README.md

We fit a exponential function for $f$: $f(V) = a \exp(V / s)$ and obtain $a \approx 0.02485702\ \mathrm{A}$ and $s \approx 0.229551831\ \mathrm{V}$. With this the deviation from the measured values from the total model $P(V) = 414\ \mathrm{mW} + V · 130 · f(V)$ is always below $8 \ \mathrm{mW}$, indeed it is atmost $\approx 3.19\ \mathrm{mW}$ and on average $\approx 1.28\ \mathrm{mW}$.

When not taking a fixed offset of $414\ \mathrm{mW}$, but instead also leave this as a variable of the fit, we obtain $\approx 412.3\ \mathrm{mW}$ for the offset, $a \approx 0.0249451511\ \mathrm{A}$ and $s \approx 0.229653915\ \mathrm{V}$ with a maximum error of $\approx 2.28\ \mathrm{mW}$ and a average error of $\approx{0.74}\ \mathrm{mW}$.

fit.py

#!/usr/bin/env python3

import numpy as np
from scipy.optimize import curve_fit

data = np.loadtxt("agx_sram_mW_mV.txt")
x = data[:,0]
y = data[:,1]

def model(x, fac, s):
    return 414 + x / 1000 * 130 * fac * np.exp(x / s / 1000)

def evaluate(x, y, fitted):
    dist = np.abs(fitted - y)
    assert len(np.where(dist > 8)[0]) == 0

def fit(model, x, y):
    popt, pcov = curve_fit(model, x, y)
    print(popt)
    fitted = model(x, *popt)
    evaluate(x, y, fitted)
    return fitted

fit(model, x, y)