Difference between revisions of "MPSK SNR Estimator"

From GNU Radio
Jump to: navigation, search
Line 1: Line 1:
A block for computing SNR of a M-PSK signal (e.g., BPSK, QPSK, 8-PSK).  It copies the input stream to the output stream, but adds a tag that contains the estimated SNR every N samples.  It may work with other types of signals, but with a certain amount of error.
+
A block for computing SNR of a M-PSK signal (e.g., BPSK, QPSK, 8-PSK).  It copies the input stream to the output stream, but adds a tag that contains the estimated SNR every N samples.  It may work with other types of signals, but with a certain amount of error. See each Type for how it works under the hood.
 
 
Under the hood, the block uses 2nd (M2) and 4th (M4) order moments. This estimator uses knowledge of the kurtosis of the signal (𝑘𝑎) and noise (𝑘𝑤) to make its estimation. We use Beaulieu's approximations here to M-PSK signals and AWGN channels such that 𝑘𝑎=1 and 𝑘𝑤=2. These approximations significantly reduce the complexity of the calculations (and computations) required.
 
 
 
Reference: D. R. Pauluzzi and N. C. Beaulieu, "A comparison of SNR estimation techniques for the AWGN channel," IEEE Trans. Communications, Vol. 48, No. 10, pp. 1681-1691, 2000.
 
 
 
  
 
== Parameters ==
 
== Parameters ==
  
'''Type''' -  
+
'''Type''' - There are currently four implemented estimators:
 
 
Below are some ROUGH estimates of what values of SNR each of these types of estimators is good for. In general, these offer a trade-off between accuracy and performance.
 
  
# Simple: >= 7 dB
+
# Simple: A very simple SNR estimator that just uses mean and variance estimates of an M-PSK constellation. This esimator is quick and cheap and accurate for high SNR (above 7 dB or so) but quickly starts to overestimate the SNR at low SNR.
# Skewness: >= 5 dB
+
# Skewness: SNR Estimator using skewness correction.  This is an estimator that came from a discussion between Tom Rondeau and fred harris with no known paper reference. The idea is that at low SNR, the variance estimations will be affected because of fold-over around the decision boundaries, which results in a skewness to the samples. We estimate the skewness and use this as a correcting term.  Best used with SNRs above 5 dB.
# 2nd and 4th Moment: >= 1 dB
+
# 2nd and 4th Moment: An SNR estimator for M-PSK signals that uses 2nd (M2) and 4th (M4) order moments. This estimator uses knowledge of the kurtosis of the signal (k_a) and noise (k_w) to make its estimation. We use Beaulieu's approximations here to M-PSK signals and AWGN channels such that k_a=1 and k_w=2. These approximations significantly reduce the complexity of the calculations (and computations) required.  Works best for SNR above 1 dB.  Reference: D. R. Pauluzzi and N. C. Beaulieu, "A comparison of SNR estimation techniques for the AWGN channel," IEEE Trans. Communications, Vol. 48, No. 10, pp. 1681-1691, 2000.
# SVR: >= 0dB
+
# SVR: Signal-to-Variation Ratio SNR Estimator.  This estimator actually comes from an SNR estimator for M-PSK signals in fading channels, but this implementation is specifically for AWGN channels. The math was simplified to assume a signal and noise kurtosis (k_a and k_w) for M-PSK signals in AWGN. These approximations significantly reduce the complexity of the calculations (and computations) required.  Works best for SNR above 0 dB.  Original paper: A. L. Brandao, L. B. Lopes, and D. C. McLernon, "In-service monitoring of multipath delay and cochannel interference for indoor mobile communication systems," Proc. IEEE Int. Conf. Communications, vol. 3, pp. 1458-1462, May 1994.
  
 
See https://www.gnuradio.org/doc/doxygen-3.6.4/group__snr__blk.html for more details
 
See https://www.gnuradio.org/doc/doxygen-3.6.4/group__snr__blk.html for more details

Revision as of 21:03, 12 March 2019

A block for computing SNR of a M-PSK signal (e.g., BPSK, QPSK, 8-PSK). It copies the input stream to the output stream, but adds a tag that contains the estimated SNR every N samples. It may work with other types of signals, but with a certain amount of error. See each Type for how it works under the hood.

Parameters

Type - There are currently four implemented estimators:

  1. Simple: A very simple SNR estimator that just uses mean and variance estimates of an M-PSK constellation. This esimator is quick and cheap and accurate for high SNR (above 7 dB or so) but quickly starts to overestimate the SNR at low SNR.
  2. Skewness: SNR Estimator using skewness correction. This is an estimator that came from a discussion between Tom Rondeau and fred harris with no known paper reference. The idea is that at low SNR, the variance estimations will be affected because of fold-over around the decision boundaries, which results in a skewness to the samples. We estimate the skewness and use this as a correcting term. Best used with SNRs above 5 dB.
  3. 2nd and 4th Moment: An SNR estimator for M-PSK signals that uses 2nd (M2) and 4th (M4) order moments. This estimator uses knowledge of the kurtosis of the signal (k_a) and noise (k_w) to make its estimation. We use Beaulieu's approximations here to M-PSK signals and AWGN channels such that k_a=1 and k_w=2. These approximations significantly reduce the complexity of the calculations (and computations) required. Works best for SNR above 1 dB. Reference: D. R. Pauluzzi and N. C. Beaulieu, "A comparison of SNR estimation techniques for the AWGN channel," IEEE Trans. Communications, Vol. 48, No. 10, pp. 1681-1691, 2000.
  4. SVR: Signal-to-Variation Ratio SNR Estimator. This estimator actually comes from an SNR estimator for M-PSK signals in fading channels, but this implementation is specifically for AWGN channels. The math was simplified to assume a signal and noise kurtosis (k_a and k_w) for M-PSK signals in AWGN. These approximations significantly reduce the complexity of the calculations (and computations) required. Works best for SNR above 0 dB. Original paper: A. L. Brandao, L. B. Lopes, and D. C. McLernon, "In-service monitoring of multipath delay and cochannel interference for indoor mobile communication systems," Proc. IEEE Int. Conf. Communications, vol. 3, pp. 1458-1462, May 1994.

See https://www.gnuradio.org/doc/doxygen-3.6.4/group__snr__blk.html for more details

Samples between tags - after this many samples, a tag containing the SNR (key='snr') will be sent on the output port.

Filter Alpha - the update rate of internal running average calculations.