I need a reference circuit for 4_wire the impedance measurement using AD5934, range 50 to 250 ohm.

You may use this post as a starting point:

4 wire Impedance measurement

In the link, it is mentioned that setting Rcal = 20 can set the impedance range from 1 to 60. Can you please explain this, I did not get this. How can Rcal resistor set the range. 

https://ez.analog.com/dds/f/q-a/28201/ad5933-4-electrode-configuration-for-human-body-thoracic-impedance-measurement-0-60-ohm

Before getting into the weeds with the frequency, when you say Vin voltage, do you mean RFB voltage? From the screenshot we know the raw Re and Im  values, but what were  the raw Re and Im values when you connected the calibration resistor of 100 Ohms? Also, is the software that calculates the impedance something you wrote?  

Let me add the pictures of the waveform.

100Ohm 

120 ohm

AD9534 frequency is 5khz and sweep increment I set to 4 Hz. The acquired signal amplitude does changes but I get 91 Ohm when I place 120 Ohm resistor after calibration.  

After calibration with 100 ohm, when I do measure 100 ohm I get 100 ohm.

Yes, I am checking the signal on RFB resistor. 

For 100 Ω the Re is 3113 and Im is 4171.

I wrote the software for this part. 

Abid said:

For 100 Ω the Re is 3113 and Im is 4171.

Output log shown in pic below.

The suspicion is that your code thinks that the output is inversely proportional to the impedance under test - which is correct when the AD5934 does not have the 4-wire analog circuits between Vout and Vin. With this 4-wire circuit the output is directly proportional to the impedance under test.

In abstract sense the circuit is converting the impedance into integers in the Re and im registers that constitute complex output Re + j*Im. The conversion factor  between physical Ohms and counts in those registers can be viewed as complex gain G. This complex gain can be calibrated with a known resistor: resistor impedance has only real component and zero imaginary component, so Zcal = 100 + j*0 Ohm. The correspondent output is Outcal = Re + j*Im = 3113 + j*4171 = G * (100 + j*0) Ohm, so G = (3113 + j*4171 ) / ((100 + j*0) Ohm). The "unknown" impedance Zx produces output,  which is proportional to Zx: Outx = 3400 + j*4574 = G * Zx, so Zx = Outx / G = (3400 + j*4574)*(100 + j*0) Ohm / (3113 + j*4171) ≈ 109.5 + j*0.21 Ohm. As expected from a resistor, the imaginary component is nearly zero. The real component is 109.5 Ohms, which is a bit far from the expected 120 Ohm, but it could still be within manufacturing tolerance of both the resistors used for calibration and as the "unknown."

Abid said:

For this part I already order 16Mhz crystal, so I will change the crystal. 

The undocumented issue with the "DFT" implemented in the AD5933/4 is that it only approximates the results given by the textbook DFT. Generally speaking, the synthesized  frequency with respect to clock frequency should be relatively high for the "DFT" on the chip to produce reasonably accurate results. To measure at lower frequencies it is often useful to run the chip at lower clock frequency.   

Abid said:

Is this because of low frequency or there can be some other issue?

Since you are running your chip on 1MHz clock the frequency should not cause this, but it could if clock frequency were 16MHz.

Snorlax said:

With this 4-wire circuit the output is directly proportional to the impedance under test.

Yes, I missed that point. Now I am getting 109 Ω when measuring 120 Ω. 

Snorlax said:

The real component is 109.5 Ohms, which is a bit far from the expected 120 Ohm, but it could still be within manufacturing tolerance of both the resistors used for calibration and as the "unknown."

I ordered more resistors (different values) I will calibrate and measure a bunch of different resistors. 

Thanks once again.

This is the result.

Yes, it would be great if you could find some 0.1% or better yet 0,05% resistors. Only avoid wire wound type - those may have too much inductance for AC measurements.

Yes, it would be great if you could find some 0.1% or better yet 0,05% resistors. Only avoid wire wound type - those may have too much inductance for AC measurements.

I did test and calibrate using different resistor values. I noted that the measurement accuracy is better near the calibrated resistance value. I think this is not a problem, this I can improve using some tricks.

The problem now I face, is, when I use 4 wires and short 2 wires together (T1 with R1 and T2 with R2) see the picture.

In this case I can get reading. See the pictures as a reference.

4 wires, 2 wires short together.

This is the waveform in the saline water.

But, WHEN I use for wires or electrodes separated from each other, then the behavior changes completely. I can not get any proper reading. please. check the pictures below.

This is 4 wires, first (white) and last (brown) are T1 and T2. Second (Blue) and  third (black) are R1 and R2 (R stands for receiver)

This is the acquire waveform,

Saline water is used in both cases, with conductivity of 7.37 mS/cm.

Can you help with this.

Thanks

Abid said:

wires or electrodes separated from each other

I am not sure it is possible to separate the excitation wire Tn from the corresponding sensing wire Rn when in saline - it is going to break the DC working point of your circuit. Roughly speaking, the wires in saline are not galvanically connected to each other due to polarization layer forming at the interface. Very crudely, you can think of it as a capacitor between the wire surface and bulk saline, so the equivalent circuit approximating corresponding behavior would be something like this:

 So, the sensing side of your circuit does not have any DC coupling to the rest of the circuit and your INAMP inputs are hanging at whatever voltage these capacitors are charged to by the input pin currents and leakages. On the excitation side this polarization "capacitance" would break the DC feedback loop for U2, that is why some kind of R11 is necessary.

I will add a capacitor on each Rn wires. This might block DC values. 

The point is that, once wires are disconnected and put into the saline, the DC voltages are blocked by the equivalent capacitance of the polarization layer between the wire and electrolyte. Adding capacitors will put those in sequence with the polarization layer and will not help with the DC bias of the INAMP - the inputs would need to be DC-biased by some additional circuits. Connecting T1 to R1 and T2 to R2 takes care of the DC bias automatically, why would it be necessary to keep those wires apart?

Snorlax said:

the DC voltages are blocked by the equivalent capacitance of the polarization layer between the wire and electrolyte.

Got it,

But, I just went through the following link.

https://www.instructables.com/Body-Composition-using-BIA/

I did not found any additional circuit. 

In this link adding capacitor is suggested on Rn wires. 

https://ez.analog.com/dds/f/q-a/28201/ad5933-4-electrode-configuration-for-human-body-thoracic-impedance-measurement-0-60-ohm

Abid said:

But, I just went through the following link.

This link does not mention additional capacitors and nebulous enough showing a rectangle in place of "Body" and providing no information on the electrodes used.

Abid said:

In this link adding capacitor is suggested on Rn wires.

Yes, but it does not explain what these capacitors are there for. It could be that the "body" is kept at some DC potential far different from Vdd/2, could be a requirement to DC-isolate the "body" from the circuit altogether (that could be the reason to also have the two more capacitors in the feedback loop of the OPAMP at the bottom of the diagram) or it could be that some non-polarizable electrodes were used (such as AgAgCl). If you want to follow this link - install all 4 capacitors and connect both INAMP inputs to the output of U3 with two larger resistors, 1MΩ or so. Thing to watch out with these 4 capacitors would be the wait time after Start Frequency command and before starting the measurements - this time should be long enough for the transients caused by these capacitors and resistors to settle and sine wave baseline to become stable.

I did solve the problem in simulation. I will test it on the PCB and I will post my results.