OPA2134 Power Supply

JCRAcoustics

Member
2019-07-26 2:31 pm
Hi All,

I have been designing a 2-way Linkwitz-Riley Crossover for a Bluetooth amplifier using TI's OPA2134s. I was wondering if anyone could explain how the datasheets supply voltage works?

I was intending on using the boards 24V supply for the positive rail and a potential divider (using 2x 10K resitors and 2x 10uF capacitors) to half that for the virtual ground (12V). On the datasheet however is says supply voltage max +/-18V which gives a swing of 36V. Does this mean I can not use 24V as the positive supply rail as it exceeds +18V even though my voltage swing would technically be 12V?

Here is the link to the datasheet:
http://www.ti.com/lit/ds/symlink/opa2134.pdf

Thanks!
Jonny
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
The voltage rating is for the total voltage the chip sees between pins 4 and 8, which is 36 volts for opamps such as this one. If you used split (dual rail supplies) you could use any combination up to that that 36V maximum such as -30 and +6 or +12 and -24 and so on.

So 24 volts supply is fine. Just remember that the inputs and outputs of the final opamp circuitry will need to be AC coupled, the virtual ground as such can not be used as an audio ground if you are combining it with circuitry already running on single rail.

In other words design your filter as a single rail design, your 10k divider is just for biasing and the audio ground is the 0V line.
 
On top of what Mooly explains, you can very well create a virtual midpoint with two resistors and two capacitors, just as you describe. You can then use this virtual midpoint as reference for the input and output signals and need no AC coupling unless you have other circuitry running from a single rail. To give the midpoint (much) less impedance, I would use the resistors and capacitors you describe as input to an OP-amp in voltage follower coupling.
 

Mooly

Administrator
Paid Member
2007-09-15 8:14 am
The resistors that go to ground (the 100k and 22k's) should go to your virtual earth point.

The device rabbitz mentions is one solution, another would be a single resistor and zener to derive a stable 12 volt reference which is better than just two series resistors. Decouple the zener with a 10uF cap.

The caps that go to ground would return to the 0V line (the true ground that the Bluetooth amp will probably use) and the two audio inputs and four audio outputs need coupling caps. 1uF is fine for the input ones and perhaps 10uF for the outputs. Each output cap should be terminated to ground via a high value resistor to define zero volts DC at each output. Same for the input caps, tie each input cap to ground via something like a 470k.
 
One thing that confuses many people is the notion of single rail v. dual rail opamps.

Its completely meaningless I'm afraid, all opamps are the same, they have a V+ and a V-, and know nothing about what voltage is ground.

Some opamps have inputs than can accept signals at and below V- (by a fraction of a volt only), and such opamps can use V- as ground in a gain stage - this can be thought of as "single rail", but its best not to think that way.

Every opamp circuit has a notional ground, which is usually somewhere between the rails, although can be coincident with one of the rails (esp for low voltage rail-to-rail devices).

For bipolar signals you'll always have ground between the rails, normally at 50%, either due to a dual supply or a virtual-ground derived from a single supply as in this case.

Once in a blue moon you'll encounter an opamp with a ground connection - read the relevant datasheet in that eventuallity!


So when you see an opamp described as having maximum supply of +/-18V, that means a maximum difference between the rails of 36V, ground can be anywhere in that 36V range that's without the input and output swing range.


The reason supplies are usually symmetrical is to maximum signal amplitude before clipping (for a symmetric signal as in audio). With undirectional pulses (such as a photo detector) there's no need for symmetric rails.
 
Last edited: