One Technology Way • P.O. Box 9106 • Norwood, MA 02062-9106 • 781/329-4700 • World Wide Web Site: http://www.analog.com
A Tamper-Resistant Watt-Hour Energy Meter
Based on the AD7751 and Two Current Sensors
by Anthony Collins and William Koon
INTRODUCTION
This application note describes a low-cost, high-accuracy
watt-hour meter based on the AD7751. The meter
described is intended for use in single-phase, two-wire
distribution systems. However, the design can easily be
adapted to suit specific regional requirements, e.g., in
the United States power is usually distributed to residential customers as single-phase, three-wire.
The AD7751 is a low-cost, single-chip solution for electrical energy measurement. The most distinctive feature
of the AD7751 is that it continuously monitors the phase
and neutral (return) currents. A FAULT condition occurs
if the two currents differ by more than 12.5%. Power calculation will be based on the larger of the two currents.
The meter calculates power correctly even if one of the
two wires does not carry any current. AD7751 provides
an effective way to combat any attempt to return the
current through earth, a very simple yet effective way of
meter tampering. The AD7751 is comprised of two
ADCs, reference circuit, and all the signal processing
necessary for the calculation of real (active) power. The
AD7751 also includes direct drive capability for electromechanical counters (i.e., the energy register) and has a
high-frequency pulse output for calibration and communications purposes.
This application note should be used in conjunction with
the AD7751 data sheet. The data sheet provides detailed
information on the functionality of the AD7751 and will
be referenced several times in this application note.
DESIGN GOALS
The International Standard IEC1036 (1996-09)—
Alternating current watt-hour meters for active energy (Classes
1 and 2)
design. For readers more familiar with the ANSI C12.16
specification, see the section at the end of this application
which compares the IEC1036 and ANSI C12.16 standards. This section explains the key IEC1036
specifications in terms of their ANSI equivalents.
The design greatly exceeds this basic specification for
many of the accuracy requirements, e.g., accuracy at
unity power factor and at low (PF = ±0.5) power factor. In
, was used as the primary specification for this
addition, the dynamic range performance of the meter
has been extended to 500. The IEC1036 standard specifies accuracy over a range of 5% Ib to I
Typical values for I
outlines the accuracy requirements for a static watthour meter. The current range (dynamic range) for
accuracy is specified in terms of Ib (basic current).
Table I. Accuracy Requirements
Current Value
0.05 Ib < I < 0.1 Ib1±1.5%±2.5%
0.1 Ib
0.1 Ib
0.2 Ib < I < I
NOTES
1
The current ranges for specified accuracy shown in Table I are expressed
in terms of the basic current (Ib). The basic current is defined in IEC1036
(1996-09) Section 3.5.1.1 as the value of current in accordance with
which the relevant performance of a direct connection meter is fixed.
I
is the maximum current at which accuracy is maintained.
MAX
2
Power Factor (PF) in Table I relates the phase relationship between the
fundamental (45 Hz to 65 Hz) voltage and current waveforms. PF in this
case can be simply defined as PF = cos( φ), where φ is the phase angle
between pure sinusoidal current and voltage.
3
Class index is defined in IEC1036 (1996-09) Section 3.5.5 as the limits of
the permissible percentage error. The percentage error is defined as:
Percentage Error =
The schematic in Figure 1 shows the implementation of
a simple, low-cost watt-hour meter using the AD7751.
Two current transformers (CTs) are used to provide the
current-to-voltage conversion needed by the AD7751,
and a simple divider network attenuates the line voltage.
The energy register (kWhrs) is a simple electromechanical counter that uses a two-phase stepper motor.
The AD7751 provides direct drive capability for this type
of counter. The AD7751 also provides a high-frequency
output at the CF pin for the meter constant (e.g.,
3200 imp/kWhr). Thus a high-frequency output is available
at the LED and opto-isolator output. This high-frequency
output is used to speed up the calibration process and
provides a means of quickly verifying meter functionality
1
< I < I
MAX
< I < 0.2 Ib0.5 Lag±1.5%±2.5%
MAX
are 400% to 600% of Ib. Table I
MAX
2
PF
1±1.0%±2.0%
0.8 Lead±1.5%—
0.5 Lag±1.0%±2.0%
0.8 Lead±1.0%—
energy registered by meter true energy
Percentage Error Limits
Class 1Class 2
true energy
—see Table I.
MAX
–
100%
×
3
REV. 0
AN-563
LOAD
FG0003
1:1800
1:1800
FG0003
PHASE
NEUTRAL
240V
K1
R5
8.2
K2
K3
R6
9.1
SOLDER JUMPERS
R7
110
K4
J16 J17 J18 J19 J20 J21
R8
330
620
J5
J4
J3
J2
J1
CALIBRATION
NETWORK 2.5%
R9 R10 R11 R12 R13
1.2k 2.4k 4.7k 9.1k
CALIBRATION
NETWORK 30%
R18
18k
R17
39k
R16
75k
R15
150k
R14
300k
K3
K4
560
1.2k
2.2k
5.1k
9.1k
R24
300k
R23
R22
R21
R20
R19
Figure 1. Simple Single-Phase Watt-Hour Meter Based on the AD7751
R25
300k
J10
J9
J8
J7
J6
Z1
C17
10nF
FAULT-RESISTANT ENERGY METER
= 40A, Ib = 10A, CLASS 1 (5%Ib TO I
I
MAX
V
CLKOUT
D2
DD
P1
F1
F2
CF
REVP
FAULT
CLKIN
G0
G1
S0
S1
SCF
DGND
POWER SUPPLY
C19
+
470F
35V
8
P24
P23
P22
P20
P19
P18
Y1
P17
P16
P15
P14
P13
P12
C13
100nF
R29
820
3.579545MHz
C10
22pF
C9
22pF
G0 = 1
G1 = 0
S0 = 1
S1 = 0
SCF = 0
1
U2
7805
2,3,6,7
C12
+
C1
33nF
C2
33nF
C3
33nF
C4
33nF
C5
33nF
C7
100nF
C18
470nF
MOV1
S20K275
C11
100nF
P10
1N4744A
220F
6.3V
R1
1k
R2
1k
R3
1k
R4A
887
R4B
887
+
C6
10F
X2
R33
10
P3P2
AVDD AC/DC DVDD
P4
V1A
AD7751
P6
V1N
P5
V1B
P7
V2N
P8
V2P
REF
IN/OUT
AGND
RESET
P9P11P21
R27
10k
V
DD
R26
470
1N4004
D2
)
MAX
C14
+
10F
6.3V
HP HLMP-D150
V
DD
V
DD
5V
R30
820
R28
10k
J15
J14
J13
J12
J11
C8
100nF
Z2
TO IMPULSE COUNTER/
STEPPER MOTOR
R32
20
R31
20
D1
FAULT
LED
D4
2
1
U3
PS2501-1
0R
K7
100IMP/kWhr
C16
C15
K8
CALIBRATION
LED
HP HLMP-D150
4
3200IMP/kWhr
3
JUMPERS USE
0 RESISTORS
0R
K9
K10
and accuracy in a production environment. The meter is
calibrated in a two-step process:
Step 1. With current passing through only Channel V1A's
CT, the meter is first calibrated by varying the line voltage
attenuation using the resistor network R14 to R23.
