building an active balanced ground?

[qusp]

Quote:

You lost me after "I've built one Bottlehead..."
 
The meier link doesn't work.
 
 
It seems to me like an awful lot of work to avoid reterminating the headphones, or using AMB style zero volts active ground.

hahaha I said those exact same words in the stepdance thread after I figured out what he was doing. seems to me it uses at least as much power and at least as many components as a 4 channel balanced amp. it may well sound realy good and its quite clever, but does seem like a lot of trouble to go to 'to travel the path less travelled' all the same it really is quite interesting, so i'll be watching on to see how it goes,
 
you may think its easier to do with discrete SS devices like jfets, but opamps are basically made for this kind of job, so use them to the fullest IMO and there are some really nice sounding ones these days too, if well implemented 

 

[qusp]

Quote:

Why would you need to double the ground channel? You're tapping two sets of wires off it, so your voltage is the same in each except now you have two of them, so it all balances out.
 

just because there is 2 wires does not mean there is double the signal strength, the way I see it you would indeed need to either double the channel, or double the signal strength and tap twice; tapping a single channel twice will simply give you half the signal strength on each wire

 

[ClieOS]

Well Dr. Meier sold a lot more Stepdance then he ever thought he will (a year of planned stock runs dry in first few 2~3 months), so he must be doing something right.

 

[qusp]

agreed, not arguing with that, i'm looking forward to having a listen at the next meet; just seems like the long way around to me thats all.  suppose there are lots of people tyhat would like some of the benefits of balanced without reterminating their headphones. BTW see edited first post, didnt come out right first time.

 

[holland]

Quote:

Why would you need to double the ground channel? You're tapping two sets of wires off it, so your voltage is the same in each except now you have two of them, so it all balances out. (I suspect we're still agreeing on the same thing but the communication is just misfiring)
 
The -(L+R)/3 was in the case where the headphone cable only has three wires running through the main length (Meier's calculations assume 4, with a separate ground wire for each channel). For the three wire case, you want:  (L+G) + (R+G) + G = 0
 
If you used -(L+G)/4 in the three wire, your audio is still fine, but now you still have an electric field.

 
Because his equations uses 4 wires.  If you want to change it to 3 wires, that's fine, but I'm not sure it meets his hypothesis.  My prior response before editing was to divide by 3 as the factor, but it doesn't meet his initial equations.
 
In order to take 4 signals and make it into 3 outputs for the TRS jack, G must be 2x the factor.  Solve the equations and it will be pretty clear. 
 
L + G + G + R + G + G = 0
L + R + 4G = 0
 
G = - (L + R) / 4 for each G.  Now *physical* G must be 2x G.  why?  
 
(L' + Gf) + Gf + (R' + Gf) + Gf = 0
(L' + Gf) + (R' + Gf) + 2*Gf = 0.
 
L = (L' + Gf)
R = (R' + Gf)
G = 2*Gf
 
That's just working off the equations he presented.
 
Now the physical reaction of the drivers seems to be off and -(L+R)/3 would seem correct to me for G, but what do I know?  It's not my theory.
 
 
 

 

[ph33lix]

Don't mind my intrusion, I'm definitely keeping an eye on this topic.
 
At first I was wondering about maybe using an additional opamp (per channel) at unity gain and use its inverting input to tap into an amp's circuit's output for a pseudo balanced system, but that would require a 4 conductor setup.
 
Being able to make a 3-wire balanced setup (hope I'm catching on with the thread's concept properly) makes it much more compatible and fuss-free for cans/buds swapping. 

 

[Armaegis]

I kinda feel like we're going to wind up hacking a mini^3 to act as a preamp "active balanced ground" stage and then hack that into a M^3 somehow

 
 
edit: actually... is that doable?

 

[Armaegis]

Quote:

Quote:

Why would you need to double the ground channel? You're tapping two sets of wires off it, so your voltage is the same in each except now you have two of them, so it all balances out.
 

just because there is 2 wires does not mean there is double the signal strength, the way I see it you would indeed need to either double the channel, or double the signal strength and tap twice; tapping a single channel twice will simply give you half the signal strength on each wire

Er... we're tapping a voltage signal correct?  Tapping it twice does not diminish that signal. The channel will need to pump out more power since it's feeding both sides, but the signal remains the same.
 
At least this is how I understand things (it's sort of like a parallel circuit... sort of).
 
An issue I see though is that since loading is complex and impedance changes with signal, things might imbalance... but I guess this is a problem inherent in a normal setup anyways and isn't something that can be accounted for (at least not with my limited knowledge of amp and headphone design).

 

[Armaegis]

I hope this shows up ok...
 
simple circuit diagrams

 
option 1:
Use a summer with feedback to get the G' signal.
We can use a differential to get the L' and R' signal.
Send these three signals into their own amp channels (like in a B22 with three amp channels instead of two plus one active ground channel).
 
