Development and Proof-of-Concept Evaluation of a Compact Autonomous Multi-Parameter Instrumentation Platform for Confined Dynamic Systems

Abstract

A compact autonomous recorder was developed for short-duration measurements in confined moving systems where event-time cabling is unavailable. The cylindrical platform combines pressure, high-range acceleration, and angular-velocity interfaces with channel-specific conditioning, FPGA/microcontroller control, SDRAM buffering, eMMC storage, internal power, and mechanical support. Its operating sequence covers baseline checks, circular buffering, threshold-initiated event capture, nonvolatile storage, and post-test retrieval. An eight-field map and count-to-unit equations describe the readout convention. Three-channel electrical checks captured 1 kHz inputs and gave voltage spans of 6–8 mV at a nominal 5 V input. The loading evaluation comprised one acceleration–angular-velocity event and one pressure–acceleration event. Both tests stored eight-field event records over 8 ms at a nominal sampling rate of 1 MS/s per recorded field. Using the count-to-unit conversions, axial acceleration exceeded 24,000 g in the first event and reached approximately 27,000 g in the second, which also recorded approximately 354 MPa pressure. Both records were retrieved after recovery, with post-event communication available. These results demonstrate autonomous recording and recovery for two paired sensing configurations within the same compact architecture. Full simultaneous pressure–acceleration–angular-velocity operation remains to be evaluated.

1. Introduction

Short-duration transient loading occurs in aerospace qualification, structural-dynamic evaluation, impact-safety assessment, and controlled laboratory testing. Pressure, translational acceleration, and rotational motion can evolve over the same millisecond-scale interval. Recording these responses involves acquiring the signals, preserving the event data, and recovering them for interpretation. Reviews of vibration-based assessment and signal processing emphasize reliable acquisition and appropriate treatment of the recorded response [1,2].

In situ recording becomes difficult when the test object moves within a restricted installation envelope. Separate instruments, external acquisition hardware, optical observation, and wired transmission are suited to accessible laboratory arrangements but can conflict with package motion or unavailable event-time communication. Wireless smart-sensor networks move sensing and acquisition closer to the monitored structure [3], while independently operated recorders may still require post-test timing alignment [4].

Autonomous recorders organize sensing and data handling around the event. A high-voltage transient recorder combined multi-channel acquisition with storage, independent power, and timing functions [5]. Shipment-impact detectors link threshold-based logging to local retention [6], and vehicle-mounted acquisition systems address embedded monitoring of electronic-control-unit variables [7]. These examples show how acquisition, storage, and retrieval are organized for different sensing tasks.

At the converter level, field-programmable gate array (FPGA) architectures support parallel handling of analog-to-digital converter (ADC) outputs [8]. Lu et al. used buffered FPGA-controlled storage and host readout in a SATA recording system [9]. Programmable multi-event triggering and timing coordination have been studied for rocket-sled tests [10], while overload-storage instrumentation has combined strain and acceleration sensing [11]. Angular-rate acquisition adds a dependence on gyro range, filtering, and scale calibration, as illustrated by the instrument developed by Radi et al. [12]. Converter throughput, trigger coordination, and calibrated sensor response are therefore distinct aspects of an integrated recorder.

Application-specific sensor platforms in other domains likewise emphasize the co-design of sensing elements, packaging, and system integration [13]. Zhou et al. examined high-overload test-device design through functional decomposition and AD-TRIZ evaluation [14]. The present work focuses on the recorder implementation: channel-specific sensing interfaces feed one acquisition, buffering, storage, and retrieval procedure inside a mechanically integrated cylindrical package. The work includes:

(1)

A functionally decoupled architecture that defines sensing, conditioning, acquisition, storage, power, and support interfaces within one compact envelope;

(2)

An autonomous event sequence linking circular buffering, threshold initiation, nonvolatile storage, and post-recovery readout;

(3)

A component-level implementation description, test-specific acquisition settings, a fixed eight-field record format, and explicit decoding and engineering-unit conversion rules;

(4)

Electrical acquisition checks and two paired loading records demonstrating event capture, local retention, retrieval, and post-recovery operation.

Here, paired denotes the two parameter types in a loading configuration: acceleration–angular velocity or pressure–acceleration. Each configuration was tested once, producing four and three displayed histories, respectively. Together, the two configurations exercise all three parameter-interface types supported by the architecture, but they do not constitute simultaneous three-parameter validation.

2. System Requirements and Instrumentation Architecture

2.1. Measurement Scenario and Design Objectives

The instrumentation travels with the test object and records local transient responses within a restricted installation envelope. Triggering, acquisition, buffering, and storage are performed internally, followed by data retrieval after recovery. The cylindrical design has a nominal envelope of approximately 110 mm in length and 65 mm in diameter.

The package must fit the host structure, operate from independent power, and retain the selected event locally. Dedicated mounts and compartments support the sensors, electronics, and battery, while shared acquisition control organizes the selected signals into a common event record.

2.2. Instrumentation Requirements

Pressure characterizes local loading, acceleration captures translational inertial response, and angular-velocity channels provide rotational-sensor outputs. Their electrical interfaces need appropriate excitation, amplification, impedance matching, scaling, and filtering before conversion. A pressure or acceleration threshold selects the event interval within the operating acquisition sequence.

