T rellis-Coded Modulation (TCM) hasevolved over

thepastdecadeascombined
modulationtechnique
a codingand
fordigitaltransmission over
band-limited channels. Its main attraction comes from
the fact that it allows the achievement of significant
codinggains over conventionaluncodedmultilevel
modulationwithoutcompromisingbandwidth effi-
ciency. T h e first T C M schemes were proposed in 1976
[I]. Following a more detailed publication [2] in 1982, an
explosion of research and actual implementations of
TCM took place, to the point where today there is a good
understanding of the theory and capabilities of T C M
methods. In Part 1 of this two-part article, an introduc-
tion into TCM is given. T h e reasons for the development
of TCM are reviewed, and examples of simple TCM
schemesarediscussed.Part I1 [I51providesfurther
insight intocode design and performance, and addresses .
recent advances in TCM.
T C M schemes employ redundant nonbinary modula-
tion in combination with a finite-state encoder which
governs the selection of modulation signals to generate
coded signal sequences. In thereceiver, the noisy signals
aredecoded by asoft-decisionmaximum-likelihood
sequence decoder. Simple four-state TCM schemes can
improve. the robustness of digital transmission against
additivenoise by 3 dB, comparedtoconventional
, uncoded modulation. Withmore
complex TCM
schemes, the coding gain can reach 6 dB or more. These
gains are obtained without bandwidth expansion or
reduction of the effective information rate as required by
traditional error-correction schemes. Shannon’s infor-
mation theory predicted the existence of coded modula-
tion schemes with these characteristics more than three
decades ago. T h e development of effective TCM tech-
niques and today’s signal-processing technology now
allow these ,gains to be obtained in practice.
Signal waveforms representing information sequences
~ are most impervious to noise-induced detection errors if
they are very different from each other. Mathematically,
~
this translates into therequirement that signal sequences
should have large distance in Euclidean signal space.
~ T h e essential new concept of TCM that led to the afore-
1 mentionedgainswastousesignal-setexpansionto
I provide redundancy for coding, and to design coding and
’ signal-mappingfunctionsjointly so astomaximize
~ directlythe “free distance”(minimumEuclideandis-
tance) between coded signal sequences. This allowed the
construction of modulation codes whose free distance
significantly exceeded the minimum distance between
uncoded modulation signals, at the same information
rate, bandwidth, and signal power. The term “trellis” is
used becausethese schemes can be described by a state-
transition (trellis) diagram similar to the trellis diagrams
of binary convolutional codes. T h e difference is that in
T C M schemes,thetrellisbranchesarelabeledwith
redundant nonbinary modulation signals rather than
with binary code symbols.
T h e basic principles of T C M were published in 1982
[2]. Furtherdescriptionsfollowed in 1984[3-61, and
coincided with a rapid transition of TCM from the re-
search stage to practical use. In 1984, a T C M scheme
with a coding gainof 4 dB was adopted by the Interna-
tional Telegraph and Telephone Consultative Commit-

