idea - partial differential output

[DCameronMauch]

Had an interesting idea the other night...

True differential output are always nice. You get many benefits. Like improved power supply rejection ratio, better handling of back emf (probably not an issue with headphones), lower distortion, blah blah blah. And I have never heard anyone say a true different output sounded anything but awesome. Like the HeadRoom BlockHead. Got wonderful reviews. But the problem is that that unit is not compatable with virtually all headphones. You have to get a special made cord. Not terribly practical.

With normal headphones, the two drivers share a single ground. So any driven signal on the ground goes to both drivers. Like a mono signal. But wait a second. What if you did drive each side with a mono signal. It could be true differential. The shared ground issue goes away. Converting stereo to mono is easy - just sum the channels. Okay. Now that we have the mono component of the sound, getting the stereo only component is also easy - subtract the channels. So we could drive each side with it's own stereo only component in the regular single ended way AND the mono component differentially at the same time. Thus giving a partial differential output. A little math...

inputs: L and R
mono component: (L+R) larger signal
stereo only component: (L-R) and (R-L) smaller signal

drive left side with (L-R)+(L+R)/2
drive right side with (R-L)+(L+R)/2
drive ground with -(L+R)/2


differential left -> (L-R)+(L+R)/2 - -(L+R)/2
-> (L-R)+(L+R)
-> 2L

differential right -> (R-L)+(L+R)/2 - -(L+R)/2
-> (R-L)+(L+R)
-> 2R

These components can be easily generated with two differential and three summing opamp configurations. I can whip up a basic schematic if anyone would like to see it. This idea presents a few questions. Like how much of the signal is mono and thus how much benefit would there be to gain. I analyzed a dozen songs of various genres and found them to be 60% mono and 80% mono in the lowest three octaves. Not bad. It also brings up the question of possible gain versus the added complexity and possible loss due to that complexity. That one I can not answer. I imagine it will come down to a personal subjective thing.

Just thought I would throw this one out to see what everyone else thought about this idea. Comments? Questions? Am I onto something or just smoking crack?

Thanks!
Cameron

 

[eric343]

You're either a genius or missing something big, and I haven't figured out which... Though I'm leaning towards the former...

But hey - try it; you never know...

 

[DCameronMauch]

Well, my grandma often tells me that I am only partially stupid...

Here is a very basic schematic for the circuit. Note that the summers also invert. Found this company, alpha, which makes super precision resistors. Would be useful for this design.

Cameron

 

[Won]

Hi. I've been lurking in this forum for a while now, and I thought this would be an interesting thread to contribute to.

So I've thought a bit about this too so you get my vote for "genius". Almost.

Say you have left, right and common.

Drive left: L-R
Drive right: R-L
Drive common: -(L+R)

So the left driver gets (L-R) + (L+R) = 2L
and the right gets (R-L) + (L+R) = 2R

So I don't know why you need all those divide-by twos...

Also, I'm a bit disappointed by your choice of using opamps to perform the summing/subtracting. A better choice might be to use differential amps. For example, you can get L-R and R-L from one diff amp, which could be as simple as a current-source biased LTP made with discrete components. The adding circuits can be made by driving matched parallel buffers at the output, which can be as simple as some discrete followers (Szekeres buffer?).

What makes this interesting is that you might be able to accomplish certain tasks simpler (crossfeed, DC servo) because you are explicitly factoring out the common and differential signals.

-Won

 

[DCameronMauch]

Hey Won,

Thanks for the feedback. Your math adds up the same, but then you are not differentially driving the mono component. That is why I split the mono component into two out of phase pieces. Only then can the mono component be driven differentially. Which was kinda the goal of the whole idea. I totally agree with you about implementation of such a design. My circuit schematic was strictly to demonstrate the idea. And the most basic way of deriving the appropriate signals. Wether or not the improvement out weighs the complexity lays in all the nasty little details of the implementation.

Cameron

 

[Won]

I thought the point was to differentially drive the headphones -- to have a quasi-bridged output. In that case, the common mode signal should be sent down the common wire and the differential mode signals should be sent down L/R.

On the other hand, I guess if the mono signal is where most of the musical information is (bass) then that is what you'd want to drive differentially...

-Won

 

[eric343]

Hrm...

I think the problem you might run into is that the sheer complexity of the circuit (opamps...) will make it quite difficult to get the kind of sonic performance you want...

Perhaps discrete would be the way to go... PPL? Kevin?

 

[kevin gilmore]

quote

I think the problem you might run into is that the sheer complexity of the circuit (opamps...)
will make it quite difficult to get the kind of sonic performance you want...

You bet. Lots of extra stuff to make the sound worse. Way
to much extra circuitry, making audiophile quality virtually
impossible, or really expensive.

The bottom line of this is that there is absolutely no benefit
to this sort of arangement. Lets look at grado's for a minute,
they have one common wire going up more than half the
way to the junction that splits between the left and right
channels. This wire has resistance. Maybe .1 or .2 ohms.
The extra added circuitry could not possibly account for this.
(without taking feedback from the common junction)

You would be a whole lot better off, modifying the headphones
to be dual mono, then drive the headphones with a balanced
bridge amp. Why go thru the effort of a quasi bridge, when a
real bridge would cost less.

And don't forget about all the common mode problems you can
get into when there is a loud sound in only one channel. It will
definitely bleed into the other channel ruining the seperation.
The resistors necessary would have to be of extremely high
precision to do something like this.