Phase-shift keying (PSK) is a digital modulation scheme that conveys data by changing, or modulating, the phase of a reference signal (the carrier wave).

Any digital modulation scheme uses a finite number of distinct signals to represent digital data. In the case of PSK, a finite number of phases are used. Each of these phases is assigned a unique pattern of binary bits. Usually, each phase encodes an equal number of bits. Each pattern of bits forms the symbol that is represented by the particular phase. The demodulator, which is designed specifically for the symbol-set used by the modulator, determines the phase of the received signal and maps it back to the symbol it represents, thus recovering the original data. This requires the receiver to be able to compare the phase of the received signal to a reference signal — such a system is termed coherent.

Alternatively, instead of using the bit patterns to set the phase of the wave, it can instead be used to change it by a specified amount. The demodulator then determines the changes in the phase of the received signal rather than the phase itself. Since this scheme depends on the difference between successive phases, it is termed differential phase-shift keying (DPSK). DPSK can be significantly simpler to implement than ordinary PSK since there is no need for the demodulator to have a copy of the reference signal to determine the exact phase of the received signal (it is a non-coherent scheme). In exchange, it produces more erroneous demodulations. The exact requirements of the particular scenario under consideration determine which scheme is used.

Introduction

---

There are three major classes of digital modulation techniques used for transmission of digitally represented data:

• Amplitude-shift keying (ASK)
• Frequency-shift keying (FSK)
• Phase-shift keying (PSK)

All convey data by changing some aspect of a base signal, the carrier wave, (usually a sinusoid) in response to a data signal. In the case of PSK, the phase is changed to represent the data signal. There are two fundamental ways of utilizing the phase of a signal in this way:

• By viewing the phase itself as conveying the information, in which case the demodulator must have a reference signal to compare the received signal's phase against; or
• By viewing the change in the phase as conveying information — differential schemes, some of which do not need a reference carrier (to a certain extent).

A convenient way to represent PSK schemes is on a constellation diagram. This shows the points in the Argand plane where, in this context, the real and imaginary axes are termed the in-phase and quadrature axes respectively due to their 90° separation. Such a representation on perpendicular axes lends itself to straightforward implementation. The amplitude of each point along the in-phase axis is used to modulate a cosine (or sine) wave and the amplitude along the quadrature axis to modulate a sine (or cosine) wave.

In PSK, the constellation points chosen are usually positioned with uniform angular spacing around a circle. This gives maximum phase-separation between adjacent points and thus the best immunity to corruption. They are positioned on a circle so that they can all be transmitted with the same energy. In this way, the moduli of the complex numbers they represent will be the same and thus so will the amplitudes needed for the cosine and sine waves. Two common examples are binary phase-shift keying (BPSK) which uses two phases, and quadrature phase-shift keying (QPSK) which uses four phases, although any number of phases may be used. Since the data to be conveyed are usually binary, the PSK scheme is usually designed with the number of constellation points being a power of 2.

Definitions

For determining error-rates mathematically, some definitions will be needed:

• $E\_b$ = Energy-per-bit
• $E\_s$ = Energy-per-symbol = $kE\_b$ with k bits per symbol
• $T\_b$ = Bit duration
• $T\_s$ = Symbol duration
• $N\_0/2$ = Noise power spectral density (W/Hz)
• $P\_b$ = Probability of bit-error
• $P\_s$ = Probability of symbol-error

$Q(x)$ will give the probability that $x$ is under the tail of the Gaussian probability density function towards positive infinity. It is a scaled form of the complementary Gaussian error function:

$Q(x)\; =\; \backslash frac\{1\}\{\backslash sqrt\{2\backslash pi\}\}\backslash int\_\{x\}^\{\backslash infty\}e^\{-t^\{2\}/2\}dt\; =\; \backslash frac\{1\}\{2\}\backslash ,\backslash operatorname\{erfc\}\backslash left(\backslash frac\{x\}\{\backslash sqrt\{2\}\}\backslash right),\backslash \; x\backslash geq\{\}0$.

The error-rates quoted here are those in additive white Gaussian noise (AWGN).

Applications

---

Owing to PSK's simplicity, particularly when compared with its competitor quadrature amplitude modulation, it is widely used in existing technologies.