February 4987-Vol. 25, No. 2

5 IEEE Communications Magazine
 tee (CCITT)for use innewhigh-speedvoiceband Conventional encoders and decoders for errorcorrec-
modems [5,7,8]. Prior to TCM, uncoded transmission tion operate on binary, or more generally Q-ary, code
at 9.6 kbit/s over voiceband channels was often con- symbols transmitted over a discrete channel. With a code
sidered asapracticallimitfordatamodems.Since of rate k/n < 1, n - k redundant checksymbolsare
1984, datamodems have appearedon themarket appended to every kinformation symbols.Sincethe
which employ TCM along with other improvements in decoder receives only discrete code symbols, Hamming
equalization, synchronization, and so forth, to transmit distance (the number of symbolsin whi.ch twocode
data reliably over voiceband channels at rates of 14.4 sequencesorblocksdiffer,regardless of how these
kbit/s and higher. Similar advances are being achieved symbols differ)is the appropriate measure of distance for
intransmission over
other
bandwidth-constrained decodingandhenceforcodedesign. A minimum
channels. The common use of TCM techniques in such Hamming distancedii,,alsocalled“freeHamming
applications, as satellite [9-1 I], terrestrial microwave, distance” in thecase of convolutional codes, guarantees
andmobilecommunications,inorder to increase that the decoder can correct at least [(dii, -1)/2] code-
throughput rate or to permit satisfactory operation at symbolerrors. If lowsignal-to-noiseratiosornon-
lower signal-to-noise ratios, can be safely predicted for stationary signal disturbance limit the performance of
the near future. the modulation system, the ability to correct errors can
justify the rate loss caused by sending redundant check
Classical Error-Correction Coding symbols. Similarly,
longdelays
in
error-recovery
procedures canbe a good reason for trading transmission
In classical digital communication systems, the func- rate for forward error-correction capability.
tions of modulation and error-correction coding are Generally, there exist two possibilities to compensate
separated.Modulatorsanddemodulatorsconvertan for the rate loss: increasing the modulation rate if the
analog waveform channel.intoa discrete channel, channel permits bandwidth expansion, or enlarging the
whereas encoders and decoders correct errors that occur signal set of the modulation system if the channel is
o n the discrete channel. band-limited. The latter necessarily leads to the use of
In conventional multilevel (amplitude and/or phase) nonbinary modulation (M > 2). However,when
modulation systems, during each modulation interval modulation and error-correction coding are performed
the modulator maps m binary symbols (bits) intoofone inthe classicalindependentmanner,disappointing
M = 2’” possible transmit signals, and the demodulator results are obtained.
recovers the m bits by making an independent M-ary As an illustration, consider four-phase modulation
nearest-neighbordecisiononeachsignal received. (4-PSK) without coding, and eight-phase modulation
Figure 1 depictsconstellations of real-orcomplex- (8-PSK) used with a binary error-correction codeof rate
valued modulation amplitudes, henceforth called signal 2/3.Bothsystemstransmittwoinformationbitsper
sets, which are commonly employed for one- or two- modulation interval (2 bit/sec/Hz). If the 4-PSK system
dimensionalM-arylinearmodulation.Two-dimen- operates at an error rate of lo-’, at the same signal-to-
sional carrier modulation requires a bandwidth of 1/T noise ratio the “raw” error rate at 8-PSK
the demodulator
Hz around the carrier frequency to transmit signals at a exceeds IO-’ because of the smaller spacing between the
modulation rate of 1 / T signals/sec (baud)without 8-PSK signals. Patternsof at least three bit errors must be
intersymbolinterference.Hence,two-dimensional 2‘”- corrected to reduce the error rate to that of the uncoded
ary modulation systems can achieve a spectral efficiency 4-PSK system.A rate-2/3 binary convolutional code with
of about m bit/sec/Hz. (The same spectral efficiency is constraint length u = 6 has the required value of dii, = 7
obtained withone-dimensional 2m/2-arybaseband [12]. Fordecoding,afairlycomplex64-statebinary
modulation.) Viterbi decoder is needed. However, after all this effort,
error performance only breaks even with ofthat uncoded

-
Amplitudemodulation Amplitude/Phasemodulation 4-PSK.
Two problems contribute to this
unsatisfactory
T situation.

4-AM Soft-Decision Decoding and Motivation for

New Code Design

w
O - 2 r - O

16-AM

Phasemodulation
LJJ
o o o o o o o : o o o o o
o o o o o o i o i o o o o o

. .lo 0 ojoiojo
: : ; ;
1j
0 01.

i i 16-PASK
.
One problem in the

of informationinthe
coded 8-PSKsystem just described
arisesfromtheindependent“hard”signaldecisions
made prior to decoding which cause anirreversible loss
receiver. T h e remedy forthis
problem is soft-decision decoding, which means that the
j j decoder operates directly on unquantized “soft” output
j i 32-CROSS
O 1 ; i samples of the channel. Let the samples be r, = a, 4-w,
j 64-OASK
(real- or complex-valued, for one- or two-dimensional
1 28-CROSS modulation, respectively), where the a, are the discrete
4-PSK 8-PSK 16-PSK
signalssent by themodulator,andthe w, represent
Fig. 1 . Signal sefs for one-dimensional amplifude modulation, samples of a n additive white Gaussian noise process.
and two-dimensional phase and ampliiudelphase modulaiion. T h e decision ruleof the optimum sequence decoder isto

