Root Raised Cosine Filter: Difference between revisions

From GNU Radio
Jump to navigation Jump to search
(adding a public domain image)
(→‎Parameters: explanation of better parameters)
Line 22: Line 22:


; Symbol Rate (''R'')
; Symbol Rate (''R'')
: Symbol rate, must be a factor of sample rate. Typically ((samples/second) / (samples/symbol)).
: The baud rate. The RRCS seems to work best setting this parameter to ''baud_rate * 2''


; Alpha (''R'')
; Alpha (''R'')
: Excess bandwidth factor, also known as alpha. (default: 0.35)
: Excess bandwidth factor, also known as alpha. (default: 0.35) With a very small alpha, say 0.001, the RRCF block produces sinusoidal-shaped symbols and culls more noise, whereas a large alpha produces more flat-topped symbols.


; Num Taps (''R'')
; Num Taps (''R'')
: Number of taps (default: 11*samp_rate)
: Number of taps. This defaults to ''11*samp_rate'', but this default is not actually in the code and such a default produces extremely high CPU costs. If the number of samples per symbol is known, a better default would be ''working_sps * num_symbols'' which extends the filter to rolloff across ''num_symbols'', say 32.


== Example Flowgraph ==
== Example Flowgraph ==

Revision as of 19:11, 1 March 2023

The Root Raised Cosine Filter blocks acts as a matched filter. Its primarily purpose is extracting a known digital signal out of noise; it does this more effectively than a low-pass filter, squelch, or other blocks.

Root Raised Cosine Filter implements match filtering to pull square pulses out of an channel containing artificial Gaussian white noise.

This filter is a convenience wrapper for a FIR filter and a firdes taps generating function.

Parameters

(R): Run-time adjustable

FIR Type
options: [Complex->Complex (Decimating), Complex->Complex (Interpolating), Float->Float (Decimating), Float->Float (Interpolating)]
Decimation or Interpolation
decimation or interpolation factor (depending on which type is chosen above).
Gain (R)
Overall gain of filter (default 1.0)
Sample Rate (R)
Sample rate in samples per second.
Symbol Rate (R)
The baud rate. The RRCS seems to work best setting this parameter to baud_rate * 2
Alpha (R)
Excess bandwidth factor, also known as alpha. (default: 0.35) With a very small alpha, say 0.001, the RRCF block produces sinusoidal-shaped symbols and culls more noise, whereas a large alpha produces more flat-topped symbols.
Num Taps (R)
Number of taps. This defaults to 11*samp_rate, but this default is not actually in the code and such a default produces extremely high CPU costs. If the number of samples per symbol is known, a better default would be working_sps * num_symbols which extends the filter to rolloff across num_symbols, say 32.

Example Flowgraph

Frequency-Shift Keying

If you have demodulated Frequency-Shift Keying signal using a Quadrature Demod block, and you know the baud rate of the signal, and thus the duration of each symbol, the RRCS is useful for pulling these symbols out of the noise. In the process, the RRCS generates sinusoidal waves in place of the original square-wave symbols, rising from and falling to 0 and peaking at the center of each symbol. This result is extremely advantageous for clock recovery, as blocks like Symbol Sync need to lock onto the center of each symbol in order to reliably extract bits. The RRCS accomplishes both steps at once.

Waveform Shaping

This flowgraph can be found at [1]. Two Root Raised Cosine Filters in series produce a Raised Cosine Filter which is used for keying waveform shaping, thereby reducing key clicks.

MorseGen xmt fg.png

Source Files

C++ files
[2]
Header files
TODO
Public header files
Taps creation
Filter definition
Block definition
[3]