Doing it this way means we tap the original L and R three times each. Resistor values (all relative) chosen for simplicity and so that all three present the same load (I think), though I have no idea if that makes a difference going into an opamp.
 
option 2
Alternatively (and I don't know if this is better or not), create the G' signal then add that into the original L and R to get -L' and -R' respectively. Invert while amping later as appropriate. The L and R signals only get tapped twice this way, but the G' gets tapped three times (once each to create L' and R', and again for the final amping).
 
Well that's it for the circuit theory (I think), and that's as far as my understanding will take me. Implementing it, I need you guys to help with that.
 
 
edit: and if going by that "3 wire" thing that I was yammering about instead of the assumed 4 wires, then change the feedback resistor to 1.333 and the ground resistor to 4.

 

[Armaegis]

More random thoughts:
- Tapping the source multiple times feels like it might cause problems. Would adding an input buffer (which could also potentially be a voltage amping stage) be overkill?
- where should the volume pot go? at the input stage (so only L/R) or after the conversion (L'/R'/G') which would require a three channel pot
 
 
 
L/R input ---> opamp buffer ---> separate taps for the L'/R'/G' channels ---> a separate amp stage for each? or combine that into the L'/R'/G' creation stage?
 
 
With option 1, it feels like a cleaner setup to me as each input is tapped evenly (I think), but I don't think a differential circuit can amplify, so there would have to be a separate amping stage for the L'/R' channels.
 
With option 2, the signal taps are a little more awkward (though does it make a difference?) but all the summing circuits can also amplify by increasing the feedback resistor. Hmm, and the Stepdance works by increasing gain rather than attenuation (I think???). I wonder if this is where he implements it. (utter speculation on my part, I'm grabbing ideas out of thin air)
 
 
Thoughts?

 

[Armaegis]

Bump? (I figure folks are still on holiday mode)

 

[Armaegis]

I feel like I'm talking to myself here...
 
Anyhow, I went shopping for supplies today. It's one of those things where I figure probably shouldn't have (and saved my money for buying an actual working amp), but sometimes I just gotta see if my wacky ideas work right?
 
So thus, proof of concept: (it's ugly, I know)

 
Using method 2:
to generate G*: opamp LT1352, input resistors 390 and feedback 150 (I also tried it with 39 and 810 as well just to see what happens)
to generate L*/R*: opamps AD8352, resistors and feedback 390
also ran G* through another AD8352 (resistors 390) to invert it again
power was supplied by 3 AA batteries (measured about 4.8V)
 
Results: well... it works, but it's not pretty right now. It works mechanically (uh, electrically?) but I've got static/buzzing and crosstalk pushing through which is disrupting things. I've also got the physical limitation of cheap parts. The breadboard is kinda spotty at times (especially the power rails) and I got a big box of cheap resistors that are 5% tolerance, so the matching is rather poor. Plus outside EM interference (heck, I can pick up radio signals if I listen closely). I'm guessing there's a bit of time delay/phase shift too, which means some of the signals don't quite perfectly cancel when they're supposed to.
 
I also compared connecting the headphones to G* vs original G. Connected to G, my channels were mixed (as expected). Connecting to G*, they were "normal" except for the static.
 
But what do you guys think? Is this worth pursuing? I'm really at the limits of my knowhow right now. I'm sure capacitors somewhere would help, or fitting it all onto a protoboard and reduce the clutter, or increase the resistors?
 
What happens when I tap an output multiple times? Does the signal decrease or does it stay the same? (I assumed that tapping voltage signals doesn't affect the signal strength, only current, which I *think* isn't a concern if we keep everything small enough)

 

[holland]

It's probably easier to play with this on a simulator.  You can try Tina-TI which is free and uses TI parts.
 
I think you'll need some caps to isolate L and R on the summer.  Other than that, I can't say because I don't know.  I haven't had the chance to think about this.

 

[Armaegis]

Thanks. I never knew about Tina-TI, so I'll have to give that a try tomorrow.
 
I've been fidgeting around with the circuit and got rid of a lot of the static with 50k ground resistors at the three outputs (still picking up radio though). I think now it's mainly trying to get the right balance so the G* properly cancels out.
 
edit: actually, I'm almost (sorta) positive that the problem is phase shift. When I play pure tones, the "tone" is the same but it's distorted, meaning the G* is overlapping but isn't canceling out

 

[Armaegis]

So I've been playing with Tina-TI today (instead of working... shame shame). Simulating the circuit, everything looks like it works out and my final output looks like my original inputs (except inverted). So I dunno... it's possible I didn't wire my circuits up correctly on the breadboard, or maybe I have an improper ground which makes the opamps misbehave (I tried making a virtual ground like on a cmoy, but that didn't work at all and my opamps wouldn't invert).
 
I wish I had some oscilliscopes or something...
 
This is discouraging