker you get the middle dark blue line in the graph. It would be very similar to hearing an actual instrument or voice from that direction. No dips, no peaks. Just a room curve for proper balance.
If we feed the mains with a sound on the left, and the same sound inverted to the right you get that top graph. This is the EQ curve of the sides.
To really "get" this you'd have to understand how splitting the sound in mid and side actually works in practice. The sides signal is just that: +L and -R.
While the mid part is L+R. And that mid signal still has some left only and right only content in it. If I were to neglect that the left ear woul get a subdued version of the mid EQ parsed on it if we restore the Stereo signal after the mid/side processing.
That's a pretty important thing to understand. If anyone knows how we can do better I'd like to hear about it. We cannot really split a Stereo stream into a pure left, a pure right and pure center panned streams.
The graph I posted here, was recorded to disk by feeding those 3 different combinations as a Dirac pulse to the mains.
* +Left and +Right
* +Left only, no right
* +Left and -Right (right side inverted)
Let's look at the 6 channels within Metaplugin one more time:
The two most left channels are the mains. L and R in other words. They get separated into mid and side in the MSED plugin, next they each enter the EQ
[EQ 6 for mid] [EQ 7 for Side] and they get formed back into Stereo in the bottom MSED.
The two most right channels are the L-R and R-L signals, feeding the ambient speakers they go trough [EQ 8] and the reverb plugin.
The two middle channels are Center and Sub. Best to call it the phantom center of the ambient mix. There only to fill in tonal holes created by the Stereo perception of the mains. Even a little goes a long way here. Mix in too much and the ambience isn't ambience anymore, it will return to a Stereo mix. It does help though, to have this mixed in, it helps make the imaging in front more stable, it creates a better sense of depth in the center part of the stage. 3D holographic like.
As said, they are added to the most right channels. The only reason for me to have them separate is that I can control their level and timing and even make adjustments to their balance should I feel the need.
I did not get there in one day. Most of this is shaped with long listening periods. It is amazing what we can hear and how we actually perceive Stereo. I can imagine ambiophonics to be popular. Fix the cross talk and you get a whole new ballgame.
If anything is unclear, let me know, I'll try again
. I almost typed: if anything is left unclear... but that could create confusion in these kind of discussions . I tried to keep it sane.
Last edited:
Pano, I don't blame your way of thinking. I had the exact same perception you had and figured I needed to make the sides less bright. As I did have tonality setup to sound best at the phantom center.
It is all a balance thing. Just look at speakers that show some of these deviations in their frequency curve, they usually get better reviews. Flat is not the answer. At least, not as long as we use 2 ears. Flat (or at least perceived flat, which results in a downward curve in a room) would be perfect for the side panned signals. Less so for the phantom parts. While this still might not be perfect, it is way better than doing nothing at all. The sense of realism is what I do it for. As well as the tonal balance to support that realism.
A single mono speaker would not suffer from all of this. Ask BYRTT. A huge center channel is not going to happen for me. I will, at some point, experiment with a small full range center driver to try and fill the perceived Stereo cross talk holes from a single front. I'll probably use a foam ball and have the mic in there at one ear position to rate the success of that solution.
Just read how many curves and dips there are advertised that can be linked back to what's happening here. The famous BBC dip, wrongly named but in wide use. A tilting point at 2 kHz leading to better perception of depth.
One really strange bird in all of that are the B&W 800 speakers that actually have a peak at ~3 kHz. Often quoted as being reference material and used as studio mastering speakers.
Wouldn't you just love music mixed on those speakers? They would have a natural dip at 3 kHz as the engineer would try to keep it natural sounding and actually correcting for all that excess energy they have there.
(too bad these on the picture are for Home Theatre, I do hope they used some form of EQ on those! 5.1 Material mixed on them is going to be a bit dull if they aren't)
It is all a balance thing. Just look at speakers that show some of these deviations in their frequency curve, they usually get better reviews. Flat is not the answer. At least, not as long as we use 2 ears. Flat (or at least perceived flat, which results in a downward curve in a room) would be perfect for the side panned signals. Less so for the phantom parts. While this still might not be perfect, it is way better than doing nothing at all. The sense of realism is what I do it for. As well as the tonal balance to support that realism.
