As known from radio engineering, when multiple frequency signals transit through a non-linear device, such as a typical broadband amplifier, new spectral components, i.e. the inter-modulation products caused by cross-modulation among the frequencies of the input signals, appear at the output of this device. Obviously, these new spectral components are in fact interferences degrading the original signal reception. This phenomenon, being “nonlinear”, occurs to quite different extents, depending on the level of the input signals and on the intrinsic characteristics of the amplifier: it is therefore necessary to characterize the latter in order to assess its behavior in the presence of high level 800 MHz-band LTE signals.

In a first stage at Rai-CRIT a software model of a nonlinear amplifier was developed, based on IMD (Inter-Modulation Distortion) up to 3^rd order, whose parameters are derived from a real device [6,9]. The model can be applied in order to predict the level of interference on television channels due to inter-modulation [3,4,8], and was published as ITU-R document BT.2298 [5].

The laboratory characterization of the non-linear behavior of an amplifier, in addition to validate the simulation model, allows to obtain the parameters used to simulate real cases. For this purpose, two different laboratory test-benches have been set up, according to the following two methods:

1. Two tone method [6], which allows to directly measure the magnitude of the 3^rd order inter-modulation products, and the rated output value of the amplifier as shown in the vendor data sheets.
2. Evaluation of the gain compression curve, which allows, by interpolation, to derive the coefficients of the polynomial (up to 3^rd order or even beyond [7]) approximating the amplifier’s input/output transfer curve.

For the first test bench, two RF signal generators have been used to obtain the two tones. The two signals were added via a hybrid combiner characterized by a high isolation between ports. Yet, it was necessary to provide an additional decoupling between the generators using two terminated circulators, to achieve the required purity of the spectrum at frequencies of third order inter-modulation products. A variable attenuator before the amplifier under test has allowed to explore the device behavior at various levels of the input signal. The levels of the spectral components of the third order were selectively measured with a high quality Spectrum Analyzer. The combined power measurement was made with a RF Power Meter.

The second test bench has been set up using a Vector Network Analyzer (VNA) configured in  power sweep  mode at the intended test frequency. From the gain/level curve provided by the VNA we could extract the 1^st, 2^nd and 3^rd order coefficients (and more, if wished) using polynomial interpolation. The measurement was then repeated for different frequencies, for each amplifier under test.

The used mathematical model and the two adopted measuring techniques have general validity for any broadband amplifier. The results of the two methods, obtained on several commercial amplifiers of different brands and types, were found to be in good agreement with the data declared by the manufacturers.  This is especially true for the first measurement method, as it is the same one used by the manufacturers;  however, this method makes it possible to evaluate the 3^rd order polynomial coefficient only. The second measurement method, based on the gain compression curve, allows to characterize the device more realistically (by choosing the appropriate order of the polynomial); on the other hand it proved quite sensitive to the input power level, which must then be carefully chosen.