RF power amplifiers can impart distortion to the amplitude and phase of the incident signal.  Some distortions (as a function of modulation type) can be categorized as AM to AM distortions or AM to PM distortions.  These types of distortions cause "spectral regrowth".

AM to AM distortions are defined as changes in the amplitude of the signal voltage at the output of the amplifier under test, due to changes in the amplitude of the incident signal.

AM to PM distortions are defined as changes in the phase of the signal voltage at the output of the amplifier under test, due to changes in the amplitude of the incident signal.

Moreover, the rate of spectral regrowth is different for AM to AM as opposed to AM to PM.

The degree to which distortions like these affect the incident signals is a function of both the characteristics of the modulated signal, the nature of the distortions and the level of overall system distortion.

• Background

If the PA non-linearities are memoryless, ie. the output signal is only a function of the present value of the input signal voltage, then the distortion products can be modeled as a Taylor series expansion.

A good example of this was given in "Regrowth of Data Spectral Sidelobes from AM/PM effects and Hard Limiting", Microwave Journal, Oct. 2001, pp. 148, by Frank Amoroso.

r(t) = k₁s(t) + k₃(s(t))³ + k₅(s(t))⁵ +...

where:

r(t) = amplifier output

s(t) = amplifier input

k are constants

If the system is bandwidth limited and filters are not an option for spectral containment due to cost/size constraints, then this "spectral regrowth" type of distortion can not be tolerated in the design.  In this case, the modulation scheme must be chosen with this in mind.

• Effects

The effect that these amplifier non-linearities have is dependent upon the modulation scheme employed in the system.

In general, modulation schemes can be classified as either linear or non-linear.

The modulation information contained in non-linear modulation schemes (like GMSK) is not particularly susceptible to PA non-linearities because non-linear modulation schemes tend to have constant envelopes.  (This is of course a function of the baseband pulse shaping employed.)  Thus, the PA can be operated in saturation (non-linear region) where the power efficiencies are higher without worrying about non-linearities.  Operating in the PA non-linear region does however generate spurious spectral sidelobes and is thus not very spectrally efficient.

On the other hand, linear modulation schemes (like pi/4 DQPSK) tend to be more spectrally efficient (as opposed to non-linear modulation schemes).  However, the PA must be operated in it's linear region in order to preserve the benefits of spectral efficiency and to accommodate the non-constant envelope waveform.

This requires the need to operate the PA in it's linear region (power back-off) which is not as power efficient, or requires a means to compensate for the PA non-linear distortions.

• Linearization Techniques

Predistortion Linearization

In this technique, a device called a "linearizer" is used to distort the signal before it is applied to the RF amplifier.  The predistortion is computed in such a way that the effects of the linearizer and the RF amplifier tend to negate each other.

Feedforward Linearization

In this technique, two control loops are used for the correction of amplitude and phase distortion.  The inner (carrier cancellation) loop cancels out the desired signal leaving only an inverted copy of the distortion (intermodulation/cross modulation products).  The outer loop then sums the original (signal + distortion) with the negative of the distortion leaving only the original signal at the output.

Note that special care must be taken to ensure that ensure that all of the carrier power is extracted in the carrier extraction loop, or the "ideal amp" in the distortion extraction loop might get saturated.

Also note that the performance of this design is very much dependent upon the distortion extraction loop's ability to correct and compensate over a wide range of power and temperature swings etc.

Envelope Elimination and Restoration (EER)

In this technique, the envelope of the incident signal is removed with a limiter, thus creating a constant envelope signal.  An envelope detector is used to capture the magnitude information.  Magnitude and phase information of the original signal can then amplified and combined to reproduce the original signal.

The next app note in this series will focus on synthesizers.