The most popular wireless LAN standard, IEEE 802.11bIEEE Std 802.11-1999: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications — the overarching IEEE 802.11 specification.IEEE Std 802.11b-1999 (R2003) — the IEEE 802.11b specification.A description of the technical features of IEEE 802.11b., uses a variety of different PSKs depending on the data-rate required. At the basic-rate of 1 Mbit/s, it uses DBPSK. To provide the extended-rate of 2 Mbit/s, DQPSK is used. In reaching 5.5 Mbit/s and the full-rate of 11 Mbit/s, QPSK is employed, but has to be coupled with complementary code keying. The higher-speed wireless LAN standard, IEEE 802.11gIEEE Std 802.11g-2003 — the IEEE 802.11g specification. has eight data rates: 6, 9, 12, 18, 24, 36, 48 and 54 Mbit/s. The 6 and 9 Mbit/s modes use BPSK. The 12 and 18 Mbit/s modes use QPSK. The fastest four modes use forms of quadrature amplitude modulation.

Because of its simplicity BPSK is appropriate for low-cost passive transmitters, and is used in RFID standards such as ISO 14443 which has been adopted for biometric passports, credit cards such as American Express's ExpressPay, and many other applications.

Bluetooth 2A description of the technical features of Bluetooth 2 — the official standards documents are only available by subscription to the Bluetooth Special Interest Group. will use $\backslash pi/4$-DQPSK at its lower rate (2 Mbit/s) and 8-DPSK at its higher rate (3 Mbit/s) when the link between the two devices is sufficiently robust. Bluetooth 1 modulates with Gaussian minimum-shift keying, a binary scheme, so either modulation choice in version 2 will yield a higher data-rate. A similar technology, ZigBeeA description of the technical features of ZigBee — the official standards documents are only available by registering with the ZigBee alliance or the IEEE. (also known as IEEE 802.15.4) also relies on PSK. ZigBee operates in two frequency bands: 868–915MHz where it employs BPSK and at 2.4GHz where it uses OQPSK.

Notably absent from these various schemes is 8-PSK. This is because its error-rate performance is close to that of 16-QAM — it is only about 0.5dB better — but its data rate is only three-quarters that of 16-QAM. Thus 8-PSK is often omitted from standards and, as seen above, schemes tend to 'jump' from QPSK to 16-QAM (8-QAM is possible but difficult to implement).

Binary Phase-shift Keying (BPSK)

---

BPSK is the simplest form of PSK. It uses two phases which are separated by 180° and so can also be termed 2-PSK. It does not particularly matter exactly where the constellation points are positioned, and in this figure they are shown on the real axis, at 0° and 180°. This modulation is the most robust of all the PSKs since it takes serious distortion to make the demodulator reach an incorrect decision. It is, however, only able to modulate at 1bit/symbol (as seen in the figure) and so is unsuitable for high data-rate applications.

The bit error rate (BER) of BPSK in AWGN can be calculated as:

$P\_b\; =\; Q\backslash left(\backslash sqrt\{\backslash frac\{2E\_b\}\{N\_0\}\}\backslash right)$

Since there is only one bit per symbol, this is also the symbol error rate.

In the presence of an arbitrary phase-shift introduced by the communications channel, the demodulator is unable to tell which constellation point is which. As a result, the data is often differentially encoded prior to modulation.

Implementation

Binary data is often conveyed with the following signals:

$s\_0(t)\; =\; \backslash sqrt\{\backslash frac\{2E\_b\}\{T\_b\}\}\; \backslash cos(2\; \backslash pi\; f\_c\; t)$ for binary "0"

$s\_1(t)\; =\; \backslash sqrt\{\backslash frac\{2E\_b\}\{T\_b\}\}\; \backslash cos(2\; \backslash pi\; f\_c\; t\; +\; \backslash pi\; )$

= - \sqrt{\frac{2E_b}{T_b}} \cos(2 \pi f_c t) for binary "1" where $f\_c$ is the frequency of the carrier-wave.

