Q-transform projections do not show chirp signature

Data Analysis
1

I am simulating a few microlensed GW signals and trying to get their QTs using gwpy, but many signals do not show the “chirp” signature even after changing various settings of the q_transform function.

For example, I have this 8s long GW signal. I use different combinations of settings in the q_transform function (changing the qrange, outseg, whiten, norm values) :

Example setting:
ts = TimeSeries(strain_values, times=time_values, sample_rate=1.0/(time_values[1] - time_values[0]))
qt1 = ts.q_transform(qrange=(4,15), frange=(20,512),whiten=True,highpass=True,outseg=(-0.5,0.05))
qt2 = ts.q_transform(qrange=(4,15), frange=(20,512),whiten=True,highpass=True, norm=‘mean’, outseg=(-0.5,0.05))
qt3 = ts.q_transform(qrange=(4,15), frange=(20,512),whiten=False, outseg=(-0.5,0.05))
qt4 = ts.q_transform(qrange=(4,15), frange=(20,512), whiten=False, norm=‘mean’, outseg=(-0.5,0.05))

ML_00041_A
ML_00041_A1200×600 247 KB

The other combinations can be seen here

Yet, the QTs show just noise and there is no chirp signature seen.
I am using the advanced LIGO H1 (O4) PSDs for noise.

There a few other microlensed signals generated in a similar way with slightly different binary parameters which show good QT projections (clear chirp sign seen). On the other hand, I face the similar ‘no-chirp seen’ problem with a few unlensed precessing binary GW signals generated.

How do I resolve this issue?

2

Adjust frange and qrange that the frequency range is wide enough to capture the full chirp signal. Consider increasing the frange to capture lower or higher frequencies, or narrowing the qrange if the signal is highly resonant.

Try Alternative Whitening Settings with whiten=False or different normalization schemes to see if it improves the visibility of the chirp.

Modify the outseg that the analysis window includes the entire duration of the signal and isn’t cutting off key parts of the chirp.

Recheck PSD toconfirm that the noise model you’re using accurately reflects the expected noise for your simulation. If necessary, experiment with other noise models or adjust the simulation parameters.

if possible, increase the duration of the signal or boost its amplitude to improve the SNR, which may help reveal the chirp signature.

3

Hi,

Thank you for the question!

Not all signals will be visible in Q-transforms. It depends both on the signal-to-noise ration (SNR) and the mass of the signal. In general, signals with higher SNR are easier to see, and signals with lower chirp mass are easier to see.

We include Q-transforms of signals in the Event Portal. By browsing events with different SNRs, you can get an idea of what SNRs are visible.

Do you know the SNR and total mass of your injected signal?

https://gwosc.org/eventapi/html/GWTC/

4

Hi, thanks for the reply.
The SNR for all the simulated signals are greater than 20 and I still don’t see QT projections for some of them, which is a bit odd (?)

Using whiten=False, norm=‘mean’ seems to work for some of these cases (when injected in gaussian noise) but not all. Similarly, changing the qrange to (4,15) with all other settings default, works for some cases (when injected in real detector O3 noise ) but not all.

There are cases where a signal with SNR 20.3 shows a neat QT but a signal with SNR 23.5 does not. So I am not sure what’s happening here.

5

Please be aware there is also a dependence on mass.

For lower mass signals (total mass < 10), the signals are spread out in the TF plane, and you don’t necessarily expect to see the track at SNR 20. Also, for lower mass signals, the Q value needs to be higher (Q > 50).

Do you know the mass and spin parameters of your injected signal?