A Costas loop is a phase-locked loop (PLL) based circuit which is used for carrier frequency recovery from suppressed-carrier modulation signals (e.g. double-sideband suppressed carrier signals) and phase modulation signals (e.g. BPSK, QPSK). It was invented by John P. Costas at General Electric in the 1950s. Its invention was described as having had "a profound effect on modern digital communications". The primary application of Costas loops is in wireless receivers. Its advantage over the PLL-based detectors is that at small deviations the Costas loop error voltage is as compared to . This translates to double the sensitivity and also makes the Costas loop uniquely suited for tracking Doppler-shifted carriers especially in OFDM and GPS receivers.
In the classical implementation of a Costas loop, a local voltage-controlled oscillator (VCO) provides quadrature outputs, one to each of two phase detectors, e.g., product detectors. The same phase of the input signal is also applied to both phase detectors and the output of each phase detector is passed through a low-pass filter. The outputs of these low-pass filters are inputs to another phase detector, the output of which passes through noise-reduction filter before being used to control the voltage-controlled oscillator. The overall loop response is controlled by the two individual low-pass filters that precede the third phase detector while the third low-pass filter serves a trivial role in terms of gain and phase margin.
The above figure of a Costas loop is drawn under the condition of the "locked" state, where the VCO frequency and the incoming carrier frequency have become the same as a result of the Costas loop process. The figure does not represent the "unlocked" state.
In the time domain
In the simplest case . Therefore, does not affect the input of noise-reduction filter. Carrier and voltage-controlled oscillator (VCO) signals are periodic oscillations with high-frequencies . Block is an analog multiplier.
From the mathematical point of view, a linear filter can be described by a system of linear differential equations
Here, is a constant matrix, is a state vector of filter, and are constant vectors.
The model of a VCO is usually assumed to be linear
where is a free-running frequency of voltage-controlled oscillator and is an oscillator gain. Similarly, it is possible to consider various nonlinear models of VCO.
Suppose that the frequency of master generator is constant Equation of VCO and equation of filter yield
The system is non-autonomous and rather difficult for investigation.
In phase-frequency domain
In the simplest case, when
the standard engineering assumption is that the filter removes the upper sideband with frequency from the input but leaves the lower sideband without change. Thus it is assumed that VCO input is This makes a Costas loop equivalent to a phase-locked loop with phase detector characteristic corresponding to the particular waveforms and of input and VCO signals. It can be proved that filter outputs in time domain and phase-frequency domain are almost equal.  
Thus it is possible  to study more simple autonomous system of differential equations
The Krylov–Bogoliubov averaging method allows one to prove that solutions of non-autonomous and autonomous equations are close under some assumptions. Thus the block-scheme of Costas Loop in the time space can be asymptotically changed to the block-scheme on the level of phase-frequency relations.
The passage to analysis of autonomous dynamical model of Costas loop (in place of the non-autonomous one) allows one to overcome the difficulties, related with modeling Costas loop in time domain where one has to simultaneously observe very fast time scale of the input signals and slow time scale of signal's phase. This idea makes it possible to calculate core performance characteristics - hold-in, pull-in, and lock-in ranges.
The classical Costas loop will work towards making the phase difference between the carrier and the VCO become a small, ideally zero, value. The small phase difference implies that frequency lock has been achieved.
QPSK Costas loop
The input QPSK signal is as follows
Inputs of low-pass filters LPF1 and LPF2 are
After synchronization outputs of LPF1 and LPF2 are used to get demodulated data ( and ). To adjust frequency of VCO to reference frequency signals and goes through limiters and cross-multiplied:
After that signal is filtered by Loop filter and forms tuning signal for VCO similar to BPSK Costas loop. Thus, QPSK Costas can be described by system of ODEs
Here - parameters of LPF1 and LPF2, and - parameters of Loop filter.
- Costas, John P. (1956). "Synchronous communications". Proceedings of the IRE. 44 (12): 1713–1718. doi:10.1109/jrproc.1956.275063.
- Costas, John P. (August 2002) . "Synchronous Communications". Proceedings of the IEEE. 90 (8): 1461–1466. doi:10.1109/JPROC.2002.800876.
- Taylor, D. (August 2002). "Introduction to `Synchronous Communications', A Classic Paper by John P. Costas" (PDF). Proceedings of the IEEE. 90 (8): 1459–1460. doi:10.1109/jproc.2002.800719.
- Feigin, Jeff (January 1, 2002). "Practical Costas loop design" (PDF). RF Design: 20–36.
- Leonov, G. A.; Kuznetsov, N. V.; Yuldashev, M. V.; Yuldashev, R. V. (August 2012). "Differential equations of Costas loop" (PDF). Doklady Mathematics. 86 (2): 723–728. doi:10.1134/s1064562412050080.
- Leonov, G. A.; Kuznetsov, N. V.; Yuldashev, M. V.; Yuldashev, R. V. (2012). "Analytical method for computation of phase-detector characteristic" (PDF). IEEE Transactions on Circuits and Systems Part II. 59 (10): 633–637. doi:10.1109/tcsii.2012.2213362.
- Leonov, G. A.; Kuznetsov, N. V.; Yuldashev, M. V.; Yuldashev, R. V. (2015). "Nonlinear dynamical model of Costas loop and an approach to the analysis of its stability in the large". Signal processing. Elsevier. 108: 124–135. doi:10.1016/j.sigpro.2014.08.033.
- Kuznetsov, N. V.; Leonov, G. A.; Neittaanmaki, P.; Seledzhi, S. M.; Yuldashev, M. V.; Yuldashev, R. V. (2012). "Nonlinear mathematical models of Costas Loop for general waveform of input signal". IEEE 4th International Conference on Nonlinear Science and Complexity, NSC 2012 – Proceedings. IEEE Press (6304729): 75–80. doi:10.1109/NSC.2012.6304729. ISBN 978-1-4673-2703-9.
- Kuznetsov, N. V.; Leonov, G. A.; Seledzhi, S. M.; Yuldashev, M. V.; Yuldashev, R. V. (2017). "Nonlinear model of the optical Costas loop: pull-in range estimation and hidden oscillations". IFAC-PapersOnLine. ELSEVIER. 50: 3325–3330. doi:10.1016/j.ifacol.2017.08.514. ISSN 2405-8963.
- Costas 1956 states, "The local oscillator must be maintained at proper phase so that the audio output contributions of the upper and lower sidebands reinforce one another. If the oscillator phase is 90° away from the optimum value a null in audio output will result which is typical of detectors of this type. The actual method of phase control will be explained shortly, but for the purpose of this discussion maintenance of correct oscillator phase shall be assumed."
- Using a loop filter with an integrator allows a steady-state phase error of zero. See PID controller § Integral term.
- Best, Roland E. (1997). Phase-Locked Loops (third ed.). New York: McGraw-Hill. pp. 44–45. ISBN 0-07-006051-7.
- US Grant 4,085,378, Carl R. Ryan & James H. Stilwell, "QPSK demodulator", published 1976-11-26, assigned to Motorola Solutions Inc
- Best, R. E.; Kuznetsov, N. V.; Leonov, G. A.; Yuldashev, M. V.; Yuldashev, R. V. (2016). "Tutorial on dynamic analysis of the Costas loop". Annual Reviews in Control. ELSEVIER. 42: 27–49. doi:10.1016/j.arcontrol.2016.08.003.