[porthillsbomber]

yeah sort of... no. The key thing to not getting cross talk is using a difference amplifier to take the difference between input signal and input ground and then referencing it to your final output ground.
 
Also the reason I would use this whole ABG method is to make clipping sound better but with beta22 you will get nowhere near clipping so I wouldn't bother, I don't quite understand Meier's reasoning so there may be a point to do what you propose provided what he says is correct.
 
Just to clear things up lets call your first circuit active balanced ground (ABG) because the ground is active just like an ordinary active ground as it seperates the headphone return current (OG) from the signal currents on the left/right channel opamps. This is a perfectly fine circuit as long as the gains are well matched.
 
Let's call mine balanced ground (BG) as the OG is connected to the signal currents on the left/right opamps, kind of like when you just have your 3rd opamp just buffering the ground in a standard opamp amp. This method does not rquire the gains to be well matched as the left/right outputs are referenced to OG.
 
Both methods also need to be feed with a low output impeadance source (which can be done with an input buffer as we both tried) to insure low crosstalk regardless of whether the gains are well matched or not.
 
I hope this makes sense, I may have to draw some diagrams to make things clearer.

 

[Armaegis]

I see what you're getting at, but I'm misunderstanding something in your implementation because the math doesn't work out the way I see it (as in, it doesn't look like you will get proper cancelation at the end).
 
I'm a little fuzzy on Meier's theory/advantages myself, but I'm really just approaching it like a proof of concept thing. An advantage I can see is that our ground is derived from the left/right channels, so this should in theory eliminate ground problems like ground loops and "dirty" ground, etc. Of course, now we're trading out ground problems for crosstalk and more circuits in the signal path... lesser of evils? *shrug*
 
I do see the advantage in reducing clipping though.
 
In some correspondance with Jan, I asked him how he accounted for 3 conductor wires vs 4 conductor wires (which his theory is based on). He said that in practice he used a value in between the two.

 

[porthillsbomber]

Which part of the math doesn't work out? I could quite easily have gotten something wrong. Yes you're right it is a bit of a trade-off.
 
The thing I don't understand with Jan's theory is that voltages are relative and when you've just got the 3 or 4 wires its the relative voltages between them that means anything but this method doesn't change that. It only changes the voltages relative to the ground in the amp. The time when it would make a difference is if you had shields on the cable that were connected to amp ground (IG) . This method would be less prone to interference in the same way a true balanced cable is (although I don't fully understand why a balanced cable is anyway).

 

[Armaegis]

 
 
Quote:

source --> volume pot--> (buffer) --> ABG ground OG gain G1 --> beta22 ground OG2 gain G2 --> ABG left/right OR/OL gain G1 --> beta22 left/right OL2/OR2 gain G2

 
Your ground component in the OL and OR channels has gone through the b22 twice, which isn't going to cancel out at the headphone end where the ground component has only been amped once... unless you're doing some juggling at the opamp to scale the proportions properly, but if that's the case why bother putting the OG through the b22... you might as well do it my way then.
 
Jan's theory on active balanced ground is to reduce the electrical field potentials generated by the wires. Any electrical fields can feedback on themselves and nearby wires, affecting the signal, although the effect on the audible frequency spectrum is open to heated debate.
 
True balanced carries a fully inverted signal in the second wire. Any external effects will in theory affect both wires in the same way, thus canceling out at the receiving end. It's used in pro audio where the long lengths of cable are more succeptible to picking up noise and interference. The fully inverted signals will also eliminate most of the electric field potential across the length of the cable.

 

[porthillsbomber]

Here's a diagram

So we have
 
OG = -G1*(IL+IR)/4
OG2 = G2*OG = -G1*G2*(IL+IR)/4
 
OL = G1*IL + OG2
OL2 = G1*G2*IL + OG2
 
Headphones see OL2 - OG2 = G1*G2*IL
 
I think that's right?

 

[Armaegis]

Ah ok, I didn't catch on that you were referencing the left and right b22's down to OG2.
 
I also don't quite fully understand your differential opamp config. I only know the config from wikipedia (link) and in your case you've switched ground for OG and replaced V1 with ground.
 
The thing about your setup that I feel a little fuzzy on is that your OG gets awfully busy. It is used to feed back into the left/right opamps and drive headphones simultaneously. In the b22 setup, the OG2 is tapped 5 times. That's a potential cause of noise/crosstalk/etc isn't it? Doing it your way, I think I would stick to just opamps (as you mentioned earlier) and put current buffer stages after them, tapping OG before the buffer to feed back into the left/right opamps.
 
In my setup  (assuming I have precisely matched resistors), IL and IR are tapped three times and none of the outputs are touched except for driving the headphones. I was thinking maybe even the summing (OG) opamp was a potential source of crosstalk, so use some dual opamps as input buffers to create a second set exclusively for the summing.

 

[porthillsbomber]

Well you can just look at as the difference between IL and IG referenced to OG as ground. That's how I looked at it using the same wiki page but what I realized is that it is just a non-inverting summing circuit G*IL + OG.
 
