Hi there,
I'd like to start a new thread about 3 front channels. I know there have already been some discussions about this in other threads, but they were mixed by many other issues and somewhat distracted.
I'm no expert in Psychoacoustics and there're lots of talks (and debates) around, I'd rather not repeat. In my simpliest thoughts, 3 front channels should be an upgrade to ordinary stereo by an important supplement in the center. And practically there's far less hassle to implement compared to full surround mult-channel (5.1 or 7.1, or whatever).
But how?
The trinaural processor mentioned in this is probably the most elaborated I've seen:
http://www.diyaudio.com/forums/multi-way/36254-try-ambiophonics-your-speakers-40.html#post2520370
However, this processor is not cheap. And it seems the circuit is quite complex. If it has to be that complex, I'd rather put it in digital. (OK, that's me.)
Ideally, I think 3 front channels should be:
true L ..... true C ..... true R
But how to get these 3 channels from stereo signal?
OTOH, there're also other proposals, such as:
L .... (L+R) ..... R
or
L-R ..... (L+R) ..... R-L
These two can be easily done in analog circuits, even between amp and speakers. Supposedly they also provide some good features to please those who are using them. ( No? )
I've briefly tried the last one. The experiement is still on going and the setup is probably far from ideal. Also, my (subjective) observation is still too preliminary to talk about. (Very interesting, though...)
So, I'd like to hear your opinions.
Thanks in advance.
I'd like to start a new thread about 3 front channels. I know there have already been some discussions about this in other threads, but they were mixed by many other issues and somewhat distracted.
I'm no expert in Psychoacoustics and there're lots of talks (and debates) around, I'd rather not repeat. In my simpliest thoughts, 3 front channels should be an upgrade to ordinary stereo by an important supplement in the center. And practically there's far less hassle to implement compared to full surround mult-channel (5.1 or 7.1, or whatever).
But how?
The trinaural processor mentioned in this is probably the most elaborated I've seen:
http://www.diyaudio.com/forums/multi-way/36254-try-ambiophonics-your-speakers-40.html#post2520370
However, this processor is not cheap. And it seems the circuit is quite complex. If it has to be that complex, I'd rather put it in digital. (OK, that's me.)
Ideally, I think 3 front channels should be:
true L ..... true C ..... true R
But how to get these 3 channels from stereo signal?
OTOH, there're also other proposals, such as:
L .... (L+R) ..... R
or
L-R ..... (L+R) ..... R-L
These two can be easily done in analog circuits, even between amp and speakers. Supposedly they also provide some good features to please those who are using them. ( No? )
I've briefly tried the last one. The experiement is still on going and the setup is probably far from ideal. Also, my (subjective) observation is still too preliminary to talk about. (Very interesting, though...)
So, I'd like to hear your opinions.
Thanks in advance.
Here's a couple various links.
Simple Surround Sound Decoder
Matrix Surround for Music
Chris Kantack's do it yourself surround sound
Surround Sound Decoder
I once played with the "Hafler passive surround" setup a bit,and it was quite good,with certain albums.
Something like this: http://sound.westhost.com/p18_fig1.gif
Except that I added another speaker in series with the 2 front speakers -neg returns,and the the amp -neg,for a center channel.
You could of course do it all in solid-state with some op-amps,and use a power amp for each channel. It's basically just a couple summing and differential amps.
Simple Surround Sound Decoder
Matrix Surround for Music
Chris Kantack's do it yourself surround sound
Surround Sound Decoder
I once played with the "Hafler passive surround" setup a bit,and it was quite good,with certain albums.
Something like this: http://sound.westhost.com/p18_fig1.gif
Except that I added another speaker in series with the 2 front speakers -neg returns,and the the amp -neg,for a center channel.
You could of course do it all in solid-state with some op-amps,and use a power amp for each channel. It's basically just a couple summing and differential amps.
Thanks for the links. I found the "Sonic Design" very interesting. Will study.
Before stepping into actual circuits, let's talk more about the channel distribution, shall we?
