Reveal hidden structure in ADC & High-Speed IO validation data

sinlPlugin.tsx — SINL: Sine INL/DNL Analysis

Plugin ID: sinl

Category: Signal / ADC Characterisation

Author: Guy Steiff

Version: 1.0.0

What It Does

Measures Integral Nonlinearity (INL) and Differential Nonlinearity (DNL) from ADC captures driven by a clipped (saturated) sine wave — the classic Okawara-T method. Designed for ADCs where the input signal clips at the rails, making standard histogram methods invalid.

> Not for ENOB. Use SMEAS for spectral metrics (SNR/SNDR/ENOB/THD/SFDR).

> Use SINL for static linearity: INL/DNL, missing codes, truncation zone.

Algorithm: Okawara-T (Truncated Sine Histogram)

Port of inl_tool.py. The algorithm:

1. Histogram raw ADC samples into 2^adcRes bins

2. Find `intMinbin` / `intMaxbin`: first/last bin with ≥ minSizeBin samples — defines the reachable code range

3. Truncation search: within ±avoidanceRadius codes of each rail, find the peak of the U-shaped PDF — this is the truncation boundary (intTruncLow / intTruncHigh). The saturation peak inflates those bins; truncating there avoids bias

4. PDF / CDF / cos(CDF): normalize histogram → cumulative → apply cos(π·CDF) to linearize the sine-wave distribution

5. DNL: −d(cosCdf)/codeSizeOverAmp − 1 over the active zone; zero-padded outside

6. INL: cumulative sum of the linearized CDF, referenced to the truncation boundary; zero-padded outside

7. 3rd-order polynomial fit to INL for smooth trend extraction and p2p metric

Key insight: the cos(CDF) transform converts the sine-wave PDF into a locally-linear function; its derivative directly gives the code-width deviation (DNL), and its integral gives the accumulated offset (INL).

Inputs

Scope: all parameters are global (shared across all files in a batch). sampleColumn is a column-select with defaultScope: 'global' — auto-seeded to the column with the most unique values when files are loaded.

Outputs (CSV row per file)

Figures (per file, interactive)

All figures are Recharts React components. Opened via the Plot button in the file table.

1. PDF — Probability Density Function

Shows the sine-wave U-shaped PDF histogram across all codes.

Annotations:

• 🟡 Shaded zone: peak-search window around codeMin (low rail)

• 🟣 Shaded zone: peak-search window around codeMax (high rail)

• Gray dashed: minbin and maxbin boundary lines

• Gray dashed: +r / −r radius markers (edge of avoidance window)

• Red solid: trunclow and trunchigh — final truncation boundaries

Layout note: all label offsets use PDF_CHART_H * fraction expressions (not magic px numbers) so label placement scales proportionally if the chart height changes.

2. DNL — Differential Nonlinearity per code

Line plot of DNL [LSB] vs code index over the active range (outside truncation zone shown as null, not plotted).

Annotations:

• y=0 reference line

• Red dashed: minimum DNL with code label

• Green dashed: maximum DNL with code label

Y-axis domain computed from finite data (immune to stale singulars).

3. INL + 3rd-order polynomial

Line plot of measured INL [LSB] vs code index, with the 3rd-order polynomial fit overlay.

Annotations:

• y=0 reference line

• Red dashed: minimum INL

• Green dashed: maximum INL

• Amber dashed: polynomial overlay (legend: "3rd-order poly")

Polynomial is restricted to the active code range — no extrapolation blowup outside the truncation zone.

Debug Tables (downloadable)

Download via the Debug button in the file table → CSV or JSON.

Error Handling

Errors surface per-file in the pipeline status column — other files in the batch continue processing.

Cohabitation Notes (running alongside SMEAS)

• sampleColumn uses key sampleColumn — distinct from SMEAS's targetColumn. No conflict.

• Output keys (inl_max, dnl_max, etc.) are all SINL-specific — no overlap with SMEAS outputs.

• Both plugins can be active simultaneously; results are merged into one CSV row per file.

Python Twin

When the platform's "Export as Python script" feature is available, it will generate:

# Auto-generated by image_site
from sinl_tool import run_sinl
import pandas as pd, glob

params = dict(
    sample_column='ADC sample [LSB]',
    input_mode='codes',
    min_code=0,
    max_code=2047,
    avoidance_radius=40,
    min_size_bin=2,
    missing_threshold=-0.9,
)

rows = [run_sinl(pd.read_csv(f), **params) for f in glob.glob('data/*.csv')]
pd.DataFrame(rows).to_csv('results.csv', index=False)

File

app/components/plugins/sinlPlugin.tsx — fully self-contained (~800 lines):

Algorithm + CSV parsing + Recharts figures + manifest + plugin export. No external dependencies beyond recharts.