Balanced "floating" output using op-amps

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

audiox

Well-known member
Joined
Feb 25, 2007
Messages
610
Everyone knows the balanced output circuit that acts like a transformer. If you ground one output pin, the level in the other rises by 6dB so that the total output level remains the same. I saw the circuit first time in HP Journal a long time ago. I don't know if it is invented by HP engineers, probably not.

HP Journal, August 1980 (pages 12-13)
http://www.hpl.hp.com/hpjournal/pdfs/IssuePDFs/1980-08.pdf

In addition to the original HP circuit, I have noticed that a version without resistor R4 and R9 is quite popular too. You can also drive it differentially (feed inverted signal to the points marked with ground symbol).

Studer has their own, relative complex version (page 11 upper left corner).
ftp://ftp.studer.ch/Public/Products/Mixing_Digital/Analog_Mixer_Assemblies/Technical_Info/1911220/1911220.pdf

I have used this circuit only once. It turned out to be unstable despite carefull layout. I have also seen several commercial audio equipment (using this circuit) which start to oscillate when driving long lines. I did some Spice simulations today and even the simulations oscillated at high frequency.

The circuit is so widely used that there must some trick to make it work properly...
 
All these circuits become tricky when you want to achieve the "infinite common mode impedance" Holy Grail. There is a delicate balance (no pun intended) of positive and negative feedback, which can easily lead to unstability. The common trick is to make sure there's a significant load that makes the thing stable when left floating. Basically, there is no problem when either of the output pin is grounded. The series resistor (R6/R10 on the HP schem) introduce enough positive feedback degeneration that stability is guaranteed. When there's no common-mode load on the output, each half stage has very little difference between positive and negative feedback, making it oscillation-prone, in particular if added component tolerances join forces with Murphy. Placing a load on each leg makes them stable in almost any conditions.
Soundcraft, before using dedicated IC's, had their EBOS (Electronically balanced Output Stage) which had two 1k resistors going to the extremities of a 2k2 trimpot with its wiper to ground. So they could trim for a perfectly balanced unloaded output (which would not guarantee a good CMRR or a particularly high common mode impedance).
 
and here is yet another for you history file... (old Barry Porter studio parametric design). http://www.collinsaudio.com/NetEQ.pdf

I have used something similar to the HP 2 opamp approach with good success..  I don't recall the details, but have seen issues (frequency response errors)  when one of two outputs is floating.  You could roll off the positive feedback path but that will degrade HF CM impedance.

JR
 
In his AES paper on balanced building blocks Bruno Putzeys states that "[...] one finds that existing floating output drivers invariably center around trying to make voltage source buffers act “un-voltage-source-like” in the common mode without compromising differential output impedance too much". He shows the design for a current-mode driver that has much better properties (page 3/4).

JDB.
[and alternatively there's the proprietary drive techniques used by THAT; haven't looked to see if there's a related patent with worthwhile information though]
 
The circuit works fine with differential load of a few kilo-ohms. Output level rises (!) when the load gets smaller (with 300 ohm load the output level is +1dB compared to 10 kohm). With 150 ohm differential load the circuit starts to oscillate. For comparison: with a simple balanced output made of two inverting stages, load of 150 ohms is not a problem.

abbey road d enfer said:
When there's no common-mode load on the output, each half stage has very little difference between positive and negative feedback, making it oscillation-prone, in particular if added component tolerances join forces with Murphy. Placing a load on each leg makes them stable in almost any conditions.

I have CM load of 2 kohms from each output pin to ground.

JohnRoberts said:
You could roll off the positive feedback path but that will degrade HF CM impedance.

Resistors R4 and R9 (HP circuit designations) are not included in the circuit, so they are not causing the instability.

The circuit is actually quite similar to SSM2142 (page 1)
http://www.analog.com/static/imported-files/data_sheets/ssm2142.pdf

JohnRoberts said:
I don't recall the details, but have seen issues (frequency response errors)  when one of two outputs is floating. 

For unbalanced use one of the output pins must be grounded. A long time ago I measured frequency response with negative output pin floating. Difference between 100 Hz and 10 kHz was something like 10 dB (highly dependent on load impedance).
 
It's perhaps worth pointing out that a real floating transformer winding will likewise be unhappy with one end unterminated.

One could argue that this design goal of benign behavior when output lines are randomly shorted may be over engineering a solution to something that shouldn't be a problem.

Blame it on the 1/4" interfaces.

But the customer is always right and the manufacturer is always wrong... if the interface doesn't work as expected.

JR
 
audiox said:
The circuit works fine with differential load of a few kilo-ohms. Output level rises (!) when the load gets smaller (with 300 ohm load the output level is +1dB compared to 10 kohm). With 150 ohm differential load the circuit starts to oscillate. For comparison: with a simple balanced output made of two inverting stages, load of 150 ohms is not a problem.
Resistors R4 and R9 (HP circuit designations) are not included in the circuit, so they are not causing the instability.
There is some positive feedback, even without R4 & 9: from output of U1 via R6 via Zload via R5; the lower Zload, the higher the positive feedback.
Silly question: have you put caps across R2/R12 ?

 
Midas uses a very similar circuit for the insert sends and all output amps in the XL series consoles.
I think they should have done this right since the XL3 or XL4 were the leading live-consoles for a long time! If you are interestet in a schematic i can mail it to you!

regards,
Wolfgang
 
wolfgang said:
Midas uses a very similar circuit for the insert sends and all output amps in the XL series consoles.
I think they should have done this right since the XL3 or XL4 were the leading live-consoles for a long time!

I have Midas Venice schematics and it has the standard circuit (which is used in thousands of other equipment too). It works fine if you have load of something like 300 ohms or higher (99.9% of practical applications). The problem occurs if you are driving very very long cable or other difficult load. 

 
This "cross-coupled balanced output stage" was touted around the audio industry in the late '70s by a Peter Leunig from Germany. That's when I first saw it, it was shown to M*I who said, "That's nice" but would not pay for it. They used their (incorrect) version of it in the JH600 series and in all the transformerless circuitry in the JH110 and JH24 recorders, and got an AES paper out of the fact that electronically balancing stuff changed the grounding rules in that you had to carry a ground through somehow to give a reference to the electronic balancing, whereas with transformers you did not carry the shield through (was one way of doing it).
Soundcraft implemented it correctly in that the "sense" should come from the outside-world side of the output coupling capacitor, MCI did not do it this way, thus defeating the purpose OF the cross-coupling. ie without properly grounding the sense, the 6dB doubling of the other stage does not happen correctly.

There is an element of futility in building this, when the SSM 2142 dine driver IS the circuit in a chip!!!
Soundcraft do it with a 5532, 10K/22pf feedback R/C, 4K7 input resistors; dead stable, works. Use SIL resistors to get some semblance of matching, but why bother.
 
Back
Top