Skip to content

Readout and multi-band scans

In lumped-element experiments, a readout circuit turns axion-induced pickup flux into a current measured by a SQUID. A capacitor tunes the pickup and input coil to resonance. The circuit combines thermal noise from resistive elements with the amplifier's measurement noise and backaction on the coil. In this guide we compute readout noise and plan multi-band scans.

From pickup flux to readout noise

The signal transfer and noise referred to the pickup's electromotive force are

\[ H_s(\omega)\equiv\frac{\Phi_{\rm SQ}}{\mathcal E_{\rm pickup}}, \qquad N_V(\omega)\equiv\frac{S_\Phi^{\rm tot}(\omega)}{\lvert H_s(\omega)\rvert^2}. \]

Here \(\Phi_{\rm SQ}\) is the SQUID flux and \(S_\Phi^{\rm tot}\) its total noise power spectral density (PSD). \(H_s\) has units of \(\mathrm{Wb/V}\) and \(N_V\) of \(\mathrm{V^2\,s}\). These PSDs are half the usual one-sided per-Hz PSDs.

With NumPy and SciPy installed, we tune to 1 MHz. Q sets the pickup loss; the default amplifier has uncorrelated noise and is matched for scan rate. We compute the effective circuit parameters and input-referred noise to characterize the readout.

from math import pi
from axionmultiphysics.circuit import series_readout

f0 = 1e6
readout = series_readout(
    L_p_H=1e-6, L_in_H=1e-7, f_Hz=f0,
    Q=1e6, T_K=0.03, M_H=1e-9,
)
eq = readout.equivalent(f_Hz=f0)
print(eq["L_eff_H"], eq["Q_loaded"], eq["T_eff_K"])
print(readout.noise_psd(2 * pi * f0)["N_V_V2_s"])

The first line gives whole-loop inductance in henries, loaded quality factor, and effective noise temperature in kelvin; the second gives input-referred noise. We use whole-loop quantities for scan reach: input-coil and lead inductances contribute even though they collect no signal. The model neglects SQUID input impedance; the resonance reduction applies locally around an isolated pole.

Tuning across several bands

The multi-band and multipole scans (axionmultiphysics.multipole) are beta.

When a detector's size approaches an electromagnetic wavelength, its impedance can contain several resonances. A multi-band scan distributes time among tunable frequency intervals. In the engineering convention \(Z=R+iX\), positive \(X\) is inductive; a series capacitor or inductor cancels the source reactance:

\[ C=\frac{1}{\omega X}\quad(X>0),\qquad L_t=-\frac{X}{\omega}\quad(X<0). \]

The effective inductance is half the slope of the total tuned reactance with respect to angular frequency, including the tuning element.

We start with an ideal 100 nH pickup and an illustrative coupling of \(10^{-6}\,\mathrm{H\,T\,m^2}\). The same axionMultiPhysics calls accept sampled full-wave responses. We tabulate the pickup response and plan a multi-band scan to find the time required for the target reach.

import numpy as np
from axionmultiphysics import multipole

f = np.linspace(10e6, 150e6, 501)
omega = 2 * np.pi * f
table = multipole.source_table(
    f, 1j * omega * 100e-9, omega * 1e-6,
    convention="exp(+jwt)", source_unit="kappa_SI",
)
plan = multipole.multiband_scan(
    table, Q=1e5, live_time_yr=1.0,
    g_target=lambda f: np.full_like(f, 1e-15), rho_GeV_cm3=0.3, Q_a=6.8e5,
)
print(plan["bands"])
print(plan["t_req_s"])

The voltage column is the open-circuit amplitude per unit SI axion source coefficient, identified by source_unit="kappa_SI". For COMSOL data, we specify source normalization and any port load. axionMultiPhysics impedance data use convention="exp(-iwt)". Here Q specifies total resonator quality factor; alternatively, Q_mqs adds intrinsic loss to the table's resistance.

bands lists intervals in Hz and their tuning-element type. Defaults allow 30–300 pF capacitors and 10–500 nH inductors, removing 2.5 MHz at each band edge. t_req_s is the time to reach \(10^{-15}\,\mathrm{GeV^{-1}}\) across accepted bands; g_target_allocation_GeV gives reach after rescaling to the requested live time.

Exporting port parameters to circuit tools

A full-wave sweep's port impedances can be written as a Touchstone file and simulated with an external circuit tool (SPICE, ADS, Microwave Office, scikit-rf). The engine's \(e^{-i\omega t}\) impedances are complex-conjugated into the \(e^{+j\omega t}\) convention these tools assume, so an inductor reads \(Z=+j\omega L\):

res = design.solve()                                   # a design with a fullwave section
report = res.write_touchstone("pickup.s2p", kind="S", z0=50)   # the solved rows
res.write_touchstone("pickup_dense.s2p", dense=True)   # an adaptive sweep's dense model curve (beta)
print(report["sigma_max"], report["passivity_violations_Hz"], report["reciprocity_error"])

From the command line, run axmp export-touchstone runs/NAME/results.json --kind S --z0 50 [--dense]. It exits with 0 when the file was written and is passive, 3 when it was written but is not passive, and 2 on an error. The file is Touchstone v1 with option line # Hz S RI R 50, increasing frequencies and a comment giving the port order and polarity (port 1 is the pickup).

The report gives \(\sigma_{\max}(S)\), which must not exceed 1 for a passive structure. A violation is reported as a warning and the data are written as computed, never clipped. The report also gives the asymmetry of \(Z\), which for a reciprocal design is zero to solver precision (about \(10^{-11}\)).

The perturbative conductor surface loss is not part of \(Z\). surface_losses=True adds its resistive part to the solved rows, but not the surface reactance. axionmultiphysics.touchstone.read_touchstone reads files back. Per-port reference impedances are written for S-parameters only, as a Touchstone v1.1 option line. Z and Y do not depend on z0: export them with one z0, or compute per-port S with z_to_s. Details are in S-, Y- and Z-parameters and Touchstone files.

multiband_scan uses local poles. axionMultiPhysics provides scan.finite_band to combine responses across resonances at each fixed tuning with a frequency-dependent readout within a capacitor-tunable interval. Further reading: dc SQUID readout chains (beta) or the API reference.