something I have´t seen in this forum - discreet balancing

[rafafredd]

Well, I´m just bringing this idea so that maybe the big fellas has something to say about it, or maybe post a link.

Why haven´t I seen a discreet balancing output stage (buffer) here?

Is it that hard?

I know we could use three discreet opamps or just two, to make a discreet balancing unity gain output, but that´s not the idea.

The idea is to have just one unity-gain stable discreet opamp with balanced outputs.

So, any thoughts?

 

[tk@halmi]

Perhaps, you have not seen it because everyone is busy adding an output transformer to whatever they make, and for that you just need a single ended output.

Some other possible reasons:
- Chip opamps, even a 5532, can do it extremely well at low cost
- Other than for some vintage equipment there is no need to have a high power output
- It does not have a transformer!

Let's brainstorm a little, you need a phase splitter, maybe a FET followed, by a couple of high current BJT emitter followers. For some downright toasty sound you could do it with three transistors. :grin:

Tamas

 

[SSLtech]

Think about it a little more...

'dumb' differential outputs are more of a problem than a help if they encounter unbalanced loads that tie a leg to ground. They cause intolerably high ground currents and thereby crosstalk, distortion and other manner of trouble.

To solve the problem, cross-coupled, output-sensing feedback is one method. it sniffs both outputs after small (50Ω or similar) resistors, then if one end is shorted to ground, half of the negative feedback to the other leg is killed, doubling the output of that leg. At the same time, the other leg's "output sniffer" sees double output, and doubles the negative input to the first leg, thereby killing the signal to that leg, and preventing any ground current. Another method is one I saw used by Neve, where they sense a mid-point between the two outputs (which in balance will be silent, then derive two opposite-polarity feedback signals from the mid point, to squirt back to the two legs. This does the correction with a similar aim in mind, but it just senses the error in a different way.

Doing this discretely to achieve a 'smart' (output load sensing) balanced output would essentially involve differential inputs, like an op-amp has. The inputs would almost certainly need to be differential in order to make the negative feedback/positive signal discrimination. The outputs would have to be decently capable in terms of current drive.... -Guess what? -You're really just making discrete op-amps: -several of them- to do a single function.

Since discrete op-amps are enough of a challenge themselves just to do the Melcor/API type of circuits, I'd say this is a job best left to a decent current-capable drive stage and a transformer.

...Besides... most people here seem to believe that "the sound is in the iron", and -you know what? -The sound that they like probably really is in the iron, so this is a lot of trouble for a task that they're not very inclined to investigate.

-My thoughts, anyway!

Keith

 

[jhaible]

[quote author="SSLtech"]Think about it a little more...

'dumb' differential outputs are more of a problem than a help if they encounter unbalanced loads that tie a leg to ground.[/quote]

Absolutely. It should be forbidden by law to market such outputs with the same words as a real, floating differential output. I just tried to enquire from the manufacturer what type of "balanced" output a certain very popular 8x8 soundcard has - just complementary signals or electronically floating. (I knew it wasn't transformer balanced.) They outright told me they would not give away such information. Which kind of also answered my question ... but I disgress.

To solve the problem, cross-coupled, output-sensing feedback is one method. it sniffs both outputs after small (50Ω or similar) resistors, then if one end is shorted to ground, half of the negative feedback to the other leg is killed, doubling the output of that leg. At the same time, the other leg's "output sniffer" sees double output, and doubles the negative input to the first leg, thereby killing the signal to that leg, and preventing any ground current. Another method is one I saw used by Neve, where they sense a mid-point between the two outputs (which in balance will be silent, then derive two opposite-polarity feedback signals from the mid point, to squirt back to the two legs. This does the correction with a similar aim in mind, but it just senses the error in a different way.

I'm familiar with the first method (SSM chips for instance), but not with the second (Neve). Can you point me to some schematics for this, please?

Thanks,

JH.

 

[Samuel Groner]

Can you point me to some schematics for this, please?

If you have a copy of Ballou's Handbook for Sound Engineers at hand, check page 738, fig. 25-52A. I think that's what SSLTech was talking about.

Samuel

 

[ijr]

Or see TI... ths41xx are low voltage balopamps,
search across ti website they have elsewhere block schem.

 

[jhaible]

[quote author="ijr"]Or see TI... ths41xx are low voltage balopamps,
search across ti website they have elsewhere block schem.[/quote]

Thanks for the hint, I just downloaded it. I'll take a look. Looks
interesting.

JH.

 

[ijr]

If someone have a hint about good implementation Vocm
in discrete transistor terms...put it here))
(actually....not simple....)
To build balanced opamp at +/-30-60 v is not very big deal;
like 14-17 transistors and you're happy.
To find good return point for Vocm signal... HEADACHE!

 

[jhaible]

[quote author="ijr"]If someone have a hint about good implementation Vocm
in discrete transistor terms...put it here))
(actually....not simple....)
To build balanced opamp at +/-30-60 v is not very big deal;
like 14-17 transistors and you're happy.
To find good return point for Vocm signal... HEADACHE![/quote]

Well Fig. 18 in the data sheet may give some hint.

I still have to figure out of this whole scheme only makes some tiny OCM corrections, or if you really can tie one of the outputs to GND and get the double signal at the other.

JH.

 

[ijr]

Actually, +/-6db compatibility (in el. bal. out terms) is not the subject of
my interest here; zero common mode, low distortion, and TRUE balanced
are the benefits of this stuff....

 

[ijr]

BTW, in case of further interest on this subject, let's refer to
http://www-s.ti.com/sc/psheets/sloa054d/sloa054d.pdf
page 4
(VERY simplified schem. just for ref.)

 

[SSLtech]

Yep That's the basic idea.

In fact this output methodology is very intolerant of imbalanced load under certain conditions. The first time I tried to drop a Neve V3 into a former SSL installation, the monitor outputs hissed like crazy... it was due to one half of the output driver going super-gain and squirting noise back into the feedback loop.

None of this is entirely simple to the extent of being good DIY material or tolerant of wide variations in transistor quality... for that reason I suggest that either a multi-op-amp or an OPA134/SSM2142 solution is the best idea.

Keith

 

[cuelist]

The later Barry Porter designed an opamp based version some 15 years ago that seems to be the best one so far.

http://savefile.com/files/7233756

Sure, a simple discrete circuit could provide you with complementary outputs (one in phase, the other inverted phase) but it would not come anywhere close to the performance level most people would expect.

It's not easy to mimic a transformer...

 

[Rochey]

As for the THS device mentioned from TI -- you may want to look at the OPA1632... it's very similar, and tested for it's audio qualities.

cheers

R