LM1893/LM2893 Carrier-Current Transceiver ²
General Description
Carrier-current systems use the power mains to transfer information between remote locations. This bipolar carriercurrent chip performs as a power line interface for half-duplex (bi-directional) communication of serial bit streams of
virtually any coding. In transmission, a sinusoidal carrier is
FSK modulated and impressed on most any power line via a
rugged on-chip driver. In reception, a PLL-based demodulator and impulse noise filter combine to give maximum range.
A complete system may consist of the LM1893, a COPSTM
controller, and discrete components.
Noise resistant FSK modulation
User-selected impulse noise filtering
Up to 4.8 kBaud data transmission rate
Strings of 0’s or 1’s in data allowed
Sinusoidal line drive for low RFI
Output power easily boosted 10-fold
50 to 300 kHz carrier frequency choice
TTL and MOS compatible digital levels
Regulated voltage to power logic
Drives all conventional power lines
Energy management systems
Home convenience control
Inter-office communication
Appliance control
Fire alarm systems
Security systems
Computer terminal interface
Typical Application
TL/H/6750 – 1
FIGURE 1. Block diagram of carrierÐcurrent chip with a complement of discrete components making a complete
FO e 125 kHz, fDATA e 360 Baud transceiver. Use caution with this circuitÐdangerous line voltage is present.
BI-LINETM and COPSTM are trademarks of National Semiconductor Corp.
² Carrier-Current Transceivers are also called Power Line Carrier (PLC) transceivers.
C1995 National Semiconductor Corporation
RRD-B30M115/Printed in U. S. A.
LM1893/LM2893 Carrier-Current Transceiver
April 1995
Absolute Maximum Ratings
Maximum continuous dissipation, TA e 25§ C,
plastic DIP N (Note 2): transmit mode
1.66 W
receive mode
1.33 W
b 40 to 85§ C
Operating ambient temp. range
b 65 to 150§ C
Storage temperature range
Lead temp., soldering, 7 seconds
260§ C
Note: Absolute maximum ratings indicate limits beyond
which damage to the device may occur. Electrical specifications are not ensured when operating the device above
guaranteed limits but below absolute maximum limits, but
there will be no device degradation.
If Military/Aerospace specified devices are required,
please contact the National Semiconductor Sales
Office/Distributors for availability and specifications.
Supply voltage
Voltage on pin 12
Voltage on pin 10 (Note 1)
Voltage on pins 5 and 17
5.6 V DC zener current
Junction temperature: transmit mode
receive mode
Electro-Static Discharge (120 pF, 1500X)
30 V
55 V
41 V
40 V
100 mA
150§ C
125§ C
General Electrical Characteristics
(Note 3). The test conditions are: V a e 18V and FO e 125 kHz, unless otherwise noted.
(Note 4)
(Note 5)
5.6 V Zener voltage, VZ
Pin 11, IZ e 2 mA
5.6 V Zener resistance, RZ
Pin 11, RZ e (VZ @ 10 mA b VZ @ 1 mA)/(10 mA b 1 mA)
V min.
V max.
Carrier I/O peak survivable
transient voltage, VOT
Pin 10, discharge 1 mF cap. charged to VOT
thru k 1X
V max.
Carrier I/O clamp voltage, VOC
Pin 10, IOC e 10 mA, RX mode
2N2222 diode pin 8 to 9
V min.
V max.
Carrier I/O clamp resistance, R10
Pin 10, IOC e 10 mA
TX/RX low input voltage, VIL
Pin 5
V max.
TX/RX high input voltage, VIH
Pin 5 (Note 9)
V min.
TX/RX low input current, IIL
Pin 5 at 0.8 V
b 20
mA min.
mA max.
TX/RX high input current, IIH
Pin 5 at 40 V
10 b 4
mA min.
mA max.
RX b TX switch-over time, TRT
Time to develop 63% of full current drive thru pin 10
TX b RX switch-over time, TTR
1 bit time, TB e 1/(2FDATA). Time TTR is user
controlled with CM, see Apps. Info.
ICO initial accuracy of FO
TX mode, RO e 6.65 kX, CO e 560 pF
F0 e (F1 a F2)/2
ICO temperature coefficient of FO
Temperature drift of FO
TX or RX mode, b 40 s TJ s TJMAX
kHz min.
kHz max.
b 100
g 2.0
g 5.0
% max.
Transmitter Electrical Characteristics
(Note 3). The test conditions are: V a e 18 V and FO e 125 kHz
unless otherwise noted. The transmit center frequency is FO, FSK low is F1, and FSK high is F2.
(Note 4)
(Note 5)
V min.
V max.
Supply voltage, V a , range
Meets test 17 spec. at TJ e 25§ C and:
l(F1 [14V] bF1 [18V])/F1 [18V] l k0.01
l(F1 [24V] bF1 [18V])/F1 [18V] l k0.01
Total supply current, IQT
Pin 15. Pin 12 high. IQT is IQ through
pin 15 and the average current IODC of the
Carrier I/O through pin 10
mA max.
Carrier I/O output current, IO
100X load on pin 10
mApp min.
Carrier I/O lower swing limit, VALC
Pin 10. Set internally be ALC.
2N2222 diode pin 8 to 9
V min.
V max.
THD of IO (Note 6)
Q of 10 tank driving 10X line
100X load, no tank
FSK deviation, F2 b F1
(F2 b F1)/([F2 a F1]/2)
% min.
% max.
Data In. low input voltage, VIL
Pin 17
V max.
Data In. high input voltage, VIH
Pin 17 (Note 9)
V min.
Data In. low input current, IIL
Pin 17 at 0.8 V
b 10
mA min.
mA max.
Data In. high input current, IIH
Pin 17 at 40 V
10 b 4
% max.
% max.
mA min.
mA max.
Receiver Electrical Characteristics (Note 3). The test conditions are: V a e 18 V, FO e 125 kHz, g 2.2%
deviation FSK, FDATA e 2.4 kHz, VIN e 100 mVpp, in the receive mode, unless otherwise noted.
(Note 4)
(Note 5)
V min.
V max.
Supply voltage, V a , range
Functional receiver (Note 7)
Supply current, IQT
IQT is pin 15 (V a ) plus pin 10
(Carrier I/O) current. 2.4 kX Pin 13 to GND.
mA min.
mA max.
Carrier I/O input resistance, R10
Pin 10
kX min.
kX max.
Max. data rate, FMD
Functional receiver (Note 7), CF e 100 pF,
RF e 0X, no tank,
2.4 kHz e 4.8 kBaud
PLL capture range, FC
CF e 100 pF, RF e 0 X
g 40
g 15
g 10
% min.
PLL lock range, FL
CF e 100 pF, RF e 0 X
g 45
g 15
Receiver input sensitivity, SIN
For a functional receiver (Note 8)
Referred to chip side (pin 10)
of the line-coupling XFMR: FO e 50 kHz
FO e 300 kHz
Referred to line side of XFMR:
(assuming a 7.07:1 XFMR) FO e 50 kHz
FO e 300 kHz
Pin 10 lower than pin 15 by VINDC
% min.
Tolerable input dc voltage offset
range, VINDC
V max.
Data Out. breakdown voltage
Pin 12, leakage I s 20 mA
V min.
Data Out. low output, VOL
Pin 12, sat. voltage at IOL e 2 mA
V max.
Impulse noise filter current, II
Pin 13 charge and discharge current
g 55
g 45
g 85
mA min.
mA max.
Offset hold cap. bias voltage, VCM
Pin 6
V min.
V max.
Offset hold capacitor max. drive
current, IMCM
Pin 6. V(pin 3) b V(pin 4) e g 250 mV
g 55
g 25
g 80
mA min.
mA max.
Offset hold bias current, IOHB
Pin 6, TX mode. Bias pin 6 as it selfbiased during test 31.
b 0.5
b 20
b 40
nA min.
nA max.
Phase comparator current, IPC
Bias pins 3 and 4 at 8.5 V
IPC e I(pin 3) a I(pin 4), TX mode
mA min.
mA max.
Phase detector output resistance,
Pins 3 and 4.
RPD e (V @ 100mA b V @ 50mA)/(50mA)
kX min.
kX max.
Phase detector demodulated output
voltage, VPD
Pin 3 to 4, measured after filtering
out the 2FO component
mVpp min.
mVpp max.
Fast offset cancel voltage ‘‘window’’
-to-VPD ratio, VW/VPD
VPIN3 b VPIN4 e g VWINDOW a DC offset
Drive for g 1 mA pin 6 current
V/V min.
V/V max.
Power supply rejection, PSRR
CL e 0.1 mF. PSRR e CMRR. 120 Hz
dB min.
Note 1: More accurately, the maximum voltage allowed on pin 10 is VOC, and VOC ranges from 41 to 50V. Also, transients may reach above 60V; see the transient
peak voltage characteristic curve.
