Linearity

[Adam Petrus]

Continued on the Drift question

It appears the relative power measured p (in dB) is quantized into either 0.01 or 0.000001 dB steps depending on whether version 0x0 or 0x1|0x2 is used. From my experience 0x2 definition is commonly used for sdatb files. 

However it appears the precision of the 0.000001 is not used but 0.02dB is used. In fact for every linearity consideration 11 inputs and dependences are created. They are spread across weighting by a distribution (looks like normal).

I assume this is done to allow the chance of uncertainty terms of similar linearity levels to be correlated against each other. 

My question is currently I cannot exactly replicate your distribution. Is a normal or is something else? Have I got this correct? 

[Michael Wollensack METAS]

Hi,

> However it appears the precision of the 0.000001 is not used but 0.02dB is used.

yes for version 2

> In fact for every linearity consideration 11 inputs and dependences are created.

yes for version 2

>  They are spread across weighting by a distribution (looks like normal).

Yes it's a normal distribution. The weights are computed using the CDF, see the following code:

               double two_sqrt_sigma = 2 * Sqrt(0.01 / 0.02);

               double error = scaled_power - rounded_scaled_power; // between -0.5 ... 0.5 

               double weight;
                // Weight (normal distribution CDF)
                if (i == 0)
                {
                    weight = Sqrt(0.5 * (Erf((x[i] + 0.5 - error) / two_sqrt_sigma) + 1));
                }
                else if (i + 1 == n)
                {
                    weight = Sqrt(0.5 * (1 - Erf((x[i] - 0.5 - error) / two_sqrt_sigma)));
                }
                else
                {
                    weight = Sqrt(0.5 * (Erf((x[i] + 0.5 - error) / two_sqrt_sigma) - Erf((x[i] - 0.5 - error) / two_sqrt_sigma)));
                }

> I assume this is done to allow the chance of uncertainty terms of similar linearity levels to be correlated against each other. 

yes

correlation of 0.5 with 0.05 dB offset

Hope this helps.

Regards

Michael

[Michael Wollensack METAS]

Hi, 

the following plots shows the correlation between two power levels.

Regards

Michael