Electronics > Beginners

Oscillations in dual-rail voltage reference

(1/1)

Melon:
Hello, I'm trying to design a dual-rail voltage reference. What I want from the circuit is the following:

* Differential outputs centered at a given common mode voltage.
* The application requires very low-noise so I'm using very-low noise op amps and regulator.
* The op amps don't support high capacitor loads (1nF from the OP209 datasheet), and I need to decouple with high capacitor values (10u) at the target chip.
The circuit I have thought of is this one (see fig1), which is a modification of a circuit (fig 2) in "Tips and tricks for designing with voltage references" from TI. Basically the regulator acts as battery and the op amp feedback sets the common mode voltage in between (in this case GND), generating +-Vreg/2 on each side (for the case of equal feedback resistors).

What I've done is put another op amp at the regulator output so I can filter the output (maybe unnecessary since it has very low noise already, but anyway) and control the regulator with a potentiometer (adjustable regulators can't reach low enough voltages so I use a voltage divider). Also I have put a BJT follower stage at the output of the opamps so they provide isolation against the load capacitance from the target circuit. However, I get pretty large oscillations at the outputs. I have tried removing the two 100u capacitors but it makes no matter. So I have the following questions:


* All the regulator decoupling capacitors, should they be referred to GND or the REG_GND (in this case VNEG)? While in the original circuit they are refered to GND, I've seen other topologies of cascade regulators where the capacitors are referred to the REG_GND (fig 3). Maybe this is why it is oscillating?
* Shouldn't the buffer stages decouple the op amps from the load capacitance of the outputs?

* Is there a better topology to achieve the functionality I want?

Kleinstein:
To really make use of the extra transistor stage, one might want an extra capacitor in direct feedback at the OP. Than some capacitance at the output can help to keep RF away from the OPs.

The OPA209 is not really happy and low noise with 10 K resistors. At only some 1.8 V, one can use lower resistance values (e.g. 300-1000 Ohms). A 10 K resistor also has significant noise, way more than the OPA209.

Pots can be tricky with drift and a not so perfect wiper resistance.  The reference is not really good at low frequencies (e.g. 10 Hz), the high frequencies are set by the filter and not so much the reference. So there may be better references instead of just a super low noise regulator.

Depending on weather the ground connection matters, one could have a reference for the sum and than an extra buffered center. This may especially help with the low voltage. A similar circuit could have the negative side of the reference at the negative side.

iMo:
Also mind you do not have a short protection at those outputs..

Navigation

[0] Message Index

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod