building an active balanced ground?

[porthillsbomber]

Hi,
 
I actually designed the same thing but for the reason of having a higher output voltage swing for the mono part of the signal. I personally don't believe/understand Meier's idea as voltages are just relative anyway aren't they?
 
I haven't read this entire thread but from what I can see is that your proposed method will have horrible cross talk unless you precisely match the resistors. Here's a schematic of how I was planning to do it.

First it creates the output ground,
OG = -G*(IL+IR)/4
where G is the desired gain. It then puts the difference between IL and IG (the unnamed ground in the schematic) relative to OG multiplied by the gain. Similarly for the right channel. So you end up with.
OL = G*(3*IL - IR)/4
which gives
OL - OG = G*IL as desired.
 
No, you can not just hook ground up to the wall. I've included a simple virtual ground circuit that will do the trick.
 
As for component values start with something like R1=R2=R4=R5=R8=R9 = ~1k and C1+=C1- = 200+ uF.

 

[Armaegis]

Have you tried it your way? Did it work for you?
 
I tried it your way and it still doesn't work. The results are pretty much the same as mine from before... if I disconnect one side, it works perfectly (ie: the ground signal cancels out what I don't want to hear), but when I have all three it doesn't. If I only have OL and OR connected (no headphone ground), then I get almost full cancellation (because I can hear just a tiny bit of the original test tones which I will chalk up to improperly matched resistors).
 
 
Is it really the cross talk I'm hearing? That's a huge amount of crosstalk if it is...
- at which point in the circuit does crosstalk happen the most? (if crosstalk can even be narrowed down that way)
- how can it be reduced?
 
p.s. I did not have a TLE426 rail splitter, so I used the simple generic opamp splitter config shown halfway down the page here: http://tangentsoft.net/elec/vgrounds.html

 

[Armaegis]

Maybe I just have really junky parts, and I still think I have some weird grounding issue. I dunno why, but my ground only works properly when I connect the input ground into the virtual ground.
 
edit: Of note though... I have now attempted three variations on the circuit and all three have had the exact same problems. Since running these in simulation works properly, perhaps the theory is fine but maybe the culprit is something fishy on my end in terms of how I'm building the circuits (and if its an error somewhere, I sure am consistent!)

 

[porthillsbomber]

Quote:

Maybe I just have really junky parts, and I still think I have some weird grounding issue. I dunno why, but my ground only works properly when I connect the input ground into the virtual ground.
 
edit: Of note though... I have now attempted three variations on the circuit and all three have had the exact same problems. Since running these in simulation works properly, perhaps the theory is fine but maybe the culprit is something fishy on my end in terms of how I'm building the circuits (and if its an error somewhere, I sure am consistent!)

You need to connect input ground to virtual ground otherwise your inputs are floating.

 

[porthillsbomber]

Quote:

Have you tried it your way? Did it work for you?
 
I tried it your way and it still doesn't work. The results are pretty much the same as mine from before... if I disconnect one side, it works perfectly (ie: the ground signal cancels out what I don't want to hear), but when I have all three it doesn't. If I only have OL and OR connected (no headphone ground), then I get almost full cancellation (because I can hear just a tiny bit of the original test tones which I will chalk up to improperly matched resistors).
 
 
Is it really the cross talk I'm hearing? That's a huge amount of crosstalk if it is...
- at which point in the circuit does crosstalk happen the most? (if crosstalk can even be narrowed down that way)
- how can it be reduced?
 
p.s. I did not have a TLE426 rail splitter, so I used the simple generic opamp splitter config shown halfway down the page here: http://tangentsoft.net/elec/vgrounds.html

No, I have not tried it, but plan to next time I've got a spare opamp. I'm going to read the whole thread now so I'll see if I can spot any problems. That virtual ground should work fine.

 

[porthillsbomber]

Ok I've read through it, I'm not really sure what you mean when you say you've got something connected, to what?
 
To use a pot you would need a very low resistance pot, like 100R, or use a non-inverting opamp as an input buffer, I think you have done the latter, correct?
 
To try and make sure you've wired it all up correctly, try and draw a schematic from what's on your board ignoring what it is supposed to be then compare the two to check for any mistakes. I would be interested if you could post a full schematic of the set up you're using at the moment

 

[Armaegis]

When I say connected it probably means to the headphone jack. I'm using alligator clips to connect the outputs directly to the tip/ring/sleeve of the headphone.
 
Right now I'm just using an opamp as input buffer. I think it's in inverting config right now, which shouldn't matter as both left and right are inverted before feeding everything else so it should all work out in the end. I've also skipped the input buffer and just wired directly into the circuit, but it doesn't make a difference.
 
I've gone over the circuits many times now. I've tried my method 1 and 2 listed on page two, and I've tried yours as well. My schematics on the board match what I have on paper as far as I can tell. I'm currently using dual channel opamps (so typically one opamps for the OG, and one opamp for both OL and OR). I don't have any single channel opamps, but I might try just using one side of each one to create my outputs. I will try to post a full schematic tomorrow, although it'll be pretty similar to what I have on page two.
 
If we can get this circuit to work properly, my goal will be to connect it to a three channel b22 amp (replacing the "active ground" board with a regular config).
 
edit: could it be that the circuits are fine and that there's just ridiculously bad crosstalk (as you mentioned earlier)? After all, I'm just slapping down opamps and resistors without really knowing how it all interacts aside from what wikipedia tells me. Judging from the volume, I think it's maybe -30dB of crosstalk (-40 if I'm generous in my estimate). If that's the case, how can we reduce it?

 

[porthillsbomber]

I'm ordering parts for big speaker amp next week so I'll get another opamp so I can have a go with this circuit. Well, you could match all your resistors with a multimeter but the circuit I drew shouldn't have any cross talk due to mismatched resistors.

 

[Armaegis]

I've got it set as unity gain so all the resistors are the same value (~4.7k), and I can't really match them since I only have ten of them and they're all in use. Even though the cheap box says 5%, they're all closer to 1-2% of each other.
 
What makes your circuit less prone to crosstalk than mine? There's gotta be more to crosstalk than just resistor matching... what about selection of opamps, signal paths, power supply, etc?

 

[porthillsbomber]

Looking at our ground channels they're the same with gain 1. Lets say R3 is 5% higher than it should be then you will have
OG = G' = -1.05/4*(L+R).
 
Then for your first method
OL = (3L - R)/4
so
OL - OG = 4.05/4*L + 0.05/4*R
 
but for my method
OL = L + OG
no matter what OG is so
OL - OG = L
 
your alternate method would do the same if it worked but it puts out -OL instead of OL.
 
There are other causes of crosstalk but the are probably small, mainly signal paths and power supply but as long as your batteries aren't flat I can't imagine they contribute to cross talk much. The breadboard could be a problem though, all those unneeded parallel conductors could be talking to each other but I don't think it'l be huge.
 
I'm going to order some parts today, so I'll give this a go next week.
 

 

[Armaegis]

I've currently got my board wired up using your method and I'm still getting bad crosstalk. Maybe it's the way I have my power supply/ground wired up? As I mentioned earlier, I don't have a TLE2426 so I've tried both a resistor divider and opamp based ground, but neither showed appreciable change.

 

[Armaegis]

Bump. Anyone else?

 

[porthillsbomber]

I should have my parts in a few days, I ordered them last Tuesday, so I will be giving this a go soon. I'll report back when I've got some results.

 

[porthillsbomber]

It works!

My order of parts turned up and I whacked this together following the schematic I posted above,
 
R1=R2=R4=R5=R8=R9=1k, R3=470 (should be 500 but that was all I had), R6=R7=R10=R11=2k all 1%
 
DC offset of 0.5mV and 1.3mV. No crosstalk heard with either of the inputs disconnected. I did notice however that with no power connected very quiet sound still passes through but I figure that happens with any inverting opamp. Ground channel is half of a opa2132 and left and right channels on a 2134. Sounds alright too.

 

[Armaegis]

How do you have your power supply set up? Do you get any crosstalk when all the inputs are connected?
 
I'm still not sure why mine had such terrible crosstalk.