Step 2. With current passing through only Channel
V1B's CT, the small gain mismatch between the CTs in
Channel V1A and V1B is calibrated by shorting the
appropriate resistors in the resistor network R8 to R13.
DESIGN EQUATIONS
The AD7751 produces an output frequency proportional
to the time average value of the product of two voltage
signals. The input voltage signals are applied at V1 and
V2. The detailed functionality of the AD7751 is explained
in the AD7751 data sheet, see
Theory Of Operation
. The
AD7751 data sheet also provides an equation that relates
the output frequency on F1 and F2 (counter drive) to the
product of the rms signal levels at V1 and V2. This equation is shown here again for convenience and will be
used to determine the correct signal scaling at V2 in
order to calibrate the meter to a fixed constant.
Frequency
5.7412
V VGain F
×× × ×
=
2
V
REF
−
14
The meter shown in Figure 1 is designed to operate at a
line voltage of 240 V and a maximum current (I
40 A. However, by correctly scaling the signals on Channel 1 and Channel 2, a meter operating from any line
voltage and maximum current could be designed.
The four frequency options available on the AD7751 will
allow similar meters (i.e., direct counter drive) with an
I
of up to 120 A to be designed. The basic current for
MAX
this meter is selected as 10 A and the current range for
accuracy will be 1% Ib to I
or a dynamic range of 400
MAX
(100 mA to 40 A). The electromechanical register (kWh)
will have a constant of 100 imp/kWh, i.e., 100 impulses
from the AD7751 will be required in order to register
1 kWhr. IEC1036 Section 4.2.11 specifies that electromagnetic registers have their lowest values numbered
in ten division, each division being subdivided into
ten parts.
–2–
(1)
) of
MAX
REV. 0
AN-563
Hence a display with five-plus-one digits is used, i.e.,
10,000s, 1,000s, 100s, 10s, 1s, 1/10s. The meter constant
(for calibration and test) is selected as 3200 imp/kWh.
The on-chip reference circuit of the AD7751 typically has
a temperature coefficient of 30 ppm/°C. However, on A
grade parts this specification is not guaranteed and may
be as high as 80 ppm /°C. At 80 ppm /°C the AD7751 error
at –20°C/+60°C would be approximately 0.65%, assuming a calibration at 25°C.
Current Transformer (CT) Selection
The CTs and their burden resistors should be selected to
maximize the use of the dynamic range on Channel V1A
and V1B (current channel). However there are some
important considerations when selecting the CTs and
the burden resistors for energy metering application.
Firstly, one need to select CTs that have good linearity in
both their gain and phase characteristics over the range
of current specified in the accuracy requirement. For
IEC1036, the range is between 5% Ib to I
. CT manu-
MAX
facturers often recommend the burden resistance to be
as small as possible to preserve linearity over large current
range. A burden resistance of less than 15 Ω is recommended. Secondly, CT introduces a phase shift between
primary and secondary current. The phase shift can contribute to a significant error at low-power factor. Note
that at power factor of 0.5, a phase shift as small as 0.1°
translates to 0.3% error in the energy reading. In this
design, the phase of the voltage channel (V2) is shifted
to match the phase shift introduced by the CT to eliminate any phase mismatch between the current and
voltage channel. This is achieved by moving the corner
frequency of the antialiasing filter in the voltage channel
input, see
and
Corrected Phase Matching between Channels
Antialias Filters
in this application note.
Design Calculations
Design Parameters:
Line Voltage = 240 V (Nominal)
I
= 40 A (Ib = 10 A)
MAX
Counter = 100 imp/kWh
Meter Constant = 3200 imp/kWh
CT Turn Ratio = 1:1800
Size of Burden Resistor (Channel 1 A) = 8.2 Ω
100 imp/hour = 100/3600 sec. = 0.027777 Hz
Meter Will Be Calibrated at Ib (10 A)
Power Dissipation at Ib = 240 V × 10 A = 2.4 kW
Frequency on F1 (and F2) at Ib = 2.4 × 0.027777 Hz
= 0.06666667 Hz
Voltage across CT at Ib (V1A) = 10 A/1800 × 8.2 Ω =
45.6 mV.
The gain setting is determined by the signal in V1 (current channel). At I
= 40 A, the rms voltage at V1 is
MAX
40 A/1800 × 8.2 Ω = 182 mV. It translates to a peak voltage
of 258 mV. From Table I of the AD7751 data sheet, it can
be seen that the gain of two provides the best utilization
of the dynamic range (±330 mV). The setting also provides more than 20% headroom in the event of surge in
the current.
To select the F
data sheet,
Application
frequency for Equation 1 see the AD7751
1-4
Selecting a Frequency for an Energy Meter
. From Tables V and VI in the AD7751 data
sheet, it can be seen that the best choice of frequency for
a meter with I
= 40 A is 3.4 Hz (F2). This frequency
MAX
selection is made by the logic inputs S0 and S1, see
Table II in the AD7751 data sheet. The CF frequency
selection (meter constant) is selected by using the
logic input SCF. The two available options are 64 × F1
(6400 imp/kWh) or 32 × F1 (3200 imp/kWh). For this design,
3200 imp/kWh is selected by setting SCF logic low. With
a meter constant of 3200 imp/kWh and a maximum current of 40 A, the maximum frequency from CF is 8.53 Hz.
