Opa + Buf circuit question

[dsavitsk]

Can somebody explain the distinction between these two circuits to me? I understand that the one on the left is in the datasheet, but is there a reason that it is superior to the one on the right?

 

[Dougigs]

The one on the right shows an op amp acting as a non-inverting amplifier, with its output hooked up to an open-loop buffer chip. The two in this case are in series, and therefore are almost independent; the buffer acts as a voltage follower. The op amp has a very large amount of feedback, which assures stability but tends to compress the sound.

The one on the left, as you can see, has the buffer included inside the op amp's feedback loop. This is a significant difference; it means that the op amp isn't as heavily loaded, it means that the amount of global feedback is lowered. It has lower distortion, lower output impedance.

The circuit on the left was pioneered by Walt Jung and is known as the Jung Multiloop; it forms the basis of many of the best headphone designs. (One slight distinction: In Jung's original description, the op amp on the left would also have its own feedback loop). In my opinion it is the best way to use op amps, if you must use op amps (and if you want something that can fit in your pocket, you must). If you bias the op amp into class A, and you operate the buffer in a class A range, then you have a truly top quality amplifier this way.

But let the designer himself speak for the merits of this circuit -- here is Walt Jung's original desctiption:

http://www.elecdesign.com/Globals/Pl...tent/1518.html

 

[~n00beR]

*deleted*

above post gives my explanation an ass whoopin

 

[tangent]

Quote:

Originally Posted by dsavitsk is there a reason that it is superior to the one on the right?

As Dougigs said, the first circuit has lower output impedance and lower distortions, when the buffer is an open-loop design like the BUF634. If you use a closed-loop buffer (or an op-amp configured that way) then the second configuration may be better.

Quote:

Originally Posted by Dougigs The op amp has a very large amount of feedback, which assures stability but tends to compress the sound.

Pfft. Distortions aside, voltage in equals voltage out in the first circuit.

If you want to argue that the distortion you get from having the buffer inside the feedback loop is less euphonious than the distortion you get from keeping the buffer outside the loop -- perhaps because the distortion types differ -- that's fine, we can argue that separately. But hand-waving generalizations like "feedback compresses the sound" (forgive me if that's an unfair restatement) are horse hockey.

Quote:

The circuit on the left was pioneered by Walt Jung and is known as the Jung Multiloop

No. You see that circuit given in Figure 1 of the Jung "Minimizing Input Errors" article, but that is only as an example of existing practice. It was probably first implemented by some forgotten engineer back when vacuum tubes were high tech.

Figure 2 is Jung multiloop.

 

[dsavitsk]

Thanks for the explanations. Not to be dense, but what is the distinction between an open and closed loop buffer?

Also, assuming that a buf634 is within the feedback loop of an opa627, does it effectively take over the job of delivering current? If so, does this lower the amount of current draw one needs to bias the opa into class a? And last, if the opa627/buf634 combination is being used for line level signals (as a phono stage with the riaa filter in the feedback loop), what is a reasonable current for class a bias?

-d

 

[tangent]

Quote:

Originally Posted by dsavitsk what is the distinction between an open and closed loop buffer?

The "loop" here is a feedback loop. So open loop means the buffer has no feedback loop. Examples are the BUF634, the Elantec EL200x series, the HA-5002, plus the discrete buffers in the PPAv2 and M3.

A "closed loop buffer" is often just a high-current op-amp with OUT tied to -IN. But not always.

Quote:

Also, assuming that a buf634 is within the feedback loop of an opa627, does it effectively take over the job of delivering current?

Yes, regardless of what op-amp you use. It is not a particular property of the BUF634 + OPA627 combination.

Quote:

If so, does this lower the amount of current draw one needs to bias the opa into class a?

Yes. In fact, buffering the op-amp is the only practical way to allow class A biasing with standard op-amps. Without this, you'd probably have to use so much bias current that the output protection circuitry would be activated.

Quote:

And last, if the opa627/buf634 combination is being used for line level signals (as a phono stage with the riaa filter in the feedback loop), what is a reasonable current for class a bias?

The same as you'd use for any other signals, because the op-amp's load -- the buffer -- is the same.

 

[Garbz]

Just an asside about the class-A bias. The bias depends on the load on the opamp correct? The BUF634 presents an extremely high load to the opamp which is why a very small bias current pushes the circuit into class-A. But what if the buffer is omitted for a preamp?

Am I right in assuming that a small 2mA bias current on the output of the opamp would still be sufficient for a class-A bias used as a phono stage or a preamp where the following stage is in the 50k-500k range for input impedance?

So bias current really depends on the following stage's impedance?

 

[stadams]

Quote:

Originally Posted by Garbz Just an asside about the class-A bias. The bias depends on the load on the opamp correct?

Not completely. The bias also depends on the maximum voltage swing that you desire into the load as well as the impedance of the load itself.

Quote:

The BUF634 presents an extremely high load to the opamp which is why a very small bias current pushes the circuit into class-A. But what if the buffer is omitted for a preamp? Am I right in assuming that a small 2mA bias current on the output of the opamp would still be sufficient for a class-A bias used as a phono stage or a preamp where the following stage is in the 50k-500k range for input impedance?

As you have suspected the current level needed for class A biasing throughout the entire voltage swing will change between different loads. Considering that as a worst case the BUF 634 has an input impedance 8 MOhms, 2 mA is more than sufficient in most cases to allow the op-amp to stay in class A.

In the case of removing the buffer and replacing it with a much lower impedance device or stage, you are effectively decreasing the bias current available for biasing the output stage of the op-amp by increasing the current that flows in the new input stage. As far as the current source/sink sees, it must supply 2 mA, the circuit the current is supplied to decides in what proportions the current flows, and for the current source/sink, the output of the op-amp and the new input stage are in parallel.

The math is easy just Thevenin's, Kirchoff's Current, and Ohm's laws to determine how the current is actually divided once the output impedance is known for the op-amp. Come to think of it, I believe this has been discussed many, many times a couple of years ago on a few other boards when this form of biasing was becoming "en vogue" in the diy audio realm.

Quote:

So bias current really depends on the following stage's impedance?

The short answer, yes.

 

[Excalibur]

Here's a nice application of this circuit.

http://www.pha.inecnet.cz/macura/buffer_en.html