Yeah it does get quite busy but a nice star ground for OG2 should sort that out, it gets back to the whole active vs non-active ground thing again.
 
You are right it is for both our circuits if the source impedance is too high. I buffered all the inputs but you might be right about just buffering the ground circuit.

 

[Armaegis]

Yeah the way your circuit is set up, it basically looks like a non-inverting circuit except with a summing bit in front of it, but I don't know how the the calculations work out. I can see where you derived your equations from, but those were based on reference to ground not OG, and I don't know enough about opamps to know if the math simplifies out the way you have it. I only tried it with unity gain everywhere which does simplifies things. I'll take your word for it that it works with other gains.
 
I think all the outputs should be buffered when driving headphones in order to maintain symmetry. In your config, I would simply tap the OG before the buffer to feed back into the OL/OR circuits. Without the buffer after the OG, the OG opamp would be feeding into your high impedance OL/OR opamps and relatively low impedance headphone simultaneously.
 
I thought of something just now too. If we take the differential circuit from wiki and make Rg = 1M (or anything suitably high), the calculation approximates nicely into:
Vout = [(Rf+R1)V2 - RfV1]/R1
 
Solving for Meier's original conditions of making all the voltages across four wires sum to zero, and setting all my input resistors to be equal, you get R1 = R2 = 2Rf, and this coincidentally applies to both the differential and summing circuit. Your eventual outputs wind up with a gain of 2, thus:
OL or OG = (3V2-V1)/2
OG = -(V1+V2)/2
 
In the case of a three wire cable and using the same assumptions, you get R1=R2=Rf and your overall gain is 3 coming out of the ABG circuit.
 
 
Man I hope I did my math correctly in there.

 

[Armaegis]

Another thought/concern: is there a way to lower the output impedance of the OG without adversely affecting the signal? Should this even be something to worry about?
 
In a regular 2 channel or active ground config, L and R run into ground because it's the path of least resistance, even though they are technically connected together at the headphone jack. With the active balanced ground though, all three channels are pushing equally, so OG might push into OL and vice versa. Obviously a higher impedance headphone will negate this somewhat since it will have that resistance along the left and right, but what about a 32 ohm headphone? Is that enough to prevent crosstalk?

 

[porthillsbomber]

No, I don't think so other than using a lower output impeadance opamp.
 
You're correct though that a non-zero output impedance on your ground channel adds crosstalk. Any amp with an active ground (ABG or ordinary active ground/3channel) has a non-zero output impeadance on ground but I don't think it will be enough to make a noticeable difference, a lot of people, including myself, use them.

 

[Armaegis]

I've got a new setup that works much better than anything else so far. Crosstalk is barely noticeable now.
 
a pair of input buffers
OG: simple inverting summing circuit, input resistors 4.7k, feedback whatever (I used 1.2k which is approx a quarter of the input resistors, we've got our (IL+IG)/4 value)
OL: non inverting summing voltage follower, summing IL and OG together... input resistors 4.7k each going into +, output short circuits to -
OR: as above except use IR
 
Headphones are tapped to OL/OR/OG
 
Here's the big thing: on the output taps, put an extra 47ohm resistor inline with the OL and OR, nothing in front of the OG2.
 
edit1: added the following which improved even further
OG2: feed OG into a voltage follower, input resistor 4.7k, output short circuits to -
tap headphones to OL/OR/OG2
 
edit2: I tried 470 in front of the OL and OR, I can't detect any crosstalk at all
doing this, I don't need the OG2 part anymore

 

[porthillsbomber]

Sounds interesting. That's quite a big output impedance does it colour the sound at all? What cans are you using? The resistors are between the opamp outputs and the headphone correct?

 

[Armaegis]

Yes, the resistors are between the opamps and the headphones.
 
Right now, I've only tested with some cheap airline clip-ons (which are surprisingly 300 ohm!) and the Phiaton MS300 (32 ohm). The airline clips are proving to have less crosstalk than the Phiatons.
 
I was surprised at how much of a difference just adding 47ohm did. 470 is probably too much, but I'm just using what I've got on hand.
 
I haven't done enough testing to make any judgement over colouration though. Sounds just fine with a few test songs though, so it's nothing overtly noticeable.
 
My dual input buffers is overkill though. A single buffer for each side is enough. It doesn't work without buffers though (though right now I'm just tapping off my laptop headphone out).

 

[porthillsbomber]

I think it would be worth bypassing the amp and plugging straight into the laptop to see if any of the crosstalk is caused by the the other components or have you already checked this?

 

[Armaegis]

I've already checked for that. I do not get any crosstalk from my laptop.
 
Truth be told, I'm not really sure why my new config works. I'm not entirely sure when some of the inputs are being summed or when they're taking a difference. I've run my current config through a software simulator (Tina TI) and it does not work, yet going by ear it is the best set up I've had so far.