Many calibration benches used to verify meter accuracy
still use optical techniques. This limits the maximum frequency which can be reliably read to about 10 Hz. The
only remaining unknown from Equation 1 is V2 or the
signal level on Channel 2 (the voltage channel).
From Equation 1 on the previous page:
0 0666667
.
=
25
.
2
mV VHz
××××
5 74 45 562 2 3 4
...
Hz
Where: V2 = 234.3 mV rms.
Therefore, in order to calibrate the meter, the line voltage needs to be attenuated down to 234.3 mV.
CALIBRATING THE METER: VOLTAGE CHANNEL
CALIBRATION
From the previous section it can be seen that the meter
is simply calibrated by attenuating the line voltage down
to 234.3 mV. The line voltage attenuation is carried out
by a simple resistor divider as shown in Figure 2. The
attenuation network should allow a calibration range of
at least ±30% to allow for CT/burden resistance tolerances and the on-chip reference tolerance of ±8%, see
the AD7751 data sheet. In addition, the topology of the
network is such that the phase matching between
Channel 1 and Channel 2 is preserved, even when the
attenuation is being adjusted, see
ing between Channels
R9
J5
J4
R8
R7
J3
R6
J2
J1
R5
in this application note.
R14
R13
R12
R11
R10
J10
J9
J8
J7
J6
Correct Phase Match-
234.3mV
R4B
R5 + R6 + ........... + R15 + R16 >> R4B
f
–3dB
C5
1/(2 R4B C5)
R15R16
240V
Figure 2. Attenuation Network for Calibrating the
Voltage Channel (V2)
REV. 0
–3–
AN-563
As can be seen from Figure 2, the –3 dB frequency of this
network is determined by R4B and C5. Even with all the
jumpers closed, the resistance of R15 (300 kΩ) and R16
(300 kΩ) is still much greater than R4B (887 Ω). Hence
varying the resistance of the resistor chain R5 to R14 will
have little effect on the –3 dB frequency of the network.
The network shown in Figure 2 allows the line voltage to
be attenuated and adjusted in the range 170 mV to
399 mV with a resolution of 10 bits or 223 µV. This is
achieved by using the binary weighted resister chain
R14 to R23. This will allow the meter to be accurately
calibrated using a successive approximation technique.
Starting with J1 each jumper is closed in order of
ascendance, e.g., J1, J2, J3, etc. If the calibration frequency on CF, i.e., 32 × 100 imp/KWh (at Ib = 10 A, CF is
expected to be 2.133 Hz) is exceeded when any jumper
is closed, it should be opened again. All jumpers are
tested, J10 being the last jumper. Note that jumper connections are made with soldering together the jumper
pins across the resistors in the network. This approach is
preferred over the use of trim pots, as the stability of
the latter over time and environmental conditions is
questionable.
Since the AD7751 transfer function is extremely linear, a
one-point calibration (at Ib) at unity power factor is all
that is needed to calibrate the meter. If the correct precautions have been taken at the design stage no
calibration will be necessary at low-power factor (e.g.,
PF = 0.5).
CALIBRATING THE METER: MATCHING THE TWO
CURRENT SENSOR INPUTS
A calibration network consisting of six parallel resistors
is used to compensate gain variation between the two
CTs used to monitor the phase and neutral currents.
However, such mismatch is often small and needs to be
compensated with a more accurate calibration network.
In this design, six resistors are used for this purpose.
The primary burden resistors for V1B, R6, and R7, combined to a 8.4 Ω burden (9.1 Ω储110 Ω = 8.4 Ω). This is
about 2.5% above the nominal burden used in V1A. The
burden is reduced by connecting the jumpers from J16
to J21. This adds more resistors to be in parallel to the
burden, thus reducing the total resistance between the
two terminals of the CT. The values of R8 to R13 are chosen carefully so that the resulting resistance values
spread out evenly across the calibration range. Closing
all jumpers, J16 to J21, represents the lower bound for
the calibration range. In our design, the lower bound is
at approximately 7.99 Ω, or 2.5% lower than the nominal
burden of V1A.
Starting from J16, each jumper is closed in order of
ascendance, e.g., J16, J17, J18, etc. If the calibration frequency on CF becomes smaller than the expected value
(at Ib = 10 A, CF = 2.133 Hz) after a jumper is closed, the
jumper should be opened again. All jumpers are tested,
J21 being the last jumper.
FG0003
1:1800
J17
J18
J19
J20
R12
J21
R13R7
J16
R8
R9
R10
R6
R11
Figure 3. Calibration Network for V1B
CORRECT PHASE-MATCHING BETWEEN CHANNELS
The AD7751 is internally phase-matched over the frequency range 40 Hz to 1 kHz. Correct phase-matching is
important in an energy metering application because
any phase mismatch between channels will translate
into significant errors at low-power factor. This is easily
illustrated with the following example. Figure 4 shows
the voltage and current waveforms for an inductive
load. In the example shown the current lags the voltage
by 60° (PF = 0.5). Assuming pure sinusoidal conditions
the power is easily calculated as V rms × I rms × cos
(60°).
PF = 1
V.I
2
PF = 0.5
V.I COS(60)
2
CURRENT
VOLTAGE
CURRENT
INSTANTANEOUS
POWER SIGNAL
VOLTAGE
INSTANTANEOUS
POWER SIGNAL
60
INSTANTANEOUS REAL
POWER SIGNAL
INSTANTANEOUS REAL
POWER SIGNAL
Figure 4. Voltage and Current (Inductive Load)
If, however, a phase error (φe) is introduced externally to
the AD7751, e.g., in the antialias filters, the error is calculated as:
[cos(δ°) – cos(δ°+φ
See
Note 3 in Table I.
voltage and current and φ
)]/cos(δ°) × 100%(2)
e
Where δ is the phase angle between
is the external phase error.
e
With a phase error of 0.2°, for example, the error at
PF = 0.5 (60°) is calculated as 0.6%. As this example
demonstrates, even a very small phase error will produce a large measurement error at low-power factor.
–4–
R3
V1B
C3
REV. 0
AN-563
The current sensor has an intrinsic phase shift of 0.1°. If
it is not compensated, it can introduce a significant error
at low-power factor. In this design, the phase is compensated by introducing a 0.1° phase shift in the voltage
channel to ensure both the current and voltage inputs
are phase matched. This is easily achieved by reducing the resistance in the antialiasing filter in the
voltage channel.
