Sending bias voltage to many single supply opamps.

[Dan Moos]

My project requires me to have sixteen opamps all running off a single twelve volt supply. They are audio buffers if it matters. Any way, I am using a voltage divider to get 6 volts which I am using as a reference voltage with which to bias the opamps at 1/2 the supply voltage. I have never fed so many opamp stages with a single reference voltage before, so I have questions.

First, my final design will run off the mains, and as such, I will easily be able to switch to a true bipolar supply for the opamps. For now, I'm running off of my bench supply, and I only have D.C. to work with. I realize I could use a D.C. to D.C. converter of some sort to get a negative voltage, but that sounds overly complicated considering I will have AC at my disposal in the eventual design. I've never used a D.C. to D.C. converter before either, so I'm open to being sold on the idea.

But if I stick to the original plan, I have a question. What is the best way to distribute this reference voltage to each opamp? Currently I have an individual 1 meg resistor for each section. So imagine an opamp wired as a follower, with the six volts fed through a 1 meg resistor to the non-inverting input. This is what I have now, and it seems to work. It just seems inefficient to have so many resistors. Also, the best value for this resistor seems hard for me to determine. I think my main purpose is to keep the audio signals from interacting. To me that says go big. Noise control seems to say the opposite. Also, I had assumed that with the massively high input impedance of the opamp input, I could use just about any value without the 6 volts being attenuated. This turned out to be incorrect. When I put a smaller resistor in just one section, that section's voltage is lower. Not sure why. My understanding of matters must be off. This seems to imply that maybe there are impedance reasons to consider also.

Any thoughts?

[Mechatrommer]

why you want the 6Vref for? without further asking your purpose, let me just assume... no need 6V line. just use 2 resistors (1-10Kohm each) and a DC block capacitor for biased audio signal. resistors connected to each opamp's power pins. job done, alot less tracks to route.

[ivan747]

You can have a voltage divider, both resistors of the same value, then have an op amp configured as a voltage follower.


Sent from my iPhone using Tapatalk

[Dan Moos]

I don't think you guys understood. I already have a voltage divider. I am just wondering the best way to get that voltage to each follower. Right now I send it through individual 1 meg  resistors. Just want to see if there is a more elegantnt way, and if not, how I should be sizing those resistors.

To the poster wondering why I need this voltage, I am currently running the opamps from a single 12 volt supply. That means that unless I bias the followers higher than the 0v they would naturally sit at,  I can't accommodate negative going signals. Since, as I mentioned, these are audio buffers, that wouldn't work. So by setting the mid point of the quiescent voltage of the buffer at 6v, I can have ~12v peak to peak signals. I present this bias voltage are one of the inputs, and the opamps natural function causes the other input to also go to six volts. Since it is a follower, the output is at six volts too. Voila, six is the new zero! This is a common technique when using opamps in battery powered devices.

[AndersJ]

Ivan747 said it all.
No need for serial resistors.

[Dan Moos]

Wait, is he saying I can use one buffer to replace all the resistors? Is it really that easy? I had considered that, but it seemed to easy. I can't see a problem though.

[JPortici]

i am doing a simillar thing. a set point made from DAC through OPA buffer and then going into four opamps (inverting configuration so it is mixed with the input signal via a resistor)

but i get a lot of crosstalk between channels

[JPortici]

Anyway, if the current consumption at the negative rail is negligible (max 1 mA or some mA) you can get away with a negative charge pump!

[IanB]

It sounds like you have an audio signal that you want to feed into each op amp (AC coupled with a capacitor?), and you want a DC bias to put the audio signal in the middle of the input range?

I think, both for isolation and to avoid cross-talk, you would want a separate voltage divider for each amplifier. Since the 6 V is being used to bias the input signal I don't think it needs buffering, so you just need a pair of resistors for each op amp.

[Mechatrommer]

my assumption is right, its just maybe its not understood. buffer, bias and amplfier all with only 3 elements, no disttribution line. using 1Mohm as distribution line is not elegant imho, as noise even from outside can enter the high impedance node easily. use 10-100Kohm is better if a reference distribution line is really needed. but then 3 elements is still needed in summing amplifier configuration for biased amplfication. so, more components there. by the time you realize each amplifier stage needs different biasing, then your distribution line idea will collapse.

[Zero999]

To maintain good power supply rejection, the potential divider should be bypassed with a suitable capacitor. It can be used to drive several op-amps, as long as its output impedance is low enough and the bypass capacitor is large enough.

A voltage follower can be connected to the potential divider, if a lower DC impedance is required but it may actually have a higher AC impedance than a passive divider and bypass capacitor, especially at higher frequencies.

[Brutte]

My project requires me to have sixteen opamps all running off a single twelve volt supply. They are audio buffers if it matters. Any way, I am using a voltage divider to get 6 volts which I am using as a reference voltage with which to bias the opamps at 1/2 the supply voltage. I have never fed so many opamp stages with a single reference voltage before, so I have questions.

The arrangement is called "noninverting". If you already have a reference voltage at (Vcc-Vss)/2 then it must have some output resistance. The current flowing through 16*noninverting inputs would affect your reference then.
If that is 1uA then on 1M resistor you will have 1V drop there.

Use voltage reference of low output resistance. For example buffered voltage divider. Or two lowish value resistors. Or use op-amps with low input bias current. Or with high bias current but low input offset current.

First, my final design will run off the mains, and as such, I will easily be able to switch to a true bipolar supply for the opamps.

The unipolar or bipolar is just a naming convention. Put a reference somewhere in between Vcc and Vss and you have a bipolar supply.

[Ian.M]

Why not simply use a rail-splitter?  e.g. http://www.ti.com/lit/gpn/tle2426

A  DIY potential divider rail splitter must either pass an impractically hih current to maintain a low output impedance, or it must have a buffer stage - which many people try to do with an OPAMP. Its *DIFFICULT* to design an OPAMP buffer stage that's unconditionally stable with a large capacitive load, so using an OPAMP buffer forces you to compromise on decoupling for the rest of the circuit.  However the TLE2426 is happy with >=1uF on its output, so it is not only much simpler to use, but eases your decoupling issues.

As the derived split rail is the signal ground for the OPAMPs, it makes NO sense whatsoever to split it up.  In fact it should be a solid ground plane with all decoupling for both supply pins of each opamp returned to it.  To reduce interaction, consider low value resistors e.g. 10R, in the +ve and -ve supply to each OPAMP feeding the decoupling cap to the ground plane at the OPAMP pin   Also, you have to be *really* careful with multiple OPAMPs on a single chip, because there is no way of providing local decoupling to each individual OPAMP on the die and their PSRR certainly isn't infinite.

[mark03]

TLE2426 doesn't quite work down to the all-important single-cell Lithium-ion threshold... minimum Vin is 4 volts, not 3.  Also, the output noise is pretty high for precision applications.  Is there a modern equivalent?