[gaboo]

Warning, armchair engineering at its best follows:

What do you guys think of fully differential op amps. A good one seems to be OPA1632.

Does it make sense to use them for a balanced input & balanced output headphone amp?

I'm thinking of getting Dr. Gilmore's amp (in its balanced config), for cheap: a fully diff op amp and two buffers.

I realize I can obtain the same outputs with a single ended op amp and two buffers, one of which would be inverting. But, unlike the Gilmore amp, it won't be differential "inside". Would it make any difference?

 

[amb]

Quote:

Originally Posted by gaboo What do you guys think of fully differential op amps. A good one seems to be OPA1632. Does it make sense to use them for a balanced input & balanced output headphone amp?

Yes, you can build a balanced headphone amp with this type of opamp.

Quote:

I'm thinking of getting Dr. Gilmore's amp (in its balanced config), for cheap: a fully diff op amp and two buffers.

In the case of the OPA1632, depending on the type of headphones, you might not even need output buffers. The opamp is rated for max 150mA output current. Of course adding a buffer is always a good idea to delegate the output drive away from the opamp. I have not used this opamp personally so can't vouch for its sonics, but the specs look pretty good.

Quote:

I realize I can obtain the same outputs with a single ended op amp and two buffers, one of which would be inverting. But, unlike the Gilmore amp, it won't be differential "inside". Would it make any difference?

I don't see why an op amp operating in inverted mode is any less "differential" than one that is non-inverting. The opamp's input stage is a differential input transistor pair. In non-inverting configuration you need the input to the + input, but in inverting configuration the input goes in the - input. The feedback circuit must be hooked up a bit differently but that's not really consequential. In the Gilmore, the fact that it's all-discrete means that you could save some parts and eliminate the input differential stage for the inverting side, and drive the voltage amplification stage from the phase inverted parts of the non-inverting side.

 

[gaboo]

This app note seems the authoritative resource on fully diff op amps. A couple of things I see obviously true:

- Reduced Even-Order Harmonic Distortion [they cancel out]
- Increased Output Voltage Swing [in a ppa-style amp the ground circuit doesn't do this]

They also claim:

- Differential Structure Rejects Coupled Noise at the Power Supply [isn't a single ended op amp doing this too?]
- Differential Structure Rejects Coupled Noise at the output [blah]

It looks like resistor matching for the two feedback paths is critical for good CMRR though.

You also get a pin that sets output common-mode voltage, not sure how useful that is for headphone amps, since you want that to be 0 anyway. Perhaps put a servo on a (third) feedback path?

And yes, you're right, 1632 already has plenty of output current. So e cheap DiMint(TM) would require only a pair of these... Definitely worth trying. From the app note it seems all diff op amps come with buffers built in (fig 2).

 

[amb]

Quote:

Originally Posted by gaboo - Reduced Even-Order Harmonic Distortion [they cancel out]

True. This is due to the symmetrical nature of the drive.

Quote:

- Increased Output Voltage Swing [in a ppa-style amp the ground circuit doesn't do this]

Also true. Under ideal conditions the voltage swing is doubled, which means four times the power output into a given load. However, this is only important if you need the additional voltage swing.

Quote:

- Differential Structure Rejects Coupled Noise at the Power Supply [isn't a single ended op amp doing this too?]

Not sure what they mean by this. I guess they're implying that the PSRR is better, but curiously I didn't see any PSRR specs in the OPA1632 datasheet.

Quote:

- Differential Structure Rejects Coupled Noise at the output [blah]

Yeah, blah.

Quote:

It looks like resistor matching for the two feedback paths is critical for good CMRR though.

Indeed.

Which leads to an interesting point (and somewhat parallel to this issue) in the Gilmore dynamic balanced amp. Since the inverting side actually runs quasi open-loop (there is no global feedback from the inverting output, but rather relies on the non-inverting side to keep things in check), it would appear that parts matching (transistor hfe, resistor values, etc) would have to be quite exact between the non-inverting and the inverting sides in order for the amp to work its best.

Quote:

You also get a pin that sets output common-mode voltage, not sure how useful that is for headphone amps, since you want that to be 0 anyway. Perhaps put a servo on a (third) feedback path?

Yeah, that's a thought, if the output offset cannot be minimized without one.

Oh and BTW, the OPA1632 seems to have BJT inputs, and has a non-trivial input bias current. This is probably not such a big problem in other applications, but where we want the output DC offset to be as close to zero as possible, this chip may be less desirable than something with FET inputs.

 

[tangent]

Quote:

It looks like resistor matching for the two feedback paths is critical for good CMRR though.

Yes, that's why diff amps are typically connected straight to the signal, with no resistors involved.

I ran into this when designing my personal monitor (headphone amp for musicians): for the balanced receiver -- an in-amp -- to be useful, it had to be the first thing in the circuit. This means the volume control had to be after it.

Interestingly, with an in-amp, gain is simpler to handle with an in-amp: it only requires one resistor with industry-standard in-amp pinouts. (Some nonstandard in-amps provide either no gain, or handle it differently.) There's no worry about degrading CMRR with the gain resistor.

 

[gaboo]

I'd better buy these before Nelson Pass sues TI into oblivion.