ZAFAR AHMAD — SYNTHETIC RECORDING-COMPARISON DEMONSTRATION
Independently authored portfolio example; created 10 October 2026.
This example uses no client recordings, report content, software or model weights.

INPUT AND CONTROLS
Each input is a two-second mono floating-point signal at 8,000 samples/s (16,000 samples).
x[n] = sum(a_f * sin(2*pi*f*n/fs)) for f in {100, 1000, 3000} Hz, with fs = 8000 samples/s and n = 0..N-1 (N = 16,000).
All tones start at zero phase. Sample rate, duration and tone frequencies are fixed.
There is no randomness, normalisation, filtering, resampling, clipping or added noise.
All frequencies complete an integer number of cycles over the record.

METRIC DEFINITIONS
RMS = sqrt(sum(x[n]^2)/N), in normalised digital-amplitude units.
Level (dBFS) = 20*log10(RMS), with digital amplitude 1 as the reference.
Under this convention a full-amplitude sine has -3.0103 dBFS RMS level.
100 Hz bin power = 2*((sum(x[n]*sin(2*pi*100*n/fs))/N)^2
                      +(sum(x[n]*cos(2*pi*100*n/fs))/N)^2).
100 Hz power share (%) = 100*(100 Hz bin power)/(sum(x[n]^2)/N).
The coherent rectangular-record projection needs no window correction in this example.
This is one exact frequency-bin fraction, not a general low-frequency band metric.

COMPUTED RESULTS
ID | Tone amplitudes at 100/1000/3000 Hz | RMS | dBFS | 100 Hz power share
A | 0.20/0.10/0.00 | 0.158114 | -16.02 | 80.00%
B | 0.40/0.10/0.00 | 0.291548 | -10.71 | 94.12%
C | 0.10/0.05/0.00 | 0.079057 | -22.04 | 80.00%
D | 0.20/0.10/0.05 | 0.162019 | -15.81 | 76.19%

READING THE COMPARISON
A is the baseline. B doubles only the 100 Hz component. C halves both baseline components.
D adds a 3000 Hz component. The table separates digital level from relative spectral composition.
C is 6.02 dB below A; both have 80.00% of their power in the 100 Hz bin.
Known generation conditions explain these differences; they do not establish a hardware effect.

REPRODUCE
Use Python 3.9 or later; no third-party dependencies are required.
python recording-comparison-demo.py --output-dir demo-output
The command generates recording-comparison-demo.csv, recording-comparison-demo.svg,
and recording-comparison-demo.txt. Displayed results are rounded; calculations use floats.

LIMITS
This report example does not recreate the private 156-combination accessory study.
It does not measure loudness, calibrated sound pressure, intelligibility, SNR or hardware quality.
Real recordings require acquisition controls, appropriate frequency analysis and uncertainty checks.
The coherent-bin assumptions here do not transfer automatically to arbitrary recordings.

Portfolio: https://zafar-ahmad-5359.github.io/audio.html#comparison-demo
Contact: ahmadzafar577@gmail.com