Hence, the signal-space can be represented by the single basis function

$\backslash phi(t)\; =\; \backslash sqrt\{\backslash frac\{2\}\{T\_b\}\}\; \backslash cos(2\; \backslash pi\; f\_c\; t)$

where 0 is represented by $\backslash sqrt\{E\_b\}\; \backslash phi(t)$ and 1 is represented by $-\backslash sqrt\{E\_b\}\; \backslash phi(t)$. This assignment is, of course, arbitrary.

The use of this basis function is shown at the end of the next section in a signal timing diagram. The topmost wave is PSK modulating a cosine wave, and is the signal that the BPSK modulator would produce. The bit-stream that causes this output is shown above the wave (the other parts of this figure are relevant only to QPSK).

Behaviour in a fading channel

In a fading channel, such as Rayleigh fading, the error rate of BPSK is proportional to the reciprocal of the signal to noise ratio.

Quadrature Phase-shift Keying (QPSK)

---

Sometimes known as quaternary or quadriphase PSK or 4-PSK, QPSK uses four points on the constellation diagram, equispaced around a circle. With four phases, QPSK can encode two bits per symbol, shown in the diagram with Gray coding to minimize the BER — twice the rate of BPSK. Analysis shows that this may be used either to double the data rate compared to a BPSK system while maintaining the bandwidth of the signal or to maintain the data-rate of BPSK but halve the bandwidth needed.

Although QPSK can be viewed as a quaternary modulation, it is easier to see it as two independently modulated quadrature carriers. With this interpretation, the even (or odd) bits are used to modulate the in-phase component of the carrier, while the odd (or even) bits are used to modulate the quadrature-phase component of the carrier. BPSK is used on both carriers and they can be independently demodulated.

As a result, the probability of bit-error for QPSK is the same as for BPSK:

$P\_b\; =\; Q\backslash left(\backslash sqrt\{\backslash frac\{2E\_b\}\{N\_0\}\}\backslash right)$.

However, with two bits per symbol, the symbol error rate is increased:

$\backslash ,\backslash !P\_s$	$=\; 1\; -\; \backslash left(\; 1\; -\; P\_b\; \backslash right)^2$
	$=\; 2Q\backslash left(\; \backslash sqrt\{\backslash frac\{E\_s\}\{N\_0\}\}\; \backslash right)\; -\; Q^2\; \backslash left(\; \backslash sqrt\{\backslash frac\{E\_s\}\{N\_0\}\}\; \backslash right)$.

If the signal-to-noise ratio is high (as is necessary for practical QPSK systems) the probability of symbol error may be approximated:

$P\_s\; \backslash approx\; 2\; Q\; \backslash left(\; \backslash sqrt\{\backslash frac\{E\_s\}\{N\_0\}\}\; \backslash right\; )$

As with BPSK, there are phase ambiguity problems at the receiver and differentially encoded QPSK is more normally used in practice.

Implementation

The implementation of QPSK is more general than that of BPSK and also indicates the implementation of higher-order PSK. Writing the symbols in the constellation diagram in terms of the sine and cosine waves used to transmit them:

$s\_i(t)\; =\; \backslash sqrt\{\backslash frac\{2E\_s\}\{T\}\}\; \backslash cos\; \backslash left\; (\; 2\; \backslash pi\; f\_c\; t\; +\; (2i\; -1)\; \backslash frac\{\backslash pi\}\{4\}\backslash right\; ),\; i\; =\; 1,\; 2,\; 3,\; 4$.

This yields the four phases $\backslash pi/4$, $3\backslash pi/4$, $5\backslash pi/4$ and $7\backslash pi/4$ as needed.

This results in a two-dimensional signal space with unit basis functions

$\backslash phi\_1(t)\; =\; \backslash sqrt\{\backslash frac\{2\}\{T\_s\}\}\; \backslash cos\; (2\; \backslash pi\; f\_c\; t)$

$\backslash phi\_2(t)\; =\; \backslash sqrt\{\backslash frac\{2\}\{T\_s\}\}\; \backslash sin\; (2\; \backslash pi\; f\_c\; t)$

The first basis function is used as the in-phase component of the signal and the second as the quadrature component of the signal.

Hence, the signal constellation consists of the signal-space 4 points

$\backslash left\; (\; \backslash pm\; \backslash sqrt\{E\_s/2\},\; \backslash pm\; \backslash sqrt\{E\_s/2\}\; \backslash right\; )$.

