Electronic – inverted input for adc

I am trying to feed a 5V signal to AD7705 which is a 16 bit ADC and expects differential inputs. Do I need to divide my 5V signal using bridge and then feed it to the ADC or can I use it as a single ended ADC.

Do not use a bridge. The result would be measuring how the bridge resistors vary with temperature, not measuring variations of the signal you want to measure.

Probably you can simply connect a fixed voltage to one of the inputs of the ADC and your signal to the other one. But watch out for whether the resulting common mode input could affect the conversion result.

Even better would be to simply switch to a single-ended ADC.

What confuses me is the need of Vref as 2.5V.

You still need to tell the ADC what voltage range you want to measure.

For example, the conversion result of a differential ADC might be given by

$$ D = 2^n \frac{V_+ - V_-}{V_ref}$$

But check the datasheet for your IC because there's other possibilities (for example to allow for both positive and negative differential inputs).

What standard grade "R"values should be selected?

If you want 16-bit accuracy (or even 16-bit precision) you should probably not be using standard grade resistors anywhere in your signal path. But again, a bridge circuit won't convert your single-ended signal to differential.

Pseudo differential takes a full positive swing on the +IN input whether it is unipolar or bipolar. If it is bipolar the -IN needs to be fed the nominal dc offset of the signal fed to +IN.

This is an advantage on analogue systems using a mid-rail generator (such as a voltage reference and op-amp buffer). If the voltage reference moves up or down a little, it doesn't affect the ADC code value produced because that offset movement will affect both input signals together.

For a unipolar input where -IN is 0V and a mid-rail generator is used, if the mid-rail generator shifts a little, then that is reflected as a dc error in the converted result.

I tend to use unipolar configurations (even when a mid-rail generator is used) because there appears to me more choice in ADCs AND I always choose really good references PLUS, in a lot of signal processing I do, I don't care about (say) a 10mV offset end-to-end.

For example, the input signal is a single ended sin function with range from 0 to 3V and a DC bias of 1.5V. Should we connect the IN– to 1.5V or to GND?

• If the dc bias of 1.5 volts and very stable it doesn't matter which you use
• If you don't care too much about small dc offset variations then use either
• If you do care about offset variations then use bipolar pseudo differential