# AD5933 is not working with external clock

I am using AD5933 impedance measurement chip. There is an option to connect an external clock. I am using a function generator to give 250 kHz, 2-volts peak square wave signal to MCLK pin, but it's not working.

The communication (I2C) between controller and AD5933 is good.

(a) I have used the internal oscillator and I can measure impedance from 5 kHz to 100 kHz frequency range. I have measured different resistor and capacitor combinations using the internal oscillator. Now, I decided to measure impedance below the 5 kHz frequency range and according to the UG-364 application note, we have to use an external oscillator to apply a clock frequency of 250 kHz to use the frequency range below 5 kHz.

(b) I used a function generator to apply an external clock frequency shown in the attached picture, but the AD5933 is not responding. In the code, you can see the while loop, which is waiting for valid real and imaginary value by reading the flag set (2nd bit) in the status register of the AD5933. The 2nd bit of the status register does not go high, which means the AD5933 is not getting valid real and imaginary values.

(c) Schematic picture of my circuit is below.

(d) I am using 3.3 VDC and my output excitation range is 1 which means output excitation voltage is 1.98 Vp-p with DC offset 1.48 V.

Please share your experience of how can I connect the external clock to the AD5933 MCLK pin.

This is my code sequence:

1. Standby mode

2. Enable external oscillator

3. Initialize sweep

4. Start sweep

This is my code:

#include "Wire.h"

#define START_FREQ_REG1 0x82
#define START_FREQ_REG2 0x83
#define START_FREQ_REG3 0x84

#define FREG_INCR_REG1 0x85
#define FREG_INCR_REG2 0x86
#define FREG_INCR_REG3 0x87

#define NUM_INCR_REG1 0x88
#define NUM_INCR_REG2 0x89

#define CYCLE_REG1 0x8A
#define CYCLE_REG2 0x8B

#define REAL_REG1 0x94
#define REAL_REG2 0x95

#define IMG_REG1 0x96
#define IMG_REG2 0x97

#define CRL_REG 0x80
#define CRL_REG1 0x81
#define STATUS_REG 0x8F

void programReg();
void runSweep();
byte getFrequency(float freq, int n);

const float MCLK = 250*pow(10,3); // AD5933 Internal Clock Speed 16.776 MHz for formula
const float start_freq = 100; // Set start freq, < 100Khz (for formula)
const float incre_freq = 10; // Set freq increment (for formula)
const int incre_num = 10; // Set number of increments; < 511

byte value;
int count=0;
char state;
double gain;
const double pi = 3.141592654;

void setup() {

Wire.begin();
Serial.begin(115200);

//nop - clear ctrl-reg
writeData(CRL_REG,0x0);

programReg();
}

void loop(){

if(Serial.available()>0) {

switch(state) {
case 'A':
programReg();
break;

case 'B':
runSweep();

break;

}
}

}

void programReg(){

// Start frequency of 1kHz
writeData(START_FREQ_REG1, getFrequency(start_freq,1));
writeData(START_FREQ_REG2, getFrequency(start_freq,2));
writeData(START_FREQ_REG3, getFrequency(start_freq,3));

// Increment by 1 kHz
writeData(FREG_INCR_REG1, getFrequency(incre_freq,1));
writeData(FREG_INCR_REG2, getFrequency(incre_freq,2));
writeData(FREG_INCR_REG3, getFrequency(incre_freq,3));

// Points in frequency sweep (150), max 511
//writeData(NUM_INCR_REG1,0x00 );
//writeData(NUM_INCR_REG2, 0x96);

writeData(NUM_INCR_REG1, (incre_num & 0x001F00)>>0x08 );
writeData(NUM_INCR_REG2, (incre_num & 0x0000FF));

// Set settling cycles
writeData(CYCLE_REG1, 0x00);
writeData(CYCLE_REG2, 0x64);

}

void runSweep() {
short re=0;
short img=0;
double freq=0;
double mag=0;
double  MAG=0;
double phase=0;

double impedance=0;

int i=0;

programReg();

// 1. Standby '10110001'
writeData(CRL_REG, 0xB0);

// Enable external oscillator
writeData(CRL_REG1, 0x8);

// 2. Initialize sweep '00010001'
writeData(CRL_REG, 0x11);

delay(20);

// 3. Start sweep '00100000'
writeData(CRL_REG, 0x21);

while((readData(STATUS_REG) & 0x07) < 4 ) {  // Check that status reg != 4, sweep not complete

delay(20); // delay between measurements

//Serial.println("");

if (flag==2)

{

re = (R1 << 8) | R2;

img = (R1 << 8) | R2;

freq = start_freq + i*incre_freq;

Serial.print(freq);
Serial.print("\t");

mag = sqrt(pow(double(re),2)+pow(double(img),2));

if (count==0)
{
gain=1.0/(mag*34000);

count=1;
}

if (count!=0)
{
impedance =1.0/(gain*mag);

Serial.println(impedance);

}

//Increment frequency
if((readData(STATUS_REG) & 0x07) < 4 ){
// if((readData(STATUS_REG) & 0x04) == 0 ){
writeData(CRL_REG,0x31);

// writeData(CRL_REG,(readData(CRL_REG) | 0x01) | 0x30);
i++;
}
}
else {
//Power down

writeData(CRL_REG,0xA0);
}
p++;
///}
}
}

void writeData(int addr, int data) {

Wire.write(data);
Wire.endTransmission();
delay(1);
}

int data;

