One of the main reasons for modulation is to appropriately match the transmitted information to the target channel.  Some channels like POTS lines are Baseband channels.  Other channels like wireless channels (Passband channels) have higher bandwidths with different propagation characteristics.  See app note on Fading.

• Background

Baseband modulation is used when the desired information is in the digital domain and bits must be mapped onto analog waveforms for transmission through baseband channels.  Without this mapping, the digital signal would undergo distortion as it passed through the bandlimited baseband channel.  One example of this is PAM (Pulse Amplitude Modulation) discussed below.  This concept is tightly coupled to, and sometimes synonymous with, baseband signaling and pulse shaping (including partial response signaling designed to mitigate ISI).   All three techniques are used to process the information signal and involve tradeoffs like transmitted signal spectral bandwidth vs. noise resiliency vs. ease of timing recovery etc.

Still another reason to use baseband modulation is the ability of multiplexing many desired information streams onto one signal, making transmission and reception of multiple information signals easier.

Carrier modulation is designed to:

1) More appropriately match the characteristics of the information signal with the channel characteristics.

2) Make physically realizable antennas a possibility for wireless transmission.

• Baseband modulation schemes

PAM

The sampled version of an analog waveform.  The discretized samples have amplitudes that follow the original waveform exactly.  "Quantized" PAM refers to a set of discretized samples that are also quantized, into a finite set of amplitudes.

Spectral content from DC to information signal bandwidth.

PCM - When a symbol represents the encoding of a quantized PAM sample into a digital word.

PCM actually belongs to the class of Source Coders known as "Waveform coders".  See app note on Coding Theory.

There are quite a few types of PCM waveforms including:

NRZ

RZ

Phase Encoded (including Manchester encoding and Miller encoding)

The particular choice of PCM depends upon tradeoffs of bandwidth, noise immunity, synchronization, etc.

• Carrier modulation schemes

1) Amplitude modulation (AM) - vary the amplitude of a fixed frequency carrier (sin wave) with the information signal.

The spectral content is centered symmetrically around carrier frequency.

Can be generated by multiplying the baseband PAM waveform with the carrier wave.

Double Sideband AM (DS-AM) has twice the transmitted bandwidth as the baseband waveform.   Double Sideband Suppressed Carrier AM (DSSC-AM) has the same bandwidth as DS-AM but consumes less power.  Single Sideband AM (SS-AM) has the same SNR as DS-AM but consumes only halve the transmitted bandwidth.

A digital counterpart of AM is ASK.

Two-level ASK is similar to On-Off Keying (OOK).  In OOK, the carrier is turned off to signify the transmission of a zero, while this is not necessarily the case in ASK.

2) Frequency modulation (FM) - vary the frequency of the carrier with the information signal.

FM is less sensitive to channel noise degradation.

For FM, the modulation index = freq deviation / f modulation.

A digital counterpart  of FM is Frequency Shift Keying (FSK).

If one bit per symbol, then called 2-level FSK.  This is standard FM.

For FSK, the modulation index = difference between frequencies / bit rate.

FSK with mod index = .5 is called Minimum Shift Keying (MSK).

In MSK a change in frequency is guaranteed every bit period.  This is the reason that MSK is also called "Fast Frequency Shift Keying".  An alternate name is Continuous Phase Shift Keying (CPFSK).

FSK with a mod index = 1 is called Orthogonal Frequency Shift Keying (OFSK).

3) Phase Modulation (PM) - vary the phase of the carrier with the information signal.

A digital counterpart of PM is Phase Shift Keying (PSK).

If one bit per symbol, then called Binary Phase Shift Keying (BPSK).

If two bits per symbol, then called Quadrature Phase Shift Keying (QPSK).

The information pulse stream can be divided into two streams.  One called the in-phase stream and one called the quadrature stream.  Even order bits can be assigned to the in-phase stream, while odd order bits can be assigned to the quadrature stream.  Each stream consists of bipolar pulses (+/- 1).  Now, the in-phase and the quadrature stream each have half the bit rate as the original information stream.