Antialias Filters
As mentioned in the previous section, one possible
source of external phase errors are the antialias filters
on Channel 1 and Channel 2. The antialias filters are lowpass filters that are placed before the analog inputs of
any ADC. They are required in order to prevent a possible distortion due to sampling called aliasing. Figure 5
illustrates the effects of aliasing.
ALIASING EFFECT
IMAGE
FREQUENCIES
0
–20
dB
–40
–60
1k10010
FREQUENCY – Hz
10k
100k1M
Figure 6. RC Filter Magnitude Response
0
–20
–40
DEGREES
–60
0
2
450
FREQUENCY – kHz
900
Figure 5. Aliasing Effects
Figure 5 shows how aliasing effects could introduce
inaccuracies in an AD7751-based meter design. The
AD7751 uses two Σ-∆ ADCs to digitize the voltage and
current signals. These ADCs have a very high sampling
rate, i.e., 900 kHz. Figure 5 shows how frequency components (arrows shown in black) above half the
sampling frequency (also known as the Nyquist frequency), i.e., 450 kHz are imaged or folded back down
below 450 kHz (arrows shown in grey). This will happen
with all ADCs no matter what the architecture is. In the
example shown it can be seen that only frequencies
near the sampling frequency, i.e., 900 kHz, will move
into the band of interest for metering, i.e., 0 kHz–2 kHz.
This fact will allow us to use a very simple LPF (LowPass Filter) to attenuate these high frequencies (near
900 kHz) and so prevent distortion in the band of interest.
The simplest form of LPF is the simple RC filter. This is a
single-pole filter with a roll-off or attenuation of –20 dB/dec.
CHOOSING THE FILTER –3 dB FREQUENCY
As well as having a magnitude response, all filters
also have a phase response. The magnitude and
phase response of a simple RC filter (R = 1 kΩ, C = 33 nF)
are shown in Figures 6 and 7. From Figure 6 it is seen
that the attenuation at 900 kHz for this simple LPF is
greater than 40 dBs. This is enough attenuation to ensure
no ill effects due to aliasing.
–80
–100
1k10010
FREQUENCY – Hz
10k
100k1M
Figure 7. RC Filter Phase Response
As explained in the last section, the phase response can
introduce significant errors if the phase response of the
LPFs on both Channel 1 and Channel 2 are not matched.
Phase mismatch can easily occur due to poor component tolerances in the LPF. The lower the –3 dB
frequency in the LPF (antialias filter) the more pronounced these errors will be at the fundamental
frequency component or the line frequency. Even with
the corner frequency set at 4.8 kHz (R = 1 kΩ, C = 33 nF)
the phase errors due to poor component tolerances can
be significant. Figure 8 illustrates the point. In Figure 8,
the phase response for the simple LPF is shown at 50 Hz
for R = 1 kΩ ± 10%, C = 33 nF ± 10%. Remember a phase
shift of 0.1°–0.2° can cause measurement errors of 0.6%
at low-power factor. This design uses resistors of 1% tolerance and capacitors of 10% tolerance for the antialias
filters to reduce the possible problems due to phase
mismatch. Alternatively the corner frequency of the
antialias filter could be pushed out to 10 kHz–15 Hz.
However, the corner frequency should not be made too
high. This could allow enough high-frequency components to be aliased and cause accuracy problems in a
noisy environment.
REV. 0
–5–
AN-563
–0.4
–0.5
–0.6
DEGREES
–0.7
–0.8
45
FREQUENCY – Hz
(50Hz, –0.481)
(R = 900, C = 29.7nF)
(50Hz, –0.594)
(R = 1k, C = 33nF)
(50Hz, –0.718)
(R = 1.1k, C = 36.3nF)
50
55
Figure 8. Phase Shift at 50 Hz Due to Component
Tolerances
Note this is also why precautions were taken with the
design of the calibration network on Channel 2 (voltage
channel). Calibrating the meter by varying the resistance of the attenuation network will not vary the –3 dB
frequency and hence the phase response of the network
on Channel 2, see
nel Calibration
Calibrating the Meter: Voltage Chan-
. Shown in Figure 9 is a plot of phase
lag at 50 Hz when the resistance of the calibration network is varied from 600 kΩ (J1–J10 closed) to 1.2 MΩ
(J1–J10 open).
when no load is connected. IEC 1036 (1996–09) Section
4.6.4 specifies the start-up current as being not more
than 0.4% Ib at PF = 1. For this design the start current is
calculated at 7.14 mA or 0.07% Ib, see
old
, AD7751 data sheet.
No Load Thresh-
POWER SUPPLY DESIGN
This design uses a simple low-cost power supply based
on a capacitor divider network, i.e., C18 and C19. Most of
the line voltage is dropped across C18, a 470 nF, 250 V
metalized polyester film capacitor. The impedance of
C18 dictates the effective VA rating of the supply. However the size of C18 is constrained by the power
consumption specification in IEC1036. The total power
consumption in the voltage circuit including power supply is specified in Section 4.4.1.1 of IEC1036 (1996–9).
The total power consumption in each phase is 2 W and
10 VA under nominal conditions. The nominal VA rating
of the supply in this design is 8.5 VA. The total power
dissipation is approximately 0.59 W. Figure 10 shows
the basic power supply design.
240V
C18
R26
V1
D2
+
C19
D3
U2
7805
5V
V2
V
DD
–0.591
–0.592
–0.593
DEGREES
–0.594
–0.595
J1–J10 OPEN
(50Hz, –0.59348)
49.950.0
FREQUENCY – Hz
J1–J10 CLOSED
(50Hz, –0.59299)
50.1
Figure 9. Phase Shift Due to Calibration
For the resistor network used for matching the CTs in
V1A and V1B, the calibration network is based on altering the burden resistance. High-precision components
are used for the RC filters in V1A and V1B. Any mismatch will be caused by the component tolerance.
Under normal operation, this mismatch is well within
the 12.5% threshold needed to cause AD7751 to indicate
a FAULT condition and perform current channel switching.
