Confusing Results when setting different fMin for SEOBNR waveforms

Data Analysis
1

Hi,
while working on some practice tasks using pyCBC, I encountered something unexpected when trying to generate several FD waveforms by converting waveforms originally in TD into FD using to_frequencies() method.
I generated several SEOBNRv4P waveforms with different fMin settings and converted them into frequency domain. While I expected them to behave in a different manner in the frequency range [10Hz,20Hz], I noticed an unusual difference at a frequency around 1024Hz, and I can’t understand why this should happen, since I have set sampling rate srate to be 16384Hz.
I suppose this is because of some issues that I didn’t noticed during my conversion?
Below is the figure of these waveforms and the code I used to generate the waveforms.
64

>>> from pycbc.waveform import get_td_waveform, fd_approximants
>>> hpe, hce = get_td_waveform(approximant="SEOBNRv4P",
...                       mass1=10,
...                       mass2=10,
...                     spin1x = 0.2,
...                           spin2y=-0.3,
...                       delta_t=1.0/16384,
...                       f_lower=20,
...                       distance=100,)
>>> hpef =hpe.to_frequencyseries()
>>> plt.loglog(hpef.sample_frequencies, hpef.squared_norm(), label=f'SEOBNRv4P')

>>> hpe, hce = get_td_waveform(approximant="SEOBNRv4P",
...                       mass1=10,
...                       mass2=10,
...                   spin1x = 0.2,
...                           spin2y=-0.3,
...                       delta_t=1.0/16384,
...                       f_lower=10,
... 
...                       distance=100,)
>>> hpef =hpe.to_frequencyseries()
>>> plt.loglog(hpef.sample_frequencies, hpef.squared_norm(), label=f'SEOBNRv4P-2')

>>> hpe, hce = get_td_waveform(approximant="SEOBNRv4PHM",
...                       mass1=10,
...                       mass2=10,
...                   spin1x = 0.2,
...                           spin2y=-0.3,
...                       delta_t=1.0/16384,
...                       f_lower=10,
... 
...                       distance=100,)
>>> hpef =hpe.to_frequencyseries()
>>> plt.loglog(hpef.sample_frequencies, hpef.squared_norm(), label=f'SEOBNRv4P-3')

Ps: the waveform labled as “Plus Polarization” refers to SEOBNRE, which should be mostly equivalent to SEOBNRv4PHM when eccentricity e0=0.
Thanks for help!