Function: time_inject_chirp()

0 void time_inject_chirp(float c0, float c90, int offset, float invMpc, float* chirp0, float* chirp90, float* data, float *response, float *work, int n)
This is a time-domain version of the previous function freq_inject_chirp() which injects chirps in the time-domain (after deconvolving them with the detector's response function). This routine injects artificial signals into the time-domain strain $h(t)$. The plane of the binary system is assumed to be normal to the line to the detector.

The arguments are:

c0: Input. The coefficient of the 0-phase template to inject.

c90: Input. The coefficient of the $90^\circ$-phase to inject. Note that $c_0^2 + c_{90}^2$ should be 1.

offset: Input. The offset number of samples at which the injected chirp starts, in the time domain.

invMpc: Input. The inverse of the distance to the system (measured in Mpc).

chirp0: Input. The time-domain phase-0 chirp (strain units) at a distance of 1 Mpc.

chirp90: Input. The time-domain phase-90 chirp (strain units) at a distance of 1 Mpc.

data: Output. The detector response in time that would be produced by the specified binary inspiral. Note that this routine adds into and increments this array, so that if it contains another ``signal" like IFO noise, the chirp is simply super-posed onto it.

response: Input. The function $R(f)$ that specifies the response function of the IFO. This is produced by the routine normalize_gw().

work: Output. A working array.

n: Input. Defines the lengths of the various arrays chirp0[0..n-1], chirp90[0..n-1], data[0..n-1], work[0..n-1], and response[0..n+1] (note that this "+" sign is not a typo!).

Note that in making use of this injection routine, you must determine the level of the quantization noise of the ADC, and be careful to inject a properly dithered version of this signal when its amplitude is small compared to the ADC quantization step size.

Author: Bruce Allen, ballen@dirac.phys.uwm.edu

Comments: A short look at the time-domain signal which is injected shows that it has a low-amplitude spike at the very start. This may be an un-avoidable Gibbs phenomenon associated with the turn-on of the waveform. A second interesting point is that for many interesting signals, the amplitude of the injected signal in the time domain is below the level of the quantization noise. Thus, a sensible injection scheme would be to add it into an appropriately dithered (float) version of the integer signal stream, then cast that back into an integer. This should be tried.