Note 2: The maximum power dissipation rating should be derated for device operation above 25§ C to insure that the junction temperature remains below the
maximum rating. Use a iJA of 75§ C/W for the N package using a socket in still air (which is the worst case). Consult the Application Information section for more
Note 3: The boldface values apply over the full junction temperature range for the specified supply voltage range. All other numbers apply at TA e TJ e 25§ C. Pin
numbers refer to LM1893. LM2893 tested by shorting Carrier In to Carrier Out and testing it as an LM1893.
Note 4: Guaranteed and 100% production tested.
Note 5: Guaranteed (but not 100% production tested) over the temperature and supply voltage ranges. These limits are not used to calculate outgoing quality
Note 6: Total harmonic distortion is measured using THD e [IRMS (all components at or above 2FO)]/[IRMS (fundamental)].
Note 7: Receiver function is defined as the error-free passage of 1 cycle of 50% duty-cycle 2.4 kHz square-wave data (2 sequential 208 mS bits), with the first bit
being a ‘‘1.’’ All of the data transitions (edges) must fall within g 10% ( g 20.8 ms) of their noise-free positions. RX time delay is minimized by using no impulse noise
filter cap. CI for this test.
Note 8: During the sensitivity check, note 7 requirements are followed with these exceptions: (1) data rate FDATA e 1.2 kHz, (2) all of the data transitions must fall
within g 20% ( g 41.6 ms) of their noise-free positions, and (3), a time-domain filter capacitor (CI) is used. The time delay of CI is (/2 bit, or 208 ms. (CI is
approximately 6200 pF).
Note 9: For TTL compatibility use a pull-up resistor to increase min. VOH to above 2.8 V.
Typical Performance Characteristics (V a e 18V, FO e 125 kHz, circuit of Figure 1 , pin numbers for
Total Current Consumption,
IQT, vs Supply Voltage
Total Current Consumption,
IQT, vs Junction Temperature
Chip Bias Current,
iQ, vs Supply Voltage
Chip Bias Current, IQ,
vs Junction Tempurature
Output Stage DC Current,
IODC, vs Output Voltage
Output Stage DC Current,
IODC, vs
Junction Temperature
Transient Voltage Survival
vs Pulse Time
Transmitter AC Output Current
vs Junction Temperature
Transmitter Sinusoid THD
vs Junction Temperature
ALC Voltage vs
Junction Temperature
ICO Frequency vs
Junction Temperature
Transmitter FSK Deviation
vs Junction Temperature
TL/H/6750 – 38
Typical Performance Characteristics
Maximum Data Rate vs
Junction Temperature
Receiver Sensitivity vs
Junction Temperature
PLL Lock Range vs
Junction Temperature and FO
PLL Capture & Lock Range vs
Junction Temperature
Receiver Sensitivity vs
PLL Lock Range and FO
Receiver Sensitivity vs
PLL Lock Range and Loop Filter
Impulse Noise Filter
Current vs Junction
Phase Detector Output
Voltage vs Junction
Offset Hold Cap. Charge
Currents vs Junction
Offset Hold Cap. Bias Current vs
Junction Temperature
Data Out. Low Voltage vs
Pull Down Current
Pin 7 Bias Voltage vs
Junction Temperature
TL/H/6750 – 39
Dual-In-Line Package
Application Information*
The BI-LINETM chip serves as a power line interface in the
carrier-current transceiver (CCT) system of Figure 3 . Figure
4 shows the interface circuit now discussed. The controller
may select either the transmit (TX) or receive (RX) mode.
Serial data from the controller is used to generate a FSKmodulated 50 to 300 kHz carrier on the line in the TX mode.
In the RX mode line signal passes through the coupling
transformer into the PLL-based receiver. The recreated serial bit stream drives the controller.
With the IC in the TX mode (pin 5 a logic high), baseband
data to 5 kHz drive the modulator’s Data In pin to generate
a switched 0.978I/1.022I control current to drive the low TC,
triangle-wave, current-controlled oscillator to g 2.2% deviation. The tri-wave passes through a differential attenuator
and sine shaper which deliver a current sinusoid through an
automatic level control (ALC) circuit to the gain of 200 current output amplifier. Drive current from the Carrier I/O develops a voltage swing on T1’s (Figure 4 ) resonant tank
proportional to line impedance, then passes through the
step-down transformer and coupling capacitor CC onto the
line. Progressively smaller line impedances cause reduced
signal swing, but never clipping-thus avoiding potential radio
frequency interference. When large line impedances threaten to allow excessive output swing on pin 10, the ALC
shunts current away from the output amplifier, holding the
voltage swing constant and within the amp’s compliance
limit. The amplifier is stable with a load of any magnitude or
phase angle.
In the RX mode (pin 5 a logic low), the TX sections on the
chip are disabled. Carrier signal, broad-band noise, transient
spikes, and power line component impinge of the receiver’s
input highpass filter, made up of CC and T1, and the tank
bandpass filter. In-band carrier signal, band-limited noise,
heavily attenuated line frequency component, and attenuated transient energy pass through to produce voltage swing
on the tank, swinging about the positive supply to drive the
Carrier I/O receiver input. The balanced Norton-input limiter
amplifier removes DC offsets, attenuates line frequency,
performs as a bandpass filter, and limits the signal to drive
the PLL phase detector differentially. The differential demodulated output signal from the phase detector, containing AC and DC data signal, noise, system DC offsets, and a
large twice-the-carrier-frequency component, passes
through a 3-stage RC lowpass filter to drive the offset cancel circuit differentially. The offset cancelling circuit works
by insuring that the (fixed) g 50 mV signal delivered to the
data squaring (‘‘slicing’’) comparator is centered around the
0 mV comparator switch point. Whenever the comparator
signal plus DC offset and noise moves outside the carefully
matched g 50 mV voltage ‘‘window’’ of the offset cancel
circuit, it adjusts its DC correction voltage in series with the
differential signal to force the signal back into the window.
While the signal is within the g 50 mV window, the DC offset
is stored on capacitor CM. By grace of the highly non-linear
offset hold capacitor charging during offset cancelling, the
DC cancellation is done much more quickly than with an AC
coupling capacitor normally used in place of the offset cancel circuit. Since impulse noise spikes normally ring the signal symmetrically around 0 V, the fully bilateral offset cancel
topology affords excellent noise rejection. The switched current output of the comparator drives the impulse noise filter
integrator capacitor that rejects all data pulses of less than
the integrator charge time. Noise appears as duty-cycle jitter
at the open collector serial data output.
Top View
TL/H/6750 – 2
Order Number LM1893N
See NS Package Number N18A
Small Outline & Dual-In-Line Package
Top View
TL/H/6750 – 41
Order Number LM2893M or LM2893N
See NS Package Number M20B or N20A
FIGURE 2. Connection Diagrams
TL/H/6750 – 3
FIGURE 3. The block diagram of a carrier-current
system using the Bi-Line chip to interface digital
controllers via the power line
*Unless otherwise noted, all pin references refer to LM1893, but hold true
for equivalent LM2893 pin.
FIGURE 4. Block diagram of a CCT system with the boost and 5V supply options shown in dashed boxes
TL/H/6750 – 4
Application Information (Continued)
Application Information (Continued)
Effect of making the component value:
CO 560 pF
RO 6.2 kX
g 5% NPO ceramic. Use low TC
Together, CO and RO Increases FO
Decreases FO
set ICO FO.
Increases FO
Decreases FO
2 k pot and 5.6 k fixed R.
k 5.6 k not recommended. l 7.6 k not recommended. Poor FO TC with k 5.6 k RO.
CF 0.047 mF
PLL loop filter pole
RF 3.3 kX
PLL loop filter zero
CC 0.22 mF
CQ 0.033 mF
Less noise immune, higher
fDATA, more PLL stability.
PLL less stable, allows
less CF. Less ringing.
More noise immune, lower
fDATA, less PLL stability.
PLL more stable, allows
more CF. More ringing.
Depending on RF value and
FO, PLL unstable with large
CF. See Apps. Info. CF
and RF values not critical.
Couples FO to line,
CC and T1 low-pass
attenuates 60 Hz.
Low TX line amplitude.
Less 60 Hz T1 current.
Less stored charge.
Drives lower line Z.
More 60 Hz T1 current.
More stored charge.
t 250 V non-polar. Use 2CC
on hot and neutral for max.
line isolation, safety.
T1 Use
Tank matches line Z,
bandpass filters,
isolates from line,
and attenuates
Tank FO up or increase
L of T1 for constant FO.
Smaller L: higher FO or
increase CC; decreased FO
line pull.
Tank FO down or decrease
L of T1 for constant FO.
Larger L: lower FO or
decrease CC; increased FO
line pull.
100 V nonpolar, low TC, g 10%
High large-signal Q needed.
Optimize for low FO line
pull with control of FO TC
and Q.