February 1987-Vol. 25, No. 2

IEEE Communications Magazine 6
determine, among the set Cof all coded signal sequences necessary forcoding wouldhave to come from expanding
which a cascaded encoder and modulator can produce, the signal set to avoid bandwidth expansion.
thesequence {%} withminimumsquaredEuclidean T o understandthepotentialimprovements tobe
distance (sum of squared errors) from {r"], that is, the expected by this approach, he computed the channel
sequence {&,} which satisfies capacity of channels with additive Gaussian noise for the
case of discrete multilevel modulation at the channel
input and unquantized signal observation at the channel
output. The results of these calculations [2] allowed
T h e Viterbi algorithm, originally proposed in 1967 makingtwoobservations:firstly,thatinprinciple
[I31 as an "asymptotically optimum" decoding tech- coding gainsof about 7-8 dB over conventional uncoded
nique for convolutional codes, can be used to determine multilevelmodulationshould be achievable, and
the coded signal sequence {^aIl) closest to the received secondly, that mostof the achievable coding gain could
unquantized signal sequence {r,,}[12,14], provided that be obtained by expandingthesignal setsusedfor
the generation of coded signal sequences {a&C follows uncodedmodulationonly by thefactor of two. T h e
the rules of a finite-state machine. However, the notion author then concentrated his efforts on finding trellis-
of "error-correction" is then no longer appropriate, since based signaling schemes that use signal sets of size 2"'"
there areno hard-demodulator decisionsto be corrected. for transmission of m bits per modulation interval. This
T h e decoder determines the most likely coded signal direction turned out to be succesful and today's TCM
sequence directly from the unquantized channel outputs. schemes still follow this approach.
Themostprobableerrorsmade by theoptimum T h e next two sections illustrate with two examples
soft-decision decoder occur between signals or signal howTCM schemeswork.Wheneverdistancesare
sequences {aIl} and {b"}, one transmitted and the other discussed, Euclidean distances are meant.
decoded, that are closest together in terms of squared
Euclidean distance. The minimum squared such dis- Four-State Trellis Code for 8-PSK Modulation
tance is called the squared "free distance:" T h e coded 8-PSK scheme described in this section was
the first TCM scheme found by the author in 1975 with a
significant codinggainover uncoded modulation. was It
designed in a heuristic manner, like other simple TCM
When optimum sequence decisions are made directly systems shortly thereafter. Figure 2 depicts signal sets
in terms of Euclideandistance,asecondproblem and state-transition (trellis) diagrams for a) uncoded
becomes apparent. Mapping of code symbols of a code 4-PSK modulation and b) coded 8-PSKmodulation with
optimized for Hamming distance into nonbinary modu- four trellis states. A trivial one-state trellis diagram is
lation signals doesnot guarantee that a good Euclidean shown in Fig. 2a only to illustrate uncoded 4-PSK from
distancestructureisobtained.Infact,generallyone the viewpointof TCM. Every connected path through a
cannot evenfindamonotonicrelationship between trellis in Fig.2 represents an allowed signal sequence. In
Hamming and Euclidean distances, no matter how code
symbols are mapped.
For a long time, this hasbeen the main reason for the
lack of good codes for multilevel modulation. Squared
Euclidean and Hamming distances are equivalent only
in the case of binary modulation or four-phase modula-
tion,which merely corresponds to twoorthogonal
binary modulations, of a carrier. In contrast to coded Redundant [I-PSK signal set
4-PSK signal set
multilevel systems, binarymodulation systems with
codes optimized for Hammingdistance and soft-decision
decoding havebeen well established since the late1960s
for power-efficient transmission at spectral efficiencies State
of less than 2 bit/sec/Hz. s!, s i
The motivation of this author for developing TCM 0 0
initiallycamefromworkonmultilevel systems that
0 1
employ the Viterbi algorithm to improve signal detection
in the presence of intersymbol interference. T h i s work 1 0
provided him with ample evidence of the importanceof
1 1
Euclideandistancebetweensignalsequences.Since
improvements over the established technique of adaptive Four-state trellis
equalization to eliminate intersymbol interference and
then making independent signal decisions in most cases
did not turn out to be very significant, he turned his
attention to using coding to improve performance. In
this connection, itwas clear to him that codes should be
designed for maximum free Euclidean distance rather Fig.2. ( a ) Uncodedfour-phasemodulation(4-PSK),(b)Four-siale
thanHammingdistance,andthattheredundancy trellis-codedeighi-phase modulaiion (8-PSK).

