Output two signals: one unipolar 0-5v, one bipolar +/-2.5v

Okay, so bear with me because I'm a complete circuitry noob. I've already created a sketch (and basic external circuitry) that can accomplish my thread title. There's a catch though....I can only get one signal working at a time - not both together. I'm currently using an Arduino Uno and some resistors. Both signals are square waves coming from the digital outputs. I've got a voltage divider that gives me a 2.5v source which I use to offset the signal I need to be bipolar; this works fine by itself. The other signal is coming directly from one of the digital outputs; this signal also works by itself. However, when I have the the bipolar signal working correctly, as soon as I connect my oscilloscope ground lead to the ground of the unipolar signal, the bipolar signal shifts back to 0-5v! I've been banging my head against the desk for a couple days now and assume it's a problem with my circuit.

I searched and found this which leads me to believe I can't do it this way (reply #3 from DVDdoug):http://arduino.cc/forum/index.php/topic,90071.0.html

A virtual ground can work if you don't connect to anything with a "real" ground to your virtual ground circuit.

It's pretty simple. Ground is just a reference point (usually connected to earth ground somewhere back at the AC power supply). If you connect a pair of equal-value resistors in series between 5V and ground, you will get 2.5V (referenced to true ground) where the two resistors connect together. If we call that "virtual ground", we can connect the common/ground lead of our voltmeter to that virtual ground. If you now connect the + lead of the meter to +5V, the meter will read +2.5V. If we connect the + lead of the meter to true-ground, the meter will read -2.5V.

Resistors usually don't work too well, because as soon as we connect something that draws current* (say an LED) you muck-up the voltage divider, and your virtual ground no-longer has 2.5V relative to true ground. The solution is to use an active circuit (transistors, or an amplifier, etc.) to hold the virtual ground at 2.5V.

Here is what I have so far:

I'm hoping this is an easy fix and is just a simple problem with my circuit. Or will I need something a little more complicated like a transistor or op-amp? Thanks in advance!

There is a TLE2426 (data sheet enclosed) that is designed to provide what you want. It is a device intended to provide a "Virtual" ground for use with Bi-Polar devices like Op-Amps. It is a great reference but capable of only +/- 20 mA. So if you need real power you will have to build or buy one.

Bob

TLE2426.pdf (670 KB)

Wow, thanks for the quick response! I'll order one of these and report back with the news.

Well, I replaced the voltage divider with a TLE2426 and got the same results. Everything looks fine on the bipolar circuit by itself. However, as soon as I connect the ground lead of my oscilloscope to the ground of the unipolar signal, the bipolar signal shifts back to 0-5V. It seems that the virtual ground of 2.5V goes to 0V as soon as I do this because the voltage potential across the ground pin and 1/2 voltage pin of the TLE2426 goes to 0. Is there something wrong with my circuit? If not, any other ideas?