Data handling separates immediate sample capture from subsequent nonvolatile writing and host communication. The mechanical arrangement provides restrained internal assemblies and an accessible recovery interface. These requirements guide the electrical and mechanical implementation described below.

2.3. Technical Specifications

Table 1 summarizes the supported interfaces, nominal event settings, and recovery outcomes. Section 4 describes the hardware, recording sequence, and readout convention.

Table 1. Implemented interfaces and nominal event settings.

Module or Interface	Reported Implementation and Event Setting
Measurement configurations	One acceleration–angular-velocity event and one pressure–acceleration event; one loading test per configuration.
Readout capacity	Fixed eight-field record: three acceleration fields, three gyro fields, one pressure-interface signal, and one electrical reference.
Coordinate convention	Axial: cylindrical longitudinal/loading direction. Radial: the measured transverse direction identified in the event plots.
Pressure input	Reported event span 0–500 MPa; IEPE-labeled signal and separate electrical-reference fields.
Acceleration input	Reported ±50,000 g configuration in both events; three instrumentation-amplifier paths.
Angular-velocity input	ADXRS649-based interface; reported sensor-side range ±50,000°/s; software readout approximately ±25,000°/s (Section 4.9).
Acquisition control	Circular buffering followed by pressure- or acceleration-threshold event recording.
Nominal time grid	1 MS/s per recorded field; 8 ms record; 1 µs nominal spacing and 8000 samples per field.
Recorded-time convention	Zero denotes the start of the displayed record; time follows the nominal sample grid.
Storage and retrieval	SDRAM buffering, eMMC retention, and serial setup/readout through a Type-C connection after recovery.
Power and support	Internal battery; separated power-conversion functions; threaded cylindrical installation with dedicated sensor, electronics, and battery supports.
Observed outcome	A record was retrieved in each configuration, with post-event communication available.

2.4. Overall Instrumentation Architecture

Figure 1 relates the measurement scenario to the supported parameter types, local recording, and post-test analysis. The selected sensor outputs remain within the autonomous package during loading and become available for analysis after recovery.

Figure 1. Measurement scenario and recording workflow. Arrows show the progression from the measurement scenario through sensing and local storage to post-test analysis. The branches identify the supported parameter types; the two evaluated configurations are described in Section 6.1.

The architecture is functionally partitioned and physically integrated. Sensing, conditioning, acquisition, storage, power, and mechanical support have distinct interfaces within one cylindrical envelope.

3. Functional Decoupling and System Implementation

3.1. Functional Decoupling Strategy

The platform assigns excitation and conditioning to channel-specific paths, separates recording control from power conversion, and uses dedicated mechanical mounts. These interfaces accommodate the different electrical and support requirements of the sensors, converters, memory, and battery within one package.

Changes to a sensor path affect excitation, gain, filtering, and conversion coefficients; changes to storage affect writing and readout. Grounding, interconnections, and the support arrangement govern the interaction of these functions in the assembled instrument.

3.2. Module Implementation and System Integration

Conditioned analog signals feed an eight-channel ADC. The corresponding software fields are defined in Section 4.9. FPGA acquisition logic and microcontroller record control link capture to temporary and retained storage. Circular buffering preserves recent samples, the completed event is committed to eMMC, and the external interface supports configuration and post-recovery readout.

The two sensing-end arrangements retain the same preparation, capture, storage, and readout sequence (Section 4.5). Dedicated mounts and electronics compartments, together with a separately accessible battery, complete the implementation.

3.3. Evaluation of the Implementation Strategy

Distributed and partially integrated arrangements offer placement flexibility but retain external interconnections or separately controlled devices. The selected integrated arrangement combines local recording and independent power with dedicated internal supports. Table 2 relates these design choices to their intended functions.

Table 2. Design constraints, implementation strategies, and intended measurement benefits.

Design Constraint	Implementation Strategy	Intended Measurement Benefit
Confined installation	Compact cylindrical integration and dedicated internal supports.	Compatibility with the host installation envelope.
Cross-parameter event association	One acquisition-control sequence and nominal record index.	Related outputs retained within the same event record.
Unavailable event-time transmission	Temporary buffering and local nonvolatile storage.	Capture and retention independent of the external readout link.
Transient mechanical loading	Housing, threaded interface, sensor mounts, and supported compartments.	A defined support path for sensing and recording functions.
Heterogeneous sensor outputs	Dedicated excitation and conditioning by channel type.	Electrical compatibility with the conversion interface.

4. Hardware Architecture and Data Acquisition System

4.1. Overall Hardware Architecture

Both loading tests used the same acquisition hardware: an AD7606C-18BSTZ ADC (Analog Devices, Wilmington, MA, USA; AD7606C-18 hereafter), a GW1N-9 FPGA (GOWIN Semiconductor, Guangzhou, China), and an STM32H743AII6TR microcontroller (STMicroelectronics, Geneva, Switzerland). The FPGA controls sampling and readout; the microcontroller handles initialization, trigger settings, event recording, memory, and communication. Three acceleration-conditioning outputs, three single-axis gyro outputs, and the pressure signal/reference pair feed the ADC. SDRAM buffers the samples, and eMMC retains the completed record. Figure 2 shows the signal path, and Table 3 lists the components.

