SAW filter in WCDMA

32 www.rfdesign.com February 2007 Tx/Rx Technology Removing the interstage transmit SAW in WCDMA Typically, in a WCDMA radio, there is a transmit (Tx) SAW filter between the transceiver Tx output and the power amplifier to reduce the transmit output from increasing the receiver’s (Rx) noise figure, as well as to reduce out-of-band spurious and noise emission from the Tx. However, if the Tx SAW filter was eliminated, it would save area on the PCB as well as cost. This article analyzes the trade-offs required for removing the Tx SAW from the transceiver and duplexer sections of the handset. By Tajinder Manku The interstage transmitter (Tx) SAW filter between the transceiver’s Tx output and the power amplifier has effectively two functions. 1. To reduce the transmit output from increasing the receiver (Rx) noise figure. 2. To reduce out-of-band spurious and noise emission from the Tx. If this Tx SAW filter can be eliminated, it would save cost and area on the PCB. This article analyzes the trade-offs involved with removing the Tx SAW filter from the perspective of improving the performance of either the transceiver and/or the duplexer. Item (1) is not a concern for GSM/GPRS/EDGE radios because these standards are half-duplex—i.e., the Rx and Tx are not on at the same time. However, item (2) is an important concern for GSM/ GPRS/EDGE. According to the 3GPP specification, Tx noise within the low bands of GSM/GPRS (850 MHz and 900 MHz) cannot ex- ceed –79 dBm over a bandwidth of 100 kHz at an offset of 20 MHz. At +33 dBm output (as per the 3GPP specification) the noise at a 20 MHz offset has to be better than –163 dBc/Hz (i.e., –79 dBm-10log (100 kHz)-33 dBm). This number is achievable for both the transceiver and the power amplifier. Consequently, a Tx SAW is not required for the GSM standard. Both functions 1 and 2, are explored with a SAW and without a SAW for WCDMA and various trade-offs are discussed. However, the article addresses item (1) in much more detail since item (2) is well understood for GSM. Transmitter to receiver desensitization Tx to Rx desensitization occurs because the transmit noise in the Rx band leaks into the Rx chain and effectively increases the noise floor of the Rx. This increase in the noise floor increases the overall noise figure of the Rx. If we assume the E b /N o =7 dB for a bit error rate (BER) of 10-3, the required noise figure (NF) at the antenna port, with the Tx on, must be designed to be less than 9 dB. There are three basic sources of noise in the transmit path that can add to the Rx noise figure: 1. The digital-to-analog converter (DAC) used to generate the ����������� ����� ������������ ��������� ������������������ ������������ �������������������� �������������������� ���������� �������������� ��������������������� ��������������������������� ����������������������������� ����������������������������������� ���������������� ��������������� ���������������� ��������������� ��������������� ��������������� �������� �������� Figure 1. System analysis diagram with a Tx SAW. 702RFDF3.indd 32 2/8/2007 11:26:13 AM 34 www.rfdesign.com February 2007 analog IQ signals. 2. The modulator (i.e., the transceiver). 3. The power amplifier. In our initial analysis, we will assume the DAC noise is much lower (at least 10 dB lower) than the transceiver and power amplifier noise. Figure 1 illustrates a generalized WCDMA system with an intra- stage Tx SAW filter. All the variables within the following analysis are defined within Figure 1. The Rx noise due to the transceiver at the output of the power amplifier, N TX-RX,PA is given by the following equation: , 10 log(3.84 ) TX RX PA TX RX PPA TX RX PA N N P MHz G G − − − = + + + + (1) Therefore, the total in-band Rx noise at the output of the power amplifier, N PA is a combination of the transceiver noise and the power amplifier noise; , /10 /1010 log(10 10 )TX RX PA PA RXN NPAN − −= + (2) Given this, the noise at the input of the Rx, N RX is identified as N PA minus the attenuation of the duplexer, plus its own noise contributions when the Tx is off (-174+10log(3.84MHz) + NF TX,eff ); , ( ) /10 ( 174 10log(3.84 ) ) /10 10 log(10 10 ) PA TX RX TX eff N D RX MHz NF N −− − + + = + (3) Therefore, the NF at the antenna input with the Tx on is given by the following equation; [ ]174 10log(3.84 ) _ ( ) RXNF N MHz RX loss switch duplexer = + − + + (4) By using equations (1) to (4), the effective noise figure of the Rx can be derived. In a typical application, the duplexer Tx to Rx isolation is approximately 42 dB, the noise from the PA is typically around –73 dBm, the noise figure of the Rx with the Tx off is typi- cally around 3 dB, and the power and noise output of the transceiver is +4 dBm and –153 dBc/Hz, respectively. If a low-cost Tx SAW is used, one would expect an in-band loss of 1 dB and an out-of-band loss of 30 dB. The in-band loss from the switch and duplexer to the antenna is typically around 2 dB and 3 dB for the Tx and Rx, respectively. Given all this, the gain of the PA would have to be 23 dB to achieve an output power of +24 dBm as per the 3GPP requirement. With all these variables, the noise figure with the Tx on would be N TX,on = 6.4 dB, which meets the spec and has about 2.6 dB of margin. For this particular case, the transmit noise in the Rx band is mainly influenced by the PA noise rather