Figure 10. Power Supply
The plots shown in Figures 11, 12, 13, and 14 show the
PSU performance under heavy load (50 A) with the line
voltage varied from 180 V to 250 V. By far the biggest
load on the power supply is the current required to drive
the stepper motor which has a coil impedance of about
400 Ω. This is clearly seen by looking at V1 (voltage on
C19) in the plots below. Figure 11 shows the current
drawn from the supply. Refer to Figure 10 when reviewing the simulation plots below.
15
10
VOLTS
5
V1
(C19)
V2
(V
DD
C19 VOLTAGE DROP DUE TO
)
STEPPER MOTOR DRIVE
NO LOAD THRESHOLD
The AD7751 has on-chip anticreep functionality. The
AD7751 will not produce a pulse on CF, F1, or F2 if the
output frequency falls below a certain level. This feature
ensures that the energy meter will not register energy
0
0
2
4
68
TIME – s
Figure 11. Power Supply Voltage Output at 220 V and
50 A Load
–6–
10
REV. 0
AN-563
24
12.5mA
MOTOR DRIVE
20
16
4 mA LED/OPTO DRIVE
12
mA
8
4
0
0
234
TIME – s
51
Figure 12. Power Supply Current Output at 220 V and
50 A Load
15
10
V2
)
VOLTS
5
0
0102
V1
(C19)
(V
DD
468
TIME – s
Figure 13. Power Supply Voltage Output at 180 V and
50 A Load
15
DESIGN FOR IMMUNITY TO ELECTROMAGNETIC
DISTURBANCE
In Section 4.5 of IEC1036 it is stated that "the meter shall
be designed in such a way that conducted or radiated
electromagnetic disturbances as well as electrostatic
discharge do not damage nor substantially influence the
meter." The considered disturbances are:
1. Electrostatic Discharge
2. Electromagnetic HF Fields
3. Fast Transience Burst
All of the precautions and design techniques (e.g., ferrite beads, capacitor line filters, physically large SMD
resistors, PCB layout including grounding) contribute to
a certain extent in protecting the sensitive meter electronics from each form of electromagnetic disturbance.
Some precautions (e.g., ferrite beads) however, play a
more important role in the presence of certain kinds of
disturbances (e.g., RF and fast transience burst). The following discusses each of the disturbances listed above
and details what protection has been put in place.
ELECTROSTATIC DISCHARGE (ESD)
Although many sensitive electronic components contain a certain amount of ESD protection on-chip, it is not
possible to protect against the kind of severe discharge
described below. Another problem is that the effects of
an ESD discharge is cumulative, i.e., a device may survive an ESD discharge, but this is no guarantee that it
will survive multiple discharges at some stage in the
future. The best approach is to eliminate or attenuate
the effects of the ESD event before it comes in contact
with sensitive electronic devices. This holds true for
all conducted electromagnetic disturbances. This test
is carried out according to IEC1000-4-2, under the following conditions:
10
V1
VOLTS
5
0
0102
(C19)
V2
(V
)
DD
468
TIME – s
Figure 14. Power Supply Voltage Output at 250 V and
50 A Load
REV. 0
• Contact Discharge;
• Test Severity Level 4;
• Test Voltage 8 kV;
• 10 Discharges.
Very often no additional components are necessary to
protect devices. With a little care those components
already required in the circuit can perform a dual role.
For example, the meter must be protected from ESD
events at those points where it comes in contact with the
"outside world," e.g., the connection to the phase wire.
For the current input, AD7751 is connected to two CTs
through antialias filters. The CTs insulate the AD7751
from outside contact. The only path for ESD comes from
the phase wire to the voltage input. Two ferrite beads
–7–
AN-563
are placed in series with the connection to the line. A
ferrite choke is particularly effective at slowing the fast
rise time of an ESD current pulse. The high-frequency
transient energy is absorbed in the ferrite material rather
than being diverted or reflected to another part of the system (the properties of ferrite are discussed later). The
PSU circuit is also connected directly to the terminals of
the meter. Here the discharge will be dissipated by the
ferrite, the line filter capacitor (C18), and the rectification
diodes D2 and D3. The analog input V2P is also protected by the large impedance of the attenuation
network used for calibration. This antialias (RC) filter
can also be enough to protect against ESD damage to
CMOS devices. However, some care must be taken with
the type of components used. For example, the resistors
should not be wire-wound as the discharge will simply
travel across them. The resistors should also be physically large to stop the discharge arcing across the
resistor. In this design 1/8 W SMD 1206 resistors were
used in the antialias filters.
Another very common low-cost technique used to arrest
ESD events is to use a spark gap on the component side
of the PCB, see Figure 15. However, since the meter will
likely operate in an open air environment and be subject
to many discharges, this is not recommended at sensitive nodes. Multiple discharges could cause carbon
build-up across the spark gap which could cause a short
or introduce an impedance that will, in time, affect accuracy. A spark gap was introduced in the PSU after the
MOV to take care of any very high-amplitude/fast rise
time discharges.
8kV ESD
EVENT
TO EXTERIOR
(I/O) CONNECTION
6–9 MILS
NO SOLDER
MASK
TRACE (TRACK)
ESD DISCHARGED
ACROSS SPARK GAP
SIGNAL
GROUND
TO CIRCUIT
103
Figure 15. Spark Gap to Arrest ESD Events
ELECTROMAGNETIC HF FIELDS
Susceptibility of integrated circuits to RF tends to be
more pronounced in the 20 MHz–200 MHz region. Frequencies higher than this tend to be shunted away from
sensitive devices by parasitic capacitances. In general,
at the IC level, the effects of RF in the region 20 MHz–
200 MHz will tend to be broadband in nature, i.e., no
individual frequency is more troublesome than another.
However, there may be higher sensitivity to certain frequencies due to resonances on the PCB. These
resonances could cause insertion gain at certain frequencies which, in turn, could cause problems for
sensitive devices. By far the greatest RF signal levels are
those coupled into the system via cabling. These connection points should be protected. Some techniques for
protecting the system are:
1. Minimize Circuit Bandwidth
2. Isolate Sensitive Parts of the System
Minimize Bandwidth
In this application the required analog bandwidth is only
2 kHz. This is a significant advantage when trying to
reduce the effects of RF. The cable entry points can be
low-pass filtered to reduce the amount of RF radiation
entering the system. The only direct connection to the
cable is at the voltage inputs. The inputs are low-pass
filtered to prevent aliasing effects which were described
earlier. By Choosing the correct components and adding
some additional components (e.g., ferrite beads) these
antialias filters can double as effective RF filters. The ferrite bead is an ideal component for this application. The
RF radiation is dissipated as heat rather than being
reflected or diverted to another part of the system. The
ferrite beads Z1 and Z2 perform very well in this respect.