A single mono speaker would not suffer from all of this. Ask BYRTT
. A huge center channel is not going to happen for me. I will, at some point, experiment with a small full range center driver to try and fill the perceived Stereo cross talk holes from a single front. I'll probably use a foam ball and have the mic in there at one ear position to rate the success of that solution.Just read how many curves and dips there are advertised that can be linked back to what's happening here. The famous BBC dip, wrongly named but in wide use. A tilting point at 2 kHz leading to better perception of depth.
One really strange bird in all of that are the B&W 800 speakers that actually have a peak at ~3 kHz. Often quoted as being reference material and used as studio mastering speakers.
Wouldn't you just love music mixed on those speakers? They would have a natural dip at 3 kHz as the engineer would try to keep it natural sounding and actually correcting for all that excess energy they have there.
(too bad these on the picture are for Home Theatre, I do hope they used some form of EQ on those! 5.1 Material mixed on them is going to be a bit dull if they aren't)
Attachments
If you want to measure effects from ear spacing, you can buy foam "heads" from either a costume store or a wig store. That way you get the face shape as well. I think you would need 2 channels to get both side corrected at the same time.
Budget stereo electret mic
P.S. Interesting setup, its funny how speaker quantity naturally increases given enough time
Budget stereo electret mic
P.S. Interesting setup, its funny how speaker quantity naturally increases given enough time
Last edited:
I'm not so sure I need two mics. I could measure one ear at a time to get a better grip on things. I could measure at the left ear, head (foam ball or other) in the way and play left speaker only. Play both speakers at once... and play Right only.
Time alignment between center and mains is the easy part. The expected combing isn't as obvious.
We do use comb patterns to scan our surroundings. Though we might not be that aware we do it at all.
I my defence, only 52 drivers used here. 2x 25 drivers and 2x one. All full range.
Time alignment between center and mains is the easy part. The expected combing isn't as obvious.
We do use comb patterns to scan our surroundings. Though we might not be that aware we do it at all
.I my defence, only 52 drivers used here
. 2x 25 drivers and 2x one. All full range.
Last edited:
Maybe time will tell. However I don't think a center will become a permanent addition to this setup. I don't think my girl would allow that. Unless I'm able to disguise it or have it do multiple purpose things..
However, if it can do what I think would be its potential, it may be very persuasive.
I won't know if it would change anything for the ambient channels until I try. Out in a real room with real reflections with a live band playing, the center part would still generate it's own reflections, right? Somehow it seems odd to only have ambient L-R / R-L content...
However, if it can do what I think would be its potential, it may be very persuasive.
I won't know if it would change anything for the ambient channels until I try. Out in a real room with real reflections with a live band playing, the center part would still generate it's own reflections, right? Somehow it seems odd to only have ambient L-R / R-L content...
No problem, Pano.
I'm just sharing this so others "could" benefit from it. I enjoy it, chances are that it might work for others too. At least two line array owners have tested the mid/side part and enjoyed it. I do not think it's line array specific, chances are there are perceived differences between different systems or for example between a near field setup and the typical home setup. The beauty of this is that it does not detract from listening pleasure off axis. A key point for me, I want at least a listening couch with good sound, even better if it sounds good anywhere.
This took me months to find out, maybe even longer because I'm still finding new ways and new things to try. I'm getting my inspiration by reading other peoples experiences, trying to figure out how this fits into my ideas of what's happening.
I'm just sharing this so others "could" benefit from it. I enjoy it, chances are that it might work for others too. At least two line array owners have tested the mid/side part and enjoyed it. I do not think it's line array specific, chances are there are perceived differences between different systems or for example between a near field setup and the typical home setup. The beauty of this is that it does not detract from listening pleasure off axis. A key point for me, I want at least a listening couch with good sound, even better if it sounds good anywhere.
This took me months to find out, maybe even longer because I'm still finding new ways and new things to try. I'm getting my inspiration by reading other peoples experiences, trying to figure out how this fits into my ideas of what's happening.
Last edited:
What I'd like to do is draw up a generic flow chart, similar to the type you see for audio mixing consoles. I'll post it and let you tell me where it's wrong.
Then correct it.With a sort of generic signal processing flow chart, it should be easier for others to try this on their own systems. I might break in down into 2 charts, Mains & Surrounds.
Can't wait to hear about it, gkh, I never got to try it on an oint source system (sorry, I couldn't resist
Won't be before summer.
Oint source LOL