GB2282290A - Radio transmitter with power amplifier linearizer. - Google Patents

Abstract

A radio transmitter is provide having a power amplifier and a linearizer arrangement (for example Cartesian feedback) for compensating for non-linearity in the power amplifier (36). A signal is fed back from an output of the power amplifier to the linearizer means, whereby the amplifier, feedback means and linearizer means form an amplifier loop having a varying loop gain. A training signal is applied to the amplifier during a training mode of operation. An automatic gain control arrangement (26) in the amplifier loop is activated during at least a portion of the training mode of operation to maintain constant closed loop gain during training. <IMAGE>

Description

RADIO TRANSMITTER WITH POWER AMPLIFIER LINEARIZER Field of the Invention This invention relates to a radio transmitter having a power amplifier and a linearizer arrangement for compensating for nonlinearity in the power amplifier. An example of such an arrangement is Cartesian feedback with training.

Background to the Invention Linear transmitters use closed loop techniques to improve linearity (for example W09208280). The open loop has serial cascaded amplifiers, mixers, attenuation and filters having an open loop DC gain of typically 90 dB. This open loop gain affects the close-in and far-out splatter, determines the stability performance and affects the clip detection and phase adjustment training.

A principal reason for using a closed loop is to improve linearity of an ordinary class AB RF power amplifier (PA). Generally the close-in splatter caused by the PA that is closed in a loop will be reduced by the loop gain relative to the splatter that is caused by the same PA without closing the loop. Far-out splatter, which is caused by forward path noise coming out through the forward path amplifiers, is directly dependent on the forward path gain. It is observed that close-in and far-out splatter are proportional to loop gain.

Gain and phase margin are directly connected with loop gain.

The open loop gain will vary the gain margin proportionally and will affect the phase margin depending on the loop phase response.

Clip detection is done by ramping the input signal and monitoring the summing junction error level, comparing it to a known level that indicates clip. The error level before clip is proportional to 1/(loop gain).

It is necessary to keep the error level before clip constant in order to detect the correct clip point. Phase adjustment is needed to ensure stability when the loop is closed. An input signal is applied at open loop and the phase is adjusted to -1800 deg. at the feedback summing junction. A power amplifier has an AM to PM response, so the input signal that is applied at the phase training should drive the PA to output power which is approximately it's average output power. For a constant input signal the output power will be directly dependent on the forward path's gain.

It can be seen that loop gain is one of the most important parameters in the closed linearizing loop. Loop gain variations, temperature, frequency and radio-to-radio variations will affect the transmitter's performance as explained above. Radios will have different splatter depending on its initial loop gain, temperature and frequency. Gain and phase margin have to be over-designed in order to ensure the required margin at worst case conditions. Phase adjustment and clip detection training signals have to be tuned at the factory.

Copending patent application No. 9317316.9 describes an improvement using a look-up table to maintain accurate training versus temperature and frequency.

Summarv of the Invention According to the present invention, a radio transmitter is provided comprising a power amplifier, linerarizer means coupled to the amplifier for compensating for non-linearity in the power amplifier, feedback means for feeding a signal from an output of the power amplifier to the linearizer means, whereby the amplifier, feedback means and linearizer means form an amplifier loop having a varying loop again, and training means for applying a training signal to the amplifier and adjusting the linearizer means during a training mode of operation.The invention is characterised by automatic gain control means in the amplifier loop and control means for activating the automatic gain control means during at least a portion of the training mode of operation to maintain constant closed loop gain during that portion and to deactivate the automatic gain control means during a transmit mode of operation.

In this manner, the automatic gain control (AGC) controls the loop gain and maintains the best performance for the transmitter in any condition of temperature, frequency, etc. The error level from the loop summing junction which drives the amplifier, can be sensed and adjusted to be maintained constant. This can be done by varying attenuators in the forward path.

The training means preferably apply a gain training signal to the amplifier during a gain training portion of the training mode, with the control means being arranged to activate the automatic control means during the gain training portion.

The gain training portion is preferably a linear portion of linearly increasing amplitude of the training signal.

Brief Description of the Drawings Fig. 1 shows a training signal suitable for use in the present invention.

Fig. 2 shows a block diagram of a power amplifier and linearizer arrangement in accordance with the invention.

Fig. 3 shows the error signal at point C in Fig. 2 during amplitude training.

Detailed Description of the Preferred Embodiment Fig. 1 shows the training signal that is applied to the loop's input.

The training starts with one period of sine wave that is used to adjust the phase. This sine wave is applied to the open loop and the total loop's phase is measured at the feedback input to the summing junction and then adjusted. After the phase has been adjusted the loop is closed at the time point marked by Tx.

The second stage of the training is clip detection. A ramp generator is now connected to the input. The power input increases linearly, driving the PA towards maximum output power. When the PA starts compressing the error voltage rises and when it reaches the threshold level clip is indicated.

Fig. 2 shows a block diagram of a Cartesian feedback transmitter modified in accordance with the invention so as to include AGC circuitry. The radio transmitter comprising a DSP (digital signal processor) 10, DIA (digital to analog converter) for I channel 11, DIA for Q channel, input attenuation for I channel 13, input attenuation for Q channel 14, loop filter for I channel 31, loop filter for Q channel 32, a loop attenuator for I channel 33 and a loop attenuator for Q channel 34, a low pass filter 35, an up converter 61, a power amplifier 36 and an antenna 50. The power amplifier 36 output signal is sampled by the coupler 55 and attenuated by the feedback attenuator 41.The feedback attenuator 41 is connected to the down converter 62, then splits to I & Q and goes to the baseband amplifier for I channel 43 and to the baseband amplifier for Q channel 44, thus closing the loop at summing junction 45, 46. The loop has also a microprocessor 21 which controls the transmitter's parameters, such as timing. It also controls a ramp generator 22 and a timer 24. The timer controls an AGC block 25 which senses the error voltage and controls the loop attenuators (for I & Q channels 33 & 34). The AGC block 25 contains the dynamics to control the AGC loop.

