Before trusting any impedance reading, identify the actual converter and its documented timing. Magnitude and phase processing can look precise while the underlying acquisition cannot support the accuracy claim.
An LCR meter is a chain: an excitation path, an acquisition front end, and a processing stage that extracts magnitude and phase. Cost and accuracy claims are easy to make and hard to support, so the useful questions are what the converter’s own documentation states and what a comparison against a reference would actually show. Related metering references include the capacitance meter reference and the auto-ranging ohmmeter reference.
Name the converter before judging the claim
Start by identifying the exact MCU and converter in the design you are evaluating. Two documented examples show why the distinction matters, and neither one stands in for the part you actually hold:
| Aspect | STM32F3 HAL ADC path (one vendor example) | TI ADS1115 (comparison only) |
|---|---|---|
| What the source documents | Vendor source implements ADC calibration entry points such as HAL_ADCEx_Calibration_Start | Product documentation describes a 16-bit, multiplexed, I2C delta-sigma ADC with programmable gain |
| Rate | Calibration and conversion behaviour are conditional on the device variant | Maximum specified data rate is 860 samples per second |
| Supply | Not stated in this source | Listed supply range is 2 V to 5.5 V |
| Fit for high-frequency FFT acquisition | Not established | Not a replacement; the slow multiplexed rate does not suit a high-frequency acquisition scheme |
The STM32F3 HAL is one vendor example, and the ADS1115 is an example of a slow multiplexed ADC, not a drop-in substitute. A programmable input range does not authorise exceeding supply-related input limits either.
Separate processing from accuracy
FFT phase discrimination is a processing method. However neatly it reports a value, it does not by itself establish that the reported component value is accurate. The mechanism that produces magnitude and phase and the accuracy of the final reading are two different claims, and only the second needs independent comparison to be credible.
Ordered evaluation process
- Identify the real hardware. Confirm the exact MCU and converter variant.
- Require documented timing. Get the sample rate and channel timing from the device’s own documentation.
- Map the processing split. Understand which stage produces magnitude and which produces phase.
- Define comparison requirements. Decide what reference instrument or known standards the readings must be compared against before any accuracy is claimed.
- Respect input limits. Programmable gain shapes the range but does not override supply-related limits.
- Scope the claim. State the measurement conditions under which any accuracy figure would be valid.
Reader questions
Does an on-chip calibration routine guarantee accurate readings?
No. A vendor source may implement calibration entry points, and calibration and conversion behaviour are conditional on the STM32F3 device variant, so check your own variant’s documentation rather than treating a routine as a guarantee.
Could a slow multiplexed ADC replace the acquisition front end?
Work out the sample rate your own scheme needs first. The 860 samples-per-second figure is a comparison point, not a licence to reuse a converter in a high-frequency path.
Why is FFT processing not enough for an accuracy claim?
Phase discrimination produces a value; it does not prove the value is right. Independent comparison against a reference is what would support accuracy.
What should an honest specification state?
The converter, the documented timing, the test conditions and the reference used for comparison, or an explicit statement that no such comparison exists.
Documentation
- STMicroelectronics STM32F3 HAL ADC extended driver source
- Texas Instruments ADS1115 product documentation
Scope note: these links document a vendor ADC driver and one commercial ADC part only. No measurement, calibration or code execution was performed, and no AN2834 figures are treated as verified; the analog front end, frequency behaviour and any claimed measurement range or accuracy still need independent testing.