The factors of $1/2$ indicate that the total power is split equally between the two carriers.

Comparing these basis functions with that for BPSK shows clearly how QPSK can be viewed as two independent BPSK signals. Note that the signal-space points for BPSK do not need to split the symbol (bit) energy over the two carriers in the scheme shown in the BPSK constellation diagram.

QPSK systems can be implemented in a number of ways. An illustration of the major components of the transmitter and structure are shown below.

QPSK signal in the time domain

The binary data that is conveyed by this waveform is: 1 1 0 0 0 1 1 0.

• The odd bits, highlighted here, contribute to the in-phase component: 1 1 0 0 0 1 1 0
• The even bits, highlighted here, contribute to the quadrature-phase component: 1 1 0 0 0 1 1 0

Offset QPSK (OQPSK)

See main article OQPSK

Offset QPSK, sometimes called Staggered QPSK (SQPSK), works by delaying one of the two components by the duration of half symbol (one bit). This limits the phase-jumps that occur at symbol boundaries to no more than 90° and reduces the effects on the amplitude of the signal due to any low-pass filtering. A disadvantage of OQPSK is that it introduces a delay of half a symbol into the demodulation process. In other words, using OQPSK increases the temporal efficiency of normal QPSK. The reason is that the in phase and quadrature phase components of the OQPSK cannot be simultaneously zero. Hence, the range of the fluctuations in the signal is smaller. An additional disadvantage is that the quiescient power is nonzero, which may be a design issue in devices targeted for low power applications.

$\backslash pi/4$–QPSK

This final variant of QPSK uses two identical constellations which are rotated by 45° ($\backslash pi/4$ radians, hence the name) with respect to one another. Usually, either the even or odd data bits are used to select points from one of the constellations and the other bits select points from the other constellation. This also reduces the phase-shifts from a maximum of 180°, but only to a maximum of 135° and so the amplitude fluctuations of $\backslash pi/4$–QPSK are between OQPSK and non-offset QPSK.

On the other hand, $\backslash pi/4$–QPSK lends itself to easy demodulation and has been adopted for use in, for example, TDMA cellular telephone systems.

The modulated signal is shown below for a short segment of a random binary data-stream. The construction is the same as above for ordinary QPSK. Successive symbols are taken from the two constellations shown in the diagram. Thus, the first symbol (1 1) is taken from the 'blue' constellation and the second symbol (0 0) is taken from the 'green' constellation. Note that magnitudes of the two component waves change as they switch between constellations, but the total signal's magnitude remains constant. The phase-shifts are between those of the two previous timing-diagrams.

One property this modulation scheme possesses is that if the modulated signal is represented in the complex domain, it does not have any zero crossings. In other words, the signal does not pass through the origin. This lowers the dynamical range of fluctuations in the signal.

Higher-order PSK

---

For the general $M$-PSK there is no simple expression for the symbol-error probability if $M>4$. Unfortunately, it can only be obtained from:

$$

P_s = 1 - \int_{\frac{\pi}{M}}^{\frac{\pi}{M}}p_{\Theta_{r}}\left(\Theta_{r}\right)d\Theta_{r}

where

$p\_\{\backslash Theta\_\{r\}\}\backslash left(\backslash Theta\_r\backslash right)\; =\; \backslash frac\{1\}\{2\backslash pi\}e^\{-2\backslash gamma\_\{s\}\backslash sin^\{2\}\backslash Theta\_\{r\}\}\backslash int\_\{0\}^\{\backslash infty\}Ve^\{-\backslash left(V-\backslash sqrt\{4\backslash gamma\_\{s\}\}\backslash cos\backslash Theta\{r\}\backslash right)^\{2\}/2\}$,

$V\; =\; \backslash sqrt\{r\_1^2\; +\; r\_2^2\}$,

$\backslash Theta\_r\; =\; \backslash tan^\{-1\}\backslash left(r\_2/r\_1\backslash right)$,

$\backslash gamma\_\{s\}\; =\; \backslash frac\{E\_\{s\}\}\{N\_\{0\}\}$ and

$r\_1\; \backslash sim\{\}\; N\backslash left(\backslash sqrt\{E\_s\},N\_\{0\}/2\backslash right)$ and $r\_2\; \backslash sim\{\}\; N\backslash left(0,N\_\{0\}/2\backslash right)$ are jointly-Gaussian random variables.

