The invention relates to a push-pull amplifier as described in the preamble of claim 1.

From the United States Patent US-A-5, 805,020 a silent start Class D amplifier is known wherein start-up noise is corrected by adding an analog switch to the integrator circuit. In an alternate embodiment a silent start switch connects a variable resistance to a comparator input. The resistance gradually increases to overcome start-up noise.

A disadvantage of this known silent start Class D amplifier is that activating the switching unit is done at a completely arbitrary moment.

It is an object of the invention to provide a push-pull amplifier and a silent start circuit, which do not have the disadvantages of the known amplifier and silent start circuit. To this end, a push-pull amplifier comprises the features of claim 1.

The invention provides the possibility to firstly bring the loop approximately at the required level and secondly to wait with the switching on moment till the moment at which the start up noise is as small as possible.

Embodiments of the invention are described in the dependent claims.

Herewith a cross-reference is made to the following co-pending applications of the same applicant and of the same date: "Carrousel handshake"applicant's ref. No. ID603908, Application No. 0 201 818. 2 "Level shifter"applicant's ref. No. ID604680, Application No. 0 201 826.5 "PWM limiter"applicant's ref. No. ID604682, Application No. 0 201828. 1 "Demodulation filter"applicant's ref. No. ID604683, Application No. 0 201 829. 9.

These and other aspects of the invention will be apparent from and elucidated with reference to the examples described hereinafter. Herein shows Figure 1 schematically an example of a push-pull amplifier, Figure 2 schematically the equivalent circuit during start-up, Figure 3 the optimal start-up moments, Figure 4 an example of the silent loop circuit, Figure 5 an example of the comparator input signals in silent mode, Figure 6 a circuit with the silent start circuits, and Figure 7 an example of the silent start logic implementation.

Figure 1 shows block-schematic an example of a push-pull amplifier PPA according to the invention. Via an input unit IU the amplifier receives the input signal. The input unit IU is coupled to a pulse-width modulator PWM, which is coupled with an output to a switching unit SU. The switching unit supplies an output signal via a demodulation filter DF to the output O of the amplifier. The pulse-width modulator PWM is coupled in a feedback loop with a feedback element RF which is coupled with one side to the output of the switching unit SU and with the other side to the input of the pulse-width modulator. The pulse-width modulator further comprises a first integrator FI and a second integrator SI and a comparator COM, the input of the first integrator is coupled to the output of the input unit IU and the input of the second integrator is coupled to an output of the first integrator FI and also coupled to an oscillator OSC.

The switching unit SU comprises a switch control unit SCU and a first and second switch SW1, SW2, respectively. The demodulation filter DF is in this example shown as an inductance L and a capacitance C can be a second order low-pass demodulation filter, or higher order demodulation filter.

Instead of using two integrators it is also possible to use only one integrator.

At the inverting input of the comparator for example a saw tooth signal can then be supplied.

A common problem with audio amplifiers is the occurrence of noises in the loudspeaker when the amplifier is switched on.

Considerable design effort is usually needed to reduce this start-up noise or'plop'. In a Class D feedback amplifier two mechanisms contribute to start-up noise.

First, when the amplifier is started, the initial condition of the integrators (FI, SI see Fig. 1) in the loop is undefined and usually not even near the steady state region.

Therefore, the loop needs some time to settle. Since the switching unit is active during this settling this can lead to audible noises in the loudspeaker. Ideally, the output of the amplifier produces a 50% dutycycle squarewave signal directly after start-up.

Second, before start-up the output of the switching unit is usually in a high- ohmic state to prevent DC current in the loudspeaker. At a certain moment in time the output starts switching. Even if the amplifier is able to produce a perfect pulse width modulated signal with 50% dutycycle directly after start-up this always results in a response in the loudspeaker because the demodulation filter needs to settle.

Assuming that the amplifier is able to produce the desired 50% dutycycle squarewave signal directly, the only degree of freedom that remains is the phase at which the signal is started.

