Analog Baseband

The analog baseband performs three kinds of analog signal processing functions:

1) low-pass filtering (LPF); 2) programmable-gain amplification (PGA), and 3) DC offset cancellation (DCOC).

3.1.2.1 PGA

In order to achieve the overall dynamic range specifications, two AGC loops controlled by the digital demodulator are utilized in the RF and analog baseband, respectively. The analog baseband provides programmable-gain ability to keep almost constant signal level at the ADC input. As derived in Ch 2.2.4, the overall receiver gain of 92－23dB is required. Since the RF front-end provides a gain range of 36－

0dB, the analog baseband should achieve a gain range of 56－23dB at least. To ensure

Chapter 3 Receiver System Analysis and Design

29

a reliable operation with a maximum SNR, however, a wider gain range from 60 to -6dB would be preferred. The extended gain range towards 0dB retains a potential to switch RF gain back-off later, alleviating a sudden SNR degradation when RF gain is changed from 15dB to 0dB. In addition, an accuracy of 0.5dB steps is specified to minimize the SNR degradation from gain changes during data reception under fading conditions [26].

3.1.2.2 Filter

To implement an analog filter for channel selection filtering, there are three important parameters that should be determined first: 1) filter prototype function; 2) 3-dB corner frequency; and 3) filter order. From DVB-T/H standard, the analog baseband must deal with 5/6/7/8 MHz channel bandwidth, i.e., 2.5/3/3.5/4 MHz cut-off frequency low-pass filtering. In addition, the selectivity pattern requires an attenuation of 30dB at 1.25MHz offset as derived in Ch 2.2.7. As a result, in this design a seventh order Chebyshev I filter is selected in a leap-frog configuration. This filter topology results in the sharpest stop-band attenuation, but contains the largest group delay. In order to compensate the group delay of the overall system, a first order all-pass filter is utilized in the analog baseband. Since the analog filter is sensitive to the process, voltage, and temperature variations, an auto-calibration circuit is needed to guarantee cut-off frequency accuracy within ±3%.

3.1.2.3 DCOC

Since direct conversion receivers down-convert the desired channel to the zero frequency, offset voltages at DC can corrupt the signal and saturate the following stages after the mixer [23]. As a result, in the analog baseband one important task to deal with is to eliminate or minimize the DC offset.

In genreal, there are two mechanisms to generate the offset voltages. The first one involves the finite matching performance of the devices in the mixer and the following stages. The mismatches of transistors and passive components will result in DC offset, which is almost constant. In order to minimize this source of DC offset, increasing the matching performance by using larger transistor size and symmetric

Chapter 3 Receiver System Analysis and Design

30

layout would be helpful.

The second one involves the finite isolation between the LO and RF port of the mixer, which causes the self-mixing effect to create the DC offset as shown in Fig. 3.2 (a) and (b). The leakage of LO signal to the input of the mixers may mix with itself to generate a constant DC offset. On the other hand, the strong interferers from the input may also leak to the LO port, but this self-mixing effect will generate a time-varying DC offset. In order to minimize this source of DC offset, improving the port isolation is helpful, which can be achieved by incorporating some shielding techniques and reducing the undesired coupling in circuit layout.

LNA

RF leakage LO (b)

LNA

LO leakage LO (a)

Fig. 3.2 LO/RF leakage and generation of DC offsets.

In addition to minimizing the sources of DC offset, an offset canceling technique is required to remove the DC offset. In general, high-pass filtering by means of AC blocking capacitors is the simplest method to remove the DC components. In order not to destroy the signal around DC, however, the cutoff frequency of the high-pass filter must be very low. This means that the required blocking capacitors will be very large and diffcult to implement on chip. On the contrary, DC servo loop is a good candidate for the purpose of high integration. As shown in Fig. 3.3, a low-pass filtering circuit which is placed between the input and output of the baseband amplifier can feedback the output DC components to the input. By subtracting these components from the input signal, a high-pass filtering can be realized. Compared with the method using the blocking capacitors, the required capacitors used in the DC servo loop is much reduced at the cost of extra power consumption. In this work, the DC servo loop has a cutoff frequency less than 1 kHz to ensure sub-carriers around DC

Chapter 3 Receiver System Analysis and Design

31

are not affected too much.

Detailed analog baseband specification is given in Table 3.3.

A V_out

0

) 0

( 



  s s k

H_lp Ak1

A A

0 (Ak1)₀

H(s) H(s)

Fig. 3.3 Servo loop for DC offset cancellations.

3.1.2.4 Noise/Linearity Trade-off

One of the main issues to address when designing the analog baseband is how to arrange the distribution of filtering attenuations and amplifications. If the amplifications are performed prior to the attenuations, the analog baseband can achieve the best noise performance but have the most stringent linearity requirement.

If the attenuations are performed prior to the amplifications, on the contrary, the linearity requirement is much relaxed at the expense of poor noise performance. To comprise the noise and linearity trade-off, the amplification and filtering procedures are repeated through the combined filter/PGA topology in the case of our particular receiver. A programmable gain of 48dB in 6dB steps is merged into the seventh-order Chebyshev filter, which is arranged as the preceding block of the ABB. The remaining variable gain range including a stage of 6dB fine gain tuning in 0.5dB steps is implemented in the latter stage of the ABB. Such arrangement ensures that the gain switching can be done from the last stage towards the front stage. Therefore, the noise figure in higher gain settings can be maintained almost in the minimum value since the gain of the preceding stages is not changed. Fig. 3.4 shows the recommended noise performance at different gain settings. The noise figure is almost constant, as pointed out earlier, less than 22dB as the gain is higher than 30dB. Then, the noise grows rapidly because the gain in the preceding stages is switched down.

Chapter 3 Receiver System Analysis and Design

32

Table 3.3 Recommended analog baseband specifications

Supply voltage (V) 1.2

Channel Bandwidth (MHz) 2－5

Passband ripple (dB) 0.5

Attenuation (dB)

Chapter 3 Receiver System Analysis and Design

33