What is the difference between Type A and Type B uncertainty evaluation in RF measurements?

What is the difference between Type A and Type B uncertainty evaluation in RF measurements? The GUM (Guide to the Expression of Uncertainty in Measurement) defines two methods for evaluating uncertainty components, and most RF uncertainty budgets use both: (1) Type A evaluation (statistical): based on the statistical analysis of a series of repeated measurements. You make N repeated measurements of the same quantity and calculate the standard deviation. The standard uncertainty (Type A): u_A = s / √N, where s is the sample standard deviation and N is the number of measurements. Example: you measure the insertion loss of a filter 10 times: values = 3.05, 3.08, 3.04, 3.07, 3.06, 3.09, 3.05, 3.07, 3.06, 3.08 dB. Mean = 3.065 dB. Standard deviation s = 0.016 dB. Standard uncertainty u_A = 0.016 / √10 = 0.005 dB. Type A captures: random effects (connector repeatability, thermal noise, instrument noise) and any variability that changes between measurements. (2) Type B evaluation (non-statistical): based on other information: manufacturer specifications, calibration certificates, published data, engineering judgment, or experience. No repeated measurements are needed. The uncertainty value is estimated from the available information and a probability distribution is assigned. Common distributions: rectangular (uniform): the value is equally likely anywhere in the range ±a. Standard uncertainty: u_B = a / √3. Normal (gaussian): used when a stated uncertainty has a coverage factor k. Standard uncertainty: u_B = U / k (typically U / 2 for k=2). U-shaped: used for sinusoidal variations (e.g., mismatch). Standard uncertainty: u_B = a / √2. Example: the power sensor calibration factor has a stated uncertainty of ±0.12 dB (k=2, 95% confidence). Standard uncertainty: u_B = 0.12 / 2 = 0.06 dB. Another example: the temperature coefficient of the sensor is not characterized, but you know it is within ±0.05 dB over the temperature range. Assuming a rectangular distribution: u_B = 0.05 / √3 = 0.029 dB. (3) Combining Type A and Type B: they are treated identically in the uncertainty budget. Combined standard uncertainty: u_c = √(u_A1² + u_A2² + u_B1² + u_B2² + ...). There is no mathematical distinction between Type A and Type B contributions in the combination. The classification is about how the value was obtained, not how it is used.