Figure 2. Component-level recording architecture. Three acceleration fields, three gyro fields, and the pressure signal/reference pair feed the conversion and record-control path. Pink arrows indicate signal and data flow; the black arrow indicates the power connection. The dashed horizontal line separates the onboard recording electronics from the post-recovery PC interface. The internal supply powers the recording electronics, and PC communication supports setup and post-test retrieval.

Table 3. Component identifications available from the design records.

Function	Design-Record Identification	Documented Role
Acceleration sensing	SENTHER462/SENTHER642, U11	Alternative identifiers in the sensor design records.
Acceleration conditioning	INA351CDSIDSGR, U1–U3	Three instrumentation-amplifier paths.
Angular-velocity sensing	ADXRS649BBGZ-RL, U15	One single-axis analog gyro per gyro board.
Pressure sensing and interface	Pressure transducer: Kistler6215; LT3092EST#TRPBF; AD8605ARTZ-REEL7	Current-source/interface circuitry with signal and reference outputs.
Analog-to-digital conversion	AD7606C-18BSTZ	Eight-channel, 18-bit simultaneous sampling; bipolar ±10 V inputs; 1 MS/s per channel.
Acquisition and system control	GW1N-9 FPGA; STM32H743AII6TR	Acquisition-side logic, record control, memory handling, and communication.
Temporary buffer	MT48LC16M16A2B4-6A IT:G	SDRAM for acquisition-time buffering.
Nonvolatile memory	KLM8G1GETF-B041	eMMC for retained event records.
Power functions	TPS54302DDCR; TPS61040DBVR; LM27761DSGR; TP5400	Local supply conversion and battery-related functions.
External communication	SIT3490EEUA; Type-C connector	Serial preparation and post-test readout using PCOMM.

ADC settings and board/firmware revisions are specified in Section 4.1; Section 4.9 defines the stored-field map.

The acquisition board was DAQ Rev. B, running FPGA firmware v1.3.2 and MCU recording software v0.9.6 in both tests. The ADC input filter operated in low-bandwidth mode, with a typical −3 dB bandwidth of 25 kHz [15]. It is distinct from the 16-bit, 800 kS/s AD7606B [16], which was not used in these tests.

The ADC operated in hardware mode with all eight inputs configured for single-ended bipolar ±10 V operation. Conversion timing and data readout were controlled by the FPGA. All eight channels used a bipolar ±10 V input range, and the sensor-conditioning outputs were kept within this range. Digital oversampling was disabled. A stable on-board reference supplied the ADC conversion reference independently of the pressure-interface electrical reference in ch8.

After power-up, the MCU initialized the system and configured sensor selection, the trigger threshold, and the recording sequence. Both tests used the same ADC settings, sampling rate, and fixed ch1–ch8 record format; only the sensor-end arrangement and trigger source changed. Section 4.5 and Section 4.9 describe the sampling sequence and raw-code conversion, respectively, and Appendix A gives the storage budget.

4.2. Pressure Measurement Channel

The IEPE-labeled pressure-interface signal and its electrical reference correspond to fields ch7 and ch8, respectively. The interface includes the programmable LT3092 current source [17] and an AD8605 amplifier. Equation (6) subtracts ch8 from ch7 at the same sample index. The reference field corrects the pressure readout; it is separate from the ADC conversion reference.

The reported event span is 0–500 MPa. Sensor-selection specifications separately list a 690 MPa element rating and a resonance of at least 400 kHz. The installed dynamic response also includes the pressure port and conditioning chain; pressure-calibration studies evaluate the delivered input, transducer, mounting interface, and transfer path together [18,19]. These component- and selection-level specifications are not used as a calibrated bandwidth or uncertainty estimate for the complete installed pressure channel.

4.3. Acceleration Measurement Channel

The acceleration front end uses three INA351 instrumentation-amplifier paths, with scaling set by the fitted gain and sensor sensitivity [20]. Both loading events used the reported ±50,000× g configuration. Sensor-selection information describes a ±50,000 g element with a DC–10 kHz response within ±5% and a DC–25 kHz −3 dB band. These are selection specifications; the design identifiers are listed in Table 3.

Sensor mounting and the acquisition path influence acceleration measurement [21,22]. A multi-axis sensitivity matrix relates main-axis response, transverse sensitivity, and mounting misalignment [23]. Modified Hopkinson-bar studies provide common-excitation triaxial shock-calibration approaches for assessing these terms [24,25]. These studies identify the assembled-channel calibration quantities required for quantitative multi-axis interpretation.

4.4. Angular-Velocity Measurement Channel

The ADXRS649 gyro has a native ±20,000°/s range. An external resistor extends the range to ±50,000°/s and changes the sensitivity [26]. The angular-velocity event used the reported extended configuration. Matching the output network to the readout scale is part of instrument-level angular-rate calibration, as illustrated by Radi et al. [12]. The corresponding software readout interval is given in Section 4.9.