I tried the simpliest 3-ch briefly: L-R speakers connected in series and linked to hot side of both channels of amp, while the C crosses the junction of L-R and the cold side of amp.
In all related diagrams I've seen, the center speaker should be in double and paralleled. I guess it's for proper summing etc. But I use only one here...
In such connection, when playing mono pink noise, the distribution (in level) is obvious center-heavy. Center channel is much louder then both sides. Interestingly, when playing music, soundstage can be very wide and not center-heavy at all.
However I found some problems in the correctness of locallizing - cymbals come to the center which should be at the right side. I guess 2 reasons for this:
1. Wrong level matching across the 3-ch. It's indeed overly center-heavy.
2. Wrong re-distribution by the crude summing. L+R is put at the center, which means all sounds are emmitting through here, including those should be far L and far R in the original signal. So in this case I heard too much right appearing at the center.
Now what?
What are your experiences?
Before stepping into actual circuits, let's talk more about the channel distribution, shall we?
I tried the simpliest 3-ch briefly: L-R speakers connected in series and linked to hot side of both channels of amp, while the C crosses the junction of L-R and the cold side of amp.
In all related diagrams I've seen, the center speaker should be in double and paralleled. I guess it's for proper summing etc. But I use only one here...
In such connection, when playing mono pink noise, the distribution (in level) is obvious center-heavy. Center channel is much louder then both sides. Interestingly, when playing music, soundstage can be very wide and not center-heavy at all.
However I found some problems in the correctness of locallizing - cymbals come to the center which should be at the right side. I guess 2 reasons for this:
1. Wrong level matching across the 3-ch. It's indeed overly center-heavy.
2. Wrong re-distribution by the crude summing. L+R is put at the center, which means all sounds are emmitting through here, including those should be far L and far R in the original signal. So in this case I heard too much right appearing at the center.
Now what?
What are your experiences?
Thinking (guessing) aloud, so don't rely on this.
The Left & Right are substantially a mono signal.
[L-R] = -[R-L]. This difference signal is what generates the "Stereo" effect. It's how they transmit a backwards compatible FM stereo signal. It's the same with Colour television signal, it too is backwards compatible to what came before (B&W)
If you sum L+R you get a signal that is twice as loud cf. L or R.
So let's half it to give 1/2[L+R] for an equal level centre channel.
What if we subtract the stereo information from the summed centre channel?
The three channels thus become L, 1/2[L+R]-[L-R], R
We have pure L and pure R and in between we have an appropriately quieter mono less the stereo information. Is this feasible?
OOPs, that does not work, the centre channel is equivalent to 1.5*R - 0.5*L.
Removing the 1/2 factor does not help much. Changing the [L-R] to [R-L] also doesn't help.
The Left & Right are substantially a mono signal.
[L-R] = -[R-L]. This difference signal is what generates the "Stereo" effect. It's how they transmit a backwards compatible FM stereo signal. It's the same with Colour television signal, it too is backwards compatible to what came before (B&W)
If you sum L+R you get a signal that is twice as loud cf. L or R.
So let's half it to give 1/2[L+R] for an equal level centre channel.
What if we subtract the stereo information from the summed centre channel?
The three channels thus become L, 1/2[L+R]-[L-R], R
We have pure L and pure R and in between we have an appropriately quieter mono less the stereo information. Is this feasible?
OOPs, that does not work, the centre channel is equivalent to 1.5*R - 0.5*L.
Removing the 1/2 factor does not help much. Changing the [L-R] to [R-L] also doesn't help.
L + R for center is the conventional way to derive it. This will tend to diminish your stereo spread soemwhat. Due to that you can play with streching the L and R speakers farther apart. A lot of people play with center channel level also. I hear -6dB as a common setting.
I would be tempted to use out of phase crosstalk to spread the left and right some. Rather than full L - R and R - L I might try L - 0.5R and R - 0.5L.
You can also pick up an older Dolby pro-logic box and see what steering adds to the effect. It looks for dominant channel direction and issolates that for better seperation between channels.