One realization of QPSK is to have the in-phase and quadrature streams amplitude modulate sin and cosine functions of a carrier wave and then summing the two orthogonal components.

The composite time domain waveform can be expressed as:

s(t) = (1/sqrt[2] ) * I(t) * cos(w₀t + pi/4) + (1/sqrt[2] ) * Q(t) * sin(w₀t + pi/4)   (see Sklar, "Digital Communications").

This is two BPSK waveforms orthogonal to each other.

The composite QPSK waveform will take on a phase value dependent upon the values of the in-phase and quadrature components of the information stream.  The four possible values are 0, +/-90 and 180 degrees.

In QPSK modulation, both the in-phase and quadrature components both transition at the same time, at a rate of (1/2)T.

In OQPSK, the in-phase and quadrature components are offset by T and never change states at the same time.  Thus the phase of the carrier can never change by 180 degrees.  Possible phase changes for OQPSK are 0 and +/-90 degrees every T sec.

This is advantageous since removal of abrupt phase transitions suppress out-of-band interference.

MSK (discussed above), can also be described as a case of OQPSK with special symbol sinusoidal weighting on each of the in-phase and quadrature data streams.  With MSK, the in-phase component has a data dependent term and a symbol weighting term, and the quadrature component has a data dependent term and a symbol weighting term.  By imposing some extra phase constraints, the in-phase and quadrature data dependent terms can only change phase at the zero crossings of the sinusoidal symbol weighting terms (which have durations of 2T).  This creates less abrupt phase transitions than would be possible with the data stream alone (without weighting).  In this case, the I and Q components are still offset by T seconds, just like OQPSK.

• Linear / Non-Linear modulation

Some modulation schemes are classified as linear modulations, while some modulation schemes are classified as non-linear modulation.

Constant envelope modulations such as FSK and MSK have transmitted carrier amplitudes that do not vary with the modulating signal and are non-linear modulations.

BPSK and QPSK modulation schemes contain AM components in the transmitted modulation envelope and are linear modulation schemes.

Linear modulation schemes can be used with power amplifiers operated near saturation, but in general they are not as spectrally efficient as non-linear modulation schemes.

Non-linear modulation schemes dictate that the PA be operated with some degree of "backoff" to ensure that the spectral efficiencies brought about with non-linear modulation are not negated with spectral splatter.

If the amplitude as well as the phase of a fixed frequency carrier is allowed to be varied, then more degrees of freedom become available for signal transmission.  This is called M-ary modulation.  M-ary PSK is called Quadrature Amplitude modulation (QAM).  Multiple bits are grouped together to form symbols, and any one of a finite number of signals is transmitted during a symbol period.

• Performance Measures

For Analog modulation, typical performance measures are the SNR, SINAD (Signal to Noise and Distortion) and THD (Total Harmonic Distortion).  For Digital Modulation, typical performance measures are Bit Error Rate (BER) and Symbol Error Rate (SER).  If Quadrature Modulation techniques are employed, then the Eye Diagram (for the I channel and for the Q channel independently) and the Signal Space Constellation Diagram are used.  The Eye Diagram is useful to explore sensitivity to timing and ISI, while the Constellation Diagram is useful to explore a given modulation scheme's sensitivity to noise and distortion.

Universal performance metrics that can be used to compare and contrast both analog and digital modulation schemes are Spectral Occupancy (Null to Null bandwidth and sidelobe levels) and Probabilities of Error.

• Summary

In general, the modulation scheme is chosen for a given channel bandwidth (which impacts receive sensitivity, data rate, etc...) as well as cost of implementation (synchronization, detection, etc...).  Or conversely, a modulation scheme is chosen based on implementation targets and a required transmission bandwidth falls out.

The coding scheme and the modulation scheme are very tightly coupled together to create a robust communication link.  Source coding (Line coding) which is designed to remove as much redundancy as possible from the information signal while maintaining fidelity, includes various types of, quantization, baseband modulations and companders (like u-law and A-law).  Channel coding  is designed to compensate for channel impairments that degrade the modulation performance.