There are various sources for measurement discrepancies sources. It is important to realize that the value that you measure is not necessarily the one you want. The testing conditions or component dependency factors (e.g., test signal frequency, test signal level, DC bias, voltage and current, and environment (temperature, humidity, etc.) affect the component behaviour and the measured values. On top of that, due to the instrument technique and the accessories you use, you introduce additional measurement errors. Finally, the choice of a given model necessarily implies errors.
Taking certain basic precautions in a test setup can eliminate many of the errors that cause inaccuracies. Similarly, an error analysis of a setup often can reveal that someone has been careless in choosing a component or its placement.
Scalar errors are those that are readily visible and that can be either removed by calibration or reduced in size so as to minimize their effect. Uncertainty errors are random and as such cannot be reduced or eliminated by calibration. They are caused by such things as connector repeatability and readability of the indicator. Vectorial errors are such that they can be represented as vectors rotating with frequency. These errors are the sum of the mismatch present at the source, connectors, junction of components, or any place where two pieces are joined. Such mismatches have different phases. These errors often start out small or insignificant but the fact of the matter is, they add up. Many of these errors can be reduced or eliminated by a careful choice of components and equipment and by proper consideration of compatible connectors. Such factors as leakage, impedance match, generator protection, and equipment stability will influence your readings and taking precautions will help improve them.
Leakage - refers to both loss of internal energy from the microwave circuit and the entrance of energy into the microwave circuit and can be attributed to poor assembly of the test setup. Leakage errors can be as high as 1 dB 
. The presence and, to some degree, effect of leakage may be detected by covering various points in the setup with your hand or with shielding material; or in severe cases, movement of the operator in the vicinity of the test setup, while observing changes in your readings. Measurements under such conditions can be meaningless, especially if you are trying to determine the loss in components or noise figure. In practice, connections should be clean, undamaged, and properly assembled, and all directional couplers or signal sampling devices should be properly terminated.
Generator protection - refers to steps required to minimize interaction of the microwave signal source with reflected power or other external factors which might cause changes in operating conditions during a measurement. Changes in match conditions presented to the signal generator may affect its output level in the measurement. Using a ferrite isolator in the line after the source will help eliminate reflected power to the source. At the expense of signal level, a well-matched attenuator of approximately 10 dB can also be used. When the loss in signal caused by a pad cannot be tolerated and an isolator is unavailable, the only solution is to minimize mismatches and discontinuities in the setup wherever possible by eliminating unnecessary adapters, tightening all connections to the proper torque, and ensuring connectors are clean.
General considerations - Coaxial cables and other flexible sections are also frequent causes of error. Flexing a cable may change its insertion loss, thus altering a reference level or calibration accuracy after a calibration has been performed. Every attempt should be made not to move them during a measurement. Keeping cables out of the way will help. Further errors can be introduced when it is necessary to break or make a connection when a component (e.g., DUT) is inserted in the line. Another source of error which sometimes goes unrecognized is using an adapter to establish a reference level or inserting a component then leaving it out in subsequent measurements.
There are no practical general methods of correcting any RF/microwave measurements for crosstalk of any type since these are unrepeatable by nature. If forced to, you can simulate the IC with different inductances or couplings, but the simplest approach is to adopt a package and fixturing with low crosstalk.
Not only is it impractical or impossible to correct for common-lead inductance crosstalk, but the inductances of typical packages or fixturing are the primary parasitic affecting an IC's performance. There will be several nanohenries of inductance in each bond wire and each lead to the circuit board. Additionally the bias connection will have some combination of reactances on the bypass network which can easily cause resonances. The effects of series inductances of the input or output leads can be added or removed from measurements relatively easily. The best solution is to carefully control the actual fixture inductances and power supply impedances.
The common-lead impedance problems for RFIC's is severe enough that first time designers are often shocked by the discrepancies between measured and modelled performance. Design success requires careful modelling of all of the inductance back to the ground plane.
Beware of Resonances in Power Bypass Networks
In SPICE simulations, it is very easy to attach a voltage source to Vdd or Vcc and not worry about impedance. In real life, the Vdd line is connected through inductances to one or more capacitors to provide a current return and broadband stability. The first capacitor and the connection inductance to the second capacitor can parallel resonate and cause a high impedance at the Vdd connection when a broadband low impedance is desired. Careful attention to the inductance and the parasitic resistances is necessary on both the board and the test fixturing, to ensure broadband low impedance even if the RFIC is narrowband. To minimize these resonances, small resistances can be inserted to minimize the Q of such reactive circuits. The effect of such a resonance on an amplifier can be to cause a gain peak or gain dip at the resonance frequency, or to cause an oscillation near that frequency in a high-gain amplifier.
Other Causes of Gain Peaking
In amplifiers with more than one stage, instability can occur if the path ground provided by the external circuit is "long" (high in impedance) compared to the path back through the ground return of another stage. This phenomenon can show up as "peaking" in the gain versus frequency response, an increase in input VSWR, or even as return gain (reflection coefficient greater than one) at the input of the RFIC.
 Thomas S. Laverghetta, Modern Microwave Measurements and Techniques. Artech House, 1988
 BCTM Short Course, Design, Measurements, and Modeling for RF and Microwave Integrated Circuits, 1996.
 Hewlett-Packard Company, Communications Components Designer's Catalog (HP publication number 5966-0895E (9/97))
Back to Test Guide Table of Contents