SHT30 Humidity Reads Low After Reflow Soldering: Recovery, PCB Cleaning and FHT30 Qualification

How do you tell temporary SHT30 offset from contamination or damage?

A post-reflow SHT30 error can come from several mechanisms. A temporary negative shift may recover as the sensing system returns to equilibrium. Chemical contamination may persist or vary by lot and process step. A warm PCB or incomplete cooldown can also lower reported relative humidity because the sensor and reference are not at the same temperature.

The table is a triage method, not a universal acceptance specification. Product limits must come from the finished product's error budget and verified manufacturing capability.

Which production records should be retained before retesting?

An SHT30 recovery investigation is useful only when the manufacturing history can be reconstructed. Retain the exact order code and lot, board serial number, paste and flux, measured reflow profile, oven lane, panel position, solder-cycle count, rework history, cleaning or coating process, packaging exposure and time from oven exit to each measurement.

The test record also needs temperature, relative humidity (RH), airflow, reference identity, raw sensor words, cyclic redundancy check (CRC) results and firmware revision. CRC checks data integrity; a valid frame does not establish humidity accuracy. Retain both RH and temperature CRC results in the acquisition log. [2]

Record the heater state as well as the reflow history. The SHT3x heater is intended for plausibility checking and changes the local sensor temperature; keep it off during the ambient accuracy comparison. [2]

How should SHT30 post-reflow recovery be tested?

Compare each SHT30 device under test (DUT) with a characterized reference over time. First record an incoming baseline where available, then use the same measurement conditions after assembly.

1. Verify the fixture. Allow the board and reference to reach thermal equilibrium. Use consistent airflow, spacing, power and enclosure state; keep the reference clear of warm PCB surfaces and oven exhaust.
2. Capture the first practical checkpoint. Record elapsed time since oven exit rather than labeling the sample only “post-reflow.”
3. Retest at planned checkpoints. For example, 24-hour and 72-hour measurements can show direction; they are observation points, not promised recovery deadlines.
4. Track signed error. Use RH_DUT − RH_reference for every unit. A lot average can hide outliers and process-position failures.
5. Keep calibration fixed. Do not mask a changing recovery curve with a new software offset.
6. Compare units, lots and positions. Retain failed samples without cleaning or rework until the cause review is complete.

A credible recovery report shows individual trajectories, the lot distribution, communication failures and missing checkpoints. If the observed change is comparable to reference uncertainty or fixture variation, report it as inconclusive rather than declaring recovery.

Track two different quantities: error = RH_DUT − RH_reference describes the current reading; shift = error_after − error_before describes the assembly-related change, provided both measurements use comparable RH and temperature conditions. Recovery toward the incoming baseline and compliance with the finished-product limit are separate decisions.

If SHT30 remains low at 72 hours: check the elapsed-time record, reference stability, temperature difference and process history before deciding on disposition. A 72-hour checkpoint alone proves neither permanent failure nor acceptable performance. Keep out-of-limit units on hold; a software correction or undocumented extra wait is not evidence of a controlled process.

Minimal code to calculate signed RH error at every checkpoint

The Python 3 function below keeps the supplied measurement record, adds signed RH and temperature error, and flags CRC failures, unverified CRC and invalid data. Only eligible_for_review rows belong in accuracy statistics; retain the other rows in the failure count.

import math

def analyse(row):
    result = dict(row, rh_error_pct="", temp_error_c="")
    crc = (row.get("crc_ok") or "").strip().lower()
    if crc != "true":
        result["analysis_status"] = (
            "crc_failed" if crc == "false" else "crc_unverified"
        )
        return result  # Keep the row; exclude it from accuracy statistics.
    try:
        fields = ("hours_since_reflow", "rh_dut_pct", "rh_reference_pct",
                  "temp_dut_c", "temp_reference_c")
        elapsed, rh, ref_rh, temp, ref_temp = [float(row[k]) for k in fields]
        if not all(math.isfinite(v) for v in (elapsed, rh, ref_rh, temp, ref_temp)):
            raise ValueError("Non-finite measurement")
        if elapsed < 0 or not (0 <= rh <= 100 and 0 <= ref_rh <= 100):
            raise ValueError("Invalid time or RH range")
        if not all((row.get(k) or "").strip() for k in ("serial", "lot", "process")):
            raise ValueError("Missing traceability")
    except (KeyError, TypeError, ValueError):
        result["analysis_status"] = "invalid_or_missing_data"
        return result
    result.update(analysis_status="eligible_for_review",
                  rh_error_pct=round(rh - ref_rh, 2),
                  temp_error_c=round(temp - ref_temp, 2))
    return result  # Eligibility is not a production pass.

Record RH as 0–100, temperature in °C and elapsed time in hours. Set crc_ok=true only when both measurement CRCs pass in the acquisition software. The function trusts that flag; it does not recalculate CRC from raw bytes. eligible_for_review confirms basic data checks, not thermal stability, reference validity or product acceptance.

When is contamination more likely than temporary recovery?

Contamination deserves closer investigation when RH error clusters around a material or process step after reference and thermal effects have been checked. Sensirion identifies solvents, outgassing adhesives and packaging among possible sources of increased drift. Persistent negative error alone does not identify the contaminant. [2]

Do not wash the assembled SHT30 as a troubleshooting shortcut. Sensirion's handling guide specifies no board wash and recommends no-clean solder paste. It also recommends placing the sensor in the last solder cycle on boards requiring multiple cycles. Quarantine suspect samples with their records before cleaning or reworking can alter the evidence. [1]

For conformal coating, keep the sensing opening clear and control curing and ventilation. A coating that feels dry may still be short of its specified full cure. Audit the material, mask and cure history together. [1]

Illustrative production pattern: recovery versus a wash-linked offset

In this hypothetical example, one lot remains low at 72 hours while a control lot moves toward baseline. The affected boards alone had an aqueous wash after sensor mounting. That association makes the wash step a suspect, but differences in incoming lots, reflow profiles or fixture positions could also explain the result.

Preserve samples, verify the reference, and compare incoming records and matched measurement conditions before assigning a cause. Any deliberate exposure experiment belongs on designated engineering samples under an approved investigation plan. The illustration does not recommend washing sensors or establish a recovery deadline.