[porthillsbomber]

I'm pretty sure you want non-inverting for the higher input impedance but I could be wrong. In a unity gain buffer nothing "touches" the ground you wouldn't be able to have variable gain though, you could do some variable attenuation pre buffer.

 

[Armaegis]

Hmm, I guess a unity gain non-inverting config is basically just a voltage follower isn't it? I haven't actually tried that as my input buffers yet.
 
*adds to the list of things to try*
 

 

[Armaegis]

Just tried it with the buffers set as voltage followers. Works just fine, no discernible improvement. Probably better this way for sheer simplicity though.
 
I just did a couple listening tests back and forth with the single and dual input buffers and couldn't really hear any improvement. If anything, with dual it might have been slightly worse. If I had to guess (and it really is just that), maybe it's because the signals aren't buffered/reproduced exactly identical, and because of that the cancellation at the headphone end will be just slightly off.
 
Hmm, so now my entire list of materials is 5 opamps and 7 resistors (9 if I count the output resistors at the headphone connection).

 

[Armaegis]

 

[Avro_Arrow]

So something like this..?
 

Edit: changed slightly for better clarity
Edit: Added spare resistor positions

 

[digger945]

^not a lot different from just summing L and R into the G channel, and sending the original L and R signals out to the headphone.

 

[Armaegis]

In theory the 1M resistors are unecessary, but I think I read somewhere that it's better to keep that reference there for stablity (or something about bias current? my memory is fuzzy on this). Can anyone clarify/expand on this?
 
On a board, I would leave the spot for them but probably leave them empty. I would also add the resistor spot on the + inputs of the OL/OR opamps, but simply jumper them in practice.
 
 
Quote:

^not a lot different from just summing L and R into the G channel, and sending the original L and R signals out to the headphone.

The whole point though is to bring the electric field potential in the wires to zero (or as close as possible) as Meier states in his Concerto and Stepdance amps.

 

[Avro_Arrow]

Done

 

[Armaegis]

Up until yesterday all my testing was done with 4AA batteries for power, producing about 5.6V. Then I finally picked up some 9V for my smoke alarms and decided to try one of them in the circuit. Plain 'ol Duracell, measured at 9.6V. Holy poop, talk about audible differences! Everything just became that much cleaner and defined. I swear it was like listening to a different headphone, no exaggeration.
 
Of note, using two 9V batteries did not make a difference (at least not a very noticeable one). I guess when I was using just the 4AA it was running near the lower limit of the opamp which did not let the opamp perform optimally.
 
edit:
opamp RC4560, voltage range +/- 2~16
with the 4AA, I had +/- 2.8 (a bit under actually)
with the 9V, I had +/- 4.8
 
I suppose current draw might be a factor too, but I don't know what the batteries are capable of supplying, and I don't think it would be a limiting factor here anyways (please correct me if I'm wrong).

 

[Armaegis]

I'm sitting here wondering if I could cram this all with a 9V (or two) into a mint tin... and if I needed the extra power I could just use dual channel opamps and tap the second channel with a voltage follower to double up. Hmm...

 

[SP Wild]

I am interested in this "balanced" ground concept as a means of common mode noise rejection at the headphone transducer without the need for recabling to XLR plugs.  As I understand it, the B22 3rd channel is an active 0 volt regenerator that would have equal impedance to chassis ground as the signal wires (infinite resistance to chassis earth?), common mode noise cancellation is possible.
 
The Meyer circuit seems more complicated it sends a combination of left and right signals, and inverted.  But isn't the end outcome exactly the same as the AMB technique...but with more chance of mismatch?  What stops the right transducer from using the inverted left channel as a point of reference...how is this "crosstalk" eliminated?  Wouldn't the third ground channel amplifier need to be able to deliver twice the power of the individual left or right channel amplifiers?
 
What advantages does the Meyer technique have over the AMB technique?  Neither of them truly isolate left and right signal paths, but it seems that the Meyer technique has potential for mismatch and crosstalk.
 