The transmitter is trained before transmitting data. The first step is phase training to adjust the loop's phase to be -180 deg. at the summing junction 45, 46. Then the loop is closed at time TX by closing switch 47 and the amplitude training detects the clip point of the power amplifier 36. During the amplitude training state, after the error settles and before the clip occurs, the loop gain is trained by sensing the error level and adjusting it to be 25 mV by changing the loop attenuators 33 & BR< 34.

On accepting a request for transmission, the power amplifier 36 is turned on and the microprocessor 21 sets the transmitter to transmit at open loop i.e. the feedback path is disabled. The DSP 10 generates one period of sine wave to the I channel and zero to the Q channel as shown in Fig. 1. The DSP's digital samples are converted to an analog signal by the D/A 13 and this passes the forward path through the power amplifier 36 to the antenna 50. Since the loop gain has been adjusted by the loop attenuators 33 & 34, the output power at the phase training stage will be constant, resulting in accurate phase adjustment and maximum phase margin.

After the loop is closed the microprocessor sets the time 24 to connect the ramp generator 22 to the loop's input.

While the input starts ramping the error steps up and it settles after a settling time of 50-80 microseconds. This is shown in Fig. 3.

After 100-150 microseconds. the AGC circuitry is activated to adjust the error level to be 25 mV by changing the loop attenuators 33 & 34. The attenuators should be set within 150-200 microseconds and then frozen until the next training. The attenuations of the loop attenuators 33 & BR< 34 are again updated at the next training period.

The input ramp increases output power of the power amplifier 36.

When the power amplifier 36 starts clipping the error increases rapidly and when it crosses the threshold level that indicates clip, the ramp is sampled. The linearizer then adjusts the input attenuators (for I channel 13 and for Q channel 14) so that maximum input level of the DSP 10 is 2-3 dB below the sampled level of the ramp. This ensures that the data's output power will be 2-3 dB below clip power.

The fact that the loop gain is adjusted before the clip is detected ensures an accurate clip detection which will allow lower overhead ratio (i.e. safety margin) and will avoid additional splatter that is caused by the amplitude training.

After the training sequence, the timer 24 connects the input attenuator 13 to the loop's input and the transmitter is ready to transmit data. The microprocessor 21 controls the DSP 10 to start sending data to the transmitter. The transmitter will operate at optimal stability conditions and the constant loop gain, achieved by the AGC operation, will result in constant adjacent splatter.

Fig. 3 describes the error signal (at point C) during the amplitude training. After a transition response that takes 50-80 microseconds the error signal is stabilized and stays constant until clip occurs. The error level before clip is about 20-30 mV. This level is inversely proportional to loop gain. This fact is used to adjust the loop gain. When the PA starts clipping the error increases rapidly. The exact clip point is detected by comparing the error level to a threshold that represents clip (approx. 75 mV).

The AGC circuitry should be located at the forward path after the main noise source and before it degrades the forward path's linearity (Fig. 2). It should be able to maintain 20 dB of AGC (6 dB radio to radio variations, 3 dB flatness and 10 dB for temperature variations). The dynamics inside the AGC block 25 (zeros, poles and gain) are dependent on location of the attenuators 33 and 34 in the forward path.

The attenuation selected by the AGC will be activated at the phase training stage as well as at the transmitting time. This AGC will keep the error level constant maintaining constant radio performance.

The invention makes use of the observation that the error signal is proportional to the loop gain. This observation is used by the AGC circuit 26 to measure the error signal and adjust the attenuation in the loop to maintain constant and consistent loop gain.

By way of explanation of the observation that error and loop gain are inversely proportional, the following mathematical analysis is provided in the Laplace plane, referring to Fig 2.

Assume the loop response is: AB(s) = ab/s (before clip); ab is the loop gain and the input is a ramp that is modeled by (at point Bj: Vinput (s) = k/s^2 ; k is the ramp's slop The error voltage will be then (at point C): Verror - Vinput * 1/(1+ AB)VinputftAB) when AB 1 Assigning AB = ab/s and Vinput = k/sA2 will result: Verror ~ k/ab*1/s (before clip) It can be seen that the error (at point C) will step up and it's level will be inversely proportional to loop gain (ab).

Claims (3)

Claims:

1. A radio transmitter comprising; a power amplifier; linearizer means coupled to the amplifier for compensating for nonlinearity in the power amplifier; feedback means for feeding a signal from an output of the power amplifier to the linearizer means, whereby said amplifier, feedback means and linearizer means form an amplifier loop having a varying loop gain, training means for applying a training signal to the amplifier and adjusting the linearizer means during a training mode of operation, characterized by: automatic gain control means in said amplifier loop and control means for activating said automatic gain control means during at least a portion of said training mode of operation to maintain constant closed loop gain during said portion and to deactivate said automatic gain control means during a transmit mode of operation.

2. A transmitter according to claim 1, wherein the training means comprise means for applying a gain training signal to the amplifier during a gain training portion of said training mode, wherein the control means are arranged to activate said automatic gain control means during said gain training portion.

3. A transmitter according to claim 2, wherein the gain training signal has a linear portion of linearly increasing amplitude and the control means are arranged to activate said automatic gain control means during said linear portion.