I am using AD5933 as a three electrode system where the 20mV peak to peak and 270mV dc shift as an excitation is given across reference and working and need to measure impedance and phase from working and counter electrode.

Can you add a diagram of the circuit you are trying to build?

You may find useful information in this post  AD5933 4-wire configuration question 

GOPANJANA said:

As this supports 2 electrode system and can generate 200mV minimum peak to peak sinusoidal waveform so is it possible to generate 20mV pk-pk?

It is not possible to program in the 20mV pk-pk voltage into the chip, but why not using something as simple as 10:1 two-resistor voltage divider to get 20mV from 200mV? Or program in 2V pk-pk excitation voltage and use 100:1 voltage divider to get the same result?
Indeed, a diagram and some more information about your your (electrochemistry?) setup would be useful.

.

The diagrams are not visible. Try uploading the file instead of pasting.

Thanks for the help. I am getting 20mVpk-pk and 270mV offset by using an external circuit of voltage divider 1: 10 with a capacitor in parallel acting as a LPF to reduce down the noise . But I am having one question as I was testing the AD5933 placing a 1Kohm resistor in Z unknown to check the response of a resistor. I am sorting the RFB and giving 1Kohm calibration impedance in the software with a given frequency range of 1KHz to 100Khz and everytime the mid point gain factor is changing.

is my understanding correct?

To implement this three eletrode system with AD5933 i will connect Vout of AD5933 to CE and Vin to WE but planning to add a 2:1 MUX with RE and CE so that after calculating the gain factor RE and CE will be connected and CE and WE will be connected then will click on sweep to get the impedance and phase angle

Again, a diagram and some more information about your your (electrochemistry?) setup would be useful - without it is very difficult to guess what your circuit is and what it does.
The 1 to 100 KHz frequency range it way too wide to rely on mid-point calibration. Also if you are running your AD5933 on a 16 MHz clock or on internal oscillator, for the frequencies below 10 KHz the built-in "DFT" produces quite a lot of systematic error.
Not sure how your divider and LPF are arranged, one thing to check would be that your 1 kOhm resistor is not too much of a load for your divider/LPF circuit.

With this configuration you get 20 mVpp at the input of the ADC, which is 0.6% of the signal range, incurring a large error due to quantization.

Try setting the PGA gain to 5, or using a larger RFB value so that a larger signal is seen at the ADC input.

With such a small excitation, the system is prone to noise and interference, which affects the precision of the measurement.