Read through articles from the 70s, you are reinventing quad matrixes, only without the rear channel.
David S.
I would be tempted to use out of phase crosstalk to spread the left and right some. Rather than full L - R and R - L I might try L - 0.5R and R - 0.5L.
You can also pick up an older Dolby pro-logic box and see what steering adds to the effect. It looks for dominant channel direction and issolates that for better seperation between channels.
Read through articles from the 70s, you are reinventing quad matrixes, only without the rear channel.
David S.
IN DSP
I use the following in Reaper JS to create the LCR.. This is similar to the Trinaural method (Dont know for sure) but done in a Reaper JS plugin....
Im no programmer. So i think there are some improvements to be made.
The degree parameters effect the gain of frequency at each speaker. In my installation i have a HP/LP at 5000hz so high freq is more centre. Low is more spread.
This method also makes sure that gain across all 3 speakers is the same as combined gain from the original L/R signal... (at least i think it is )
This works very well. You need to either delay the mid or bring the L/R speakers closer to you to get things right.
JS Code:
I use the following in Reaper JS to create the LCR.. This is similar to the Trinaural method (Dont know for sure) but done in a Reaper JS plugin....
Im no programmer. So i think there are some improvements to be made.
The degree parameters effect the gain of frequency at each speaker. In my installation i have a HP/LP at 5000hz so high freq is more centre. Low is more spread.
This method also makes sure that gain across all 3 speakers is the same as combined gain from the original L/R signal... (at least i think it is
)This works very well. You need to either delay the mid or bring the L/R speakers closer to you to get things right.
JS Code:
Code:
//=================================================
desc: LCR Processor, Simple
//slider1:0<-25,25,0.05>Output (dB)
slider2:1<0.1,2.5,0.1>Width
slider3:35<15,75,1>Degree (Below Cutoff)
slider4:54<15,75,1>Degree (Above Cutoff)
slider5:80<0,100,0.05>Cutoff
//=================================================
@slider
sx = 16+slider5*1.20103;
cutoff = floor(exp(sx*log(1.059))*8.17742);
//coeff for 6db
cutoff = min(cutoff,20000);
lp_cut = 2*$pi*cutoff;
lp_n = 1/(lp_cut+ 3*srate);
lp_b1 = (3*srate - lp_cut)*lp_n;
lp_a0 = lp_cut*lp_n;
//outgain
width = slider2;
degree = slider3 * 0.017453;
degree2 = slider4 * 0.017453;
//=================================================
@sample
s0 = spl0; s1 = spl1;
//0.663214
//OutL = (0.5*sin(degree))*(s0+s1)+0.5*width*(s0-s1);
//OutR = (0.5*sin(degree))*(s0+s1)-0.5*width*(s0-s1);
//OutCC = (0.7071*cos(degree))*(s0+s1);
//create MS from Left/Right
m2 = 0.7071*(s0+s1);
s2 = 0.7071*(s0-s1);
//On m2 create a 6db hp/lp
inlp = m2;
inhp = m2;
lp_pre = 2*inlp*lp_a0 + lp_pre*lp_b1;
hp_pre = 2*inhp*lp_a0 + hp_pre*lp_b1;
lp = lp_pre;
hp = inhp-hp_pre;
c3a = lp*cos(degree);
c3b = hp*cos(degree);
c3 = c3a+c3b;
m3a = lp*sin(degree2);
m3b = hp*sin(degree2);
m3 = (m3a+m3b);
s3 = s2*width;
l3 = 0.7071 * (m3+s3);
r3 = 0.7071 * (m3-s3);
sr= 0.7071 * tan(degree) * (s1-s0);
sl= 0.7071 * tan(degree) * (s0-s1);
spl2=l3; // Left
spl3=r3; // Right
spl4=c3; // Centre
spl6=sl; // Left Back
spl7=sr; // Right Back
@gfx 100 10
//draw freq scale numbers
gfx_x=gfx_y=5;
gfx_lineto(gfx_x, gfx_y,0);
gfx_r=gfx_b=0;
gfx_g=gfx_a=1;
gfx_drawchar($'F');
gfx_drawchar($' ');
gfx_drawchar($'=');
gfx_drawchar($' ');
gfx_drawnumber(cutoff,0);
gfx_drawchar($' ');
gfx_drawchar($'H');
gfx_drawchar($'z');
//=================================================Attachments
Wow!
