Designing interface electronics for zirconium dioxide

Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
1. CIRCUIT DESIGN
If not using one of First Sensors ZBXYA interface
boards for sensor control and conditioning, this
section describes the basic building blocks required
to create an interface circuit. Before continuing a
good understanding of application note AN_XYAO2_E_11154 is required.
Noise on the buffered amplified signal should be
filtered by a low pass filter with a cut-off frequency of
around 250 Hz. It is important not to filter the mV Nernst
Voltage as this can load the cell. To improve common
mode noise rejection a small value capacitor (~10 nF)
can be placed across the input terminals of the amplifier.
1.1 HEATER CONTROL
The sensor requires 4.35 VDC to create the correct
operating temperature for the sensing cell. This should
be measured as close to the sensor as possible
because due to the high current requirement of the
low resistance heater there will be voltage drops
across connections and wiring. The designed
adjustable voltage supply should be capable of
providing at least 2 A and emit minimal noise.
1.7 VOLTAGE REFERENCE AND COMPARISON
The amplified sense signal should be compared to
voltage references which are the specified pump
reversal voltages scaled by the same gain factor as
the output amplifier. Each time either the upper or lower
reference is met the constant current source should
be reversed. This part of the circuit should always start
up in the condition that applies the constant current
source between PUMP and COMMON as this begins
the evacuation necessary to start the pumping cycle
i.e. PUMP should be positive with respect to COMMON.
1.2 CONTROL CIRCUIT VOLTAGE REGULATION
Step down and control of input supply voltage.
1.3 START UP DELAY
Zirconium dioxide only becomes operational above
650 °C and as the temperature decreases below this
threshold the cell impedance increases dramatically.
It is therefore important that the sensing cell is not
pumped when cold. Doing so may damage the sensor
as the constant current source will try and drive
whatever voltage is necessary, this has been found to
create an effect similar to when there is zero ppO2.
It is recommended that the sensor is warmed up for a
minimum 60 s before the sensor control circuitry
becomes active. This delay is usually achieved in
software but could also be implemented in hardware .
1.4 CONSTANT CURRENT SOURCE
A typ. 40 µA DC constant current source is required to
drive the pump side of the sensing cell. It is recommended
that an op amp configured as a constant current
source is used. A single resistor and reference voltage
are chosen to set the current with the sensor cell
being the variable load placed in the feedback loop.
1.5 CONSTANT CURRENT SOURCE REVERSAL
Connection of the constant current source between
PUMP and COMMON has to be able to be reversed
whenever either of the reversal voltages are met.
1.8 SIGNAL CONDITIONING
A suitable microprocessor is required to monitor the
amplified sense signal and continually calculate td or
tp. Averaging will reduce natural sensor noise with the
amount of averaging set to suit the response time needs
of the application. Adaptive filtering is the best solution
where the amount of averaging is changed depending
on the amount of variation in the calculated values.
1.9 OUTPUT CONDITIONING
The microprocessor output should then be scaled or
transformed into the required output i.e. voltage,
current loop, serial etc. This may involve the use of a
DAC and output drive circuitry. Filtering and resolution
should also be taken into consideration.
Sensor cell
Constant
current source
reversal
Constant
current source
Sensor output
amplification
and filtering
Signal
conditioning
Voltage
reference and
comparison
Output
conditioning
Start up delay
1.6 OUTPUT AMPLIFICATION AND FILTERING
As the sensed Nernst voltage is a mV signal it is
practical to amplify this to a more sensible operating
range before analysis. Input impedance of the chosen
amplifier should be as high as possible to avoid
loading the cell. Input offset should be less than 0.5 mV.
E / 11163 / B
Control circuit
voltage
regulation
Fig. 1:
Sensor heater
Power supply
Heater voltage
regulation
Sensor interface block diagramm
1/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
2. AMPLIFYING AND SAMPLING THE SENSORS SENSE SIGNAL
This section describes the hardware required to amplify
the generated Nernst voltage from the sensor and also
the ADC requirements to correctly sample the signal.
2.1 ADC minimum resolution
To accurately sample the sensor SENSE signal
(Nernst Voltage) using the recommended hardware
solution in Section 2.3 the ADC resolution must be at
least 12 bits. Two ADC channels are required as the
signal is a differential signal (SENSE with respect to
COMMON).