The available gyro design specifies a 120 kΩ resistor and a 0.1 µF capacitor between RATEOUT and SUMJ. Using the manufacturer’s nominal 180 kΩ internal output resistance [26], the equivalent resistance and first-order pole are

$$R_{\mathrm{eq}}={\displaystyle \frac{R_{\mathrm{int}}{R}_{\mathrm{ext}}}{R_{\mathrm{int}}+R_{\mathrm{ext}}}}=72\hspace{0.17em}\mathrm{k}\mathsf{\Omega },$$

(1)

$$f_{\mathrm{c}}={\displaystyle \frac{1}{2\pi R_{\mathrm{eq}}C}}\approx 22.1\hspace{0.17em}\mathrm{Hz}, \tau =R_{\mathrm{eq}}C=7.2\hspace{0.17em}\mathrm{ms}.$$

(2)

The documented network therefore has an estimated time constant of 7.2 ms and a 10–90% rise time of approximately 15.8 ms. Because event-specific network and calibration records are unavailable, the angular-velocity histories are treated as recorded channel outputs rather than as calibrated high-frequency rotational motion.

4.5. Triggering and Data Acquisition

Before loading, the procedure checks baseline behavior, channel response, threshold initiation, storage, retrieval, and the internal supply. The package is installed, and the external cable is disconnected for the event. Circular acquisition retains recent samples in SDRAM; pressure- or acceleration-threshold detection initiates the selected event interval. Acquisition completion is followed by commitment to eMMC. Table 4 relates the operating states to record handling.

Table 4. Operating sequence and event-record handling.

State	Operation	Record-Handling Function
A0	Entry to the main sequence.	Precedes circular acquisition after preparation and checks.
A1	Circular buffering.	Retains recent samples in temporary memory.
A2	Event recording.	Captures the selected threshold-initiated event interval.
A3	eMMC write.	Commits buffered data to nonvolatile storage.
A4	Write complete; readout available.	Provides access to the retained record through PCOMM after recovery.

The FPGA issues conversion-start commands at a nominal 1 MHz rate, giving one eight-channel sample group every 1 µs. Conversion results are read in channel order, converted to the unsigned fields described in Section 4.9, and placed in the SDRAM circular buffer. Threshold detection selects an 8 ms event window; the MCU then commits the buffered frames to eMMC.

The B1 and B2 modes provide separate diagnostic checks. PCOMM self_check_value_0 through self_check_value_7 are B2 diagnostic limits, and cnt is a diagnostic frame counter. Event-trigger thresholds are separate from these diagnostic limits. The trigger sample index for the two archived loading records is not available and is therefore not assigned to the plotted time origin.

Let f_s denote the nominal sampling rate per recorded field, T_s the nominal sample interval, T the record duration, and n the sample index. The nominal time grid is

$$T_{\mathrm{s}}={\displaystyle \frac{1}{f_{\mathrm{s}}}}, t_n=n{T}_{\mathrm{s}}, N_{\mathrm{nom}}=f_{\mathrm{s}}T.$$

(3)

Each 8 ms event stores 8000 samples per field at a nominal rate of 1 MS/s, giving 1 µs nominal spacing between sample positions. Each position holds one eight-field frame. Indices n = 0, …, 7999 span 0–7.999 ms. The event plots express time in seconds, with zero denoting the start of the displayed record.

4.6. Embedded Storage and Power Management

SDRAM separates event-time capture from subsequent eMMC writing. After recovery, PCOMM v0.1.2 uses the serial port associated with the Type-C connection. The operator selects the port and destination and retrieves the completed record in state A4. The host interface serves preparation and readout; the buffer, memory, and internal supply serve event-time operation.

Both tests used the same fixed-length frame: eight unsigned 16-bit fields ordered ch1–ch8, including fields not used in the displayed results. Each frame contains 16 sample bytes. An 8000-frame event carries 128,000 sample bytes (125 KiB), at a nominal acquisition-side data rate of 16,000,000 bytes/s (16 MB/s). The complete event file is approximately 128–132 kB; the sample payload is 128,000 bytes.

The internal battery powers sensing, conditioning, conversion, and memory until the buffered record is committed. Capacity and endurance values vary among the design descriptions. Power conversion is separated from signal handling. Board grounding, protection, and layout affect electromagnetic compatibility [27].

4.7. Mechanical Integration of the Instrumentation Package

The hardware is integrated in a threaded cylindrical package. Dedicated interfaces support the sensors, acquisition electronics, storage, battery, rubber pad, sleeve, and end components. Two sensing-end arrangements accommodate acceleration–angular-velocity and pressure–acceleration recording while retaining the common storage and power architecture (Figure 3). Integrated high-g MEMS–ASIC work likewise treats packaging and calibration as coupled development issues [28].

Figure 3. Mechanical integration models for the paired experimental configurations. (a) 1—Housing I; 2—sensor-mounting platform I; 3—sensor; 4—data-storage compartment; 5—battery compartment; 6—rubber pad; 7—sleeve. (b) 1—Housing II; 2—pressure sensor; 3—sensor-mounting platform II; 4—acceleration sensor; 5—data-storage compartment; 6—battery compartment; 7—rubber pad; 8—sleeve.