CA 0.1 mF
RA 10 kX
ALC pole
ALC zero
Noise spikes turn ALC off.
Less stable ALC.
Slower ALC response.
More stable ALC.
RA optional. ALC stable
for CAt100 pF.
CL 0.047 mF
Limiter 50 kHz pole,
60 Hz rejection.
Higher pole F, more 60 Hz
reject. FO attenuation?
Lower pole F, less 60 Hz
reject, more noise BW.
Any reasonably low TC cap.
300 pF guarantees stability.
CM 0.47 mF
Holds RX path VOS
Less noise immune, shorter More noise immune, longer Low leakage g 20% cap.
VOS hold, faster VOS aqui- VOS hold, slower VOS aqui- Scale with fDATA.
sition, shorter preamble.
sition, longer preamble.
CI 0.047 mF
Rejects short pulses Less impulse reject, less
like impulse noise.
delay, more pulse jitter.
More impulse reject, more CI charge time (/2 bit nom.
delay, less pulse jitter.
Must be k1 bit worst-case.
RC 10 kX
Open-col. pull-up
Less available sink I.
Less available source I.
RCt1.5 kX on 5.6 V
RZ 12 kX
5.6 V Zener bias
Larger shunt current,
more chip dissipation.
Smaller shunt current,
less V a current draw.
1kIZk30 mA recommended.
(Chip power-up needs 5.6 V)
ZT t44 V BV
k 60 V peak
Transient clamp
Transient I limit
Over-drive Clamp
ZT costly, lower series
R gives enhanced
transient clamp,
more ruggedness.
Excessive TX attenuation.
Recommend Zener rated
for t500 W for 1 ms.
RT 4.7 X
DT t44V BV
ZT failure, higher series
R-excess peak V, Zener
and chip damage,
less ruggedness.
Damage ZT, pull up V a .
Failure on Transient
RB 180 X
QB Power NPN
RG 1.1 X
Base bleed
Boost gain device
Current setting R
Faster, lower THD IO.
Excessive TJ and VSAT.
More IO, need higher hfe.
Inadequate turn-off speed. Boost optional. QB F(b3 dB)
More rugged, but costly.
of l200 MHz. RB l 24 Ohm.
Less IO, lower min. hfe.
IO e 70[(10 a RG)/RG] mApp.
CB t47 mF
Supply bypass
Transients destroy chip.
Less supply spike.
V a never over abs. max.
ZA 5.1V
Stop ALC charge
in RX mode
Excess ALC
current flow
ALC RX charging
not inhibited over TJ
ZA optional - 5.1V
g 20% low leakage type
Carbon comp. recommended.
IRF 11DQ05 or 1N5819
FIGURE 5. A quick explanation of the external component function using the circuit of Figure 4 . Values given are for V a e
18 V, FO e 125 kHz, fDATA e 360 Baud (180 Hz), using a 115 V 60 Hz power line
Component Selection
Assuming the circuit of Figure 4 is used with something other than the nominal 125 kHz carrier frequency, 180 Hz data
rate, 18V supply voltage, etcetera, the component values
listed in Figure 5 will need changing. This section will help
direct the CCT designer in finding the required component
values with emphasis placed on look-up tables and charts. It
is assumed that the designer has selected values for carrier
center frequency, FO; data rate, fDATA; supply voltage, V a ;
power line voltage, VL; and power line frequency, FL. If one
or more of those parameters is not defined, one may read
the data sheet and make an educated guess.
Maxims to keep in mind, based on CCT electrical perform-
ance considerations only, are: 1) the higher the FO the better, 2) the lower the maximum data rate the better, and 3)
the more time and frequency filtering the better.
Use Figure 5 as a quick reference to the external component function.
Central to chip operation is the low TC of FO emitter-coupled oscillator. With proper CO, the FO of the 2VBE amplitude triangle-wave oscillator output may vary from near DC
to above 300 kHz. While CO may have any value, CO should
Component Selection (Continued)
be made above 10 pF so that parasitic capacitance is not
dominant. Excessive or unbalanced common-mode-toground capacitance should be avoided. A low temperature
coefficient (TC) of capacitance (k100 PPM/§ C), such as a
monolithic NPO ceramic multilayer type, preserves low TC
of FO. Figure 6 finds a CO value given FO.
At this point, the CCT system designer may choose to use
one of the recommended transformers or to design custom
T1. Consult ‘‘The Coupling Transformer’’ section to help
with the design of T1 if a new or boost-capable transformer
is needed. The recommended 125 kHz transformer functions with an IO of up to 600 mApp.
It is recommended that CCT systems use the recommended
transformers, described in Figure 7 , for T1. The 3 transformers are optimized for use in the ranges of 50 – 100 kHz, 100 –
200 kHz, and 200 – 400 kHz with unloaded Q’s (QU) of about
35, and loaded Q’s (QL) of about 12. Three secondary taps
are supplied with nominal 7.07, 10, and 14.1 turns ratios (N)
to drive industrial and residential power line impedances of
3.5, 7, and 14X respectively. All are inexpensive, all have
the same pin-outs for easy exchange in a PC board, and all
are small - on the order of 10 mm diameter at the base.
Resistor RO is used by the IC to generate a VBE/R related
current that is multiplied by 2 to produce the 200 mA ICO
control current that sets FO. The control current TC ‘‘bucks’’
the VBE related tri-wave amplitude across CO to effect a low
TC of FO. Vary RO to trim FO, within limits. Raising FO more
than 20% above its untrimmed value by means of decreasing RO more than 20% is not recommended. Low RO, and
so high control current, risks ICO saturation and poor TC
under worst-case conditions. Raising RO reduces the demodulated signal amplitude from the phase detector; raising
RO by more than a factor of 2 (1 octave) is not recommended.
Since lower TC pots are relatively costly, it is recommended
that RO be made up of a 5.6 k fixed (k100 PPM/§ C) resistor
with a 2 kX (k250 PPM/§ C) series pot.
Tank resonant frequency FQ must be correct to allow passage of transmitter signal to the line. Use Figure 8 to find
CQ’s value. Trimming FQ to equal FO is done with T1’s trimming slug. The inductance of T1 has a TC of a 150 PPM/§ C
which may be cancelled by using a b150 PPM/§ C cap such
as polystyrene. Since circulating current in the tank is (/4
ARMS, CQ should have a low series resistance (a 1 X series
resistance is too much). Polypropelene caps are excellent,
‘‘orange drop’’ mylars are adequate, while many other mylars are inadequate. A 100V rating is needed for transient
CA and RA
Components CA and RA control the dynamic characteristics
of the transmitter output envelope. Their values are not critical. Use the values given in Figure 5 . CA and RA are functions of loaded T1 tank Q, RO, fDATA, and line impulse
noise. Any changes made in CA and RA should be made
based on empirical measurements of a CCT on the line.
Roughly, CA acts as an ALC pole and RA an ALC zero.
TL/H/6750 – 10
TL/H/6750 – 5
FIGURE 8. Find CO’s value given FO
FIGURE 6. Find CO’s value knowing FO
Bottom View
TL/H/6750 – 7
TL/H/6750 – 8
TL/H/6750 – 9
125 kHz
50 kHz
300 kHz
Toko 707VX-A042YUK
Toko 707VX-A043YUK
Toko 161XN-A207YUK
FIGURE 7. The recommended T1 transformers, available through:
Toko America, 1250 Feehanville Drive, Mount Prospect, IL, 60056, (312) 297-0070
Component Selection (Continued)
neous power of greater than 1 kW has been measured using the recommended transformers). For self protection, the
Carrier I/O has an internal 44V voltage clamp with a 20X
series resistance. A parallel low impedance 44V external
transient suppression diode will then conduct the lion’s
share of any current when transients force the Carrier I/O to
a high voltage.
Capacitor CC’s primary function is to block the power line
voltage from T1’s line-side winding. Also, CC and T1’s lineside winding comprise a LC highpass filter. The self-inductance of T1 is far too low to support a direct line connection.
CC must have a low enough impedance at FO to allow T1 to
drive transmitted energy onto the line. To drive a 14X power
line, the impedance of CC should be below 14X.
Use Figure 9 to find the reactive impedance of CC to check
that it is less than the line impedance. Then check Figure 10
to see that the power line current is small enough to keep
T1 well out of saturation; the recommended transformers
can withstand a 10 Amp-turn magnetizing force (1 Amp
through the worst-case 10 turn line-side winding).
Caution is required when choosing CC to avoid series resonance of the series combination of CC, the transformer inductance, and the reflected tank impedance. The low resistance of the network under series resonance will load the
line, possibly decreasing range. For your particular line coupling circuit, measure for series resonance using some expected line impedance load.
TL/H/6750 – 12
FIGURE 10. The AC line-induced current passed by CC
This base-bleed resistor turns QB off quickly - important
since the amplifier output swing is about 200V/ms. An RB
below about 24X will conduct excessive current and overload the chip amplifier and is not recommended.