Figure 16 shows how the impedance of the ferrite beads
varies with frequency.
260
240
220
200
180
160
140
120
.R.X –
I
Z
I
100
80
60
40
20
0
1101001k
IZI
R
X
FREQUENCY – MHz
Figure 16. Frequency Response of the Ferrite Chips
(Z1 and Z2)
From Figure 16 it can be seen that the ferrite material
becomes predominately resistive at high frequencies.
Also note that the impedance of the ferrite material
increases with frequency, causing only high (RF) frequencies to be attenuated.
ISOLATION
On the current channels (V1A and V1B), current transformers are used to isolate the line from the system. The
system is connected to the phase and neutral lines for
the purpose of generating a power supply and voltage
channel signal (V2). The ferrite bead (Z1) and line filter
capacitor (C18) should significantly reduce any RF radiation on the power supply.
–8–
REV. 0
4kV
90%
50%
10%
TIME
50ns
5ns
Another possible path for RF is the signal ground for the
system. A moating technique has been used to help isolate the signal ground surrounding the AD7751 from the
external ground reference point (K6). Figure 18 illustrates the principle of this technique called partitioning
or "moating."
"MOAT" – NO
POWER OR
GROUND PLANE
I/O
CONNECTION
POWER CONNECTION
MADE USING FERRITE
"BEAD ON LEAD"
Figure 17. High-Frequency Isolation of I/O
Connections Using a "Moat"
Sensitive regions of the system are protected from RF
radiation entering the system at I/O connection. An area
surrounding the I/O connection does not have any
ground or power planes. This limits the conduction
paths for RF radiation and is called a "moat." Obviously
power, ground, and signal connections must cross this
moat and Figure 17 shows how this can be safely
achieved by using a ferrite bead. Remember that ferrite
offers a large impedance to high frequencies—see Figure 16.
ELECTRICAL FAST TRANSIENCE BURST TESTING (EFT)
This testing determines the immunity of a system to
conducted transients. Testing is carried out in accordance with IEC1000-4-4 under well-defined conditions.
The EFT pulse can be particularly difficult to guard
against because the disturbance is conducted into the
system via external connections, e.g., power lines. Figure 18 shows the physical properties of the EFT pulse
used in IEC1000-4-4. Perhaps the most debilitating
attribute of the pulse is not its amplitude (which can be
as high as 4 kV), but the high-frequency content due to
the fast rise times involved. Fast rise times mean highfrequency content which allows the pulse to couple to
other parts of the system through stray capacitance, etc.
Large differential signals can be generated by the inductance of PCB traces and signal ground. These large
differential signals could interrupt the operation of
sensitive electronic components. Digital systems are
generally most at risk because of data corruption. Analog electronic systems tend to be affected only for the
duration of the disturbance.
AN-563
Figure 18. Single EFT Pulse Characteristics
Another possible issue with conducted EFT is that the
effects of the radiation will, like ESD, generally be cumulative for electronic components. The energy in an EFT
pulse can be as high as 4 mJ and deliver 40 A into a 50 Ω
load, see Figure 21. Therefore continuous exposure to
EFT due to inductive load switching etc., may have implications for the long term reliability of components. The
best approach is to protect those parts of the system
which could be sensitive to EFT.
The protection techniques described in the last section
(Electromagnetic HF Fields) also apply equally well in
the case of EFT. The electronics should be isolated as
much as possible from the source of the disturbance
through PCB layout (i.e., moating) and filtering signal
and power connections. In addition, a 10 nF capacitor
(C17) placed across the mains provides a low-impedance
shunt for differential EFT pulses. Stray inductance due
to leads and PCB traces will mean that the MOV will not
be very effective in attenuating the differential EFT
pulse. The MOV is very effective in attenuating high
energy, relatively long duration disturbances, e.g., due
to lightening strikes, etc. The MOV is discussed in the
next section.
MOV Type S20K275
The MOV used in this design was of type S20K275 from
Siemens. An MOV is basically a voltage-dependant
resistor whose resistance decreases with increasing
voltage. The MOV is typically connected in parallel with
the device or circuit being protected. During an overvoltage event it forms a low-resistance shunt and thus
prevents any further rise in the voltage across the circuit
being protected. The overvoltage is essentially dropped
across the source impedance of the overvoltage source,
e.g., the mains network source impedance. Figure 19
illustrates the principle of operation.
REV. 0
–9–
AN-563
C16
10nF
MOV
50
L1
5nH
SW1SW2
R2
50
C2
10nF
C1
6F
R1
0.01
0.5kV
TO 5kV
+
SOURCE
Z
S
ELECTRONIC
CIRCUIT
MOV
V
t
i*
LEAKAGE
CURRENT >> 0
TO BE
PROTECTED
"LOAD LINE" OF
THE OVERVOLTAGE
V/I CHARACTERISTIC
CURVE OF MOV
i*
SURGE
CURRENT
VB, V
V
S
OVERVOLTAGE
V
B
V
V
S
MOV
TAKEN FROM SIEMENS MATSUSHITA COMPONENTS
*
SIOV METAL OXIDE VARISTOR CATALOG
Figure 19. Principle of MOV Overvoltage Protection
The plot in Figure 19 shows how the MOV voltage and
current can be estimated for a given overvoltage and
source impedance. A load line (open-circuit voltage,
short-circuit current) is plotted on the same graph as the
MOV characteristic curve. Where the curves intersect,
the MOV clamping voltage and current can be read.
Note, care must be taken when determining the shortcircuit current. The frequency content of the overvoltage
must be taken into account as the source impedance
(e.g., mains) may vary considerably with frequency. A
typical impedance of 50 Ω is used for mains source
impedance during fast transience (high-frequency)
pulse testing. The next section discusses IEC1000-4-4
and IEC1000-4-5 which are transience and overvoltage
EMC compliance tests.