Figure 4 shows the fabricated modules, threaded installation interface, and sensing end. The sensor and circuit assemblies are potted, while the battery remains separately accessible. Together with the models in Figure 3, these photographs document the implemented support arrangement.

Figure 4. Fabricated instrumentation platform: (a) disassembled modules; (b) threaded installation; and (c) installed sensing end.

4.8. Shared Acquisition Sequence and Synchronization Scope

The FPGA assigns the eight simultaneous ADC results to one sample index. This provides a shared acquisition sequence and time grid for the stored fields. The timing of the physical signals also depends on sensor and conditioning responses and converter aperture offsets. For pressure measurements, the sensor and transfer path can affect timing interpretation [29]. Channel delay affects comparisons within a record; clock-frequency error scales the time axis; and trigger latency and jitter locate the record relative to the event. These timing quantities were not measured in the two loading tests.

At 1 MS/s per recorded field, the nominal Nyquist frequency is 500 kHz. Oversampling was disabled; the measurement bandwidth is set by the sensor, analog-conditioning, and converter responses, including filter cutoff and out-of-band attenuation. Table 5 summarizes the recording functions and time convention.

Table 5. Recording functions and their interpretation in the present evaluation.

Function or Setting	Reported Implementation	Interpretation in This Study
Shared acquisition sequence	Eight fields retained in each event frame.	Associated histories on a common record index; synchronization accuracy was not independently characterized.
Nominal temporal grid	1 MS/s per recorded field; 8 ms interval.	1 µs nominal spacing; 8000 samples per field.
Conversion and transfer	AD7606C-18 conversion and FPGA fixed-order field readout.	Eight 16-bit fields per frame; SDRAM buffering and MCU-controlled eMMC storage.
Threshold initiation	Pressure- or acceleration-threshold event selection.	The plotted origin follows the displayed-record convention in Section 4.5.
Sensor and analog response	Outputs retained on the existing engineering-unit scales.	Numerical conversions in Section 4.9.
Local record retention	One retrieved record per configuration.	Local retention and post-recovery communication.

4.9. Engineering-Unit Conversion and Calibration Scope

Table 6 defines the eight-field readout convention. Software X/Y/Z labels identify the inertial fields; the event figures retain the axial and radial labels used for the installed package.

Table 6. Eight-field readout map and numerical use.

Readout Field	Electrical Quantity	Numerical Use
ch1, ch2, ch3	Software X-, Y-, and Z-axis acceleration fields.	Inertial fields DA,i in Equation (4).
ch4, ch5, ch6	Software X-, Y-, and Z-axis gyro fields.	Inertial fields DG,i in Equation (5).
ch7	IEPE-labeled pressure-interface signal.	Signal field DIEPE in Equation (6).
ch8	Pressure-interface electrical reference, separate from the ADC conversion reference.	Reference field DVOS in Equation (6).

The fixed frame contains seven measurement fields and one electrical-reference field. All eight positions are stored in both configurations; Section 6.1 identifies the histories presented from each event.

The FPGA converts each signed 18-bit ADC output C to a stored field D by an arithmetic right shift of three bits followed by addition of 16,384. Equivalently, D = floor(C/8) + 16,384, where floor denotes rounding toward negative infinity. The field spans 0–32,767 and is stored as a 16-bit unsigned integer; C = 0 maps to D = 16,384. This representation retains 15 code bits, not all 18 ADC output bits.

The inertial-channel conversions subtract the software offset of 16,384 from the stored fields. For inertial axis i and record index n, the engineering-unit conversions are

$$A_i\left[n\right]={\displaystyle \frac{D_{\mathrm{A},i}\left[n\right]-\mathrm{16,384}}{0.196608}}\hspace{0.17em}\mathrm{g},$$

(4)

$$G_i[n]={\displaystyle \frac{D_{\mathrm{G},i}\left[n\right]-\mathrm{16,384}}{0.65536}}\hspace{0.17em}°/\mathrm{s},$$

(5)

P[n] = 0.05087(D_IEPE[n] − D_VOS[n]) MPa.

(6)

D denotes the stored unsigned field, A acceleration in units of conventional standard gravity, G angular-velocity channel output, and P pressure. The pressure coefficient is 0.05087 MPa per count difference, equivalent to 50.87 kPa. Subtraction of ch8 from ch7 at the same sample index removes the common software offset and applies the electrical-reference correction.

For D = 0–32,767, Equation (5) gives a numerical angular-velocity readout interval from −25,000 to approximately 24,998°/s. The reported ±50,000°/s setting in Section 4.4 describes the sensor-side configuration, not the range preserved by this software scale. The angular-channel peaks reported in Section 6.3 fall within the stored-code interval.

A stored-count increment corresponds to approximately 5.0863 g for acceleration, 1.5259°/s for angular velocity, and 0.05087 MPa for the pressure-field difference. These are the numerical scales used in Equations (4)–(6). For a locally linear channel, y = K(d − d₀), K combines sensor sensitivity, conditioning gain, and converter/readout scaling. A traceable uncertainty budget also requires trial-specific sensitivity, gain and offset, dynamic response, mounting and cross-axis effects, and timing [30]. No combined measurement uncertainty is assigned to the event histories.