This may be approximated for high $M$ and high $E\_b/N\_0$ by:

$P\_s\; \backslash approx\; 2Q\backslash left(\backslash sqrt\{2\backslash gamma\_s\}\backslash sin\backslash frac\{\backslash pi\}\{M\}\backslash right)$.

The bit-error probability for $M$-PSK can only be determined exactly once the bit-mapping is known. However, when Gray coding is used, the most probable error from one symbol to the next produces only a single bit-error and

$P\_b\; \backslash approx\; \backslash frac\{1\}\{k\}P\_s$.

The graph on the left compares the bit-error rates of BPSK, QPSK (which are the same, as noted above), 8-PSK and 16-PSK. It is seen that higher-order modulations exhibit higher error-rates; in exchange however they deliver a higher raw data-rate.

Any number of phases may be used to construct a PSK constellation but 8-PSK is usually the highest order PSK constellation deployed. With more than 8 phases, the error-rate becomes too high and there are better, though more complex, modulations available such as quadrature amplitude modulation (QAM). Although any number of phases may be used, the fact that the constellation must usually deal with binary data means that the number of symbols is usually a power of 2 — this allows an equal number of bits-per-symbol.

Bounds on the error rates of various digital modulation schemes can be computed with application of the union bound or union bhattacharya bound to the signal constellation.

Differential Encoding

---

As mentioned for BPSK and QPSK there is an ambiguity of phase if the constellation is rotated by some effect in the communications channel the signal passes through. This problem can be overcome by using the data to change rather than set the phase.

For example, in differentially-encoded BPSK a binary '1' may be transmitted by adding 180° to the current phase and a binary '0' by adding 0° to the current phase. In differentially-encoded QPSK, the phase-shifts are 0°, 90°, 180°, -90° corresponding to data '00', '01', '11', '10'. This kind of encoding may be demodulated in the same way as for non-differential PSK but the phase ambiguities can be ignored. Thus, each received symbol is demodulated to one of the $M$ points in the constellation and a comparator then computes the difference in phase between this received signal and the preceding one. The difference encodes the data as described above.

The modulated signal is shown below for both DBPSK and DQPSK as described above. It is assumed that the signal starts with zero phase, and so there is a phase shift in both signals at $t\; =\; 0$.

Analysis shows that differential encoding approximately doubles the error rate compared to ordinary $M$-PSK but this may be overcome by only a small increase in $E\_b/N\_0$. Furthermore, this analysis (and the graphical results below) are based on a system in which the only corruption is additive white Gaussian noise. However, there will also be a physical channel between the transmitter and receiver in the communication system. This channel will, in general, introduce an unknown phase-shift to the PSK signal; in these cases the differential schemes can yield a better error-rate than the ordinary schemes which rely on precise phase information.

Example: Differentially encoded BPSK

At the $k^\{\backslash textrm\{th\}\}$ time-slot call the bit to be modulated $b\_k$, the differentially encoded bit $e\_k$ and the resulting modulated signal $m\_k(t)$. Assume that the constellation diagram positions the symbols at ±1 (which is BPSK). The differential encoder produces:

$\backslash ,e\_k\; =\; e\_\{k-1\}\backslash oplus\{\}b\_k$

where $\backslash oplus\{\}$ indicates binary or modulo-2 addition.

So $e\_k$ only changes state (from binary '0' to binary '1' or from binary '1' to binary '0') if $b\_k$ is a binary '1'. Otherwise it remains in its previous state. This is the description of differentially-encoded BPSK given above.

The received signal is demodulated to yield $e\_k=$±1 and then the differential decoder reverses the encoding procedure and produces:

$\backslash ,b\_k\; =\; e\_\{k\}\backslash oplus\{\}e\_\{k-1\}$ since binary subtraction is the same as binary addition.

Therefore, $b\_k=1$ if $e\_k$ and $e\_\{k-1\}$ differ and $b\_k=0$ if they are the same. Hence, if both $e\_k$ and $e\_\{k-1\}$ are inverted, $b\_k$ will still be decoded correctly. Thus, the 180° phase ambiguity does not matter.