TL/H/6750 – 13
FIGURE 11. Output amplifier current and required min.
QB hfe versus gain-setting resistor RG
FIGURE 9. CC’s impedance should be,
as a rule-of-thumb, smaller than the lowest
expected line impedance
This resistor, in parallel with the internal 10X resistor, fixes
the current gain of the output amplifier, and so the output
current amplitude. Figure 11 gives output current and minimum AC current gain hfe for QB when RG is used to boost
output current.
TL/H/6750 – 14
FIGURE 12. Boost transistor power dissipation versus
amplifier output current
ZT must be used unless some precaution is taken to protect
the Carrier I/O pin from line transients or transients caused
when stored line energy in CC is discharged by the random
phase of power line connection and disconnection. Worst
case, CC may discharge a full peak-to-peak line voltage into
the tuned circuit. Another way to reduce the need for ZT is
by placing another magnetic circuit in the signal path that
relies on a high, but easily saturated, permeability to couple
a primary and secondary winding - a toroidal transformer for
example. Toroids cost more than ZT.
Use an avalanche diode designed specifically for transient
suppression Ð they have orders of magnitude higher pulse
The boost gain transistor QB must be fast. Double-diffused
devices with 50 MHz FT’s work, slower transistors (epi-base
types) do not preserve a sinusoidal waveform when FO is
high or will cause the output amp. to oscillate. QB must have
a certain minimum hfe for given boost levels, as shown in
Figure 11 . Figure 12 shows the power QB must dissipate
continuously operating with a shorted output. BVCER (R e
RB) must be 60V or greater and QB must have adequate
SOA for transient survival.
Unfortunately, potentially damaging transient energy passes
through transformer T1 onto the Carrier I/O pin (instanta10
Component Selection (Continued)
differential inputs of the Norton amp. equally, while the single-ended input signal swings only the positive input. Overall
PSRR consists of the input CMRR (set by the input stage
component matching) and the ripple-frequency attenuation
of the input amplifier bandpass response that passes carrier
frequency but stops low frequencies. A typical 1% resistor
and 1 mV n-p-n mirror offsets give 26 dB of attenuation, the
bandpass gives 54 dB 120 Hz attenuation, for an overall 80
dB PSRR to allow tens of volts of ripple before impacting
ultimate sensitivity.
power capability than standard avalanche diodes rated for
equal DC dissipation. Metal oxide varistors have not proven
useful because of their inferior clamping coefficient and are
not recommended. Specifications for an example minimum
diode are given in Figure 13 .
Breakdown Voltage
44–49V @ 1 mA
Maximum Leakage
1mA @ 40V
300 pF @ BV
Maximum Clamp Voltage
64.5V @ 7.8A
Peak Non-Repetitive Pulse Power
10 kW for 1 ms
(REA Standard Exponential Pulse)
Surge Current
70A for 1/120s
A value was chosen earlier. Knowing T1’s secondary inductance allows a check of LC line attenuation using Figure 14 .
FIGURE 13. Key specifications for a recommended
transient suppressor ZT available from General
Semiconductor, 2001 West Tenth Place, Tempe, AZ
85281, 602–968-3101, part no. SA40A
The Norton input limiter amplifier has a bandpass filter for
enhanced receiver selectivity, noise immunity, and line frequency rejection. The nominal response curve for FO e 50
kHz is shown in Figure 15 . The 300 kHz pole is fixed. The 50
kHz pole is set by CL’s value. After CL is found, the resulting
line frequency attenuation is found for the bandpass filter.
Use Figure 15 to find a CL value given for FO. The approximate line frequency attenuation of the bandpass filter may
then be found in Figure 16 . Figure 15 returns a value for CL
33% larger than nominal, giving a low frequency pole 33%
low to allow for component tolerances.
RT acts as a voltage divider with ZT, absorbing transient
energy that attempts to pull the Carrier Input pin above 44V.
Make the resistor a carbon composition 1/4W. When experiments discharging CC charged to the peak-to-peak 620V
AC thru a 1X power line were carried out, film resistors blew
This Schottky diode is placed in parallel with the CCT chip’s
substrate diode to pass the majority of the current drawn
from ground when the Carrier Input or Carrier Output is
pulled below ground by a larger-than-twice-the supply-swing
on the tank. Note that ZT is in parallel with the substrate
diode, but is ineffective due to its high forward voltage drop
and high diffusion capacitance caused by its low forward
speed. Tests proved that a 1N5818 kept a receive-path
functional with a 20X boost transmitter with a 7:1 transformer attempted to swing the receiver’s Carrier I/O to g 100V
(300 mA peak ground current in the receiver). Without DT,
the receiver momentarily stops functioning at a 100 times
lower ground current.
This diode is not needed if the Carrier I/O never swings
below ground. If your CCT systems all run on the same
regulated voltage with all matched transformers and turns
ratios, it is not needed. Otherwise, it is.
TL/H/6750 – 15
FIGURE 14. The 60 Hz line rejection of the highpass
filter made up of CC and T1’s line-side winding
(neglecting capacitive coupling)
The receiver and transmitter share components CC, T1, CQ,
RT, ZT, CO, RO, and peripheral supply and bias components
that are not in need of change for RX mode operation. Values for the balance of the components are now found.
TL/H/6750 – 16
Line-Frequency Rejection
To use the ultimate sensitivity of the device, fully 110 dB of
115 V, 60 Hz attenuation is required between the line and
the limiter amplifier output. Using the circuit topology of Figure 4 , the combined attenuation of the CC/T1 highpass, the
tuned transformer, and the bandpass filter attenuation of
the limiter amplifier give far more line rejection than the
above-stated minimum. However, if some other CCT line
coupling circuit is used, line rejection will become important
to the system designer.
Receiver input power supply rejection (PSRR) and commonmode rejection (CMRR) are one-in-the-same using the supply-referenced signal input of Figure 4 . Ripple swings both
TL/H/6750 – 17
FIGURE 15. Given FO, CL is found. Also shown is the
input amplifier’s small signal amplitude response
Component Selection (Continued)
obvious way out is to then reduce the unfiltered loop bandwidth. That bandwidth is approximately proportional to the
value of CO. For a fixed FO, unfiltered loop bandwidth reduction requires a larger CO and larger control current. With this
chip, changing the control current is not allowed. So one is
forced to choose a CF/RF combination with some minimum
capture range, say g 20%, that is within some guardband
from the point of loop instability. Happily, impulse noise
tends to last only fractions of a millisecond so that the lack
of low bandwidth loop response with low data rates is not a
heavy penalty. As long as there is adequate capture range,
the impulse noise filter performs admirably. Note that reducing FO will reduce the no-filter loop bandwidth, and indeed
the maximum data rate falls below the limit set by the RC
lowpass filter as FO falls below 100 kHz (Figure 19 ).
The tuned transformer characteristics will affect the demodulated data waveform more than CF and RF at low data
rates. Tank Q and off-tuning will affect overshoot during the
FSK frequency steps. This is a property of tuned circuits.
The maximum data rate of Figure 19 is measured from the
receiver input to the Data Out and does not include the data
bandwidth reducing effects of TI.
CF and RF
These phase-locked loop (PLL) loop filter components remove some of the noise and most of the 2FO components
present in the demodulated differential output voltage signal
from the phase detector. They affect the PLL capture range,
loop bandwidth, damping, and capture time. Because the
PLL has an inherent loop pole due to the integrator action of
the ICO (via CO), the loop pole set by CF and the zero set by
RF gives the loop filter a classical 2nd-order response.
FIGURE 16. The Norton-input limiter amplifier bandpass
filter line-frequency signal attenuation given CL
Capacitor CM stores a voltage corresponding to a correction
factor required to cancel the phase detector differential output DC offsets. The stored voltage is ±/6 of the DC offset
plus some bias level of about 2.2 V. A large CM value increases the time required to bias-up the receive path at the
beginning of transmission. A large CM does filter well and
store its bias voltage long. Because of the initial random
charge of CM, the receiver must be given a data transition to
charge to the proper bias voltage. Therefore, reducing CM’s
value to one that may be charged in less than 2 bit-times will
not save biasing time and is not recommended.
FIGURE 17. Find CF given FO. Figure 19
gives the maximum data rate
No CF and RF give the most stable PLL with the fastest
response. Large CF’s with a too-small RF cause PLL loop
instability leading to poor capture range and poor step response or oscillation.
Calculation of CF and RF is quite difficult, involving not only
the 2nd-order loop step response, but also the PLL nondominant poles, the tuned transformer stepped-frequency
response, and the RC lowpass step response (for data rates
approaching 1 kHz). CF and RF values are best found empirically. Tolerance is not critical. Component values are selected to give the best possible impulse noise rejection
while preserving a g 20% capture range and wide stability
margin. Figures 17 and 18 give CF and RF values versus FO,
where ‘‘fDATA kk MAX DATA RATE’’ means that fDATA
should be less than the maximum data rate, in kHz, from
Figure 19 divided by 10.