5. Mechanical Integration and Post-Test Functional Evidence

5.1. Mechanical Integration

The assembly in Figure 3 connects the instrument to the host structure through the housing and threaded interface. Dedicated platforms locate the sensors, and the storage and battery compartments support recording and power. The sleeve, rubber pad, and end components complete the internal arrangement. The housing is specified as 35CrMnSi steel, the mounting platforms and end cap as 7075 aluminum alloy, and the storage compartment as galvanized steel.

5.2. Post-Test Functional Observations

Table 7 relates the mechanical supports to the two event outcomes. Recovered histories and post-test communication provide evidence that the assembled recording path operated through the reported loading events.

Table 7. Mechanical interfaces and post-test functional observations.

Interface	Implemented Support	Observation and Interpretation
Housing and installation	Cylindrical housing and threaded host connection.	The recovered package was described as intact after each event.
Sensor mounts	Dedicated platforms and potted sensor assemblies.	Presented channels produced event-related histories in their respective configurations.
Recording electronics and interconnects	Supported circuit assemblies and a dedicated storage compartment.	Stored event records were retrieved after recovery in both configurations.
Battery and internal supports	Battery compartment, sleeve, rubber pad, and end components.	Post-recovery communication remained available in both configurations.

The observations refer to the assembled package in the two reported events.

5.3. Scope of Structural Qualification

The structural evidence concerns the assembled recorder in the two loading events. Stored records were retrieved after recovery, and communication remained available for both configurations (Section 5.2 and Table 7). These observations demonstrate recorder-level function after loading, but do not measure the internal structural response or establish component-level safety margins.

Component-level qualification requires documented finite-element assumptions and measurements of PCB strain, connector displacement, relative motion of internal assemblies, and pre- and post-test condition. Related studies address electronic assemblies, potting materials, and MEMS shock behavior [31,32,33,34]. These indicators were not measured in the present tests; they define the structural qualification work outlined in Section 7.4.

6. Proof-of-Concept Tests of Paired-Parameter Autonomous Acquisition

6.1. Scope and Experimental Setup

Two paired-parameter configurations were evaluated once each under room-condition laboratory loading. The first presents axial and radial acceleration together with axial and radial angular-velocity channel outputs. The second presents pressure together with axial and radial acceleration. Both followed pre-event checks, internally powered event recording, package recovery, and external readout. Figure 5 shows the controlled transient-loading apparatus.

Figure 5. Laboratory transient-loading apparatus used for the two paired proof-of-concept tests.

Both loading tests used the DAQ Rev. B hardware and firmware configuration described in Section 4.1. Table 8 lists the presented histories, acquisition settings, and sensor-side ranges. Both events used the reported ±50,000 g acceleration setting. The gyro sensor range and software readout interval are described in Section 4.9.

Table 8. Paired loading configurations and nominal acquisition settings.

Configuration	Test Count	Presented Histories	Nominal Record Settings	Reported Spans
Acceleration–angular velocity	1 event	Axial/radial acceleration; axial/radial angular-velocity outputs.	1 MS/s per recorded field; 8 ms event window; eight-field frames; threshold initiated.	Acceleration ±50,000 g; gyro sensor ±50,000°/s.
Pressure–acceleration	1 event	Axial/radial acceleration; pressure.	1 MS/s per recorded field; 8 ms event window; eight-field frames; threshold initiated.	Acceleration ±50,000 g; pressure 0–500 MPa.

Both event formats store all eight fields in fixed frame positions. In the acceleration–angular-velocity configuration, ch1–ch6 carry the inertial outputs, while ch7 and ch8 remain in the file layout without contributing to the displayed results. In the pressure–acceleration configuration, ch1–ch3 provide acceleration, and pressure is calculated from ch7 and ch8. Fields not used in the reported results keep their frame positions. Four and three engineering-unit histories are presented, respectively.

Axial denotes the package longitudinal/loading direction; radial denotes the measured transverse direction identified in the plots. The quoted peaks refer to the full displayed interval on the existing engineering-unit scales.

The event histories are presented on the retained engineering-unit scales. No additional filtering, coordinate transformation, or amplitude rescaling was applied.

6.2. Pre-Test Verification

Pre-test checks covered baseline response, channel response, triggering, local storage, retrieval, and the internal supply. Electrical checks also exercised the three acceleration-input paths with 1 kHz sinusoidal signals over a stated 0.1–5 V peak-to-peak range. Individual-channel acquisitions and a three-channel acquisition with deliberately different input phases produced the corresponding stored waveforms, confirming channel identification and signal capture.

A separate check applied a nominal input V₀ = 5 V. Table 9 lists the recorded extrema and two descriptive metrics: the peak-to-peak span and the largest extremum deviation from the nominal input, calculated as

ΔV_pp = V_max − V_min,

(7)

$$E_{\max }={\displaystyle \frac{\max \left(\right|V_{\max }-V_0|,|V_{\min }-V_0\left|\right)}{V_0}}\times 100\%.$$

(8)

Table 9. Recorded electrical extrema and calculated descriptive metrics at a nominal 5 V input.