February 1987-Vol. 25. No. 2

7 IEEE Cornrnunlcatlonr Magarlne
both systems, starting from any state, four transitions can throughthe codetrelliswiththeminimumsumof
occur, as required to encode two information bits per squared distances from the sequence of noisy channel
modulation interval (2 bit/sec/Hz). For the following outputs received. Only the signals already chosen by
discussion, the specific encoding of information bits into subset decoding are considered.
signals is not important. Tutorial descriptions of the, Viterbi algorithm can be
The four “parallel” transitions in the one-state trellis foundin severaltextbooks,forexample, [12]. T h e
diagram of Fig. 2a for uncoded 4-PSK do notrestrikt the essential points are summarized here as follows: assume
sequences of 4-PSK signals that canbe transmitted, that that the optimum signal paths from the infinite pastto
is, there is n o ‘sequence coding. Hence, the optimum alltrellisstatesattimenareknown;thealgorithm
decoder can make independent nearest-signal decisions extends these paths iteratively from the states at time to n
+
foreachnoisy4-PSKsignal
distance between the4-PSK signals is
We callitthe“freedistance”
a,
received. T h e smallest
denoted asA,,.
of uncoded 4-PSK
the states at time n 1 by choosing onebest path to each
new state as a “survivor” and “forgetting” all other paths
that cannot be extended as the best paths to the new
modulation to use common terminology with sequence- states; looking backwards in time, the “surviving” paths
codedsystems.Each4-PSK signalhastwonearest- tend to merge into the same “history path” at some time
neighbor signals at this distance. n - d; with a sufficient decoding delay D (so that the
In the four-state trellis of Fig. 2b for the coded 8-PSK randomly changing value of d is highly likely to be
scheme, the transitions occur in pairs of two parallel smallerthan D), theinformation associated witha
transitions. (A four-state code with four distinct transi- transition on the common history path atn - D can
time
tions from each state to all successor states was also be selected for output.
considered; however, the trellis as shown with parallel Let the received signals be disturbed by uncorrelated
transitions permitted the achievement of a larger free Gaussian noise samples with variance u’ in each signal
distance.) Fig. 2b shows the numbering of the 8-PSK dimension. The probability that at any given time the
signalsandrelevantdistancesbetween these signals: decodermakesa wrongdecisionamongthesignals
A, = 2 sin(rr/8), A , = fi, a n d A, = 2. The 8-PSK sig- associated with parallel transitions, or starts to make a
nalsareassigned to thetransitionsinthefour-state sequence o f wrong decisions along some path diverging
trellis in accordance with the following rules: for more than one transition from the correct path, is
calledtheerror-eventprobability. At highsignal-to-
a) Parallel transitions are associated with signals with noise ratios, this probability is generally well approxi-
maximum distance A2(8-PSK)= 2 between them,
mated by
the signals in the subsets (0,4), (1,5), (2,6), or (3,7).
b) Four transitions originating from or merging in
onestatearelabeledwithsignalswithatleast W e ) N,,,, Q[d,,d(Pu)l,
distance Al(8-PSK) =@between them, that is, the
signals in the subsets (0,4,2,6) or (1,5,3,7).
c) All 8-PSK signals are used in the trellis diagram
where,a(.) represents the Gaussian error integral
with equal frequency.
Any two signal paths in the trellis of Fig. 2(b) that Q(x) = -
J2rr JEp(-y’/2)dy,
diverge in one state and remerge in another after more
thanonetransitionhaveat least squareddistance
A: + + +
& A: = At & between them. For example, the and Nfrrrdenotesthe(average)numberofnearest-
paths with signals 0-0-0 and2-1-2 have this distance. The neighbor signal sequences with distance dlrrrthat diverge
distance between such pathsis greater than the distance at any state from a transmitted signal sequence, and
betweenthesignalsassigned to paralleltransitions, remerge with it after one or more transitions. The above
A2(8-PSK)= 2, which thus is found as thefree distance approximateformula expressesthefactthatathigh
inthefour-state8-PSKcode:dflcr = 2. Expressed in
decibels, this amounts to an improvement of 3 dB over
the minimum distance 42 betweenthesignals of
uncoded1-PSKmodulation.Foranystatetransition
alonganycoded8-PSKsequencetransmitted,there
exists only one nearest-neighbor signal atfree distance,
which is the 180’ rotatedversion of thetransmitted DIFFERENTIAL ENCODER SPSK SIGNAL
signal. Hence, the codeis invariant to a.signal rotation MAPPING
by 180”, but to no other rotations (cf., Part11). Figure 3
illustrates one possible realizationof an encoder-modu-