Note that CF and RF are a function of data rate only for high
data rates and are not plotted against data rate - as one
might expect. The reason for this is important to understand
if the CCT system designer wishes to find CF and RF empirically. Data signal is, loosely speaking, passed through the
PLL loop and is therefore potentially attenuated if the loop
bandwidth is on the order of the 3rd harmonic of the data
rate, or less. Overall loop bandwidth is held as low as possible for maximum noise rejection while passing the data.
Loop bandwidth is roughly proportional to the geometric
mean of the unfiltered loop bandwidth and the filter pole set
by CF. Therefore, CF is related to data rate. Unfortunately,
the loop capture range falls to critically low values when
large enough values of CF are used to reduce loop bandwidth down to the 100’s of Hz range, for low data rates. The
TL/H/6750 – 20
FIGURE 18. Find RF given FO with FDATA a parameter
TL/H/6750 – 21
FIGURE 19. The maximum data rate versus FO using
loop filter components optimized for max. noise
performance while retaining a min. g 20% capture
range (large signal)
Use Figure 20 to find CM’s value knowing fDATA, assuming
the standard 2 bit receive charge time is desired. The cap.
value and TC are not critical, but the capacitor should have
low leakage.
Component Selection (Continued)
The 5.1V silicon zener diode ZA is required when a short
RX-to-TX switch-over time is needed at the same time that
the chip is operating in the RX mode with a pin 10 input
signal swing approaching or exceeding twice the supply
voltage. Predominant causes of these large swings impinging on the RX input are: 1) a transmitter’s supply voltage
higher than the receiver’s supply voltage, 2) a TX and RX
pair that are electrically close, or, 3) a higher RX T1 step-up
turns ratio than the TX T1 step-down ratio.
Normally, when in the RX mode with small incoming signal
on pin 10, the ALC remains off with pin 7 at a 6V
(VZb2VBE) bias voltage. CA is then charged to 6V. TX
mode may then be selected with 6V on CA allowing 100%
TX power to pump T1’s tuned circuit, and so the AC line,
quickly for fast RX-to-TX switch time. As TX output swing
increases so that pin 10 swings below VALC (4.7V typically),
that ALC activates to charge CA to about 6.6V to reduce TX
output drive. However, if in the RX mode pin 10 ever swings
below VALC, CA will charge to above 6.6V. Now, when the
TX mode is selected with CA at 6.6V, somewhere from 0 to
100% TX output drive is available to pump T1’s tuned circuit
resulting in a slower rising line signal - effectively reducing
the RX-to-TX switch time.
Use a 5.1V ZA driven by a 0 to 0.8V logic low signal to
guarantee over-temp. operation. RA must be in series with
ZA to limit current flow and should never fall below 1 kX. If
RA is less than 1 kX, then put a 2 kX resistor in series with
ZA. Logic high voltages above 10V will cause current flow
into pin 7 that must be limited to 1 mA (with RA or a
series R).
TL/H/6750 – 22
FIGURE 20. Size CM assuming a 2 bit-time
receive bias time
The impulse noise filter integrator capacitor CI is used to
disallow the passage of any pulse shorter than the integrator charge time. That charge time, set to a nominal (/2 bit
time, is the time required for a g 50 mA charge current to
swing CI over a 2 VBE range. Charge time under worst case
conditions must never be greater than a bit time since no
signal could then pass. Using a g 10% capacitor, full junction temperature range, and full specified current range, a
maximum nominal charge time of (/2 bit is recommended.
Figure 21 gives CI versus data rate under those conditions.
The collector pull-up resistor is sized to supply adequate
pull-up current drive and speed while preserving adequate
output low current drive.
Breadboarding Tips
During CCT system evaluation, some techniques listed below will simplify certain measurements.
Ð Use caution when working on this circuit - dangerous
line voltages may be present.
Ð When evaluating PLL operation, offset cancel circuit operation, and loop filter values, use the filter of Figure 22
to view the demodulated signal minus the 2FO and noise
components. This filter models the RC lowpass filter on
TL/H/6750 – 24
FIGURE 21. Impulse noise filter cap. CI versus FDATA
where the charge time is (/2 bit time
TL/H/6750 – 25
FIGURE 22. Circuit to view the differential demodulated data signal, minus the noise and 2FO components,
conveniently with a single-ended gain-of-one output
Breadboarding Tips (Continued)
representing an average line impedance may be connected
to the line side of T1. The circuit of Figure 23 should then be
used to defeat the leveling effect of the ALC.
Ð When evaluating CCT system noise performance on a
real power line, it is desirable to vary the signal amplitude to the receiver. This is not easy. An in-line lineproof L-pad is fine except that the line impedance is unknown and variable and so the L-pad will rarely match.
Instead, the power output of a chip transmitter may be
controlled using the circuit of Figure 23 . This circuit controls the ALC.
Ð It is sometimes desirable to place impulse noise on the
line. A simple light dimmer with a 100 W light bulb load
produces representative impulse noise.
Ð Do not allow peak currents of over 1 A through the 5.6 V
Zener. In other words, don’t short charged capacitors
into this low-impedance device. Take care not to momentarily short pins 10 and 11 - chip damage may result.
Ð Figure 24 shows some typical signals beginning with serial data transmitted to received signal.
TL/H/6750 – 26
FIGURE 23. A means of transmitter output amplitude
control is shown
Thermal Considerations
It is desirable to place the largest possible signal on the
power line for maximum range, limited only by the chip power dissipation and maximum junction temperature TJ. The
falling output power at elevated TJ allows a more optimal
power output - high power at low TJ and lower power at high
TJ for chip self-protection. However, it is still possible to
exceed the maximum TJ within the specified ambient temperature limit (TA e 85§ C) under worst case conditions of
100% TX duty cyle, high supply, shorted load, poor PC
board layout (with small copper foil area), and an above
nominal current part. Under those conditions, a part may
dissipate 2140 mW, reaching a TJ e 170§ C worst-case (admittedly a rare occurrence). Proper system design includes
the measurement or calculation of TJ max. to guarantee
function under worst-case operation. Like all devices with
failure modes modeled by the Arrhenius model, the high
chip reliability is further enhanced by keeping the die temperature mercifully below the absolute maximum rating.
A direct method of measuring operating junction temperature is to measure the VBE voltage on pin 18, which is always available under all operating modes. The graph of Figure 25 may be used to find TJ, knowing VBE at the operating
point in question and VBE at TA e TJ e 25§ C. VBE is found
by powering up a chip (in RX mode) that has been dissipating zero power at some TA for some time and measuring
VBE in less than 1 s (for better than 5§ C accuracy).
Alternately, TJ may be calculated using:
where iJA is 75§ C/W for the plastic (N) package using a
socket. That iJA value is for a high confidence level; nomi-
Tuning Procedure
This procedure applies to circuits similar to Figure 4 LM1893
or LM2893 circuit.
First, trim FO by putting the chip in the TX mode, setting a
logical high data input, and measuring the TX high frequency, 1.022 FO, on the Carrier I/O using these steps:
1. Take pin 17 to a logic low.
2. Take pin 5 to a logic high.
3. Place a counter on pin 10.
4. Adjust RO on pin 18 for F e 1.022FO.
Second, the line transformer is tuned. The chip is placed in
the TX mode, a resistive line load is connected to disable
the ALC by reducing tank voltage swing below its limit. FSK
data is then passed through the tank so that the tank envelope may be adjusted for equal amplitude for high and low
data frequency.
1. Take pin 5 to a logic high.
2. Place a logic-level square wave at or below the receiver’s
maximum data rate on pin 17.
3. Temporarily place a 330 X resistor across the tank.
4. Place a scope on pin 10.
5. Adjust the transformer slug for the least envelope modulation.
In lieu of the 330 X resistive load, T1 may be coupled to the
power line to better simulate actual load and tank pull conditions during tank tuning. Alternatively, a passive network
TL/H/6750 – 23
FIGURE 24. Oscillogram revealing signals at several important nodes under weak signal (0.5 mVRMS) conditions with
SCR spikes on an otherwise quiet 115 V, 60 Hz power line. The signals are: 1) transmitted data, 2) RX carrier on the
tuned transformer, 3) demodulated signal from the PLL after passing thru circuit of Figure 22 , 4) signal after RC
lowpass, 5) data at impulse noise filter integrator, and 6) received data. Horizontal scale is 10 ms per div.
mended CF and RF (47 nF and 6.2 kX) with a g 4.4% DFO
(a g 100 mV DC offset on CF and RF), lock was measured
to take less than 50 cycles of FO. That is a 0.40 ms delay
(proportional to 1/FO).
