Conditioning or splitting bipolar signal for microcontroller ADC

[Sigmoid]

Hey

I'm trying to figure out the best course of action. I'd like to get a reliable measurement of a bipolar signal input on a microcontroller ADC pin.

I have found some solutions for pulling the voltage up, so the entire domain lies within the input range of the ADC, like this: http://masteringelectronicsdesign.com/measure-a-bipolar-signal-with-an-arduino-board/

...however, it would be pretty important to be entirely and completely sure that "zero" IS zero. Zero being zero is slightly more important than the waveform. I wonder, is there a reliable way to split the bipolar signal into a "positive" and a "negative" signal? That way I could use two ADC pins, and get double the resolution, too.

I haven't found any tutorials or circuits on-line that do this. Regular op amps hate having an actual voltage across their inputs, so they are out of the question, but I guess it would be a reasonably simple circuit with ideal amps and diodes... Am I trying to do something really messed up and unnecessary, or am I on the right track?

[katzohki]

Yeah, you could split it into positive and negative halves. Just use a rectifier. There's a handful of ways to actually implement it. Actually you could set it up so that the negative voltages just flip into the positive and then you only need one ADC. That's assuming you can do something useful with that.

I guess you could also separate out positive half and give it to an ADC and then separate and flip negative and give it to another ADC, but that seems inefficient to me.

[dentaku]

I've been looking at ways of offsetting AC signals lately and when I see something like Figure 1 from http://masteringelectronicsdesign.com/measure-a-bipolar-signal-with-an-arduino-board/ I wonder about this because you're connecting an external voltage (2.5V in this case) through two 10k resistors to the output of whatever is creating your -2.5V to 2.5V signal.
Couldn't this be dangerous in certain situations?
I'd feel nervous applying +2.5V to the headphone jack of an audio device, for example. I'm assuming the audio player would have coupling caps on the output blocking DC though.

[T3sl4co1l]

A rectifier will only make things worse.  At best, you're trading an ill-defined offset voltage for a combination of input offset error plus saturation error.

So, what to do?

Simple: measure the ill-defined offset voltage, then measure the offset-ed signal.  Subtract (or use a "differential" ADC or input mode -- but these are either extra hardware or poor performing) and you have your answer.

If you have many signals, you only need to measure the offset once per set; an example I've used before is reading three phase AC line current transformers.  All three have one side biased to VCC/2 (which goes to ADC0), the other side (with suitable burden resistor attached) to ADC1, 2 and 3.  Read all four during each sample interval, and subtract: give or take a few microseconds or so, there lies your answer(s).

Tim

[codeboy2k]

If you have an op-amp signal chain in there somewhere (likely) why not make the first one into a precision full-wave rectifier ?

Then you have the offset error mentioned by T3sl4co1l but that's easily nulled out.

[Sigmoid]

If you have an op-amp signal chain in there somewhere (likely) why not make the first one into a precision full-wave rectifier ?

Then you have the offset error mentioned by T3sl4co1l but that's easily nulled out.

Thanks all. I think this is what I'll try to do. I found some pretty cool tutorials on-line for analog computing.