-
lator for the four-state coded 8-PSK scheme.

I
Soft-decision decoding is accomplished in two steps: T
In the first step, called “subset decoding”, within each
subset of signalsassignedtoparalleltransitions,the
&z:j 01010101

signal closest to the received channel output is deter- CSTATE CONVOLUTIONAL

-t
0 1234567
a”
mined.Thesesignalsare storedtogetherwiththeir Signal No.
ENCODER
squared distances from the channel output. In the second
step, the Viterbi algorithm is used to find the signal path Fig. 3. Illushales an encoder for the four-slate 8-PSK code

February 1987-Mi. 25, NO. 2

IEEE
Magazine
Communications 0
signal-to-noiseratiostheprobability of errorevents
associatedwithadistancelargerthen d,,,.,. becomes
negligible.
Foruncoded4-PSK, we have = & ? a n d N,,cc= 2,
and for four-state coded 8-PSK we found d,,,,. = 2 and N,,ce
= 1 . Since in both systemsfree distance is found between
parallel transitions, single signal-decision errors are the
dominating error events. In the special case of these
simple systems, the numbersof nearest neighbors do not
depend on which particular signal sequence is trans-
mitted.
Figure 4 shows the error-event probability of the two
systemsasafunction of signal-to-noiseratio.For
uncoded 4-PSK, the error-event probability is extremely
well approximated by the last two equations above. For
four-state coded8-PSK, these equations provide a lower
bound that is asymptotically achievedat high signal-to-
noise ratios. Simulation results are included in4 Fig. for
the coded 8-PSK system to illustrate the effect of error
w i t h distance larger than free distance* whose Fig. 5. Nor.q~four-slale coded 8-PSK signal ai complex baseband
probability of occurrenceisnotnegligibleatlowsignal- ulilh a signal-lo-noue rnlio of E,No = 12.6 dB.
to-noise ratios.
Figure 5 illustrates a noisy four-state coded 8-PSK
signal as observed at complex basebandbefore sampling

in the receiver of a n experimental 64 kbit/ssatellite

modem [9]. At a signal-to-noise ratio of E,/No = 12.6 dB
(E,: signal energy, No: one-sided spectral noise density),
II the signal is decoded essentially error-free. At the same
signal-to-noise ratio, the error rate wit.h uncoded 4-PSK
modulation would be around
I n TCM schemes with more trellis states and other
signal sets, d,,<.?
is not necessarily found between parallel
transitions, andN,,,,,will generallybe an average number
larger than one, as will be shown by the second example.

