Headphones: Driving the Transducers at Both Ends Using M/S

[mediatechnology]

This is something I've wanted to try for along time just to see if I could.
I wasn't sure if it had ever been done.

Drive the shield with mono, and tip and ring with difference.
This de-matrixes the Mid Side/Sum Difference in the transducer wiring and can be used with conventional 1/4" or 3.5 mm headphone cables.

I tried it and it works...


Driving Headphones with Sum and Difference

Believe it or not this actually comes out stereo.
You're probably asking why.
Good question.

There are obvious reasons to do this in portable single-supply applications to eliminate the coupling caps.
The mono/mid/sum output can be floated to 1/2 V+ along with the other outputs.
If you have to build a voltage source to hold the sleeve at 1/2 V+ why not make it do something more useful?
Curious, I Googled to see if anyone else had done this.

The point did not escape Freescale Semiconductor in their patent: http://www.ka-electronics.com/images/pd ... 876911.pdf

In the patent they talk about an improvement in voltage swing capability as if this is some form of bridge-tied-load "BTL."
I don't see that happening here since most of the peak level is in the mono channel.
In mono, the Side + and - outputs are held at AC ground.
It's only when something is panned midway between hard left and center (or hard right and center) before any BTL-type advantage occurs.
I think they may have missed something.
Still, kinda cool, offbeat, and maybe useful.
A nice diversion.

One thing that may provide some overall MS insight is that inverting Mid swaps left and right.
Reversing Side Drive to the transducers also swaps left and right.
In this drawing mono/mid/sum drives the - side of the transducer.
To make left and right correct, - Side drives the left transducer; + Side drives the right.

What are the practical benefits for those of us running off +/-15V rails?
Not sure yet. But it sounds really good.
This circuit proves that spare time is a dangerous thing.

[JR.]

They could have just asked me... I did this in a headphone booster amp I designed for peavey's AMR line back in the '80s. where I used that trick, drive the sleeve with -(L+R) then drive L with (L-R), and drive R with (R-L).. This was attractive for simple headphone amps running from general purpose +/-15v rails that could really kick ass into 600 ohms. 2X voltage is 4X power. It works pretty well because most of the energy in stereo program is mixed mono. Strong out of phase material between L and R will not make this happy, but that isn't common in recorded music.

Peavey still sells an HB-1 but it has been repackaged and is only rated at 100mW into 600 ohms so they probably replaced my old design with some cheesy chip driver (Losers- that used to be a hip product).. I used t0-220 transistor buffers on fast opamps and could drive a loudspeaker reasonably well. I ended up using a resistor in series with the rectifier in the PS to current limit it to a level lower than what would melt the thermal fuse in the 1A wall wart. it got louder than you would expect for a headphone amp.

JR

PS: If you ever come across an old HB-1 (grey painted aluminum extrusion) in a junk pile, it may be worth picking up for a dollar.

[mediatechnology]

Cool John. Would you have a resource for an old manual for that product showing it to be prior art? The Freescale patent has a lot of claims that include the filters and ground detection. But it looks like from the patent that they may have convinced the examiner sum/difference drive was unique.

It works pretty well because most of the energy in stereo program is mixed mono. Strong out of phase material between L and R will not make this happy, but that isn't common in recorded music.

I've been looking at that with real music and yes most of the energy is in mono. No big surprise there. The peak level ratios are all over the map for stereo material but I see differences in the L+R and L-R ratios that go from perhaps 6-20 dB higher for mono.

I'm wondering if it would be worthwhile to take some of the -(L+R) fed to the sleeve and inject some of it as +(L+R) into -(L-R) and (L+R) to more equally distribute the peak level. With most music, the maximum peak level of the difference outputs isn't used. As I mentioned earlier for pure mono the tip and ring connections are held at AC ground.

[emrr]

Makes sense, I'd made part of the leap when repairing headphones, but not thought through it all the way. It does seem like the only advantage would occur with a perfect test signal, and never in the real world. Dunno, keep playing with it. I was hearing about all these Parisian clubs and house parties that no longer use sound systems, and instead give everyone a set of headphones. Probably some efficiency tricks could be done with dance music using this sort of headphone drive.

[JR.]

Cool John. Would you have a resource for an old manual for that product showing it to be prior art? The Freescale patent has a lot of claims that include the filters and ground detection. But it looks like from the patent that they may have convinced the examiner sum/difference drive was unique.

No I don't have any docs, this is just one of many products I designed (or re-designed) during my 15 years at Peavey.

