Op amp Rail splitter buffered with transistors

[saint gillis]

Here are two rail spliting circuits using an op amp and transistors to provide more power :

https://www.sonelec-musique.com/images/electronique_masse_virtuelle_004.gif

https://groupdiy.com/attachments/qoj8c-jpg.85379/



Please comment the differences between the two topologies !

 

[Bo Deadly]

First one is troubling because the negative input will see a relatively high impedance and slew harder at the zero crossing because there's such a difference in load. That could equate to noise and distortion. It would be much better if there were also a resistor in series with the output of the op amp and overall output to provide a lower impedance feedback path.

The second one is ok.

However, I would argue that these sort of buffers are not really necessary anymore. When op amp power was limited to +-20mA they were useful. But you can find op amps that put out +-50mA no problem for less than $1. An NJM4556 is the go-to cheap but powerful and yet relatively high fidelity op amp. Or parallel just about any amp by connecting the outputs together through 10R resistors (outside the feedback loop). An NJM4556 wired in parallel is a dandy headphone driver. Although in that particular case I would add a little series resistance inside the feedback loop to limit power so that you don't accidentally destroy your ears but without compromising output impedance and thus fidelity.

 

[abbey road d enfer]

The first one is plagued with cross-over distortion. That means that its output impedance close to 0V (which is where it should stand) is quite high. It is compensated by capacitors C2 & C3, though.
As Bo explains, this kind of buffered circuit is often not necessary, however it depends on the particular load requirements of the circuit.
The second circuit should be fine, but I would strongly suggest adding capacitors.

 

[saint gillis]

Thank you for the replies.
About paralleling 2 op amps (whatever the op amp in question) , something like that ?

 

[abbey road d enfer]

Thank you for the replies.
About paralleling 2 op amps (whatever the op amp in question) , something like that ?

That's exactly it!

 

[Bo Deadly]

Thank you for the replies.
About paralleling 2 op amps (whatever the op amp in question) , something like that ?

Note that for a virtual ground / rail-splitter it is very important that the 10R are inside the feedback loop. Normally you see them outside the feedback loop to stabilize driving a line or headphones. A virtual ground is the exception where you want the virtual ground impedance to be as low as possible. Otherwise, if you put the 10Rs on the outside of the feedback loop, your ground impedance would be 5 ohms which could easily cause noise and probably all sorts of problems.

But the 10Rs are important because any offset between the amps will result in a current between them and without the 10Rs that current could be unnecessarily large. If the offset were 10mV for example, then 10mV / 20R = 0.5mA which is tolerable (although having the amps biased slightly might actually help distortion like an amp biased class A).

Also note that the 10Rs define how much current your Vg can handle. If max output swing is +-4V, that's 4V / 5R = 800mA. That would be absurd for a little rail splitter like this. You could probably increase those resistors to 47R and still handle 170mA. That could also reduce distortion because the op amp would be using more voltage to deliver power and not just current which is a lot easier on the amp.

Of course I'm being a little pedantic. In practice, the different rail-splitter circuits will work just fine. Carry on.

 

[saint gillis]

Of course I'm being a little pedantic. In practice, the different rail-splitter circuits will work just fine.

No you're not, this is very educational !