than the transceiver noise. For example, if the transceiver noise increased by 8 dB, N TX,on = 6.5 dB, which is only a 0.1 dB increase. Consequently, with a Tx SAW filter, the affects of the transceiver noise in the Rx band can be effectively ignored. Now, let’s assume the SAW is removed from the Tx chain and the transceiver noise is –153 dBc/Hz; see Figure 2. Under this scenario, the NF at the antenna port is 10.9 dB, which fails the NF spec of 9 dB. From the analysis, it becomes apparent that there are only two variables that dramatically influence the amount of Tx to Rx desen- sitization; (a) the duplex isolation between the Tx and Rx (D TX-RX ) and (b) the transceiver noise in the Rx band (N TX-RX ). Better duplexer isolation usually implies increased cost and size. Furthermore, duplexer technology generally moves toward producing smaller scales, thus one would expect the Tx to Rx isolation to become worse. In any case, let’s make the assumption the duplex isolation is now 48 dB (as opposed to 42 dB). With 48 dB of isolation, N TX,on = 7.8 dB, which meets the spec of 9 dB, but only has 1.2 dB of margin. ��������������� ���������������� ���������������� ������������� ������������������ ������������ �������������������� ����������������� ������������������� ���������������� ����������������� �������� ���������� �������������� �������� Figure 2. System analysis diagram without a Tx SAW. Noise Figure (NF) Specification Requirement = 9 dB Scenario Total NF Margin Pass / Fail With SAW 6.4 dB 2.6 dB Pass Conventional duplexer without SAW 10.9 dB N/A Fail Better duplexer without SAW 7.8 dB 1.2 dB Pass* Better Tx noise without SAW 7.8 dB 1.2 dB Pass* Table 1. Trade offs without the SAW filter. 702RFDF3.indd 34 2/8/2007 11:26:14 AM 36 www.rfdesign.com February 2007 The other variable is the transceiver noise in the Rx band. To im- prove this usually implies increasing the current consumption within the transceiver, which would in turn reduce battery life. In any case, let’s assume N TX-RX = -160 dBc/Hz (opposed to –153 dBc/Hz) and the duplex isolation is still only 42 dB. Under this scenario, N TX,on = 7.8 dB that meets spec, but only has 1.2 dB of margin. So far in our analysis we have assumed that the noise due to the DAC generating the base- band IQ signals can be ignored. However, this assumption may not be valid if the noise required from the transceiver needs to be better than –160 dBc/Hz. For the DAC noise to be neglected it would need to Bands Uplink (MHz) Downlink (MHz) Duplexer Close in Spurs Measured Min Power Lower Upper Lower Upper Spacing (MHz) Upper Lower BW (kHz) dBm IMT, I 1920 1980 2110 2170 190 1805 1880 100 -71 PCS, II 1850 1910 1930 1990 80 1930 1990 3840 -60 DCS, III 1710 1785 1805 1880 95 1805 1880 3840 -60 850, V 824 849 869 894 45 869 894 3840 -60 be better than –170 dBc/Hz. If we assume a 1 V swing at the DAC output, the SNR required by the DAC output would be 170 dBc/Hz- 10log(3.84 MHz), or 104 dB. Assuming 45 dB of quantization noise filtering, the DAC itself requires about 59 dB of SNR. This would result in a 9-bit DAC at a clocking rate of 26 MHz. In summary, Table 1 depicts the trade offs without the SAW filter. Transmit out-of-band spurious requirements Table 2 illustrates the closest spurious requirements for four popular WCDMA bands. These numbers can be translated into a noise require- ment at the output of the transceiver if no Tx SAW is used; _ ( / ) min_ 10*log( ) 24 = − − noise TX dBc Hz power BW dBm (5) The hardest requirements in terms of noise are bands I and V. These requirements are –145 dBc/Hz at 40 MHz offset and –150 dBc/ Hz at 45 MHz offset, respectively. However, not only does the noise have to be below these settings, care has to be taken to make sure the spurious components are below these values. However, for band I, five exceptions are allowed within the band limits of 1805 MHz and 1880 MHz. Conclusion When an interstage Tx SAW is used, the Rx sensitivity is optimized for best performance. By removing the Tx SAW there becomes a design trade off between duplexer technology, the transmit noise in the Rx band, and the Rx noise figure. If the Tx SAW is removed, the specification requirements of either/both the transceiver and duplexer have to be better. The Tx must be designed to have less noise in the Rx band and less spurious power out-of-band. The duplexer, on the other hand, has to have better isolation characteristics between the Tx and Rx. However, increasing the duplexer requirements may not be the right choice given better duplexer performance may come at a higher cost and more PCB area. RFD Table 2. Spurious requirement for all the key WCDMA bands. ABOUT THE AUTHOR Tajinder Manku is chief technology officer and founder of Sirific Wireless Corporation, Richardson, Texas. Prior to Sirific, Manku served as an associate professor at the University of Waterloo in Canada, where he earned his B.S in solid-state physics and a PhD. l o c k o n w i t h R a k o n find the best product to suit your needs at: www.rakon.com ��������������������� ����������������������������� ����������������������������� � ������������������������������������ � ������������������������������������������ � ����������������������������������������������������������� ������������������������� � ����������������������������������������� 702RFDF3.indd 36 2/8/2007 11:26:16 AM