ALEVA

Insights

Ultrasound Element Crosstalk: Define the Measurement Before Setting a Limit

Define the measured quantity, normalization and loading before accepting an ultrasound-array crosstalk number, with recent research and an exact educational two-channel model.

By ALEVA · Orange County, California, USA

What does a supplier's ultrasound-array crosstalk number establish? Before setting an acceptance limit, an OEM team needs to identify the driven element, the observed quantity, the reference response and the loading conditions. A voltage ratio at a connector, displacement of a neighboring membrane and acoustic pressure at a field point are different measurements.

Start with the physical quantity

Recent work illustrates why these boundaries matter. Boujenoui and colleagues' 2026 study models lateral and thickness behavior, losses and coupling in a transducer array, comparing a two-dimensional Mason model with measurements on a seven-element array. Its validation is specific to that studied configuration; it does not establish a universal crosstalk limit for other arrays. [1]

A November 2026 PMUT study by Jin and colleagues uses interacting cells within a channel as part of its frequency-merging design, while also characterizing coupling between channels. Its mechanical and underwater characterization makes an important distinction: interaction can be intentional inside one effective channel and unwanted between independently controlled channels. A measurement plan therefore needs the channel definition as well as the element geometry. [2]

For an OEM specification, label the measured response explicitly: terminal voltage, surface displacement or pressure at a defined observation point. Record whether inactive channels are open, terminated, connected to the receiver or held in another documented state. Those conditions belong in the result, rather than in a separate assumption that the next laboratory may miss.

Use a defined transfer matrix

For a linear, time-invariant test at frequency f, write y(f) = H(f)x(f). Each column describes the measured outputs when one input is excited. The input and output units, reference planes and channel ordering must be declared; the matrix is not automatically dimensionless.

One possible normalized amplitude measure is Cij(f) = 20 log10 |Hij(f) / Hjj(f)|: output i relative to output j while driving input j. This definition requires comparable output quantities and a usable reference response. It is not interchangeable with normalizing by the drive voltage or by each receiving channel's own sensitivity. Near a zero in Hjj, the ratio can become large even when the absolute coupled response is small. Report the absolute responses and the measurement floor alongside the ratio.

Retain complex phase where the intended application combines channels coherently. A magnitude-only nearest-neighbor number does not specify the complete matrix, its frequency dependence or the response to a particular excitation vector.

An exact educational example

Assume two identical, normalized channels at one frequency, linear behavior and real, symmetric coupling:

H = [[1, 0.1],
     [0.1, 1]]

The off-diagonal amplitude ratio is 20 log10(0.1) = -20 dB. An in-phase input [1, 1] produces [1.1, 1.1]; an opposite-phase input [1, -1] produces [0.9, -0.9]. Relative to an uncoupled identity matrix, these responses differ by approximately +0.83 dB and -0.92 dB respectively.

These values are direct algebra, not a simulation of a real probe or measured ALEVA performance. They show that the same coupling coefficient can affect different excitation patterns differently. Frequency-dependent phase, unequal elements, multiple neighbors and acoustic propagation are absent from this example. It cannot predict an imaging artifact or set a product acceptance threshold.

A proposed measurement sequence

  1. Define the boundary. Record the probe configuration, cable and electronics, calibration plane, drive waveform, inactive-channel loading, medium, temperature and observation geometry. Associate every physical element with its data channel.

  2. Characterize the electrical fixture. Use suitable reference loads and isolated-channel checks to identify feedthrough and instrument limitations. A probe-disconnected measurement is a fixture baseline; it is not the assembled probe's total coupling.

  3. Measure the assembled configuration. Excite one channel at a time within the verified linear range and record the selected response on neighboring and more distant channels. Keep frequency, amplitude, phase and uncertainty information where measurable. State any time gate and show what it excludes.

  4. Test the intended excitation patterns. Compare relevant steering or focusing conditions under the same loading and geometry. An element-level matrix and an acoustic field measurement answer complementary questions; neither should be silently substituted for the other.

Leipner, Wolf and Kupsch's 2025 study demonstrates S-parameter measurements combined with time and frequency analysis for separating electrical and acoustic responses in high-frequency transducer characterization. Its scanning-acoustic-microscopy setting is not a ready-made multichannel medical-array acceptance protocol. The transferable lesson is to establish calibration and separation conditions before assigning a physical cause to a feature in the response. [3]

Turn the evidence into an OEM decision

A full matrix across operating conditions requires more measurement time than a single adjacent-channel check. A practical compromise is to establish detailed characterization during development, then justify a smaller production test against that evidence. Recheck the relationship when the stack, cable, loading or channel grouping changes.

The acceptance document should connect a specified measurement to an intended system behavior and its uncertainty. It should not promise that one crosstalk number proves image quality. For related integration boundaries, see ALEVA's probe-termination discussion and custom ultrasound probe design page.

Sources and scope

All three sources below are journal publications, not preprints. This article is an engineering interpretation and proposed test framework. The numerical example is independently reproducible algebra. No supplier qualification, clinical claim or use of these methods in an ALEVA product is asserted.

  1. Boujenoui, A., et al. Development and validation of a 2D Mason model for ultrasonic transducer arrays: A simplified approach simulating crosstalk and dissipation mechanisms. Applied Acoustics 241, 110991; 5 January 2026. doi:10.1016/j.apacoust.2025.110991. Discussion here is limited to the publicly available abstract.

  2. Jin, T., et al. Low-voltage, ultra-broadband wearable MEMS ultrasound: a frequency-merged, wafer-scale framework for continuous in vivo cardiovascular imaging and monitoring. Microsystems & Nanoengineering 12, 356; 2 November 2026. doi:10.1038/s41378-026-01451-6. Only the publicly accessible design and characterization descriptions are used here.

  3. Leipner, E., Wolf, M., and Kupsch, C. Characterization of high-frequency ultrasonic transducers using S-parameters: Separation of electrical and acoustic responses. Mechanical Systems and Signal Processing 238, 113172; 1 September 2025. doi:10.1016/j.ymssp.2025.113172.