Acquisition is incomplete until the second order PLL loop
settles. For the above-mentioned CF and RF, the loop natural frequency FN and damping factor are found to be
2.3 kHz and 1.0 respectively. Settling to within g 25 mV of
the g 100 mV DC offset change requires 2.7 periods of FN,
or 1.2 ms (a function of CF and RF).
Third, the RC lowpass filter introduces a 0.12 ms delay.
Fourth, CM must charge up to g (±/6)100 e 83 mV depending on the polarity of FO. Borderline data squaring with zero
noise immunity is possible with only g (±/6) 50 mV of charging. CM charge current is an asymptotic function approximated by assuming a 50 mA charge current and the full 83
mV charge voltage. CM charge time is then 1.7 ms (proportional to 1/fDATA).
Fifth, the impulse noise filter adds a (/2 bit-time delay. Total
TTR is 3.9 ms plus (/2 bit-time for a total of 1.9 bit-times at
360 Baud.
Thermal Considerations (Continued)
nal iJA for an N package is 60§ C/W, lower with good PC
board layout. Since PD is a relatively strong function of TJ,
an iterative solution process starting with an initial guess for
TJ is used. With the estimated TJ, find the total supply current found in the typical performance characteristics.
TL/H/6750 – 27
FIGURE 25. TJ may be found by using the temperature
coefficient of pin 18 VBE if VBE is known at 25§ C
Switch-Over Time
Switch-Over Time
An important figure-of-merit for a half-duplex CCT link, affecting effective data rate, is the TX-to-RX switch time TTR.
Using the recommended component values gives this part a
nominal 2 bit-time (1 bit time e 1/[2fDATA]) over a wide
range of operating conditions, where the receiver requires 1
data transition. TTR cannot be decreased significantly but
does increase as noise filtering, especially via CM, is increased. Impulse noise at switch, signals near the limiting
sensitivity, poor FO match between receiver and transmitter
because of poor trim or worst-case conditions, and the statistical nature of PLL signal acquisition may all contribute to
increase TTR to possibly 4 bit-times.
TTR is lower when a pair of LM1893’s handshake rapidly.
The receiver was designed to ‘‘remember’’ the RX-mode
DC operating points on CM and CF while in the TX mode.
Under noisy worst case conditions, CM will discharge to the
point of false operation after 35 bit-times in the TX mode
(1400 bit times with no noise and a nominal part, fDATA e
180 Hz). TTR is about 0.8 ms (proportional to the selected
FO) plus (/2 bit-time.
The major components of TTR are described below for a
nominal 125 kHz FO, 180 Hz fDATA, lightly-loaded tank with
a Q of 20, and the circuit of Figure 4 . The remote CCT has
been operating in the TX mode with a 26.6 VPP tank swing
and is now selected as a receiver. An incoming signal requiring the ultimate receiver sensitivity immediately is placed
on the line.
First, the tank stored energy at the transmit frequency must
decay to a level below the 2.8 mVPP swing caused by the
0.14 mVRMS incoming line signal containing the information
to be received.
#V J
125 000
# 0.0028 J
decay time e
Assume the chip has been in the RX mode and the TX
mode is now selected. In less than 10 ms, full output current
is exponentially building tank swing. 50% of full swing is
achieved in less than 10 cycles - or under 80 ms at 125 kHz.
In the same 10 ms that the output amp went on, the phase
detector and loop filter are disconnected and the modulator
input is enabled. FSK modulation is produced in 10 ms after
switching to TX mode.
Power Line Impedance
Irrespective of how wide the limits on power line impedance
ZL are placed, there are no guarantees. However, since the
CCT design requires an estimate of the lowest expected line
impedance ZLN encountered for the most efficient transmitter-to-line coupling, line impedance should be measured
and ZL limits fixed to a given confidence level. Reasonable
values for T1 turns ratio, loaded Q, and tank resonant frequency pull FQ may be found to enable a CCT system design that functions with the overwhelming majority of power
A limited sampling of ZL was made, during the LM1893 design, of residential and commercial 115V 60 Hz power line.
Data was also drawn from the research of Nicholson and
Malack (reference 1), among others, to produce Figures 26
and 27 . All measured impedances are contained within the
shaded portions of Figure 27 . A nominal 3.5, 7.0 and 14 X
ZLN is used throughout the application information with a
nominal 45§ phase angle (0§ is sometimes used for simplicity).
e 0.466 ms
That is 0.47 ms of delay (proportional to I/FO and Q).
Second, the PLL must acquire the signal; it must lock and
settle. Acquisition time is statistical and may take any length
of time, but average acquisition time depends on the loop
filter components CF and RF and the difference in center
frequencies, DFO, of the TX/RX pair. Using the recom-
TL/H/6750 – 28
FIGURE 26. Measured line impedance range for
residential and commercial 115V, 60 Hz lines
Power Line Impedance (Continued)
TL/H/6750 – 30
TL/H/6750 – 31
FIGURE 27. Complex-plane plots of measured 115V, 60 Hz line impedance where ZL e RL a jXL
Power Line Attenuation
The wiring in most US buildings is a flat 3 conductor cable
called Amerflex, BX, or Romex. All referenced line impedances refer to hot-to-neutral impedances with a grounded
center conductor. The cable has a 100 X characteristic impedance, a 125 kHz quarter-wavelength of 600 m (250 m at
300 kHz), and a measured 7 dB attenuation for a 50 m run
with a 10 X termination. Generally, line loads may be treated as lumped impedances. Instrument line cords exhibit
about 0.7 mH and 30 pF per meter.
Limited tests of CCT link range using this chip show extensive coverage while remaining on one phase of a distribution transformer (100’s of m), with link failure often occuring
across transformer phases or through transformers unless
coupling networks are utilized. Total line attenuation allowed
from full signal to limiting sensitivity is more than 70 dB.
Typically, signal is coupled across transformer phases by
parasitic winding capacitance, typically giving 40 dB attenuation between phased 115 V windings. Coupling capacitors
may be installed for improved link operation across phases.
Power factor correcting capacitor banks on industrial lines
or filter capacitors across the power lines of some electronic
gear short carrier signal and should be isolated with inductors. Increasing range is sometimes accomplished by electing to install the isolating inductors (Figure 28 ) and coupling
capacitors, as well as by electing to use the boost option.
Frequency translating or time division multiplexed repeaters
will also increase range.
T1 with a stable resonant frequency, FQ, that is little affected by the de-tuning effect of the line impedance ZL, and of
2) building a tightly line-coupled transformer for transmitted
carrier with loose coupling for transients, are somewhat mutually exclusive. The tradeoffs are exposed in the following
example for the CCT designer attempting a new boost-capable, or different core, transformer design.
The compromises are eased by separating the TX output
and RX input in the LM2893. An untuned TX coupling transformer with only core coupling (not air-coupled solenoid
windings) would employ a high permeability, high magnetic
field, low loss, square saturating, toroidal core. The resonant RX path would be isolated from line-pull problems by a
unilateral amplifier that operates at line voltages with much
more than 110 dB of dynamic range, or by a capacitively
coupled pulse transformer driving a unilateral amplifier and
filter, for increased selectivity. See the LM2893-specific applications section.
For a LM1893-style transformer application, first, choose
the turns ratio N based on an estimated lowest ZL likely
encountered, ZLN. Figure 29 shows graphically how N affects line signal. N should be as large as possible to drive
ZLN with full signal. If T1 has an unloaded Q, QU, of well less
than 35, a guess of N somewhat high should be used and
later checked for accuracy. The recommended transformers
have secondary taps giving a choice of N e 7.07, 10, and
14.1 (nominally) for driving ZLN’s of 14, 7.0, and 3.5 X respectively (at TJ e 25§ C, V a e 18V, and QU e 35).
The resonating inductance of the tuned primary, L1, is
sought. Note that, while standard transformer design gives a
transformer self-inductance with an impedance at operating
frequency well above load impedance, the tuned transformer requires a low L1 for adequate QU and minimum line pull.
Result: relatively poor mutual coupling.
It is known that resonant frequency FQ e FO and some
minimum bandwidth, or maximum Q, will be required to pass
signal under full load conditions.
L1 e
FIGURE 28. An isolation network to prevent: 1) noise
from some device from polluting the AC line, and 2) to
stop some low impedance device (measured at Fo)
from shorting carrier signal. Component values given
as an example for Fo e 125 kHz on
residential power lines
RQ ll lZLNlÊ
lZLNlÊ is the reflected ZLN, QL is the loaded Q, and parallel
resistance RQ models all transformer losses and sets QO.
RQ ll lZLNlÊ is found knowing that it absorbs full rated power.
L1 e
The Coupling Transformer
The design arrived at for T1 is the result of an unhappy
compromise - but a workable one. The goals of 1) building
The Coupling Transformer (Continued)
Line pull DFQ was calculated (reference 3) for a ZL magnitude of 14X and up with any phase angle from b90§ to 90§ .
