Ideal-Spectrum RV Recovery
Recover an injected RV from the rest-frame BT-Settl lte040 (4000 K)
template using the packaged full-band *.ccfopt.delta.mas mask (~887 lines)
and recover_stellar_rv(). The same
-50…+50 km/s grid (1 km/s, 101 samples) is run on three independent paths.
Run the preamble and section 1 first, then any of sections 2–4. Each path
prints the same recovery table and residual plot (cm/s; mean and np.std
with ddof=0).
Native L2 (section 2):
build_l2()onto the echelle WAVE grid. Fast; no KOA cache.Rate map → Stage 1 (section 3): native rate map, wrap as L0, extract and blaze.
Rate map → UTR → Stages 0–2 (section 4): same as (2) through a simulated ramp and Stage 0. Slowest.
Pipeline logs are suppressed during each sweep. For an LSF-matched closed loop,
set APPLY_LSF = True and pass a full-band *.lsf.ccfopt.delta.mas from
ensure_lsf_ccfopt_mask(). The packaged
lte040...lsf.ccfopt.delta.mas is a two-order smoke file and is not used
here.
The HISPEC DRP is in active development. Please submit issues and suggestions on the GitHub.
Imports and output directory
from pathlib import Path
import hispecdrp
print(f"hispecdrp {hispecdrp.__version__} @ {hispecdrp.__file__}")
OUT = Path("/path/to/ideal_spectrum_rv_recovery")
OUT.mkdir(parents=True, exist_ok=True)
print("output:", OUT.resolve())
1. Configure template and mask
Packaged lte040 template, native *.ccfopt.delta.mas mask, and the -50…+50
km/s grid. run_injected_rv_grid prints the recovery table and residual plot
used by sections 2–4.
import numpy as np
from hispecdrp.simulations.plotting import plot_ideal_spectrum_rv_recovery
from hispecdrp.simulations.spectra import quiet_pipeline_logging
from hispecdrp.utils.line_mask_utils import (
paired_line_mask_path_for_template,
read_mask_generation_meta,
)
FIBER = 1
BULK_RV_KM_S = np.arange(-50.0, 51.0, 1.0)
TEMPLATE = "lte040.0-4.5-0.0a+0.0.BT-Settl.spec.7.h5"
APPLY_LSF = False
MASK_PATH = paired_line_mask_path_for_template(TEMPLATE, mask_style="ccfopt.delta")
HDR = (
f"{'bulk_km/s':>10} {'injected_m/s':>14} {'recovered_m/s':>16} "
f"{'rv_err_m/s':>12} {'bias_m/s':>12} {'bias_mm/s':>10}"
)
meta = read_mask_generation_meta(MASK_PATH) or {}
print(f"mask: {MASK_PATH}")
print(
f"n_lines={meta.get('n_lines')} n_seed={meta.get('n_seed_lines')} "
f"n_drop={meta.get('n_dropped_lines')}"
)
print(
f"n_bulk={len(BULK_RV_KM_S)} "
f"grid={BULK_RV_KM_S[0]:.0f}..{BULK_RV_KM_S[-1]:.0f} km/s step 1"
)
def run_injected_rv_grid(recover_one, *, title, out_path):
inj_m_s, rec_m_s, err_m_s = [], [], []
print(HDR)
print("-" * len(HDR))
for bulk in BULK_RV_KM_S:
with quiet_pipeline_logging():
inj, rec, err = recover_one(bulk)
bias = rec - inj
inj_m_s.append(float(inj))
rec_m_s.append(float(rec))
err_m_s.append(float(err))
print(
f"{bulk:10.1f} {inj:14.4f} {rec:16.6f} {err:12.6f} "
f"{bias:12.6e} {bias * 1e3:10.4f}"
)
inj_m_s = np.asarray(inj_m_s, dtype=float)
rec_m_s = np.asarray(rec_m_s, dtype=float)
err_m_s = np.asarray(err_m_s, dtype=float)
bias_m_s = rec_m_s - inj_m_s
max_abs = float(np.max(np.abs(bias_m_s)))
rms = float(np.sqrt(np.mean(bias_m_s**2)))
d_bias = bias_m_s - bias_m_s[0]
max_d = float(np.max(np.abs(d_bias[1:]))) if bias_m_s.size > 1 else 0.0
print()
print(f"max|bias| = {max_abs:.6e} m/s ({max_abs * 1e3:.4f} mm/s)")
print(f"rms bias = {rms:.6e} m/s ({rms * 1e3:.4f} mm/s)")
print(
f"max|differential bias| (vs RV=0) = {max_d:.6e} m/s "
f"({max_d * 1e3:.4f} mm/s)"
)
summary = plot_ideal_spectrum_rv_recovery(
inj_m_s,
rec_m_s,
err_m_s,
title=title,
out_path=out_path,
)
if summary.get("out_path"):
print("wrote", summary["out_path"])
print(
f"mean bias = {summary['mean_bias_cm_s']:.3f} cm/s "
f"std = {summary['std_bias_cm_s']:.3f} cm/s"
)
return summary
2. Native L2 injection
build_l2() paints the Doppler-shifted
template onto a native BSPEC L2, then
recover_stellar_rv() recovers the RV.
Requires the preamble and section 1 only.
from hispecdrp.simulations.spectra import build_l2, recover_stellar_rv
def recover_native(bulk):
l2 = build_l2(
bulk,
template=TEMPLATE,
fiber=FIBER,
apply_lsf=APPLY_LSF,
)
return recover_stellar_rv(
l2,
bulk,
template=TEMPLATE,
mask_path=str(MASK_PATH),
spectrograph="BSPEC",
fiber=FIBER,
)
summary_native = run_injected_rv_grid(
recover_native,
title="BT-Settl lte040 (native L2): injected vs recovered RV",
out_path=OUT / "lte040_native_l2_simulated_vs_recovered_rv.png",
)
3. Rate map → Stage 1
Paint a native rate map, wrap it as L0, box-extract and blaze with simulated
Stage-1 calibrations, then recover RV. The first point builds masters under
OUT; later points reuse them. Requires the preamble and section 1 only.
from hispecdrp.simulations.ideal_spectrum_rv import recover_rv_from_rate_map
def recover_stage1(bulk):
return recover_rv_from_rate_map(
bulk,
template=TEMPLATE,
mask_path=str(MASK_PATH),
output_dir=OUT / "rate_map_stage1",
fiber=FIBER,
)
summary_s1 = run_injected_rv_grid(
recover_stage1,
title="BT-Settl lte040 (rate map -> Stage 1): injected vs recovered RV",
out_path=OUT / "lte040_rate_map_stage1_simulated_vs_recovered_rv.png",
)
4. Rate map → UTR → Stages 0–2
Same grid through a simulated up-the-ramp exposure and Stage 0, then the section 3 extract / blaze / CCF path. Poisson noise is off. Each iteration deletes its working directory after the RV is recorded. Requires the preamble and section 1 only.
from hispecdrp.simulations.ideal_spectrum_rv import recover_rv_from_ramp
def recover_stage0(bulk):
return recover_rv_from_ramp(
bulk,
template=TEMPLATE,
mask_path=str(MASK_PATH),
output_dir=OUT / "ramp_stage0",
fiber=FIBER,
)
summary_s0 = run_injected_rv_grid(
recover_stage0,
title="BT-Settl lte040 (rate map -> UTR -> Stages 0-2): injected vs recovered RV",
out_path=OUT / "lte040_rate_map_stage0_simulated_vs_recovered_rv.png",
)