Eight-State Trellis Code for

\ Amplitude/Phase Modulation
I 1
\
\
I T h e eight-state trellis code discussed in this section
I 1,
I
, 3 dB
wasdesignedfortwo-dimensionalsignal sets whose
I s signals are located o n a quadratic grid, also knownas a
1 0.5 s
\ lattice of type “22”. T h e code can be used with allof the
I \ 4 signal sets depictedinFig. 1 foramplitude/phase
I modulation. To transmit m information bits per modula-
I tion interval, a signal set with 2”’+’ signals is needed.
I trellis- coded Hence, for m= 3 the 16-QASK signal =
set is used, for m
I 4 the 32-CROSS signal set, and soforth. For any m, a
I coding gain of approximately 4 dB isachievedover
I uncoded modulation.
I
Figure6illustrates a “set partitioning” of the 16-
I\ QASK and 32-CROSS signalsets into eightsubsets. T h e
I Channel partitioning of larger signalsets isdone in the same way.
I capacity The signalset chosen is denotedby AO, and its subsets by
I of 8-PSK
- 2 bit/sec/Hz
DO, D l , . . . D7. If thesmallestdistanceamongthe
I signals inA0 is A,,, then among the signals in the union
I of the subsets DO,D4,D2,D6 or Dl,D5,D3,D7 the mini-
I I m u m distance is fi A, in the union of the subsets
5 6 7 8 9 10 11 12 dB DO,D4; D2,D6; Dl,D5; or D3,D7 it is & &, and within
SNR (E s/NO> the individual subsets it fib. is (Aconceptually similar
Fis partitioning of the 8-PSK signal set into smaller signal
F I g . 4. Error-even! probnbrlrly ver.src.c srgnal-lo-norse rulro for sets with increasingintra-set distances.wasimplied in the
uncoded 4 - P S K nnd four-sinie coded R-PSR. example of coded 8-PSK. The fundamental importance

February 1987-Voi. 25, No. 2

9 IEEE Communications Magazine
 Slgnal sets: 1bPASK and 32-CROSS
sequence and from any state along this sequence, there
A0
are four such paths, two of length three and two of length
four. The most likely error events will correspond to
these error paths, and will result in bursts of decision
errors of length three or four.
Signal subsets The coding gains asymptotically achieved at high
I I I I signal-to-noise ratios are calculated in decibels by
DO D4 D2 D6 Dl D5 D3 07
G, = 10 loglo [(d~~,,,,/d~~,,,,,)/E,,,-/~.,,,,)l,
where d:,??,<and d:r,.,.,,,are the squared free distances, and
E,,<and E,,,,denote the average signal energies of the coded
and uncoded schemes, respectively. When the signalsets
Fig.6. Selparlilioningofll~e16-QASKnnd3-3-CROSSsignalsels.
havethesameminimumsignalspacing &, dylrr,c/
d:l,,,,, = 5, and E,JE,,,, 2: 2 for all relevant values of m.
Hence, the coding gain is 10 lOgl0(5/2) = 4 dB.
of this partitioning for TCM codes will be explained in The number of nearestneighborsdependsonthe
Part 11.) sequence of signals transmitted, thatis N,,,, represents an
Intheeight-state trellisdepictedinFig. 7, four averagenumber.This is easy to see foruncoded
transitions diverge from and merge into each state. T o modulation, where signals in the center of a signal set
eachtransition,one of thesubsets DO, . . . D7 is have more nearest neighbors than the outer ones. For
assigned. If A0 contains 2”’” signals, each of its subsets uncoded 16-QASK, N,,,, equals 3. For eight-state coded
will comprise 2‘”-’ signals. This means that the transi- 16-QASK, Nc,,.,, is around 3.75. Inthelimit of large
tionsshowninFig. 7 in factrepresent 2‘“-‘ parallel “Z,”-type signal sets, these values increase toward4 and
transitions in the same sense as there were two parallel 16 foruncodedandeight-statecoded systems,re-
transitions in the coded 8-PSK scheme. Hence, 2’” signals spectively.
can be sent from each state, as required to encode m bits
per modulation interval. Trellis Codes of Higher Complexity
Theassignment of signal subsets to transitions
satisfies the same three rules as discussed for coded 8- Heuristic code design and checking of code properties
PSK, appropriately adaptedto the present situation.T h e by hand, as was done during the early phases of the
four transitions from or to the same state are always development of T C M schemes, becomes infeasible for
assigned either the subsets DO,D4,D2,D6 or Dl ,D5,D3,D7. codes with many trellis states. Optimum codes must then
This guarantees a squared signal distance of at least 2& be found by computer search, using knowledge of the
when sequences diverge and when they remerge. If paths general structureof T C M codes and anefficient method
remerge after two transitions, the squared signal distance to determinefree distance. T h e search technique should
is at least 4Ai between the diverging transitions, and alsoincluderulesto rejectcodes withimproperor
hence the total squared distance between such paths will equivalent distance properties without having to evalu-
be at least 6Ai. If paths remerge after three or more ate free distance.
transitions,atleastoneintermediatetransitioncon- In Part 11, the principles of T C M codedesignare
tributes an additional squared signal distance &, so the outlined, and tables of optimum TCM codes given for
squared distance between sequences is at least & &. one-, two-, and higher-dimensional signal sets. T C M
Hence, the free distance of this code is 6
&. T h i s is encoder/modulators are shownto exhibit the following
smaller than the minimum signal distance within in the general structure: (a)of the m bits to be transmitted per
subsets DO, . . . D7, which is fib. For one particular encoder/modulator operation, mIm bits are expanded
code sequence DO-DO-D3-D6, Fig. 6 illustrates four error into fi +1 codedbits by abinaryrate-i%/(i%+l)
paths at distance 6
& from that code sequence; all convolutional encoder; (b) ii~ +
the 1 coded bits selectone
of 2,Gi+l subsets of a redundant 2’”+’-arysignal set; (c) the
starting at the same state and remerging after three or
fourtransitions.Itcan be shownthat foranycode remaining m-x% bitsdetermineone of2”-’ signals
within the selected subset.