It is possible to use a single 10 bit ADC but this
involves two stage amplification to firstly amplify the
signal then a second stage to remove the offset and
scale the signal to use the entire 10 bit ADC input
range. Due to the requirement for instrumentation
amplifiers it is preferred to use higher resolution ADCs
which are now common in most microprocessors and
a lower cost amplifier setup.
2.2 ADC acquisition time
The acquisition time required to convert the analogue
signal should be keep to a minimum. If the ADC is
serviced by an interrupt it is important to keep its
frequency equal to or greater than the maximum
sample frequency (see Section 4.1).
2.4 ADC averaging
To help reduce noise in the sampled signal the ADC
results should be placed into a rolling average filter
(see Section 8).
2.5 ADC step voltage
Knowing the step voltage is important when
calculating the voltage level thresholds of the
amplified SENSE signal (see Fig. 3).
To calculate the step voltage, the following equation
should be used:
ADC SV =
VS
2N
(1)
ADCSV = ADC step voltage
VS
= ADC voltage supply
N
= ADC bit resolution
Example:
If our ADC is connected to a 3.3 V supply and the
resolution is 12 bits, then:
ADC SV =
3 .3
212
= 0.00080566 Volts per bit
2.3 Nernst signal amplification
The recommended circuit for amplifying the sensor
Nernst voltage generated across the SENSE
connection with respect to the COMMON connection
is shown in Figure 2. The circuit provides two buffered
and filtered outputs to be sampled by the ADC channels.
The key characteristics of the amplifier design are:
1. Good common mode noise rejection.
2. Biased for low frequency operation. The SENSE
signal is typically less than 15 Hz.
3. Op amp gain bandwidth product of 10 kHz ideal for
low frequency operation.
4. Low input offset voltage ±150 µV maximum.
5. Single ended power supply operation coupled with
high power supply rejection ratio (88 dB typical).
6. Ultra low input bias current avoids loading of the
SENSE signal.
7. Rail to rail input and output.
8. Low cost surface mount components used, X7R/X5R
ceramic capacitors and 1 % tolerance resistors.
E / 11163 / B
2/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
Fig. 2:
Sensor SENSE signal amplification and filtering with buffered COMMON reference
E / 11163 / B
3/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
3. SENSOR PUMP CONTROL
This section describes the relationship between the
direction of the constant source supplied between the
sensor PUMP and COMMON connections and the
generated Nernst voltage (see Fig. 3).
This simple constant current source uses a very low
cost amplifier, X7R/X5R ceramic capacitors and 1 %
tolerance resistors. A digital output from the
microprocessor connects to the terminal CCS reverse
in the schematic.
3.1 Pump current minimum requirements
3.2 Controlling the waveform
The minimum options required in software for
controlling the direction of the pump current are:
To successfully run the sensor the pump current
direction needs to alternated at fixed points V1 and V5
as illustrated in Fig. 3. To calculate V1 to V5 refer to
Section 4.3.
• 40 µA PUMP to COMMON
• 40 µA COMMON to PUMP
• No pump current (sensor disabled)
It is important to have the capability to remove the
pump current as this prevents the sensor being
operated before the appropriate start routine is
applied.
The process for controlling the direction of the pump
current is described in Fig. 4. When the sensor is first
activated the 40 µA PUMP to COMMON must be
applied to the sensor (CCS LOW). It should remain in
this state until the sampled SENSE voltage reaches
the threshold V5.
Figure 5 on page 5 shows the recommended
hardware to provide a true 40 µA constant current
source. This is very important for correct sensor
operation. Note the voltage across the cell cannot
exceed 1.65 V as excess voltage will damage the
sensor !
The pump current direction can now be reversed and
40 µA COMMON to PUMP is applied to the sensor
(CCS HIGH). The system should remain in this state
until the sampled SENSE voltage reaches the
threshold V1.
The system will continue to switch between states
until the pump current is disabled (Pump idle, CCS
high impedance or tri-stated) or power is removed
from the microprocessor/system.