Differential schemes for other PSK modulations may be devised along similar lines. The waveforms for DPSK are the same as for differentially-encoded PSK given above since the only change between the two schemes is at the receiver.

The BER curve for this example is compared to ordinary BPSK on the right. As mentioned above, whilst the error-rate is approximately doubled, the increase needed in $E\_b/N\_0$ to overcome this is small. The performance degradation is a result of noncoherent transmission - in this case it refers to the fact that tracking of the phase is completely ignored.

Differential Phase-shift Keying (DPSK)

---

For a signal that has been differentially encoded, there is an obvious alternative method of demodulation. Instead of demodulating as usual and ignoring carrier-phase ambiguity, the phase between two successive received symbols is compared and used to determine what the data must have been. When differential encoding is used in this manner, the scheme is known as differential phase-shift keying (DPSK). Note that this is subtly different to just differentially-encoded PSK since, upon reception, the received symbols are not decoded one-by-one to constellation points but are instead compared directly to one another.

Call the received symbol in the $k$^th timeslot $r\_k$ and let it have phase $\backslash phi\_k$. Assume without loss of generality that the phase of the carrier wave is zero. Denote the AWGN term as $n\_k$. Then

$r\_k\; =\; \backslash sqrt\{E\_s\}e^\{j\backslash phi\_k\}\; +\; n\_k$.

The decision variable for the $k-1$^th symbol and the $k$^th symbol is the phase difference between $r\_k$ and $r\_\{k-1\}$. That is, if $r\_k$ is projected onto $r\_\{k-1\}$, the decision is taken on the phase of the resultant complex number:

$r\_kr\_\{k-1\}^\{*\}\; =\; E\_se^\{j\backslash left(\backslash theta\_k\; -\; \backslash theta\_\{k-1\}\backslash right)\}\; +\; \backslash sqrt\{E\_s\}e^\{j\backslash theta\_k\}n\_\{k-1\}^\{*\}\; +\; \backslash sqrt\{E\_s\}e^\{-j\backslash theta\_\{k-1\}\}n\_k\; +\; n\_kn\_\{k-1\}$

where superscript * denotes complex conjugation. In the absence of noise, the phase of this is $\backslash theta\_\{k\}-\backslash theta\_\{k-1\}$, the phase-shift between the two received signals which can be used to determine the data transmitted.

The probability of error for DPSK is difficult to calculate in general, but, in the case of DBPSK it is:

$P\_b\; =\; \backslash frac\{1\}\{2\}e^\{-E\_b/N\_0\}$,

which, when numerically evaluated, is only slightly worse than ordinary BPSK, particularly at higher $E\_b/N\_0$ values.

Using DPSK avoids the need for possibly complex carrier-recovery schemes to provide an accurate phase estimate and can be an attractive alternative to ordinary PSK.

In optical communications, the data can be modulated onto the phase of a laser in a differential way. The modulation is a laser which emits a continuous wave, and a Mach-Zehner modulator which receives electrical binary data. For the case of BPSK for example, the laser transmits the field unchanged for binary '1', and with reverse polarity for '0'. The demodulator consists of a delay interferometer which delays one bit, so two bits can be compared at one time. There is also a photo diode to transform the optical field into an electric current, so the information is changed back into its original state. This technology is used in the optical in/out ports of some sound cards.

The bit-error rates of DBPSK and DQPSK are compared to their non-differential counterparts in the graph to the right. The loss for using DBPSK is small enough compared to the complexity reduction that it is often used in communications systems that would otherwise use BPSK. For DQPSK though, the loss in performance compared to ordinary QPSK is larger and the system designer must balance this against the reduction in complexity.

See also

---

• Differential coding
• Modulation — for an overview of all modulation schemes
• Phase modulation (PM) — the analogue equivalent of PSK

Notes

---

References

---

The notation and theoretical results in this article are based on material prsented in the following sources:

Radio modulation modes | Data transmission

Quadrature Phase Shift Keying | QPSK | Quadrature phase-shift keying | Kvadratúra fázisbillentyűzés | デジタル変調 | BPSK | QPSK