Create Simulated HISPEC Data

This example shows how to create simulated HISPEC data using hispecdrp.simulations.simulate_rate_map and hispecdrp.simulations.simulate_up_the_ramp. For the full end-to-end pipeline (ramp → Stage 0–2 → RV), see Basic Usage.

Note

Older examples referenced a benchmarks/ package at the repository root; the same APIs now live under hispecdrp.simulations.

Simulated rate maps can be created using a wrapper for the pyechelle package, or using a native DRP simulator. pyechelle uses raytracing of individual photons and naturally includes photon noise, whereas the native simulator creates noiseless rate maps. pyechelle takes longer to run for brighter targets (more photons to trace), and the native simulator is expected to be faster for all but the faintest sources.

All simulations require downloading model PSF and trace data (~2 GB). This will be done automatically the first time you run either simulator. By default, the data is stored in the same path as the testing data.

To create simulated ramp data suitable for input to the stage0 pipeline:

  • Set each fiber with a desired spectrum. Note that for continuous sources, the spectral sampling must be greater than HISPEC’s spectral resolution.

  • Create rate maps using either the pyechelle wrapper or the native DRP simulator.

  • Create ramp data from the rate maps using the create_ramp function. To create ramp data, you will also need input flat, dark, gain maps etc.

Below is an example using the native DRP simulator to create a rate map for a flat spectrum source on fiber 1, and an emission line source on fiber 2.

import numpy as np
import astropy.units as u
import hispecdrp.simulations as simulations

srm = simulations.SimulateRateMap()

fiber_1_wavelengths = np.arange(0.9,2.5,2.5e-6)* u.micron
fiber_1_flux = np.ones_like(fiber_1_wavelengths.value)*1e-12 * u.erg/u.s/u.cm**2/u.micron
srm.set_fiber_spectrum(fiber_number=1,
                   spectrum_wavelengths=fiber_1_wavelengths,
                   spectrum_flux=fiber_1_flux,
                  continuous=True)

fiber_2_wavelengths = np.arange(0.9,2.5,2.5e-4)* u.micron
fiber_2_flux = np.ones_like(fiber_2_wavelengths.value)*1e-9 * u.erg/u.s/u.cm**2/u.micron
srm.set_fiber_spectrum(fiber_number=2,
                    spectrum_wavelengths=fiber_2_wavelengths,
                    spectrum_flux=fiber_2_flux,
                    continuous=False)

native_rate_map = srm.create_native_rate_map(spectrograph='BSPEC', method='prf_convolution')

If you then want to create ramp data from the rate map:

max_cores = 1
max_slope = 50  # e-/s
readtime = 1.7  # seconds
n_reads = 10
ny, nx = 4096, 4096 # size of the detector
nonlin_coeffs = [-1e-6, 1, 0]  # Example coefficients for a cubic non-linearity

gain = 1.0  # e-/ADU
gain_map = np.full((ny, nx), gain, dtype=np.float32)
detflat_map = np.random.normal(loc=1, scale=0.1, size=(ny, nx)).astype(np.float32)

bias = 1000  # e-/s
bias_map = np.full((ny, nx), bias, dtype=np.float32)

dark_current = 10  # e-/s
dark_map = np.random.normal(loc=dark_current, scale=dark_current / 10, size=(ny, nx)).astype(np.float32)

ramp_data = simulations.create_ramp(native_rate_map,  # in e-/s
                    readtime=readtime, n_reads=n_reads,
                    nonlin_coeffs=nonlin_coeffs,
                    gain=gain_map,  # in e-/s
                    flat=detflat_map,
                    dark=dark_map,  # in e-/s
                    bias=bias_map,  # in e-/s
                    kTC_noise=50,  # in e-/s
                    poisson_noise=False, read_noise=0,  # in e-/s
                    convert_to_uint16=False, clip_ramps=True,
                    max_cores=max_cores)