IEC1000-4-4 and the S20K275
While the graphical technique just described is useful,
an even better approached is to use simulation to obtain
a better understanding of MOV operation. Siemens
Matsushita Components provides SPICE models for all
their MOVs and these are very useful in determining
device operation under the various IEC EMC compliance
tests. For more information on S&M SPICE models and
their applications see:
http://www.siemens.de/pr/index.htm
The purpose of IEC1000-4-4 is to determine the effect
of repetitive, low-energy, high-voltage, fast rise time
pulses on an electronic system. This test is intended to
simulate transient disturbances such as those originating from switching transience (e.g., interruption of
inductive loads, relay contact bounce, etc.).
Figure 20 shows an equivalent circuit that is intended to
replicate the EFT test pulse as specified in IEC1000-4-4. The
generator circuit is based on Figure 1 IEC1000-4-4 (1995-
01). The characteristics of operation are:
• Maximum Energy of 4 mJ/Pulse at 2 kV into 50 Ω
• Source Impedance of 50 Ω ± 20%
• D.C. Blocking Capacitor of 10 nF
• Pulse Rise Time of 5 ns ± 30%
• Pulse Duration (50% Value) of 50 ns ± 30%
• Pulse Shape as Shown in Figure 18
i
Figure 20. EFT Generator
The simulated output of this generator delivered to a
purely resistive 50 Ω load is shown in Figure 21. The
open-circuit output pulse amplitude from the generator
is 4 kV. Therefore the source impedance of the generator
is 50 Ω as specified by the IEC1000-4-4, i.e., ratio of peak
pulse output unloaded and loaded (50 Ω) is 2:1.
The plot in Figure 21 also shows the current and instantaneous power (V × I) delivered to the load. The total energy
is the integral of the power and can be approximated by
the rectangle method as shown. It is approximately 4 mJ
at 2 kV as per specification.
4kV
3kV
2kV
1kV
100kW
50A
ENERGY = 80kW 50ns = 4mJ
40A
80kW
30A
20A
10A
60kW
40kW
20kW
0W
0A0V
3.00
POWER
CURRENT
VOLTAGE
TIME – s
Figure 21. EFT Generator Output into 50
3.203.043.083.123.16
Ω
(No Protection)
Figure 22 shows the generator output into 50 Ω load with
the MOV and some inductance (5 nH). This is included
to take into account stray inductance due to PCB traces
and leads. Although the simulation result shows that the
EFT pulse has been attenuated (600 V) and most of the
energy being absorbed by the MOV (only 0.8 mJ is delivered to the 50 Ω load), it should be noted that stray
–10–
REV. 0
inductance and capacitance could render the MOV unless.
3.83.03.23.43.6
TIME – s
–80A–100V
0V
100V
200V
300V
4.0
–60A
–40A
–20A
0A
20A
CURRENT (INTO C16)
VOLTAGE (ACROSS 50 LOAD)
For example, Figure 25 shows the same simulation with
the stay inductance increased to 1 µH, which could easily happen if proper care is not taken with the layout. The
pulse amplitude reaches 2 kV once again.
8kW
20A
800V
600V
400V
200V
15A
10A
6kW
4kW
2kW
5A
POWER (INTO 50)
CURRENT (INTO 50)
VOLTAGE
AN-563
Figure 24. EFT Generator Output into 50 Ω with MOV
in Place, Stray Inductance of 1
Place
µ
H and C16 (10 nF) in
0V
0W
0A
3.00
TIME – s
Figure 22. EFT Generator Output into 50 Ω with MOV
in Place
2.0
1.6
1.2
0.8
VOLTS – kV
0.4
0
–0.4
VOLTAGE
TIME – s
Figure 23. EFT Generator Output into 50 Ω with MOV
in Place and Stray Inductance of 1
µ
H
When the 10 nF capacitor (C17) is connected a lowimpedance path is provided for differential EFT pulses.
Figure 24 shows the effect of connecting C17. Here the
stray inductance (L1) is left at 1 µH and the MOV is in
place. The plot shows the current through C17 and the
voltage across the 50 Ω load. The capacitor C17 provides
a low-impedance path for the EFT pulse. Note the peak
current through C17 of 80 A. The result is the amplitude
of the EFT pulse is greatly attenuated.
3.203.043.083.123.16
IEC1000-4-5
The purpose of IEC1000-4-5 is to establish a common
reference for evaluating the performance of equipment
when subjected to high-energy disturbances on the
power and interconnect lines. Figure 25 shows a circuit
that was used to generate the combinational wave
(hybrid) pulse described in IEC1000-4-5. It is based on
the circuit shown in Figure 1 of IEC1000-4-5 (1995-02).
Such a generator produces a 1.2 µs/50 µs open-circuit
voltage, which is why it is referred to as a hybrid generator. The surge generator has an effective output
impedance of 2 Ω. This is defined as the ratio of peak
open-circuit voltage to peak short-circuit current.
L
SW1SW2
+
0.5kV
3.203.003.053.103.15
TO 4kV
C1
20F
1.9
R1
3.9
R2
10H
R3
50
5nH
C17
10nF
L1
MOV
S20K275
Figure 25. Surge Generator (IEC1000-4-5)
Figure 26 shows the generator voltage and current output wave forms. The characteristics of the combination
wave generator are:
Open Circuit Voltage
• 0.5 kV to at least 4.0 kV
• Waveform as shown in Figure 26
• Tolerance on open-circuit voltage is ±10%.
Short-Circuit Current
• 0.25 kA to 2.0 kA
• Waveform as shown in Figure 26
• Tolerance on short-circuit current is ±10%
Repetition rate of at least 60 seconds
REV. 0
–11–
AN-563
I
NOISE
GROUND
V
NOISE
= I
NOISE
Z
COMMON
IMPEDANCE
Z
+
+
ANALOG CIRCUITRY
DIGITAL CIRCUITRY
4kV
2.0kA
3kV
1.5kA
2kV
1.0kA
1kV
0.5kA
0V
0A
0
TIME – s
Figure 26. Open-Circuit Voltage/Short-Circuit Current
The MOV is very effective in suppressing these kinds of
high-energy/long-duration surges. Figure 27 shows the
voltage across the MOV when it is connected to the generator as shown in Figure 25. Also shown are the current
and instantaneous power waveform. The energy absorbed
by the MOV is readily estimated using the rectangle
method as shown.