Thanks a lot, folks !
I'll study those papers first, as much as I can (which means not so much, actually ).
As to the DSP provided by optic, it's now far beyond me, sigh~ In the diagram, I can't even guess how it works. (why is C3 irrelevant to R2, and made out of L2 only?)
---------
As I'm fiddling with the crude experiment with questionable new speakers, I have some interesting (subjective) observations:
Although on paper, it seems the stereo width (separation) is reduced by the summed center, in my ears, the soundstage is still very wide. The imaging stretches to the outer edges of the L/R speakers at least. And in some recordings, the feel of ambience would be far beyond. For example, On The Night, Dire Straits. That is immensely spacious! (Or, am I cheating here?) Orchestra in other classical records are also very good in the sizes of soundstages with spacious ambience.
In some simpler records with fewer 'sources', it's interesting to hear sounds of the opposite side when I'm very near in front of one speaker. When I get back to the couch, the image goes back to the correct position. It's amazing how the differential and summed sounds steer the perception.
In the negative side, I encountered wrong positioning of certain HF sounds as mentioned above... I'll be working on them.
To be continued....
Thanks a lot, folks !
I'll study those papers first, as much as I can (which means not so much, actually
). As to the DSP provided by optic, it's now far beyond me, sigh~ In the diagram, I can't even guess how it works. (why is C3 irrelevant to R2, and made out of L2 only?)
---------
As I'm fiddling with the crude experiment with questionable new speakers, I have some interesting (subjective) observations:
Although on paper, it seems the stereo width (separation) is reduced by the summed center, in my ears, the soundstage is still very wide. The imaging stretches to the outer edges of the L/R speakers at least. And in some recordings, the feel of ambience would be far beyond. For example, On The Night, Dire Straits. That is immensely spacious! (Or, am I cheating here?) Orchestra in other classical records are also very good in the sizes of soundstages with spacious ambience.
In some simpler records with fewer 'sources', it's interesting to hear sounds of the opposite side when I'm very near in front of one speaker. When I get back to the couch, the image goes back to the correct position. It's amazing how the differential and summed sounds steer the perception.
In the negative side, I encountered wrong positioning of certain HF sounds as mentioned above... I'll be working on them.
To be continued....
Last edited:
Hi CLS, this nice description would be better if we could know what kind of speakers you use in this unusual set.
Same efficiency ? same phase shifts (I think of possible + or - interferences) ? same FR ? same distance from the listener ? same orientation ? same radiation ?...terra incognita.
Same efficiency ? same phase shifts (I think of possible + or - interferences) ? same FR ? same distance from the listener ? same orientation ? same radiation ?...terra incognita.
oops how about the patent i got this from then..... This should help you understand my madness...
Sound reproduction system having a ... - Google Patents
and small mistake in my posted code.... Pasted old file... If anyone is interested i will post an update....
BTW Happy international Beer Day....
Cheers,
Optic
Hi, it's just the same as what the paper is saying below:
Wow!
Radugazon,
I take that as a compliment. Thank you.
An Austrian guy is currently building a "trinaural decoder". You might want to get in contact with him.
Thanks again Markus.
In the paper of your previous post, the 'partially' differential left and right signals play an important roll in steering the images. Why don't you use this method in your experiment? In that omni thread, I remember you use original R/L for the B200. Any reason for that?
In the paper of your previous post, the 'partially' differential left and right signals play an important roll in steering the images. Why don't you use this method in your experiment? In that omni thread, I remember you use original R/L for the B200. Any reason for that?