Pump current
+i
3.3 Timeout health check
-i
Nernst voltage
V5
V4
V2
V1
t1
t2
t4
t5
time
Fig. 3:
Relationship between the applied pump
current and the generated Nernst voltage
(measured between COMMON and SENSE)
E / 11163 / B
Pump idle
(0)
Start
Stop
V3
Stop
A pump current timeout should be introduced as a
fault detector. This can help indicate a faulty sensor
or a problem with the interface. This can be achieved
introducing a timeout of approximately 30 sec. The
timeout should be reset at each pump current
reversal. When a timeout occurs the stop routine
should be implemented (see section 6).
SENSE voltage V5
40 µA
COMMON to
PUMP
SENSE voltage V1
Fig. 4:
40 µA
PUMP to
COMMON
Controlling the direction of the pump current
4/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
Fig. 5:
Microprocessor controlled constant current source
E / 11163 / B
5/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
4. SIGNAL PROCESSING
4.1 Sample frequency
For the best possible accuracy a minimum sample
frequency of 10 kHz should be implemented in the
system. Higher frequencies up to 30 kHz can be used
to marginally increase accuracy but the benefits are
minimal and not normally required for the majority of
applications.
4.2 Timer requirement
To sample the amplified SENSE signal correctly a
timer is required to be set up to measure t1, t2, t4 and
t5. If an interrupt timer is used it is important to make
sure a high priority is assigned to the interrupt to
prevent inaccurate measurements. The time resolution
needed has to be equal to or greater than the chosen
sample frequency, although it should be noted that
having greater time resolution will yield no extra benefits.
Example:
If using a 10 kHz sample frequency a time resolution
of 0.1 ms will be sufficient.
4.3 Voltage level calculations
To calculate the SENSE voltage levels (V1 to V5)
correctly a good understanding of the SENSE
amplification and the ADC step volts are required.
Taking into account all amplification gains (x15 for the
recommended circuit) and the common reference
voltage (if applicable) the following equation should be
used to calculate each threshold in ADC steps:
V
− VCOMMON
Threshold = SENSE
ADC SV
(2)
Threshold = Digital threshold voltage level (ADC steps)
VSENSE
= Each amplified SENSE voltage, V1 to V5
(SENSE AMP from Fig. 2)
V COMMON = COMMON reference voltage
(COMMON REF from Fig. 2)
ADCSV
= ADC volts per step as calculated in (1)
The recommended Nernst voltages at the sensor level
versus the corresponding ADC thresholds for 12 bit
ADCs using the recommended circuit from Figure 2
can be found in Table 1.
The system should sample both ADC channels
applying the rolling average described in Section 2.4
and Section 8. Every measurement should be
VSENSE minus VCOMMON and this result should be
compared to the ADC thresholds in Table 1.
Threshold
V1
V2
V3
V4
V5
Nernst voltage
at the sensor
40 mV
45 mV
64 mV
85 mV
90 mV
12 bit ADC threshold
(amplified SENSE - COMMON)
745
838
1191
1583
1676
Table 1: Maximum water vapour pressure (WVPMax)
4.4 Signal sampling
To illustrate the sampling of the SENSE signal the
waveform can be split up into six unique steps. The
following steps describe the process and operations
required.
Individually each step has it own process to perform
in order to obtain the timing values (t1, t2, t4 and t5)
required to calculate td and subsequently %O2.
Idle state
Current direction:
No Pump Current
In idle state the system should not be trying to
sample the SENSE signal. Once the sensor pump
current is activated the system should begin at
Step 1: Peak detection.
The pump current should always initialise in the state
40 µA PUMP to COMMON.
The calculated thresholds in ADC steps can be saved
in a lookup table for system reference.
E / 11163 / B
6/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
Step 1: Peak detection
Current direction:
40 µA PUMP to COMMON →
40 µA COMMON to PUMP
Step 3 : t5
Current direction:
40 µA COMMON to PUMP
voltage
V5
V4
voltage
V5
V4
V3
V3
V2
V1
V2
V1
t1
t2
t4
t2
t4
t5
time
time
In this section the system should be looking to detect
the first peak when the sampled SENSE voltage is
≥ V5. When this occurs the pump current should be
reversed as described above.
Once the sampled SENSE voltage is ≤ V4,
Step 2: t4 is activated.
Step 2 : t4
Current direction:
t1
t5
When entering this section the timer should be reset.
This is done when the sampled SENSE voltage is ≤ V3.
