Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- C:\Python>C:\Python\madev_plot_V622.py C:\Python\7V_long_full.txt
- Loading C:\Python\7V_long_full.txt
- Loaded 434140 samples (0 lines skipped)
- tau0 = 0.506287 s (median timestamp interval; timing jitter ~0.06%)
- Computing MADEV (N=434140, max 60 points)...
- 56 tau values [0.5063s ... 7.327e+04s]
- Computing ADEV...
- 56 tau values
- Computing HDEV...
- 56 tau values
- Computing statistics...
- ======================================================================
- MADEV/ADEV/HDEV Statistical Report -- 7V_long_full.txt
- ======================================================================
- N=434,140 tau0=0.5063 s Duration=2.198e+05 s (3663.29 min / 61.0548 h)
- ----------------------------------------------------------------------
- SANITIZATION
- Spikes flagged : 1674
- Spikes replaced : 1674
- Non-finite removed : 0
- Duplicate times rm : 0
- Timing jitter (MAD) : 0.06% of tau0
- Note : replaced 1674 isolated spikes using local median (window=9, nsigma=6.0, diff_nsigma=8.0)
- ----------------------------------------------------------------------
- MOST PRECISE VALUE (grand mean of all N samples)
- Best estimate : 7.05652203 V
- Uncertainty u(V) : 29.453 nV (k=1, MADEV-derived)
- Relative precision : 0.004174 ppm
- 95% CI (t-dist) : 7.05652202 .. 7.05652204 V
- 99% CI (t-dist) : 7.05652201 .. 7.05652204 V
- ----------------------------------------------------------------------
- BASIC STATISTICS
- Mean : 7.05652203 V
- Median : 7.05652210 V
- Std dev sigma : 3.516 uV
- Std error SEM : 5.3363 nV
- Peak-to-peak : 56.6 uV
- Min : 7.05648540 V
- Max : 7.05654200 V
- Skewness : -0.1000 (0=symmetric)
- Kurtosis (excess) : -0.0328 (0=Gaussian)
- ----------------------------------------------------------------------
- NORMALITY / TAIL DIAGNOSTICS (large-sample strategy)
- Strategy : 100k deterministic subsample for JB/AD; full tails
- Test sample N : 100,000 (QQ N=20,000)
- Jarque-Bera p : 1.551e-46 (large N: sensitive)
- Anderson-Darling : A2=22.72, 5% crit=0.787 (fail)
- QQ correlation : 0.999475 (closer to 1 is straighter)
- |z| > 3 sigma : 0.2244% (Gaussian ideal ~0.27%)
- |z| > 4 sigma : 0.007371% (Gaussian ideal ~0.0063%)
- Robust |z| > 3 : 0.208% (MAD-scaled)
- Practical flag : small deviation from Gaussian
- ----------------------------------------------------------------------
- ROBUST / CORRELATED-SAMPLE STATISTICS
- Trimmed mean (10%) : 7.05652206 V
- Robust sigma (MAD) : 3.5582 uV
- Winsorized std : 2.9212 uV
- Effective N : 596.5 (autocorr lag used: 1000)
- Corr. SEM : 143.96 nV
- ----------------------------------------------------------------------
- RAW VS CLEAN COMPARISON
- Raw mean : 7.05652191 V
- Clean mean : 7.05652203 V
- Mean delta : 118.11 nV
- Raw std sigma : 63.1 uV
- Clean/raw std ratio: 0.056x
- ----------------------------------------------------------------------
- MADEV NOISE FLOOR (NRC slides 11-13; n=2 for ModADev)
- Single-sample MADEV: 1.4352 uV
- Noise floor MADEV : 422.74 nV at tau=1067 s
- 95% chi2 CI (low) : 385.4 nV
- 95% chi2 CI (high) : 468.15 nV
- Degrees of freedom : 204.0 (NRC slide 28: (N-2m)/m)
- Optimal m : 2107 samples
- Indep. windows : 206
- Averaging gain : 3.40x vs single sample
- Floor ppm : 0.05991 ppm
- SNR vs floor : 16692484.85 (144.45 dB)
- SNR single sample : 4916772.16 (133.83 dB)
- TDev at tau_opt : 260.36 uVs (tau/sqrt3 x MADEV)
- Dominant noise : Flicker freq./1f
- ----------------------------------------------------------------------
- ADEV NOISE FLOOR
- ADEV floor : 521.77 nV at tau=1305 s
- Dominant noise : Flicker freq./1f
- ----------------------------------------------------------------------
- HDEV NOISE FLOOR
- HDEV floor : 432.59 nV at tau=1596 s
- Dominant noise : Flicker freq./1f
- NOTE: If HDEV << ADEV at long tau, linear drift is significant.
