Insights
Ultrasound ADC Test Patterns: Separate Link Integrity from Analog Fidelity
What does a passing digital ADC test pattern establish in an ultrasound receiver? A source-led OEM review of injection points, link and transport checks, channel ambiguity and a reproducible periodic-sequence counterexample.
Engineering question: When an ultrasound receiver passes an ADC-generated digital test pattern, what has actually been verified—and which parts of the acquisition path remain untested?
A recognizable ramp on a host screen is useful evidence, but its meaning depends on where that ramp enters the system and where the comparison occurs. The proposed acceptance method below separates digital transport, sample interpretation and analog acquisition. It is an engineering test design, not a report of measured ALEVA hardware performance.
Start with the injection point
The public TI AFE5808A datasheet, revision D from November 2015, documents digital patterns that replace normal ADC output on the LVDS interface. Its modes include fixed words, alternating patterns and a repeating ramp. This is an established device example, not a new product announcement. A passing comparison downstream of that replacement point cannot establish the noise, linearity or overload recovery of the bypassed analog signal path. See sections 8.4.4.1 and 8.6.2.10–11 of the TI datasheet.
For an OEM test record, draw the actual injection-to-checker path. A checker immediately after the FPGA receiver covers a different boundary from a checker reading the saved host acquisition. If the delivered product includes buffering, transfer and file conversion, retain a second check at that delivered boundary. Record which processing blocks were enabled or bypassed rather than treating every successful pattern test as an end-to-end result.
Test the layer you intend to qualify
Analog Devices’ October 17, 2023 interface guide separates physical-link, data-link and transport-layer debugging. It describes PRBS testing for physical-interface integrity and converter-generated patterns for investigating sample reconstruction in the transport layer. These tests answer different questions: a clean serial-lane test does not by itself establish that the receiving application assigns the correct bits to each sample. This layered distinction is documented in the ADI guide; the required modes remain device- and implementation-specific.
A recent ultrasound example is Bhattacharjee and colleagues’ ListenToJESD204B. The author manuscript submitted to arXiv on August 20, 2025 reports simulation against Xilinx JESD204C IP operating in JESD204B mode and hardware testing with TI AFE58JD48 converters. The arXiv record states acceptance for IEEE IWASI proceedings and supplies a related publication DOI. Here the cited evidence is the public author manuscript, not an independently repeated experiment. Its interface tests do not establish the analog or imaging performance of a different OEM receiver. Read the manuscript and publication-status record.
Two patterns can hide two different faults
Channel ambiguity: suppose two received channels carry exactly the same sequence at every sample. Swapping their channel labels leaves the comparison unchanged. This is a mathematical ambiguity, even when every received bit is correct. Use distinguishable channel stimuli or a separate identity test where the hardware permits it; do not assume a converter supports independently programmable patterns. Physical element identity requires its own channel-mapping evidence.
Period ambiguity: consider an illustrative software sequence r[n] = n mod 16. Removing 16 consecutive samples from the stream preserves the expected modulo progression at the join. A checker that only compares adjacent values can therefore miss the loss. This example uses a synthetic four-bit sequence; it is not a claim about a particular converter’s operating mode. Independent sample counts, capture identifiers or timing evidence are needed to detect losses that the chosen pattern cannot distinguish. The same reasoning applies to a sequence with any finite repetition period.
Define a bounded acceptance record
For each supported configuration, retain the converter mode, FPGA/software revision, expected word format, injection point, checker location, captured sample count and mismatch count. Specify startup alignment explicitly so that a permitted initial phase is not confused with a continuing error. Include transitions between configurations and repeated resets in the proposed test matrix; a long uninterrupted capture does not exercise those transitions.
Then disable the digital replacement pattern and run a separately specified analog-input test. Define the stimulus, gain, bandwidth, loading and measurement procedure for the property being evaluated. An injected electrical signal still does not qualify the probe’s acoustic response or an imaging task. Keep those evidence boundaries visible in the handover.
The trade-off is additional test time and fixtures in exchange for clearer fault isolation. OEM acceptance should identify the tested path, conditions and unresolved gaps. A useful pass statement is narrow enough that another team can reproduce it and understand what remains to be measured.
Sources and scope
Texas Instruments, AFE5808A datasheet, SLOS729D, revised November 2015, sections 8.4.4.1 and 8.6.2.10–11. Public manufacturer PDF.
Nikhil Ahuja, Analog Devices, Unlock Higher Data Transfer Speeds by Upgrading to the JESD204B/JESD204C Interface, October 17, 2023. Manufacturer technical article.
Soumyo Bhattacharjee et al., ListenToJESD204B: A Lightweight Open-Source JESD204B IP Core for FPGA-Based Ultrasound Acquisition systems, arXiv v1, August 20, 2025. Public author manuscript; the record notes acceptance for IEEE IWASI proceedings. arXiv:2508.14798; related proceedings DOI: 10.1109/IWASI66786.2025.11122015.
ALEVA — Orange County, California, USA. The examples are conceptual and algebraic. No use of a named component or IP core in an ALEVA commercial product, measured device result, or regulatory approval is asserted.