1.0kV 2.0kA
0.8kV
0.6kV
0.4kV
0.2kV
0V
1.5kA
1.0kA
0.5kA
0A
1.5MW
1.0MW
0.5MW
0W
ENERGY = 1.3MW 30s = 40 JOULES
POWER
CURRENT
VOLTAGE
0
TIME – s
Figure 27. Energy Absorbed by MOV During 4 kV Surge
Derating the MOV Surge Current
The maximum surge current (and therefore energy
absorbed) that an MOV can handle is dependent on the
number of times the MOV will be exposed to surges
over its lifetime. The life of an MOV is shortened every
time it is exposed to a surge event. The data sheet for an
MOV device will list the maximum nonrepetitive surge
current for an 8 µs/20 µs current pulse. If the current
pulse is of longer duration, and if it occurs more than
once during the life of the device, this maximum current
must be derated. Figure 28 shows the derating curve for
the S20K275. Assuming exposures of duration 30 µs and
a peak current as shown in Figure 27, the maximum
number of surges the MOV can handle before it goes out
of specification is about 10. After repeated loading (10
times in the case just described) the MOV voltage will
change. After initially increasing, it will decay rapidly.
250
SIOV-S20K275
4
10
A
3
10
2
10
MAX
I
1
10
0
10
10020406080
–1
10
10100100010,000
SIEMENS MATSUSHITA COMPONENTS
4
10
1
x
2
10
10
5
10
6
10
t
r
t
2
3
10
I
MAX
r
Figure 28. Derating Curve for S20K275
PCB DESIGN
Both susceptibility to conducted or radiated electromagnetic disturbances and analog performance were
considered at the PCB design stage. Fortunately, many
of the design techniques used to enhance analog and
mixed-signal performance also lend themselves well to
improving the EMI robustness of the design. The key
idea is to isolate that part of the circuit which is sensitive
to noise and electromagnetic disturbances. Since the
AD7751 carries out all the data conversion and signal
processing, the robustness of the meter will be determined to a large extent by how protected the AD7751 is.
In order to ensure accuracy over a wide dynamic range,
the data acquisition portion of the PCB should be kept as
quiet as possible, i.e., minimal electrical noise. Noise
will cause inaccuracies in the analog-to-digital conversion process which takes place in the AD7751. One
common source of noise in any mixed-signal system is
30050100150200
the ground return for the power supply. Here highfrequency noise (from fast edge rise times) can be
coupled into the analog portion of the PCB by the common impedance of the ground return path. Figure 29
illustrates the mechanism.
Figure 29. Noise Coupling via Ground Return Impedance
–12–
REV. 0
AN-563
AMPS
0.50
0.011000.1110
ERROR – %
0.40
0.30
0.20
0.10
0.00
–0.10
–0.20
–0.30
–0.40
–0.50
PF = 0.5
PF = 1
AMPS
0.50
0.011000.1110
ERROR – %
0.40
0.30
0.20
0.10
0.00
–0.10
–0.20
–0.30
–0.40
–0.50
PF = 0.5
PF = 1
One common technique which is used to overcome these
kinds of problems is to use separate analog and digital
return paths for the supply. Also every effort should be
made to keep the impedance of these return paths as low
as possible. In the PCB design for the AD7751, separate
ground planes were used to isolate the noisy ground
returns. The use of ground plane also ensures the impedance of the ground return path is kept as very low. The
AD7751 and sensitive signal paths are located in a "quiet"
part of the board which is isolated from the noisy elements
of the design like the power supply, flashing LED etc. Since
the PSU is capacitor- based a substantial current (approximately 35 mA at 240 V) will flow in the ground return back
to the wire (system ground). This is shown in Figure 29. By
locating the PSU in the digital portion of the PCB this
return current is kept away from the AD7751 and analog
input signals. This current is at the same frequency as the
signals being measured and could cause accuracy issues
(e.g., crosstalk between the PSU as analog inputs) if
care is not taken with the routing of the return current.
Also part of the attenuation network for the Channel 2
(voltage channel) is in the digital portion of the PCB.
This helps to eliminate possible crosstalk to Channel 1
by ensuring analog signal amplitudes are kept as low as
possible in the analog ("quiet") portion of the PCB. Figure 30 shows the PCB design which was eventually
adopted for the watt-hour meter.
The partitioning of the power planes in the PCB design
shown in Figure 30 also allows us to implement the idea
of a "moat" for the purposes of immunity to electromagnetic disturbances. The digital portion of the PCB is
the only place where both phase and neutral wires are
connected. This portion of the PCB contains the transience suppression circuitry (MOV, ferrite, etc.) and
power supply circuitry. The ground planes are connected via a ferrite bead which help to isolate the
analog ground from high-frequency disturbances, see
Design For Immunity to Electromagnetic Disturbances
The ANSI standard governing Solid-State Electricity
Meters is ANSI C12.16-1991. Since this application note
refers to the IEC 1036 specifications when explaining the
design, this section will explain some of those key
IEC1036 specifications in terms of their ANSI equivalents. This should help eliminate any confusion caused
by the different application of some terminology contained in both standards.
Class—IEC1036
The class designation of an electricity meter under
IEC1036 refers to its accuracy. For example, a Class 1
meter will have a deviation from reference performance
of no more than 1%. A Class 0.5 meter will have a maximum deviation of 0.5% and so on. Under ANSI C12.16
Class refers to the maximum current the meter can
handle for rated accuracy. The given classes are: 10, 20,
100, 200 and 320. These correspond to a maximum
meter current of 10 A, 20 A, 100 A, 200 A, and 320 A,
respectively.
Ibasic (Ib)—IEC1036
The basic current (Ib) is a value of current with which the
operating range of the meter is defined. IEC1036 defines
the accuracy class of a meter over a specific dynamic
range, e.g., 0.05 Ib
< I < I
. It is also used as the test
MAX
load when specifying the maximum permissible effect
of influencing factors, e.g., voltage variation and frequency variation. The closest equivalent in ANSI C12.16
is the Test Current. The Test Current for each meter class
(maximum current) is given below:
Class 10 : 2.5 A
Class 20 : 2.5 A
Class 100 : 15 A
Class 200 : 30 A
Class 320 : 50 A
I
–IEC1036
MAX
I
is the maximum current for which the meter meets
MAX
rated accuracy. This would correspond to the meter
class under ANSI C12.16. For example a meter with an
I