- ----------------------------------------------------------------------
- DRIFT-AWARE STATISTICS
- Detrended mean : 7.05652203 V
- Detrended std : 3.2875 uV
- Detrended MADEV : 422.97 nV at tau=1067 s
- Detrended ADEV : 522.03 nV
- Detrended HDEV : 432.59 nV
- MADEV drift ratio : 0.999x (raw/detrended floor)
- ADEV drift ratio : 1.000x
- HDEV drift ratio : 1.000x
- Drift influence : no
- ----------------------------------------------------------------------
- DRIFT & SYSTEMATIC ERRORS
- Drift rate : 70.748 nV/h (0.01003 ppm/h)
- Drift per 24 h : 1.6979 uV (0.2406 ppm/24h)
- Total drift : 4.3195 uV over 2.198e+05 s (2 days 13 hours 3 min)
- R2 of linear fit : 0.12578
- Drift uncertainty : 0.28308 nV/h
- Bootstrap drift CI : 62.086 nV/h .. 79.907 nV/h
- ----------------------------------------------------------------------
- STATIONARITY / BOOTSTRAP
- Bootstrap mean CI : 7.05652186 .. 7.05652219 V
- Bootstrap block : 659 samples
- Split mean shift : 0.820 sigma
- Split std ratio : 1.057x
- Nonstationary flag : YES
- ----------------------------------------------------------------------
- ADVANCED DIAGNOSTICS (recommendations 10-15)
- Fitted MADEV slope : 0.059 (Flicker freq./1f, confidence=low, R2=0.375)
- Noise-floor plateau: 389.8 .. 5344 s (14 tau points overlap CI)
- Residual std : 3.2875 uV
- Residual lag-1 rho : 0.8094
- Residual MADEV : 422.72 nV at tau=1029 s
- Robust drift fit : 71.503 nV/h (Huber IRLS on deterministic subsample, N=20,000)
- Robust-vs-OLS drift: 0.75532 nV/h
- 24h periodic fit : good (2.54 cycles observed)
- 24h sine amplitude : 967.71 nV peak, 684.28 nV RMS
- 24h sine p-p : 1.9354 uV (R2=0.0385, variance=3.85%)
- 24h hour-bin p-p : 2.8289 uV
- 24h peak phase : 12.60 h after run start
- 24h equiv drift : 80.643 nV/h
- 24h-corrected std : 3.4475 uV (raw/corrected=1.020x)
- 24h-corrected mean : 14.11304405 V
- 24h-corrected MADEV: 421.02 nV at tau=1067 s
- 24h MADEV impact : 1.011x at tau=7.327e+04 s
- 24h influence : moderate (flag=YES)
- ----------------------------------------------------------------------
- STABILITY SUMMARY
- Mean +/- corr SEM : 7.05652203 V +/- 143.96 nV
- Bootstrap mean CI : 7.05652186 .. 7.05652219 V
- MADEV floor : 422.74 nV at 1067 s
- Drift / 24 h : 1.6979 uV
- 24h periodic effect : moderate (1.9354 uV p-p, R2=0.0385)
- Quantization flag : no
- ----------------------------------------------------------------------
- QUANTIZATION
- Detected LSD : 50 nV (mode-of-positive-differences)
- LSD percentile : diagnostic only / ignored (8.882e-16 V)
- LSD mode estimate : 50 nV
- Levels hit : 393 of 1133 possible
- Quantized range : 7.05648539 .. 7.05654199 V
- Quant. origin : 7.05648539 V (common grid)
- Missing levels : 740
- Occupancy : 34.69% of codes in sanitized range
- Occupancy entropy : 0.833 (1=evenly used occupied codes)
- Quantization RMS : 14.434 nV (=LSD/sqrt(12))
- Sigma / LSD : 70.321
- Quantization limit : no
- ----------------------------------------------------------------------
- RECOMMENDATIONS
- - Use corrected SEM or bootstrap CI instead of naive SEM when samples are correlated.
- - Compare raw and cleaned MADEV results before trusting the cleaned estimate.
- - Nonstationarity flag is set; avoid quoting one grand mean without segment analysis.
- - LSD percentile estimate was ignored as a floating-point artifact; mode-based LSD was used.
- - 24h periodic/environmental cycle detected; correlate voltage with lab temperature/day-night HVAC and report amplitude/R2/MADEV impact.
- - Quote the MADEV floor as a plateau range, not a single exact tau, because neighboring tau values overlap statistically.
- - Noise-type slope confidence is low; avoid over-interpreting the color/noise labels near the floor.
- - Residuals remain strongly autocorrelated after linear drift removal; inspect environmental or periodic effects.
- ======================================================================
- Rendering plot...
- Saved -> 7V_long_full.madev.png
- C:\Python>
Advertisement