DFQ was 6.4% - well above the 3.3% estimate. Referring to
(11), an 11.8% bandwidth is required, forcing L1 to be reduced to reduce Q. That fix was not implemented; some
signal attenuation under worst-case drift and DFQ is allowed. L1 is already so small that the 31 gauge winding
conducts a (/4 ARMS circulating current.
Line Carrier Detection
While the addition of a carrier detection circuit (for a mute or
squelch function) will only decrease receiver ultimate sensitivity, there is sometimes good reason to employ it to free
the controller from watching for RX signal when no carrier is
incoming, or to employ it to reduce the probability of line
collisions (when multiple transmitters operate simultaneously to cause one or more transmissions to fail). Unless the
detector is heavily filtered or uses a high carrier amplitude
threshold, there will be false outputs that force the controller
to have Data Out data checking capability just as is required
when using no carrier detector. If false triggering is minimized, the probability of line collisions is increased due to
the inability to sense low carrier amplitudes and because of
sense delay. The property of the LM1893 to change output
state infrequently (although the polarity is undefined) when
in the RX mode, with no incoming carrier, reduces the desire
to implement carrier detection and preserves the full ultimate sensitivity. Also, many impulse-noise insensitive transmission schemes, like handshaking, are easily modified to
recover from line collisions.
Regarding this, it should be stated that for very complicated
industrial systems with long signal runs and high line noise
levels, it is probably wise to use a protocol which is inherently collision free so that no carrier detect hardware or software is needed. A token passing protocol is an example of
such a system.
TL/H/6750 – 32
FIGURE 29. Impressed line voltage for a given ZL
for each of the 3 taps available
on the recommended transformers
IOPP 2(bVALC a V a )
(b4.7 a V a )IO
2 02
2 02
where IO is in amps peak-to-peak at an elevated TJ
(18 b 4.7) 0.06
e 0.200 W
PO e
(bVALC a V a )02
e 442 X
RQ ll lZLNlÊ e
RQ is found using ZLN and the value for N found when assuming QU e 35.
lZLNlÊ e N2 ZLN e (7.07)2 13.9 e 695 X
e 1210 X
RQ e
RQ ll lZLNlÊ
lZLNlÊ 442 695
1 a QU2
1 a 352
Only QL remains to be found to calculate L1. QL is related to
the b3 dB (half-power) bandwidth by
QL e
BW (% of FO)
An iterative solution is forced where line pull, DFQ, must be
guessed to find QL and L1. L1 is then used to check the line
pull guess; a large error requires a new guess. Try a BW of
8.7% - that is 4.4% for deviation, 1% for TC of FO, and
3.3% for DFQ - giving QL e 11.5.
e 49.0 mH
L1 e
2q c 125 000 c 11.5
Knowing the core inductance per turn, L, and L1, the number of turns is found.
T1 e
0 L 020 nH/T
49.0 mH
e 49 (/2 turns
Figure 30 shows a low cost carrier amplitude detection circuit.
Audio Transmission
The LM1893 is designed to allow analog data transmission
and reception. Base-band audio-bandwidth signals FM
modulate the carrier passing through the tuned transformer
(placing a limit on the usable percent modulation) onto the
power line to be linearly demodulated by the receiver PLL.
Because the receiver data path beyond the phase detector
will pass only digital signal, external audio filtering and amplification is required. Figure 31 shows a simple audio transmitter and receiver circuit utilizing a carrier detection mute
circuit. A single LM339 quad. comparator may be used to
build the carrier detect and mute. Filter bandwidth is held to
a minimum to minimize noise, especially line-related correlated noise.
T is normally an integer, but these transformers require so
few turns that half-turns are specified, remembering that the
remaining (/2 turn is completed on the P.C. board and is
loosely coupled. The secondary turns are calculated
Communication and System
e 7.00 e 7 turns
T2 e
giving an L2 of 0.98 mH. Note that the recommended 125
kHz transformer mirrors these specifications. The resonating capacitor is
CQ e
e 33.1 c 10 b 9 e 33 nF
(2qFQ)2 L1
The development of communication and system protocols
has historically been the single most time consuming element in design of carrier current systems. The protocols are
defined as the following:
1. Communication protocol : a software method of encoding
and decoding data that remains constant for every transmis-
TL/H/6750 – 33
FIGURE 30. A simple carrier amplitude detector with output low when carrier is detected
TL/H/6750 – 34
FIGURE 31. A simple linear analog audio transmitter and receiver are shown.
The carrier and 1.6V inputs are derived from the carrier detector of Figure 30 .
The remaining 2 LM339 comparators may be used to build the carrier detector circuit.
Communication and System
Protocols (Continued)
ensure message retransmission to correct errors (handshake). Secondly it coordinates messages for maximum utilization and efficiency on the network. Lastly, it ensures that
messages do not collide on the network. Common system
protocols include master-slave, carrier detect multiple access, and token passing. Token passing and master slave
have been found to be the most useful since they are inherently collision free.
Both protocols usually reside as software in a single microcontroller that is connected to the LM1893/2893 I/O. In any
case, some sort of intelligence is needed to process incoming and outgoing messages. UARTs have no usefulness in
sion in a system. Its first purpose is to put data in a baseband digital form that is more easily recognized as a real
message at the receive end. Secondly, it incorporates encoding techniques to ensure that noise induced errors do
not easily occur; and when they do, they can always be
detected. Lastly, the software algorithms that are used on
the receive end to decode incoming data prevent the reception of noise induced ‘‘phantom’’ messages, and insure the
recovery of real messages from an incoming bit stream that
has been altered by noise.
2. System protocol : the manner in which messages are coordinated between nodes in a system. Its first purpose is to
Communication and System
Protocols (Continued)
transmission, using a random number of bits delay or a delay based on each transmitter’s address, since each transceiver has a unique address.
carrier current applications since they do not have the intelligence needed to distinguish between real messages and
noise induced phantoms.
The difficulty in designing special protocols arises out of the
special nature of the AC line, an environment laden with the
worst imaginable noise conditions. The relatively low data
rates possible over the AC line (typically less than 9600
baud) make it even more imperative that systems utilize the
most sophisticated means available to ensure network efficiency.
With these facts in mind, the designer is referred to a publication intended to aid in the development of carrier current
systems. This is literature Ý570075 The Bi-Line Carrier Current Networking System, a 200 pp. book that functions as
the ‘‘bible’’ of Bi-Line system design. It has sections on
LM1893 circuit optimization, protocol design, evaluation kit
usage, critical component selection, and the Datachecker/
DTS case study.
An example of a simple transmission data packet is shown
in Figure 32 . The 8 bit 50% duty-cycle preamble is long
enough to allow receiver biasing with enough bits left over
to allow the receiver controller to detect the square-wave
that signals the start of a transmission. If there had been no
transmission for some time, the receiver would simply need
to note that a data transition had occurred and begin its
watch for a square-wave. If the receive controller detected
the alternating-polarity data square-wave it would then use
the sync. bit to signal that the address and data were immediately following. The address data would then be loaded,
assuming the fixed format, and tested against its own. If the
address was correct, the receiver would then load and store
the data. If the address was not correct, either the transmission was not meant for this receiver or noise has fooled the
receiver. In the former case, when the transmission was not
meant for the receiver, the controller should immediately
return to watching the incoming data for its address. If the
later case were true, then the receive controller would continue to detect edges, tieing itself up by loading false data
and being forced to handshake. The square-wave detection
and address load and check routines should be fast to minimize the time spent in loops after being false-triggered by
noise. If the controller detects an error (a received data bit
that does not conform to the pre-defined encoding format) it
should immediately resume watching the LM1893’s Data
Out for transmissions, the next bit would be shifted in and
the process repeated.
A line-synchronous CCT system passing 3 bits per half-cycle may replace the long 8 bit preamble and sync pulse with
a 2 bit start-of-transmission bias preamble. The receive controller might then assume that preamble always starts after
bit 1 (the first bit after zero-crossing) so that any data transition at a zero crossing must be the start of the address bits
and is tested as such. The line synchronous receiver operates with a simpler controller than an asynchronous system.
Discussion has assumed that the controller has always
known when the Data Out is high or low. The controller
must sample at the proper time to check the Data Out state.
Since noise shows itself as pulse width jitter, symmetrically
placed about the no-noise switch-points, optimum Data Out
sampling is done in the center of the received data pulse.
The receive data path has a time delay that, at low data
rates, is dominated by the impulse noise filter integrator and
is nominally (/2 bit. At a 2 kHz data rate, an additional delay
of approximately (/10 bit is added because of the cumulative
delay of the remainder of the receiver. Figure 33 shows that
Data Out sampling occurs conveniently at the transmitted
Basic Data Encoding (please refer to the previously mentioned publications for advanced techniques)
At the beginning of a received transmission, the first 0 to 2
bits may be lost while the chip’s receiver settles to the DC
bias point required for the given transmitter/receiver pair
carrier frequency offset. With proper data encoding,
dropped start bits can be tolerated and correct communication can take place. One simple data encoding scheme is
now discussed.