Channel	Vmax (V)	Vmin (V)	ΔVpp (mV)	Emax (%)
1	5.004	4.998	6	0.08
2	5.002	4.994	8	0.12
3	5.005	4.997	8	0.10

The voltage span and maximum extremum deviation are calculated from the recorded extrema using Equations (7) and (8).

The three paths gave voltage spans of 6–8 mV and maximum extremum deviations of 0.08–0.12% at the nominal 5 V input. These electrical checks confirm basic acquisition-stage signal capture and complement the subsequent integrated loading tests.

6.3. Acceleration–Angular-Velocity Results

The first event produced the four histories in Figure 6. Axial acceleration rises above 24,000 g and then decays, while radial acceleration is smaller and oscillatory. The displayed angular-channel peaks are approximately 7000°/s axially and 2500°/s in radial magnitude. Both acceleration histories and both angular-velocity outputs were recovered in the same event record.

Figure 6. Acceleration–angular-velocity event: (a) axial acceleration; (b) radial acceleration; (c) axial angular-velocity channel output; and (d) radial angular-velocity channel output. Nominal sampling rate: 1 MS/s per recorded field; displayed interval: 8 ms. Time is expressed in seconds, with zero denoting the start of the displayed record rather than the trigger instant. Angular-channel scaling and interpretation are described in Section 4.4 and Section 4.9.

The four retrieved histories show that the acceleration and angular-velocity interfaces were captured and retained together through the shared acquisition sequence.

6.4. Pressure–Acceleration Results

The second event produced the three histories in Figure 7. Pressure rises to approximately 354 MPa and then decays. Axial acceleration reaches approximately 27,000 g, while radial acceleration remains below 100 g in displayed peak magnitude. The pressure and axial-acceleration envelopes rise and decay within the same nominal event interval.

Figure 7. Pressure–acceleration event: (a) axial acceleration; (b) radial acceleration; and (c) pressure. Nominal sampling rate: 1 MS/s per recorded field; displayed interval: 8 ms. Time is expressed in seconds, with zero denoting the start of the displayed record rather than the trigger instant. The displayed peaks are approximately 27,000 g axially, below 100 g in radial magnitude, and approximately 354 MPa in pressure.

The recovered record contains the pressure and acceleration histories from the second sensing-end arrangement. This configuration used the same threshold-initiated recording, local storage, and post-recovery readout procedure as the first event.

6.5. Record Retrieval and Post-Test Operation

Each configuration yielded one retrieved record from one loading event. The package remained operational after recovery, with post-test communication available in both cases. These observations demonstrate event capture, local retention, and post-test readout for the two proof-of-concept tests. File-level completeness, sample ordering, checksum integrity, and sensor-calibration retention were not independently audited. Table 10 summarizes the results and recovery outcomes of the two events.

Table 10. Reported results and recovery outcomes for the paired events.

Item	Acceleration–Angular Velocity	Pressure–Acceleration
Loading events/records retrieved	1/1	1/1
Engineering-unit histories presented	4	3
Samples per field	8000; eight-field event format.	8000; eight-field event format.
Recording workflow	Threshold initiation, local recording, commitment, and post-test readout.	Threshold initiation, local recording, commitment, and post-test readout.
Axial-acceleration peak	>24,000 g.	Approximately 27,000 g.
Radial-acceleration response	Lower-amplitude oscillatory response; Figure 6b.	Reported magnitude below 100 g.
Axial angular-channel readout peak	Approximately 7000°/s.	-
Radial angular-channel readout peak magnitude	Approximately 2500°/s.	-
Pressure peak	-	Approximately 354 MPa.
Post-event outcome	Record retrieved; post-event communication available.	Record retrieved; post-event communication available.

7. Discussion

7.1. What the Experiments Demonstrate

The main result is an acquisition-to-retrieval procedure that operates with two different sensor combinations in a confined cylindrical installation. Channel-specific analog interfaces feed common record control, temporary buffering separates capture from nonvolatile writing, and the recovered files preserve the association between outputs from each event. Changing the sensing-end arrangement leaves the autonomous record-handling sequence unchanged.

The electrical checks and loading tests address successive parts of this process. The 1 kHz and nominal 5 V checks exercise signal capture through the acquisition stage. The two loading records then show that the sensing, power, storage, and communication functions can support event recording and later retrieval in the assembled package.

The practical benefit is access to related sensor outputs after a short event in which an external acquisition link is unavailable. The demonstrated contribution therefore lies in autonomous recording and recovery across two sensor combinations within the same installation architecture. The present evidence supports recorder-level feasibility across two paired configurations rather than repeated-condition reliability or simultaneous three-parameter validation.

7.2. Comparison with Related Instrumentation

Table 11 compares the measurement tasks, relevant capabilities, and study scope of related instrumentation. It places the present package alongside autonomous recorders, converter and storage systems, triggering systems, and sensor-calibration studies. The comparison distinguishes the evidence supplied by each implementation from the recorder-level evaluation reported here.

Table 11. Measurement purpose, relevant capability, and comparison with the present study.

