Circuit Note

CN-0349

Temperature Compensation

Conductivity measuring system accuracy is only as good as its

temperature compensation. Because common solution temperature

coefficients vary in the order of 1%/°C to 3%/°C, or more,

measuring instruments with adjustable temperature compensation

must be used. Solution temperature coefficients are somewhat

nonlinear and usually vary with the actual conductivity as well.

Thus, calibration at the actual measuring temperature yields the

best accuracy.

The circuit shown in Figure 1 yields accurate conductivity

measurements, extending from the low µS range to several

hundred mS, and it also optimizes the overall accuracy of the

AD5934 over a wide range of admittance. Temperature

measurement using a Pt100 resistance temperature detector

(RTD) is also incorporated. The circuit can be connected to the

evaluation microprocessor board using the 8-pin IMOD (I2C

interface) connector for rapid prototyping (Digilent, Pmod™

Specifications).

The circuit consists of four major blocks. The first block is an

impedance-to-digital converter containing the AD5934 (U3)

impedance converter, a follower (one half of the AD8606, U2A)

to bias the ac signal to VDD/2, and a current-to-voltage

converter configuration using the second half of the AD8606,

U2B. Details of the operation of the AD5934 can be found in

Circuit Note CN-0217 and in the AD5934 data sheet.

The second block is the programmable resistor feedback (R6,

R8, R9) and calibration (R3, R4, R7) circuitry and the octal

SPST switches ADG715 (U1). The measurement range and

calibration procedures are realized by controlling the ADG715

via the I2C serial interface.

The third block is the ADuM1250 (U5) hot swappable, digital

isolator for transmitting the serial data from the AD5934 (U3)

to the IMOD CON (J2). The fourth block is the ADuM5000

(U4) isolated dc-to-dc converter with a 3.3 V output voltage to

supply the power for the circuit.

The third and fourth blocks are optional; however, they provide

galvanic isolation between the circuit and the evaluation micro-

processor board. They are not required unless isolation is needed.

The circuit is connected to the conductivity cell with a built-in

Pt100 RTD temperature sensor using the CON1 (J1) connector.

The frequency of the clock applied to the MCLK pin is set to

1 MHz using a stable, low jitter, FXO-HC536R-1 (U6) quartz

crystal oscillator. This oscillator allows the AD5934 to excite the

conductivity cell with a frequency of 2 kHz, which is well suited

for conductivity measurements.

Circuit Design

Figure 2 shows the optimized signal chain for conductivity and

temperature measurements used in the circuit. The AD5934 has

four programmable output voltage ranges. Each range has an

associated output impedance. For example, the output impedance

for a 1.98 V p-p output voltage is typically 200 Ω (see the AD5934

data sheet). The output impedance affects the impedance

measurement accuracy, particularly in the low ohms range. A

simple buffer in the signal chain prevents the output impedance

from affecting the unknown impedance measurement. A low

output impedance amplifier, with sufficient bandwidth to

accommodate the AD5934 excitation frequency, is selected. An

example of the low output impedance that can be achieved is

shown in Figure 2 for the AD8605/AD8606/AD8608 family of

CMOS op amps. The output impedance for this amplifier, for a

gain of 1 is less than 1 Ω up to 100 kHz, which corresponds to

the maximum operating range of the AD5934.

Figure 2. Optimized Signal Chain for Conductivity and Temperature

Measurements

The four programmable output voltage ranges in the AD5934

have four associated bias voltages (see the AD5934 data sheet).

For example, the 1.98 V p-p excitation voltage requires a bias of

1.48 V. However, the current-to-voltage (I-V) receiver stage of

the AD5934 is set to a fixed bias of VDD/2. Therefore, for a

3.3 V supply, the transmit bias voltage is 1.48 V, and the receive

bias voltage is 3.3 V/2 = 1.65 V. This potential difference polarizes

the measured solution, YX, and can cause significant inaccuracies in

the conductivity measurement. One solution is to add a simple

high-pass filter with a corner frequency in the low Hz range

(see the Circuit Note CN-0217). Removing the dc bias from the

transmit stage and rebiasing the ac signal to VDD/2 keeps the dc

level constant throughout the signal chain. Use precision 0.1%

resistors for both of the biasing resistors, R1 and R5 (10 kΩ), to

reduce inaccuracies.

The I-V amplifier stage of the AD5934 can also add minor

inaccuracies to the signal chain. The I-V conversion stage is

sensitive to the amplifier bias current, to the offset voltage, and to

the common-mode rejection (CMR). Accuracy is improved by

selecting the proper external discrete amplifier (U2B) to perform

the I-V conversion. The AD8606 is chosen because of its low offset

voltage (65 μV maximum), low bias current (1 pA maximum), high

CMRR (95 dB typically), and low noise (12 nV/√Hz maximum).

The internal amplifier can then be configured as a simple inverting

gain stage. Selection of the RFB resistor depends on the gain through

the system, as described in the AN-1252 Application Note. The

input and output of the I-V converter must be accurately biased

to VDD/2. Use precision 0.1% resistors for both of the biasing

resistors, R12 and R13 (10 kΩ).

RCAL

RFB

RTD

1

3

2

YX

R11

20kΩ

3.3V

R10

20kΩ

RFB

VIN

VOUT

U2A

AD8606

U2B

AD8606

C1

470nF

U3

AD5934

R13

10kΩ

R2

100Ω

R12

10kΩ

U1

ADG715

R1

10kΩ

R5

10kΩ

Rev. 0 | Page 3 of 10