Generally, a CCT system consists of many transceivers that
normally listen to the line at all times (or during predetermined time windows), waiting for a transmission that directs
one or more of the receivers to operate. If any receiver finds
its address in the transmitted data packet, further action
such as handshaking with the transmitter is initiated. The
receiver might tell the transmitter, via retransmission, that it
received this data, waiting for acknowledgement before acting on the received command. Error detecting and correcting codes may be employed throughout. The transmitter
must have the capability to retransmit after a time if no response from the receiver is heard - under the assumption
that the receiver didn’t detect its address because of noise,
or that the response was missed because of noise or a line
collision. (A line collision happens when more than 1 transmitter operates at one time - causing one or more of the
communications to fail). After many re-transmissions the
transmitter might choose to give up. Collision recovery is
achieved by waiting some variable amount of time before re-
TL/H/6750 – 35
FIGURE 32. A simple encoded data packet, generated by the transmit controller is shown.
The horizontal axis is time where 1 bit time is 1/(2fDATA)
Because the coupling transformer is used as a filter, the
LM1893 circuit is susceptible to pulling of the center frequency under conditions of changing line impedances or
when several LM1893 circuits are close in proximity on the
AC line. Because the tuned transformer has a high value of
‘‘Q’’, ringing also occurs in the presence of impulsive noise.
This ringing occurs at the center frequency and increases
the error rate of transmissions, especially at relatively high
data rates (l2000 baud). Because it is the only tuned circuit
in the system, the selectivity characteristics leave a lot to be
The LM2893, having separate receive input and transmit
output pins, removes the limitations on coupling transformer
design, allowing the design of circuits devoid of the previous
The first enhancement that can be made with the LM2893
circuit is the use of a high permeability ferrite toroid for line
coupling along with a separate filter. The transformer would
be of broadband design (untuned) with two secondaries,
one for coupling to the transmit output and one for coupling
to the receive input. This allows impedance matching of
both the transmitter and receiver, with the result of quite a
bit more receive sensitivity.
Because of the increased signal and separate receive signal
path, a 3 or 6 db pad can be used before the selective
stages to eliminate pulling of the center frequency due to
changes in line impedance.
Another advantage of the toroidal transformer is that it can
be designed for use at very low line impedances due to its
inherent tight coupling.
Basic Data Encoding (Continued)
FIGURE 33. Operating waveforms of a linesynchronized transceiver pair are shown. The diagram
shows how the transmitted data transitions may be
used as received data sampling points
data edges for the line synchronous data transmission
scheme mentioned in the previous paragraph. With the
asynchronous system suggested, the receive controller
must sample the Data Out pin often to determine, with several bits of accuracy, where the square-wave data transitions take place, average their positions assuming a
known data rate, and calculate where the center of the data
bits are and will continue to be as the address and data are
read. A long preamble is helpful. Software that continuously
updates the center-of-bit time estimate, as address and
data are received, works even better. Alternatively, a coding
scheme employing an embedded clock can be used.
Because of the separate receive path of the LM2893, a relatively high quality bandpass filter can be used for selectivity.
Inexpensive ceramic filters are available that have bandpass and center frequency characteristics compatible with
carrier current operation. Futhermore, the use of these filters allows multichannel operation, previously made difficult
by the single tuned network of the LM1893. These filters are
easily cascaded for even more off-frequency rejection. If the
pad is added before the filter, there will be negligible pulling
due to changes in line impedance reflected through the coupling transformer.
Alternatively, a Butterworth/Chebyshev bandpass LC filter
or an active filter can be used in place of the ceramic filter.
LM2893 Application Hints
The LM2893 is intended for advanced applications where
special circuitry is used in the transmit and receive paths.
The LM2893 makes this possible by featuring separate
transmit output and receive input pins.
Examples of enhancements that can be added to the basic
LM1893/2893 circuit include separate transmit and receive
windings on the coupling transformer, high quality ceramic
or LC filters in the receive path, and simple impulse noise
blanking circuits.
In many applications, the additional performance to be
gained outweighs the extra cost of the additional circuitry.
More than likely, high performance industrial applications
such as building energy management will fit into this category, since they require the utmost in reliability.
Because of the specialized nature of individual LM2893 applications, it is not possible to give one circuit that will satisfy
all requirements for performance and cost effectiveness.
Therefore no specific application examples will be given.
Instead the subsequent text describes in general terms the
types of circuits that can be used to increase performance
along with their advantages and disadvantages. It is intended to be a springboard for ideas.
Although the LM2893 has adequate impulse noise rejection
for most applications, there is reason to employ impulse
blanking to improve error rates in severe AC line environments. Typically, errors occur due to pulse jitter in the
LM1893/2893 data output that originates when the internal
time domain filter smooths out an incoming noise pulse.
The solution involves removing the impulse completely and
not simply trying to filter it. Moreover, the pulse should be
removed in the receive signal path before the selective portions of the circuit to eliminate ringing. This also allows the
receiver filter to smooth out the blanks that also occur in the
desired incoming carrier signal.
If a carrier detect circuit is desired in conjunction with the
LM2893 it can be located after the filter and impulse blanker. Because impulse noise is removed, the false triggering
that plagues these circuits will be greatly reduced.
The main disadvantages of the typical LM1893 coupling
network are that it functions as the bandpass filter, has
loose coupling between primary and secondary, and has a
single secondary. The LM1893 coupling network was designed this way mainly because of the restraint that the carrier input and output are tied together.
TL/H/6750 – 37
Simplified Schematic
4. FCC, ‘‘Notice of Proposed Rule Making,’’ Docket 20780,
adopted Apr. 14, 1976, (Proposed regulation)
1. Nicholson, J.R. and J.A. Malack; ‘‘RF Impedance of Power Lines and Line Impedance Stabilization Network in
Conducted Interference Measurements;’’ IEEE Transactions on Electromagnetic Compatibility; May 1973; (line
impedance data)
2. Southwick, R.A.; ‘‘Impedance Characteristics of SinglePhase Power Lines;’’ Conference Rec.; 1973 IEEE Int.
Symp. on Electromagnetic Compatibility; (line impedance
3. Hayt, William H. Jr. and Jack E. Kemmerly; ‘‘Engineering
Circuit Analysis;’’ McGraw-Hill Books; 1971; pp. 447–
453; (linear transformer reflected impedance)
5. Monticelli, Dennis M. and Michael E. Wright; ‘‘A Carrier
Current Transceiver IC for Data Transmission Over the
AC Power Lines;’’ IEEE J. Solid-State Circuits; vol. SC-17;
Dec. 1982; pp. 1158-1165; (LM1893 circuit description)
6. Lee, Mitchell; ‘‘A New Carrier Current Transceiver IC;’’
IEEE Trans. on Consumer Electronics; vol. CE-28; Aug.
1982; pp. 409 – 414; (Application of LM1893)
Physical Dimensions inches (millimeters)
Molded Small Outline Package (M)
Order Part LM2893M
NS Package Number M20B
Molded Dual-In-Line Package (N)
Order Part LM1893N
NS Package Number N18A
LM1893/LM2893 Carrier-Current Transceiver
Physical Dimensions inches (millimeters) (Continued)
Lit. Ý 107664
Molded Dual-In-Line Package (N)
Order Part LM2893N
NS Package Number N20A
1. Life support devices or systems are devices or
systems which, (a) are intended for surgical implant
into the body, or (b) support or sustain life, and whose
failure to perform, when properly used in accordance
with instructions for use provided in the labeling, can
be reasonably expected to result in a significant injury
to the user.
National Semiconductor
1111 West Bardin Road
Arlington, TX 76017
Tel: 1(800) 272-9959
Fax: 1(800) 737-7018
2. A critical component is any component of a life
support device or system whose failure to perform can
be reasonably expected to cause the failure of the life
support device or system, or to affect its safety or
National Semiconductor
Fax: (a49) 0-180-530 85 86
Email: cnjwge @ tevm2.nsc.com
Deutsch Tel: (a49) 0-180-530 85 85
English Tel: (a49) 0-180-532 78 32
Fran3ais Tel: (a49) 0-180-532 93 58
Italiano Tel: (a49) 0-180-534 16 80
National Semiconductor
Hong Kong Ltd.
13th Floor, Straight Block,
Ocean Centre, 5 Canton Rd.
Tsimshatsui, Kowloon
Hong Kong
Tel: (852) 2737-1600
Fax: (852) 2736-9960
National Semiconductor
Japan Ltd.
Tel: 81-043-299-2309
Fax: 81-043-299-2408
National does not assume any responsibility for use of any circuitry described, no circuit patent licenses are implied and National reserves the right at any time without notice to change said circuitry and specifications.