Learning objectives
- Propagate independent fractional calibration terms in quadrature.
- Separate thermal noise from multiplicative and spatially correlated systematics.
- Design residual and jackknife checks for a calibrated image.
Prerequisites in use: ASTR 340; PHYS 341 electromagnetism is recommended for receiver and radiation physics.
Terms Developed in Context
These terms connect the measured radio product to the physical model and delimit what the observation can support.
- Thermal noise
Random receiver and sky fluctuations that average down approximately with the square root of independent samples. In Calibration II: Error Budgets and Imaging Systematics, it helps resolve the question
Two sources share the same 5% flux-scale error but have independent 1% thermal errors. Which uncertainty cancels in their flux ratio, and which does not?
Use in the reasoning chain: The term links the observable or diagnostic
Independent small fractional errors combine in quadrature, whereas shared calibration errors correlate many image pixels
to the model \(\left(\frac{\sigma_S}{S}\right)^2=\sum_j f_j^2+2\sum_{i<j}\rho_{ij}f_i f_j\). Compare it directly with Flux-scale uncertainty rather than treating either quantity in isolation.Scope: Quadrature is justified only for independent zero-mean errors; common-mode flux-scale uncertainty requires covariance-aware comparisons. The term therefore supports an inference only when its measurement convention, calibration stage, and stated approximation are retained.
- Flux-scale uncertainty
A multiplicative calibration uncertainty shared by targets using the same amplitude reference. In Calibration II: Error Budgets and Imaging Systematics, it helps resolve the question
Two sources share the same 5% flux-scale error but have independent 1% thermal errors. Which uncertainty cancels in their flux ratio, and which does not?
Use in the reasoning chain: The term links the observable or diagnostic
Independent small fractional errors combine in quadrature, whereas shared calibration errors correlate many image pixels
to the model \(\left(\frac{\sigma_S}{S}\right)^2=\sum_j f_j^2+2\sum_{i<j}\rho_{ij}f_i f_j\). Compare it directly with Primary-beam correction rather than treating either quantity in isolation.Scope: Quadrature is justified only for independent zero-mean errors; common-mode flux-scale uncertainty requires covariance-aware comparisons. The term therefore supports an inference only when its measurement convention, calibration stage, and stated approximation are retained.
- Primary-beam correction
Division by the antenna sensitivity pattern to estimate off-axis sky brightness. In Calibration II: Error Budgets and Imaging Systematics, it helps resolve the question
Two sources share the same 5% flux-scale error but have independent 1% thermal errors. Which uncertainty cancels in their flux ratio, and which does not?
Use in the reasoning chain: The term links the observable or diagnostic
Independent small fractional errors combine in quadrature, whereas shared calibration errors correlate many image pixels
to the model \(\left(\frac{\sigma_S}{S}\right)^2=\sum_j f_j^2+2\sum_{i<j}\rho_{ij}f_i f_j\). Compare it directly with Covariance rather than treating either quantity in isolation.Scope: Quadrature is justified only for independent zero-mean errors; common-mode flux-scale uncertainty requires covariance-aware comparisons. The term therefore supports an inference only when its measurement convention, calibration stage, and stated approximation are retained.
- Covariance
A quantitative description of how errors in different measurements vary together. In Calibration II: Error Budgets and Imaging Systematics, it helps resolve the question
Two sources share the same 5% flux-scale error but have independent 1% thermal errors. Which uncertainty cancels in their flux ratio, and which does not?
Use in the reasoning chain: The term links the observable or diagnostic
Independent small fractional errors combine in quadrature, whereas shared calibration errors correlate many image pixels
to the model \(\left(\frac{\sigma_S}{S}\right)^2=\sum_j f_j^2+2\sum_{i<j}\rho_{ij}f_i f_j\). Compare it directly with Jackknife test rather than treating either quantity in isolation.Scope: Quadrature is justified only for independent zero-mean errors; common-mode flux-scale uncertainty requires covariance-aware comparisons. The term therefore supports an inference only when its measurement convention, calibration stage, and stated approximation are retained.