Sorry for the OT, but I am learning these foundations as it fascinates me, but have no idea where I should ask such a question.

 

[Armaegis]

I don't claim to understand this stuff as it's all still pretty new to me as well, but this is how I see it...
 
- In a normal config, your L and R are referenced to a zero ground. The cleaner that "zero ground", the better your signal. Any fuzziness in your ground will show up as noise, which has plagued audio enthusiasts for decades
- An active ground creates a zero and feeds it into a unity gain amp channel, so now it's dependent on how well that zero is maintained (I've seen some arguments both good and bad for this)
- An active balanced ground does away with that zero altogether, instead generating your ground from L and R. This eliminates the zero ground fuzzies, but now you are highly dependent on proper matching in order for everything to cancel out at the end
 
 
Quote:

Wouldn't the third ground channel amplifier need to be able to deliver twice the power of the individual left or right channel amplifiers?

 
The ground channel needs to deliver twice as much power for both the active ground and balanced ground configs.
 

 

[Armaegis]

Just for fun I've cobbled together a circuit using INA134 and INA137 (to generate OG and OL/OR respectively) which have resistors built in, so my entire circuit has no external resistors and overall gain is 3x (inverted). Clarity seems better too, as I would assume the internal resistors are better matched than anything I could have done by hand.
 
If I used a balanced line driver like DRV134 as my input buffers (to generate + and - of my input L/R), then I could use three INA134 to generate my OL/OR/OG with gain of 3x. I will assume it's better to use the same chip to generate all three signals since we want the OG to cancel out; but I don't have the line drivers to try that out yet. I faked it using an extra set of input buffers set as unity-inverting and it all worked out ok.
 
The downside to using these chips is that the math only works out for the 3-wire solution and sticks you at 3x gain (or -3x). You also can't roll opamps, well, except for very specific line drivers and receivers. The upside is no need to match resistors (which is a somewhat crucial part of this circuit) and perhaps a cleaner layout overall.
 
In terms of sound when tapping the circuit directly with headphones, compared to using the set of RC4560... the RC4560 were more "fun" sounding while the INA134/137 combo felt more neutral.

 

[Armaegis]

So I was reading through some old threads and saw this interesting project: http://www.head-fi.org/forum/thread/377413/balanced-tpa6120-amp-photos-thoughts
 
The TPA6120 chip seems to have been pretty popular a few years ago, and while working with it is currently well beyond my skill level, one can always theorize right? Seeing as doobooloo was able to create a balanced amp, it shouldn't be too far-fetched to create one with an active balanced ground topology right?
 
Assuming all resistor values the same and omit the ground resistor on the + input, feeding L and R into the - and + inputs for each side will generate (2L-R) and (2R-L) respectively. To generate the ground channel, I'm thinking use the second chip to invert L and R separately, then tap the outputs together to create -(L+R). I think that works... either that or sum them before feeding into the unity inverter and tap both sides for more power. Anyhow, this should give an overall gain of 3, not counting any gain generated in the input buffer.
 
edit: or in the case of a DRV134 as input buffer, then then use the negative outputs and the ground channel chip doesn't have to be inverting
 
Hmm, just random thoughts. I know it's a tricky chip to implement.

 

[jcx]

early A/DSL driver chips were designed with a goal of extreme low distortion into 12-100 Ohm loads at 100KHz-1MHz - kind of a good match for dynamic headphone drive
 
the 1st few generations of chips used high internal bias, including output Q bias so these amps can be "deep(A)" Class AB for dynamic headphone loads - and can be externally baised for full Class A output stage operation with many headphone loads
 
more recently the industry has concluded more modest distortion goals are acceptable and low power is a bigger selling point so recent A/DSL driver chips specs don't look as great for audio
 
the TPA6120 is a "re-badged" THS6012 - so reading both data sheets can be informative
 
as 100 MHz CFA op amps they do have some different requirements than most hobby project diy amp op amps - but at ppm distortion, >1000V/us slew rate, 100 MHz BW into 25 Ohm load its hard to argue that they do anything "wrong" to the sound