Question about cross-coupled balanced output circuits

[SSLtech]

Hey Jaakko,

That version doesn;t work all that well. SSL and (I think) soundcraft used that version (using only 2 op-amps) and the output balance stability is pretty awful.

You really need a 3-opamp version (with a unity-gain inverter to split the signal at the front end) to make things more reliable. -Basically, this version is indeed very fussy about what it sees.

Keith

 

[Mark Burnley]

Yep,

There's a really good paper lying around somewhere that has the layout and details for this improved design. I'll see if I can find it...

Mark

 

[SSLtech]

Hmmmf,

In the 4000 E-series they found on the multitrack buss outputs that this circuit tended to go out of balance a lot. The solution was to add a third op-amp to generate a polarity inversion and start from there. Later G-series cards (from memory) had this version of the circuit. At my old studio we went so far as to piggy-back polarity inverters onto the old boards to generate better balance. The main problem was that if one leg went hotter and the other leg went colder, the headroom got a bit messy and actually varied from channel to channel theoretically by up to as much as 6dB...

I'll see if I can sketch something up.

Keith

 

[TomWaterman]

Hey,
this is what Fred Forssell mentioned to me when I asked him about the G.D Pontis x-coupled O/P stage....I hope he doesn't mind me copying/pasting.....

Me:
Does it suffer from possible instability because of the mixture of positive and negative feedback?

FF:
No, it is quite stable. Gain can be dialed into the topology, but at a cost
of comprimise to the "floating" quality of the design.

You MUST think of this topology as a transformer secondary.
You would not (and could not) take a signal from only on side of a transformer's secondary, right? Don't do it with this design either, even though you could and have it "work". Common mode performance is handled properly when the design is used just like a transformer.

Me:
I notice a few TBA values on your schematic - trim pots for CMRR etc. I
would really like to know how the values in the circuit are determined.

FF:
The actual values can be fairly randomly chosen, BUT the ratio of values and the matching is very important. I think the ratio of input resistor value to the feedback and cross-couple R value should be no greater than 0.5. Equal value input and feedback R's, (ratio=1) are ok (I think) but be careful about increasing the ratio. Also important is that the circut see a low source impedance.

I think that you can eliminate the cmrr trim pots that I drew, use equal 0.1% matched resistors on the input side of the design and equal 0.1% matched resistor on the feedback/x-coupling side of the design and be fine.

Treat this design EXACTLY like a transformer coupled output. Let it drive
balanced loads or unbalanced (single-ended) loads but ALWAYS take the signal between pins 2 and 3... just like a transformer, never between pins 1 and 2 or 3.

I don't remember how various load type effects cmr, but I don't think it
does at all. It should be the same, but don't quote me!

Well as Fred says, don't quote him....just passing the info along....

I haven't had time to go back to this design and have a play with it, but I mean too...doesn't sound too hot after reading Keiths post above!

Keith, is it essential to have all 3-opamps the same type when you go for the 3-stage design? i.e could the invverter be OPA604 and the main O/P amps Forssells?

There is a four-stage design kicking about too. Crookwood use a variation of it in their mastering consoles.

The Barry Porter EQ that NYD posted has an example of it. Supposedly a very solid O/P stage:

http://www.diyfactory.com/data/misc/NetEQ.pdf

-Tom

 

[SSLtech]

[quote author="TomWaterman"]Keith, is it essential to have all 3-opamps the same type when you go for the 3-stage design? i.e could the invverter be OPA604 and the main O/P amps Forssells?[/quote]
By all means use a TL07x, Ne55xx or preferred flavour op-amp for the inverter.

These things look at an unbalanced load (in the sense of 600Ω from pin 2 to ground and open circuit from pin 3 to ground) as if the 600Ω was a comparative short to ground compared to the open leg. That 'slews' the output balance and consequently the feedback difference starts the rest.

I'll see if I can dig up a 3-op-amp version for ya.

Keith

 

[adrianh]

This circuit as discussed in the Rane note
suffers from loss of headroom and as stated above cannot drive
a good signal if one side of the balanced signal is grounded.
It will dump signal into ground.
Real the Rane not I am explaining it very poorly.


Funny good ole transformers never had all
these problems. And they sound better too.

 

[TomWaterman]

Dunno if anyone has a copy of the sound engineers handbook by Glen Balou, but there two different examples of the three op-amp diff.out in there.

-Tom

 

[Bjorn Zetterlund]

The Analog Devices SSM 2142 line driver chip is of the 3-opamp variety...sorry, too tired and lazy to link properly

Bjorn