Figure 2 shows schematically the equivalent circuit during start up.

In order to determine the optimal starting phase consider the situation shown in Figure 3. Here the voltage source Vs generates a square wave with amplitude As and a frequency cos much higher than the cut-off frequency (Ho of the filter. Initially, switch Sl is open and no energy is stored in the filter, i. e. both inductor current IL and capacitor voltage Vc are zero.

After switch Si is closed and the filter has settled to steady state the inductor current IL and capacitor voltage Vc change periodically. The energy stored in the inductor and capacitor are expressed as: EL =I2. L Ec = 1/2VC2.C In steady state the total energy stored in the filter also changes periodically.

The optimal time to close the switch is in the phase of the source signal where the stored energy reaches the minimal value.

In the stopband the energy stored in the components in a LC-filter decreases rapidly from source to load and is dominated by the element nearest to the source to load and is dominated by the element nearest to the source. In the second order filter shown in Figure 2. the total stored energy is dominated by the inductor. Therefore the stored energy is nearly minimal when the inductor current IL is zero. The source signal Vs and inductor current IL during steady state are shown in Figure 3.

As can be seen in Figure 3 the inductor current IL becomes zero twice each in each period of the source. Consequently, the optimal start-up moment is at 1/4 or 3/4 of the period. For the output response both moments are equivalent. However, since in is preferred that the output is switched to the lowside first in order to charge the bootstrap. Therefore, the optimal phase to start switching is at 3/4 period. This derivation is also valid for higher order demodulation filters.

Now two problems have to be solved to realize the optimal startup behavior.

First, the control loop needs to be in steady state before the output is enabled and second the output has to be enabled at the optimal moment.

The first problem can easily be solved by using a secondary feedback loop. As explained earlier, the signal that is fed back to the virtual ground of the first integrator in the loop is a squarewave current with amplitude Ifb=Vp/Rl. This current can be emulated by a switched current source that is controlled by the same comparator output as used to control the switching unit.

This results in the system shown in Fig. 4. The pulse width modulator PWM4 operates as follows. As long as the switching unit (SU, see Fig. 1) is not enabled, no current is fed back through resistor Ri. In this case the switched current source (SCU, see Fig. 1) is enabled and a current +Isilent or-Isilent is fed back to the virtual ground. As far as the loop is concerned this situation is equivalent to the situation where the switching unit is enabled and the switched current source SS4 is disabled as long as Isslent equals Ifb. Consequently, the loop converges to steady state. After steady state is reached, which is within a few clock-cycles, the switched current source is disabled and the switching unit is enabled simultaneously. This secondary feedback configuration is called the silent loop since it operates only if the switching unit is disabled.

The second problem is to determine the optimal moment to switch between the silent loop and the main loop. For this purpose the internal signals can be exploited. Since the

silent loop is equivalent to the main loop the internal signals in silent mode are identical to those in normal mode. These signals are shown in Fig. 5.

In Fig. 5 can be seen that the optimal moment to enable the switching unit is when the plus signal crosses zero. This moment can easily be detected by a second comparator COM62 as shown in Fig. 6. The output signal of this comparator is called the sync signal. In order to have optimal startup behavior the signal that enables the switching unit has to be synchronized with the rising edge of the sync signal. For this purpose a simple logic circuit SSLOG6 can be used.

Note that the inverting input of the second compurgator is not connected to the signal ground but to a low pass filtered version of the plus signal. This has been done to compensate for offset errors which cause a DC component to the internal signals which influences the timing.

The signal path contains two switched current sources. The function of these sources is to either sink or source a current of constant magnitude depending on a logic control signal.

Figure 7 shows an implementation of the silent start logic circuit SSLOG7.

It is to be noticed that above the invention has been described on the basis of an example but that the man skilled in the art is well aware of amendments without departing from the invention.