Wire.endTransmission();
delay(1);

if (Wire.available() >= 1){
}
else {
data = -1;
}

delay(1);
return data;
}

byte getFrequency(float freq, int n){
long val = long((freq/(MCLK/4)) * pow(2,27));
byte code;

switch (n) {
case 1:
code = (val & 0xFF0000) >> 0x10;
break;

case 2:
code = (val & 0x00FF00) >> 0x08;
break;

case 3:
code = (val & 0x0000FF);
break;

default:
code = 0;
}

return code;
}

double phase_sweep(double img, double re)
{
double phase;
if ((re>0)&(img>0))
{
phase = atan(double(img)/double(re));
phase = (phase*180)/pi;
}
else if((re>0)&(img<0))
{
phase = atan(double(img)/double(re));
phase = (phase*180)/pi+360;
}
else if ((re<0)&(img<0))
{
phase = -pi + atan(double(img)/double(re));
phase = (phase*180)/pi;
}
else if ((re<0)&(img>0))
{
phase = pi + atan(double(img)/double(re));
phase = (phase*180)/pi;
}

return phase;
}


I am using a function generator to give 250 kHz, 2-volts peak square wave signal to MCLK pin

According to the oscilloscope trace, that's not what you are actually supplying to the MCLK pin!

The clock shown on the oscilloscope trace is invalid as MCLK.

The trace shows a bipolar signal, measured by the scope as 3.88 Vp-p centred around 0 V i.e. approximately +1.94 V to -1.94 V. That -1.94 V exceeds the allowed negative voltage on the MCLK pin. The datasheet for the AD5933 shows that the voltage supplied to the MCLK pin must not be below -0.3 V:

(Image source: AD5933 datasheet, page 7)

From a quick look at the instruction manual for your SFG-2010 function generator, I guess you are using the normal waveform output. You need to either:

1. make that signal output unipolar within the voltage range explained below, if you know how to configure the signal generator's normal output to do that, or

2. use the TTL/CMOS output on the function generator instead, remove the 50 Ω load resistor shown on your schematic, and ensure that this unipolar output signal is between 0 V and your Vdd, which you said was +3.3 V.

According to the function generator user manual, when switched to "TTL" output, the output voltage is 3V or above - that would be a problem if it exceeds 3.3 V. And when switched to "CMOS" output, the output voltage can be adjusted, but the lowest mentioned is 4 V with a plus or minus 1 V tolerance. It therefore seems possible that the "CMOS" output voltage might also be too high and exceed your 3.3 V Vdd. For balanced rise/fall times, the CMOS output setting would be my preferred option in this case.

So you will need to do some measurements of the TTL and CMOS mode output voltages from your signal generator and, depending on what you see on your scope, you might need to add external attenuation or push-pull level translation (with fast rise/fall times) to ensure a clock signal amplitude not exceeding 0 V to your Vdd, suitable for the MCLK input.

Or, depending on your requirements for things like frequency accuracy, stability etc., you could investigate using a different clock generator to produce the MCLK signal, which has a suitable CMOS output stage.

In the worst case, connecting that incorrect bipolar signal to the MCLK pin might have already damaged that part of the AD5933 e.g. by shorting the ESD input protection on that pin, due to excessive current. If you generate a suitable MCLK signal as I suggested above, which shows correctly (0 V to 3.3 V max) on the scope when not connected to the MCLK pin, but which changes significantly when it is connected to the MCLK pin, then that would be an indication that you have already damaged the MCLK input on that AD5933 IC.

One more point from your schematic: The required power supply bypass capacitors (explained on page 34 of the datasheet) are missing from your schematic, and their physical location is also important, as explained in the datasheet. Omitting those bypass capacitors may also affect the device behaviour.

(I haven't investigated your code - until the MCLK signal is fixed, there's no point looking at other possible problems. I assume that the same code (except the change to select the external clock) was used in your experiments with the internal clock, which worked).

• SamGibson thank you so much for your reply, you right according to datasheet MCLK signal should be between (-0.3 to VDD+0.3), but at end of datasheet there is schematic of evaluation board that is developed using AD5933, they used external oscillator and applied the signal directly to MCLK pin. I saw the output of this oscillator on oscilloscope its output is (6.08)Vp-p at VDD 3.3volt. – Ati Apr 23 '19 at 9:07
• @Ati - "output of this oscillator on oscilloscope [...] is (6.08)Vp-p at VDD 3.3volt." I have never seen a standard crystal oscillator produce a higher o/p voltage than its Vdd. I cannot explain your results, without having the board in my hands. I've done more research (too long to list here) and everything confirms that MCLK i/p expects a unipolar digital signal ≤ Vdd. I see that you have now asked this question on the Analog Devices forum, so hopefully you will get an answer there. Good luck! – SamGibson Apr 23 '19 at 13:36
• SamGibson thank you so much for your time with your help I have solved my problem, the output of four leg oscillator is unipolar and I have used this 16MHZ oscillator that is mounted on evaluation board and now my AD5933 chip is working with external oscillator thank you so much for your guidance. – Ati Apr 24 '19 at 2:27
• @Ati - Thanks for the update. "the output of four leg oscillator is unipolar" Good, as I said, that's exactly what I expected. Can you explain why you reported the bipolar 6.08 Vp-p oscillator output signal on the oscilloscope yesterday, please? What was the cause? Then I can add those details into my answer, to make it more complete. Thanks. – SamGibson Apr 24 '19 at 2:45
• SamGibson I was choosing AC coupling in an oscilloscope that's the problem, after reading your comment I have change oscilloscope setting to DC coupling then visualize the data, the output waveform is unipolar and its amplitude is 3.3V – Ati Apr 24 '19 at 3:01