Once the sampled SENSE voltage is ≤ V2, the results
from the timer can be stored as t5.
Step 4: Trough detection is now activated.
Step 4 : Trough detection
Current direction:
40 µA COMMON to PUMP →
40 µA PUMP to COMMON
40 µA COMMON to PUMP
voltage
V5
V4
voltage
V5
V4
V3
V3
V2
V1
V2
V1
t1
t2
t4
t2
t4
t5
time
t5
time
When entering this section the timer should be
initialised/reset. This is done when the sampled
SENSE voltage is ≤ V4.
Once the sampled SENSE voltage is ≤ V3, the results
from the timer can be stored as t4.
Step 3: t5 is now activated.
E / 11163 / B
t1
In this section the system should be looking to detect
the waveform trough when the sampled SENSE
voltage is ≤ V1. When this occurs the pump current
should be reversed as described above.
Once the sampled SENSE voltage is ≥ V2 ,
Step 5: t1 is activated.
7/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
Step 5 : t1
Current direction:
40 µA PUMP to COMMON
voltage
V5
V4
V3
V2
V1
t1
t2
t4
t5
time
When entering this section the timer should be reset.
This is done when the sampled SENSE voltage is ≥ V2.
Once the sampled SENSE voltage is ≥ V3, the results
from the timer can be stored as t1.
Step 6: t2 is now activated.
Step 6 : t2
Current direction:
40 µA PUMP to COMMON
voltage
V5
V4
V3
V2
V1
t1
t2
t4
t5
time
When entering this section the timer should be reset.
This is done when the sampled SENSE voltage is ≥ V3.
Once the sampled SENSE voltage is ≥ V4, the results
from the timer can be stored as t2.
Step 1: Peak detection is now activated and the
continuous loop begins again.
E / 11163 / B
8/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
5. START ROUTINE
6. STOP ROUTINE
The start routine is required every time the sensor is
switched off or power cycled. This helps prevent
irreversible damage to the oxygen sensor which can
occur if the sensor is pumped when the zirconium
dioxide sensing cell is cold.
Some applications may require the sensor to be
stopped during operation for safety, maintenance or
for energy efficiency reasons.
On system initialisation it is important to make sure
the pump current and signal processing are
deactivated.
The first process should be to deactivate the pump
current and signal processing. Minimal delay should
be present between each process shutdown. The
heater may then be turned off.
6.1 Stop routine description
5.1 Start routine description
The first process should be to make sure the heater
is enabled to heat up the sensor. After the heater is
applied the system should then begin a warm up
delay period with a minimum of 60 sec.
On delay completion the pump current and signal
processing can be activated to allow the sensor to
begin its pump cycle.
The system cool down delay is a optional process
depending on the application requirements. If used a
minimum of three minutes should be applied. It may
be necessary for a longer delay to be implemented to
allow the application to fully cool down before the
sensor heater is turned off. The delay should be
determined by the application and it’s purpose is to
prevent condensation forming on the sensor in humid
environments during the shutdown process.
The following stop routine should be applied to
shutdown the sensor operation correctly.
Start of start routine
Start of stop routine
Enable heater
Deactivate pump current
Start 60 sec. delay
Deactivate signal process
Start system cooldown delay
Delay
completed?
No
Delay
completed?
Yes
Activate pump current
No
Yes
Activate signal process
Disable heater
End of start routine
End of stop routine
Fig. 6:
Start routine
E / 11163 / B
Fig. 7:
Stop routine
9/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
7. CALCULATIONS
The calculations needed to calculate td and for
diagnostics are not time dependant and can be
managed during the processors free time.
tp is time dependant and will need to be calculated
using a timer which is reset at each peak or trough
detection.
The frequency of the signal can also be calculated
using Equation (4). tp should be non zero before this
calculation is made.
Freq =
1
tp
(4)
7.3 Asymmetry
The equation for calculating the sampled SENSE
voltage asymmetry is displayed in Equation (5):
7.1 td
The following equation is used to calculate td:
t d = (t1 − t 2 ) + (t 5 − t 4 )
Asymmetry =
(3)
The time values (t1, t2, t4 and t5) are obtained during
the signal processing routine. Therefore td only needs
to recalculated after every new t value.