I do recall how it came about... I was in a small hole-in-wall recording studio in MS messing around. IIRC we were making a few recordings using mics and gear from the AMR product line we were making and selling. During a break between takes, I decided to check out the monitor mix the drummer was getting through closed back cans... I got them up to about chest level when I stopped because they were sooo loud... Dudley (a really great drummer) asked me if I could make them louder, Of course I could and a product was born...

It works pretty well because most of the energy in stereo program is mixed mono. Strong out of phase material between L and R will not make this happy, but that isn't common in recorded music.

I've been looking at that with real music and yes most of the energy is in mono. No big surprise there. The peak level ratios are all over the map for stereo material but I see differences in the L+R and L-R ratios that go from perhaps 6-20 dB higher for mono.

I'm wondering if it would be worthwhile to take some of the -(L+R) fed to the sleeve and inject some of it as +(L+R) into -(L-R) and (L+R) to more equally distribute the peak level. With most music, the maximum peak level of the difference outputs isn't used. As I mentioned earlier for pure mono the tip and ring connections are held at AC ground.

I guess you can trade some headroom around,,, if you drop the sleeve to -(L+R)/2 you change L and R feeds to L-(R/2), etc... Whatever... this is ancient history to me.. and there may be new crosstalk issues with common impedance in wiring and stuff for tweaky .00x measurements. I haven't thought about this seriously in years, but IIRC there were errors associated with buildout impedance and things...

========
The patent system has always been "leaky" where they let stuff like this through, and expect the litigants to fight it out in court later. The lawyers working in this field are rewarded for getting any patent issued, not to discourage an inventor from filing a weak one.

Dorrough got a rather narrow improvement patent over mine, for making it a curved display instead of straight line, but even they played word games by not using peak and average terminology for the two display metrics. Like "persistence" and some other nonsense.

I haven't heard anything from the Peavey litigation with Behringer over my feedback LED patent lately... I really don't look forward to having to defend my drum tuner patent... My long term hope is to partner with some deep pockets so they can be the brawn to back up my brains.

JR

PS: Not to dime on myself, but I have discovered that one of my other old peavey patents was actually in the public domain at the time I patented it, but I didn't know at the time and the PTO is pretty worthless for finding such stuff. Now I am conflicted, since I assigned the patent to Peavey, and nobody is disputing it, what do I do? I am not going to spend my own money, or advise Peavey to, to spend money to challenge it, while that is the right thing to do. I have talked to the person with the prior art and they are aware, but unconcerned. I should probably advise Peavey to let the patent expire, but even that discussion is awkward (since I plan to use the IP myself).,

[mediatechnology]

I decided to try out an "equal mono power" version to distribute the mono energy equally between the Mid, and "Side" outputs.

Mono (L+R) drives the headphone transducers in common mode.
Difference (L-R) drives the headphone transducers differentially.

The Side outputs operate differentially: +Side drives left, - Side drives Right. (Or vice versa depending on the sign of Mid.)
To be able to split the Mid/Mono between power amps, the Side amp has to be driven in common mode.
The mono signal in both tip and ring has to go "up" as the sleeve goes "down."

Here's how I did it. The circuit keeps getting more "trick."


Headphone Amplifier Mid Side/Sum Diff Equal Mono Drive

The lower L+R stage has it's inverting input fed back to the output.
This provides a Mono/Mid signal that is (L+R)/2.

For proof of concept only I used THAT1646s to directly drive the 'phones.
This allows me to easily flip polarity and provide both CM and Differential drive to an amplifier.

(L+R)/2 feeds both the Mid/Mono PA tied to sleeve and the "Side+Mid" driver in common mode tied to the transducer + terminals.
The use of the inverting output makes the drive to the sleeve -(L+R)/2.
The top 1646, formerly "Side," or Difference is, now Side in differential mode and in common mode, (L+R)/2.

Thus, 1/2 Mono is fed to both ends of the transducer in anti-polarity.

Doing this for mono signals - feeding both ends - allows the peak-to-peak drive voltage to be doubled for mono signals.
Signals panned hard left or right are also made "more equal."
The level ratios between amp outputs are lower and the signals are more equally shared.

For most of the stereo material I'm seeing the "Side" levels - which are actually Side+ 1/2Mid - are running about 6 dB hotter than 1/2 Mono sleeve PA.

This doesn't have to be done with 1646s and the 25R build-out in the PA outputs reduce separation a bit.
This is a proof-of-concept.
With low Zout PAs the limiting factor on separation is likely to be headphone cable resistance.
If anything the I*R drop in the common conductor is likely to widen things a bit.

Bottom line is that under most conditions the peak-peak swings are made more equal thus providing twice the available voltage drive for highly-correlated mono signals.