Post Go back to editing

# How to correctly calculate the Rfb value in AD5933

I was reading the datasheet of AD5933, and I did not understand a few things about the calculation for the Rfb value and other things.

This is the Rfb equation:

How do I find the value of Zmin?

I intend to power the AD5933 with 5V, use the 1st Output Range (3Vpp in this case) and PGA gain = 1.
With a frequency sweep of 1K Hz up to 100K Hz:
Internal oscillator: MCLK = 16.776 MHz
fSTART = 1000 Hz
∆f = 1000 Hz
Increments = 99

In addition, I will use a high pass filter with a cutoff frequency of 100Hz, a circuit with a current source controlled by voltage (1mA), and an instrumentation amplifier INA118 (Gain = 2). This circuit will use 4 Electrodes. Similar to this:

In AN-1252 APPLICATION NOTE, page 3 of 12, it is said that for range 1 (Adding external op amp) Zout is equal to "> 100 Ohms". And on page 5, The maximum ratio, ZMAX / ZMIN, for my case is x45. Right? If it is correct. How to calculate Zmax and Zmin for "> 100"?

Furthermore. What is "RCal" for? If I do not use it in any equation.

In the example on page 7, the Zmin value is 4.7K and the Zmax value is 47K. Why? I dont understand.

• The formula just helps calculating the RFB value when Zmin is known. From the formula it follows that RFB is always lower than Zmin, so the chip is not supposed to work when RFB is 10k and Z is 5k and even less so when it is 1k. It works backwards as well: for given RFB you can always solve it for Zmin and see that Zmin is always larger than RFB. So for your 10K and other parameters Zmin is about 15k (provided that Zmin is just a resistor, for complex impedance things are more, well, complex).

• There are various differential, ratiometric and bridge methods to perform such measurements, including good old Wheatstone bridge.

• Hi Snorlax,

I'm using my AD5933 with AFE which I found in AN-1252. On Page 7 in table 6 the RFB Value is 6.8 kOhms and it is bigger than ZMin in table 5 is 4.7 kOhms. it makes no sense for me. because ZMin should be bigger than Rfb...

• With the AFE it is a bit different story as VDCOFFSET = VDD/2, so solving the Rfb equation above for Zmin in this case yields: Zmin = Rfb *Gain * Vpk/(Vdd/2 -0.2). Depending on Gain and Vpk Zmin can be smaller than Rfb.

• Thank you for your help. I didn´t have a clue. Now it makes sense for me.

• Hi Snorlax,

only for right understanding.

I have a question again according to calculation of RFB and RCAL

Using AD 5933 with AFE like in CN0217 I calculate the Values of each resistor with the following equations:

I determine ZMIN to calculate RFB and RCAL. ZMAX is with AFE max 45*ZMIN right?

Whats about GAIN? Could I use 5x when I´m using AFE?

• Hi,

Digipots will add some parasitic capacitance and no resistance linearity due to the CMOS switches.

depends on your C values, this may add a non negligible errors.

Regards,

Miguel

• Using the first equation one calculates the Rfb at given Vpk and Gain values (programmed into the chip). At Gain=1, the Rfb is typically 15-20% higher than Zmin.
The second equation is not very useful as Zmax is either arbitrary or defined by one’s specific application. There is nothing wrong with infinitely high Zmax: measuring it should produce zeroes in both Re and Im output registers, but the equation does not yield any practical value for Rcal. The suggestion would be to use Rcal = Zmin to begin with and worry about the subtleties later.

Yes, you can certainly use Gain=5 with the AFE.

• Ok. Could you look at this table please?

 VDD 3.3 VPK 1.98 GAIN 1 Z_min 5 35 245 1715 46305 RFB 3.66 25.63 179.41 1255.93 33910.22 Z_max = 10 * Z_min Z_max 50 350 2450 77175 2083725 R_cal 3.66 35 245 1715 46305

I would use these calculations to cover a measuring range from 5 Ohm until 1 MOhm. I like to measure parallel RC Elements. What do you think? Should it work?

What do you think about digipots? Do I get more inaccurate measures or do you think it could be accurate? Whats the difference between using digipots for RFB und ZCAL  or using resistors?

• It is difficult to discuss this matter without knowing what you application really is. On the surface, yes, it should work, but, depending on what you are shooting for, may or may not be accurate enough.

Digipots are not terribly reproducible device-to-device, not that linear and flat with frequency compared to discrete resistors, which, again, may or may not be an issue for your application. Parallel RC networks should be measurable, generally speaking. A potential issue you may or may not run into at the lower impedance values is relatively high current across Rfb, for example, at Vpk = 1.98 V and Zx =  5 Ohm this current is ~0.4 A, which your excitation voltage source must be able to deliver, Rfb must be able to pass across and the OPAMP with Rfb must be able to sink.