New Ground Covered by Trellis-Coded

Modulation
T C M schemesachievesignificantcodinggains at
values of spectral efficiency for which efficient coded-
modulation schemes werenot previously known, that is,
above and including 2 bit/sec/Hz. Figure 8 shows the
free distances obtained by binary convolutional coding
with 4-PSK modulation for spectral efficiencies smaller
than 2 bit/sec/Hz,and by T C M schemes withtwo-
dimensional signal sets for spectral efficiencies equal to
Fzg. 7 . Eight-stale frellis code for nmplitudelphnse modulation orlargerthan 2 bit/sec/Hz. T h e free distances of
w i f h “Zz”-type srgnal sets; dlrr. = 641. uncodedmodulationattherespectivespectral effi-

February 1987-Vol. 25, No. 2

IEEE Communications Magazine 10
 BINARY CONVOLUTIONAL CODES
WITH CPSK MODULATION
coding to nonbinarymodulationwithsignal sets of
I (4 - 256 stater) I / I I I arbitrary size. It allows the achievement of coding gains

t of 3-6 dB at spectralefficiencies equal to or larger than 2

bit/sec/Hz. These are the values at which one wants to
operate on many band-limited channels. Thus, a gap in
thetheoryandpractice
closed.
of channelcodinghas been

References
[I] G. Ungerboeck and I . Csajka, “On improving data-link
performance by increasing the channel alphabetand
introducing sequence coding,” 1976 Int. Symp. Inform.
Theory, Ronneby, Sweden, June 1976.
[2] G. Ungerboeck, “Channel coding with multileveVphase
signals,” I E E E T r a n s . I n f o r m a l i o n T h e o r y , vol. IT-28,
pp. 55-67, Jan. 1982.
[3] G. D. Forney, Jr., R. G. Gallager, G. R. Lang, F. M.
Longstaff, and S. U. Qureshi, “Efficient modulation for
band-limited channels,” IEEE Trans. Selecled Areas in
C o m m . , vol. SAC-2, pp. 632-647, Sept. 1984.
[4] L. F. Wei,“Rotationally invariant convolutional channel
coding with expanded signal space-Part I: 180 degrees,”
I E E E T r a n s . S e l e c t e d A r e a s i n C o m m . , vol. SAC-2, pp.
659-672, Sept. 1984.
[5] L. F. Wei,“Rotationally invariant convolutional channel
1 2 3 4 5 b4-uA:
coding with expanded signal space-Part 11: nonlinear
codes,” IEEE Trans. Selecled Areas in Comm., vol. SAC-2,
Spectral efficiency [bit/sec/Hz]
pp. 672-686, Sept. 1984.
Fig. 8. Free distance of binaryco~tuolutional codes with f - P S K [6] A. R. Calderbank and J. E. Mazo, “A new description of
modulation, and TChf with a unrirty of two-dimenszonal modula- trellis codes,” I E E E T r a n s . I n f o r m a l i o n T h e o r y , vol. IT-
tzon .scheme.s, for spectral efficiencies from 213 to 6 b i t l s e c l H z . 30, pp. 784-791, NOV.1984.
[7] CCITT Study Group XVII, “Recommendation V.32 fora
ciencies are also depicted. T h e average signal energy of family of 2-wire, duplex modems operating on the general
switched telephone network and on leased telephone-type
all signal sets is normalized to unity. Free distances are circuits,” Document AP VIII-43-E, May 1984.
