Data underlying the publication: Ultrasound Transparent Neural Interfaces for Multimodal Interaction - Immersion experiments

<strong>Recorded data of immersion experiments</strong><br>The experiments were conducted in a 580 x 450 x 300 mm^3 water tank. Water was used as the intermediate medium as its density, sound velocity and sound impedance are near those of soft tissue. An acoustic-damping material was attached to the sides of the tank to minimise acoustic reflections. An ultrasonic transducer (V309-SU, Olympus, Japan) with a diameter of 0.5 inch (12.7 mm) was driven by an arbitrary function generator (33622A, Keysight, USA) and a 50 W linear power amplifier (350L, Electronics &amp; Innovation, USA). An ultrasonic pulse train with a duration of 15 μs and 4 μs rise and fall times was employed, sweeping frequencies from 1.25 MHz to 25 MHz in 1.25 MHz increments. The radiated ultrasonic signal passed through the fabricated samples, aligned and positioned 10 mm away from the transducer surface. To perform measurements of through-transmitted energy we employed a needle hydrophone (NH1000, Precision Acoustics, UK) with a diameter of 1 mm at a distance of 35 mm away from the transducer. The hydrophone was coupled with a submersible preamplifier and DC coupler. The samples were tested one at a time. By measuring the pressure with and without the sample placed between the transducer and the hydrophone, transmittance can be calculated by comparing the pressure collected through the sample and without the sample in the path. Therefore each transmission coefficient for each sample is calculated by comparing the two measurements. Both recordings and the corresponding transmission coefficient are saved in a hdf5 file. A custom-made 3D print was used to align the transducer, sample and hydrophone. The data acquired from the hydrophone were recorded using a digital storage oscilloscope (RTA4004, Rohde &amp; Schwarz, Germany). All applied and monitored signals were time-synced between the function generator and oscilloscope with trigger inputs. The monitored signals were averaged over 100 cycles and recorded with 1.25 GSa/s. The logged signals were streamed from the data acquisition hardware to a workstation PC using custom Python code to control the triggers, channels, and function generator output. The compiled C code which called the oscilloscope and function generator commands was in turn called within a Python wrapper.