System	Measurement Purpose	Relevant Capability	Difference from Present Study
Autonomous electrical recorder [5]	Electrical transient recording.	Autonomous event acquisition, storage, power, and retrieval.	Electrical rather than pressure and inertial sensing.
Smart impact detector [6]	Shipment-impact monitoring.	Compact threshold-based event logging.	Shipment monitoring rather than the present high-g paired-sensing tests.
FPGA DAQ [8]	Multi-channel data acquisition.	Converter interfacing and real-time data handling.	DAQ implementation rather than evaluation of a recovered autonomous package.
SATA recorder [9]	High-speed storage and host readout.	Buffered storage and separation of acquisition from readout.	SATA storage evaluation rather than the present short-record SDRAM–eMMC sequence.
PMT-TC [10]	Programmable multi-event triggering.	Event initiation and timing coordination.	Event-control coordination rather than calibration of heterogeneous analog-channel delays.
Multi-sensor storage system [11]	Strain and acceleration acquisition.	Static calibration and an air-cannon test reporting 53 g.	Different sensing quantities and loading conditions.
Analog-gyro recorder [12]	High-spinning angular-rate measurement.	Gyro acquisition and instrument-level rotational calibration.	Provides instrument-specific angular-rate calibration rather than recorder-level multi-sensor evaluation.
Present platform	Pressure and inertial interfaces in two paired configurations.	One retrieved record per configuration and post-event communication.	Event-level feasibility; simultaneous three-parameter and repeated-condition qualification remain future work.

DAQ = data acquisition. The table summarizes the aspects relevant to the present recorder rather than every capability of each cited system. The reported loading level refers to the individual study and is not a common-condition comparison.

The present platform combines heterogeneous sensing interfaces with a shared record controller, temporary buffering, nonvolatile retention, and a recoverable cylindrical assembly. Its electrical checks and paired loading records provide complementary evidence at acquisition-stage and integrated-recorder levels.

7.3. Scope of Measurement and Data Interpretation

The acquisition hardware, ADC settings, and fixed-frame format are described in Section 4.1, Section 4.2, Section 4.3, Section 4.4, Section 4.5 and Section 4.6. Section 4.9 specifies the conversion from ADC codes to stored fields and engineering units. Figure 6 and Figure 7 use these conversions without additional filtering or amplitude rescaling. Signal-processing choices can affect transient features used in high-g calibration [35,36].

The electrical checks establish acquisition-stage signal capture. Quantitative interpretation of the loading records also depends on installed-channel sensitivity, bandwidth, relative delay, clock accuracy, and trigger latency/jitter. For pressure, pressure-port dynamics, mounting vibration, and cross-sensitivity act in addition to the electrical-reference correction [37,38]. The smaller radial-acceleration response is more sensitive to baseline behavior, transverse sensitivity, and mounting alignment; common-excitation multi-axis calibration can quantify these effects [23,24,25]. End-to-end calibration against defined references is needed to assign measurement and timing uncertainties.

One record was recovered from each paired configuration. These two events demonstrate capture, local storage, and post-test retrieval, but do not establish repeatability or statistical reliability. The fixed-frame format defines the expected sample count and payload; sample ordering and checksum integrity were not independently audited. Simultaneous three-parameter operation and repeated loading require further tests.

7.4. Future Experimental and Metrological Qualification

Future records should embed the board and firmware versions, input configuration, channel assignment, trigger settings, word format, and sample sequence in the event metadata.

End-to-end calibration should characterize the installed sensitivity, bandwidth, cross-axis response, filtering, and dynamic transfer characteristics of the pressure and inertial channels, with pre- and post-loading checks used to assess calibration retention. Quantitative timing qualification should use a common reference input to measure inter-channel delay and an external timing reference to characterize clock error and trigger latency/jitter.

A simultaneous pressure–acceleration–angular-velocity experiment is required to validate full three-parameter operation, with the complete channel-enable mask and event metadata retained. Repeated matched-condition tests at representative load levels should quantify peak-value variability, trigger/recording success, retrieval success, and any data loss, providing a statistical basis for repeatability and reliability assessment.

Structural qualification should combine documented finite-element assumptions with component-level indicators such as PCB strain, connector displacement, internal relative motion, and pre/post-test inspection. Environmental testing would further determine whether the demonstrated recorder operation extends beyond the present laboratory conditions.

8. Conclusions

A compact autonomous platform was developed for transient measurements in confined moving systems. Pressure and inertial interfaces, FPGA/microcontroller control, SDRAM buffering, eMMC storage, internal power, and mechanical support form one integrated recorder. Its operating sequence links circular buffering to threshold-initiated capture, nonvolatile writing, and post-test retrieval. The component description, fixed eight-field format, decoding rule, and conversion equations document the implementation and readout convention.

Three-channel electrical checks captured 1 kHz inputs and gave voltage spans of 6–8 mV at a nominal 5 V input. The loading evaluation included one event per paired configuration. Each test stored an 8 ms record at a nominal sampling rate of 1 MS/s per recorded field. The count-to-unit conversions give axial acceleration above 24,000 g in the first event and approximately 27,000 g in the second, with approximately 354 MPa pressure in the latter. Both records were retrieved, and post-event communication remained available.

The two proof-of-concept tests demonstrate autonomous paired-parameter recording and post-test recovery in the compact package. Further evaluation should address simultaneous three-parameter operation, end-to-end calibration, quantitative synchronization, and repeated-condition reliability.