expressed in decibels relative to the value dfrct, = 2 of 81 CCITT Study Group XVII, “Draft recommendation V.33
uncoded 4-PSK modulation. The binary convolutional for 14400 bits per second modem standardized for useon
codes of rates 113, 1/2, and 314 with optimum Hamming point-to-point 4-wire leased telephone-type circuits,”
distances are taken from textbooks, such as, [12]. T h e Circular No. 12, COM XVII/YS, Geneva, May 17, 1985.
T C M codes and their properties are found in the code 91 G. Ungerboeck, J . Hagenauer, and T. Abdel Nabi,
tables presented in Part I1 (largely reproduced from [2]). “Coded 8-PSK experimental modem for the INTELSAT
All coded systems achieve significant distance gains SCPC system,” Proc. 7th Int. Conf. on Digital Satellite
with as few as 4,8, and 16 code states. Roughly speaking, Communications (ICDS-7), pp. 299-304, Munich, May
12-16,1986.
it is possible to gain 3 dB with4 states, 4 dB with 8 states, IO] R. J. F. Fang, “A coded8-PSKsystemfor140-Mbit/s
nearly 5 dB with 16 states, a n d u p to 6 dB with 128 or information ratetransmissionover80-MHz nonlinear
more states. The gains obtained with two-state codes transponders,” Proc. 7th Int. Conf. on Digital Satellite
usually are very modest. With higher numbers of states, Communications (ICDS-’I), pp. 305-313, Munich, May
theincrementalgains become smaller.Doublingthe 12-16,1986.
number of states does not always yield a code with larger111 T. Fujino, Y. Moritani, M. Miyake, K. Murakami, Y.
free distance. Generally, limited distance growth and Sakato. and H . Shiino, “A 120 Mbit/s 8PSK modem with
increasing numbersof nearest neighbors, and neighbors soft-Viterbi decoding,” Proc.7th Int. Conf. on Digital
with next-larger distances, are the two mechanisms that Satellite Communications (ICDS-7), pp. 315-321, Mu-
prevent real coding gains from exceeding the ultimate nich, May12-16,1986.
[I21 G. C. Clark and J. B. Cain, Error-Correction Coding for
limit set by channelcapacity.Thislimitcan be Digital Communications, Plenum Press, New York and
characterized by the signal-to-noise ratio at which the London, 1981.
channel capacity of a modulation system with a 2””-ary [I31 A. J. Viterbi, “Error bounds for convolutional codes and
signal set equals m bit/sec/Hz [2] (see also Fig. 4). an asymptotically optimum decoding algorithm,” I E E E
T r a n s . I n f o r m a l i o n T h e o r yvol.
, IT-13, pp. 260-269, April
Conclusion 1967.
[I41 G. D. Forney, Jr., “The Viterbi algorithm,” Proc. of l h e
Trellis-coded modulationwas invented as a method to I E E E , vol. 61, pp. 268-278, March 1973.
improvethe noise immunity of digitaltransmission [ 151 G. Ungerboeck. “Trellis-coded modulation with redun-
systems without bandwidth expansion or reduction of dant signal sets, Part 11: State of the art.” I E E E
data rate. TCM extended the principles of convolutional C o m m u n i c a t i o n s M a g a z i n e , vol. 25, no. 2, Feb. 1987.

February 1987-Vol. 25, No. 2

11 IEEE Communications Magazine