- Jackknife test
A repeat analysis on deliberately split data used to expose systematic differences. In Calibration II: Error Budgets and Imaging Systematics, it helps resolve the question
Two sources share the same 5% flux-scale error but have independent 1% thermal errors. Which uncertainty cancels in their flux ratio, and which does not?
Use in the reasoning chain: The term links the observable or diagnostic
Independent small fractional errors combine in quadrature, whereas shared calibration errors correlate many image pixels
to the model \(\left(\frac{\sigma_S}{S}\right)^2=\sum_j f_j^2+2\sum_{i<j}\rho_{ij}f_i f_j\). Compare it directly with Thermal noise rather than treating either quantity in isolation.Scope: Quadrature is justified only for independent zero-mean errors; common-mode flux-scale uncertainty requires covariance-aware comparisons. The term therefore supports an inference only when its measurement convention, calibration stage, and stated approximation are retained.
Conceptual entry point
A formal image rms is not a complete uncertainty when the flux scale, primary beam, and deconvolution add correlated errors. Begin with the receiver product: a spectrum, time series, image, or set of complex visibilities. The object we ultimately discuss is several inference steps removed from that product.
Radio astronomy's history makes this distinction concrete. Karl Jansky's 1930s rotating antenna detected persistent celestial interference while he investigated radio communication noise; Grote Reber then built a purpose-designed reflector and mapped the radio sky. Instrumental curiosity became a new window on the Galaxy.
Evidence and measurement layers
- Independent small fractional errors combine in quadrature, whereas shared calibration errors correlate many image pixels.
- Primary-beam correction amplifies both signal and noise toward the field edge.
- Residual images, calibrator closure quantities, split-data comparisons, and injection tests probe different failure modes.
A calibrated brightness scale is not equivalent to a flux measurement, and a catalog parameter is not equivalent to a direct observable. State the processing stage whenever comparing results.
Derivation and assumptions
For a derived flux S=f(x_i), first-order propagation gives sigma_S^2=sum_i (partial f/partial x_i)^2 sigma_i^2 plus covariance terms. For independent fractional terms f_j, (sigma_S/S)^2=sum_j f_j^2. A multiplicative flux-scale term remains correlated among sources calibrated with the same solution and should not be averaged down as independent pixel noise.
The approximation is useful because it compresses a full electromagnetic/instrument problem into a scale relation. It is not a substitute for the response model: explicitly retain the beam, bandpass, convention, sampling, or plasma assumptions relevant to this lecture.
Worked example
Inputs: S=120 mJy; independent 5%, 3%, and 2% terms.
Programmatically evaluated result: sigma_S = 7.397 mJy; fractional uncertainty = 0.06164; sigma_S = 7.397 mJy.
Quadrature is justified only for independent zero-mean errors; common-mode flux-scale uncertainty requires covariance-aware comparisons. Recompute independently from the equation before using the result to compare facilities or source models.
Observational interpretation
Primary-beam correction amplifies both signal and noise toward the field edge. A defensible interpretation records the competing hypotheses, checks whether the instrument response can mimic the feature, and identifies a measurement that would separate the alternatives. The next layer of evidence should be chosen for discriminating power, not simply because it is available.
Common conceptual traps
- Confusing total flux with flux per beam, or a brightness temperature with a thermodynamic temperature.
- Assuming a narrow thermal-noise estimate includes flux-scale or selection uncertainty.
- Treating spatial resolution, field of view, and largest recoverable scale as interchangeable.
- Using a Doppler or calibration convention without recording its reference frame.
Lecture-specific limitation: Quadrature is justified only for independent zero-mean errors; common-mode flux-scale uncertainty requires covariance-aware comparisons.
Study and transfer questions
- Two sources share the same 5% flux-scale error but have independent 1% thermal errors. Which uncertainty cancels in their flux ratio, and which does not?
- Change the dominant input by 10%. Predict the direction of the output change before computing it.
- Name one raw diagnostic and one independent data product that could falsify the preferred interpretation.
Reading
NRAO VLA calibration and imaging-pipeline documentation; exact error propagation is developed here beyond introductory textbook coverage.
See materials/ASTR472/reference-log.md for sources, verification dates, exact OpenStax anchors, and human spot-check caveats.