It is recommended td is put into a rolling average filter
to reduce noise and stabilise the td output. We
recommend a buffer size of between 4 to 400. This
value is very application dependant with a small buffer
size best for fast sensor response and a large buffer
size optimal for output stability.
Therefore the maximum buffer size is ideal for
systems with slowly drifting O2 levels and the
minimum buffer size is ideal for applications with
rapidly changing O2 levels.
For a balance between response and stability a buffer
size of 100 is ideal.
For applications where both response and stability
are critical an adaptive filtering method may be used.
This can be achieved by monitoring the variance in
each new recorded td value and when the variance
exceeds a predetermined level the buffer is flushed
and the buffer size reduced to its minimum value.
When the td values begin to stabilise again the buffer
size can be gradually increased until it reaches its
maximum value.
7.2 tp
(t1 + t 2 )
(t 5 + t 4 )
(5)
Asymmetry need only be recalculated on each new t
value at the same time as td.
To help avoid divide by zero fault conditions during the
start-up cycle it is good practice to only calculate
asymmetry if t4 or t5 are not equal to zero.
The asymmetry value should also be placed into a
rolling average filter to reduce noise and add stability.
A buffer size of 10 to 100 is recommended.
7.4 O2
To transform the calculated and buffered td values into
the corresponding O2% in the atmosphere a
calibration scalar (CS) is required (see Section 9).
The O2% value can then be obtained using Equation (6):
O 2 (%) = t d (Ave ) × CS
(6)
It should be noted that this is an averaged O2 value as
the buffered td value is used in the calculation. If an
instantaneous O2 value is required then td (Ave) can be
replaced with each newly calculated td. This is often
referred to td raw and the calculated oxygen level as
O2% raw.
If using a barometric pressure sensor to compensate
for pressure changes please refer to Section 3.5 of
application note AN_XYA-O2_E_11154 for guidance.
The tp calculation can be made by measuring the
period of the sampled SENSE voltage waveform. The
recommended way to perform this calculation is to
measure the time between the waveform peak to peak
as this generally more repeatable than measuring the
time between the trough to trough.
E / 11163 / B
10/11
www.first-sensor.com
www.sensortechnics.com
Designing interface electronics for zirconium dioxide
oxygen sensors of the XYA series
8. ROLLING AVERAGE FILTER
9. CALIBRATION CONSIDERATIONS
8.1 Filter principle
To calculate the calibration scalar used in Section 7.4,
the following equation should be used:
A basic rolling average filter is defined as the sum of
all the last N number of data points divided by the
number of results:
Average =
x1 + x 2 + ... + x N−1 + x N
N
(7)
Where:
N = Buffer size
x = Data
This simple filter is extremely useful in reducing noise
in a signal or system. It can also be quickly
implemented into a system to improve the stability of
sampled signals.
8.2 Processor Overhead
In some applications this approach can be
problematic depending on the platform and compiler.
The process of division can take a large amount of
processing power and therefore time.
As the measurement of oxygen in this system is very
time dependant all efforts should be made to avoid
any unnecessary overheads.One option to reduce the
overhead is by replacing the intensive division
calculations present in the averaging filters, with a
less intensive process.
CS =
O 2 (%)
t d (Ave )
(9)
Where:
CS
= Calibration scalar
O2(%) = Known O2 % in the calibration environment
td (Ave) = Average td value
Before a calibration process it is vital to make sure
the sensor output is stable and the environment only
comprises of the calibration gas. It is for these
reasons that the sensor is normally calibrated in
normal air to 20.7 %O2 and the sensor is given 10 min.
after powering the heater before proceeding with
calibration.
If the heater has been on for more than ten minutes
then the sensor only requires 5 min. in the calibration
gas before a calibration can proceed.
If a calibration gas of another known oxygen
concentration is available then this may be used by
replacing O2 (%) in the equation above.
A division of two can be easily implemented by
shifting the value right by one.
Example:
Binary 00001000
which equals decimal value 8 becomes
Binary 00000100
which equals decimal value 4.
Using this principle we can carefully select N such
that it equates to 2 to the power of y :
(8)
N = 2y
Where:
N = Chosen buffer size
y = Number of places to shift to the right
It is recommended N should be between 16 and 32,
when the ADC is sampled at 10 kHz.
E / 11163 / B
11/11
www.first-sensor.com
www.sensortechnics.com