Example for a Gain Tracking Algorithm Required in a GSM Terminal Based on the Lucent Technologies Sceptre Platform

Transcription

1 Application Note Example for a Gain Tracking Algorithm Required in a GSM Terminal Based on the Lucent Technologies Sceptre Platform Introduction This note gives an example implementation of the gain tracking algorithm required in a GSM physical layer software as part of a complete solution using the Lucent Technologies GSM Sceptre Platform. Initially the fading channel is discussed to form a basis for understanding the requirements. The Lucent Technologies GSM Sceptre Platform details with regard to the gain setting are described next. An algorithm for gain tracking is presented. The algorithm uses various output parameters from the Lucent Technologies GSM Sceptre digital receiver. Finally, some implementation aspects are discussed. It is assumed that the user is familiar with the architecture of the CSP1088, the functionality of the digital receiver, as well as the GSM requirements as listed in GSM Rec Fading Channel Mobile radio communication in GSM takes place between a fixed base station (BS) and a number of roaming mobile stations (MSs). As the distance between a BS and a MS increases the received mean signal level tends to decrease. Over relatively short distances the received mean signal is essentially constant but the received signal level can vary rapidly by amounts typically up to 40 db. These rapid variations are known as fast fading. Whilst the MS is travelling the MS does not only receive one version of the transmitted carrier, but a number which have been reflected and diffracted by buildings and other urban paraphernalia. Indeed in most environments, each version of the transmitted signal received by the MS is subjected to a specific time delay, amplitude, phase and Doppler shift depending on its path from the BS to the MS. As a consequence the constant amplitude carrier signal transmitted may be substantially different from the signal the MS receives. When the signals from the various paths sum constructively at the MS antenna, the received signal level is enhanced. A serious condition occurs when the multipath signals, i.e., the transmitted signal arriving via many paths, vectorially sum to a small value. When this occurs the received signal is said to be in a fade, and the phenomenon is called multipath fading. As the MS travels it passes through an electromagnetic field that results in the received signal level experiencing fades approximately every half wavelength along its route. When a very deep fade occurs the received signal is essentially zero and the receiver output depends on the noise. If each multipath component in the received signal is independent then the probability density function of its envelope is Rayleigh. A typical received signal's fading envelope as a function of time is shown in Figure 1

2 Example for a Gain Tracking Algorithm Required in a GSM Terminal Based on the Lucent Technologies Sceptre Platform Application Note Fading Channel (continued) Figure 1. Typical Received Signal s Fading Envelope as a Function of Time Gain Setting With the Sceptre Chipset Mobile stations based on the Sceptre chipset allow gain setting at various places. Figure 2 shows a typical block diagram of the receive section. LNA W2020 PGA CSP1088 PGA duplexer filtermixer filter mixer filter ADC CSP 1088 timing control unit automatic gain control algorithm Figure 2. Typical Block Diagram of Sceptre Receive Path 2 Lucent Technologies Inc.

3 Application Note Example for a Gain Tracking Algorithm Required in a Terminal Based on the Lucent Technologies Gain Setting With the Sceptre Chipset (continued) The gain variance of the LNA is dependent on the device and RF design. A typical value for LNA_on vs. LNA_off is 20 db. The W2020 allows a gain setting in steps of 4 db over a range of 60 db. The gain in the CSP1088 receive section can be modified by 18 db in steps of 2 db. With the dynamic range of all three components, a signal can be tracked over a range of round 100 db. The MS has to cope with signals at the antenna of less then 102 dbm (GSM Rec ). In a GSM system, the receive level is expected not to be above 20 dbm. This relates to a dynamic range of the receive signal of at least 80 db. It should be noted that the Equalizer as part of the digital receiver requires a minimum SNR (SNRmin) to work properly. For the digital receiver provided by Lucent Technologies SNRmin = 9 db. The baseband analog digital converters (ADC) of the CSP1088 have a resolution of 10bit. According to the data sheet, a SNR+distortion ratio of SNDRtypical = 59 db can be achieved (SNDRmin = 55 db). The minimum requirement of the gain tracking algorithm is to make sure that the signal into the baseband AD converters of the CSP1088 is between SNRmin and SNDRmin. As described before, the received signal is subject to fading. Considering the nature of Rayleigh fading, the optimum signal level at the baseband ADC of the CSP1088 should be SNRopt = 40 db. Figure 3 summarizes these assumptions. 10dB 10dB head room usabl e range SNR opt = 55dB min margin for 26dBdeep fading gaps SNDR 9dB SNR mi n for Equaliser optimum receive signal strength at the CSP 1088 baseband ADC Figure 3. Summary of Signal Level at the CSP1088 RX ADC One function of the digital receiver provided by Lucent Technologies is to calculate the signal strength of the baseband signal captured with the CSP1088. The function performed is: Im Q RXLEV LSB 2 + m symbol = = 128 The scaling by 8 is performed to avoid overflow in the arithmetic of the DSP and to keep the calculated RXLEV within 16 bit. The highest possible RXLEV reported by the digital receiver can be which corresponds to full scale at the baseband ADC. The dynamic range of RXLEV is 48 db. In the following chapter, full scale at the ADC is used as a reference level. As shown in Figure 3 the optimum receive level is between -20 db to -10 db relative to full scale. This corresponds to sampled I and Q data of ±50LSB ±158LSB. A good value for the following calculations is ±100. This equates to a signal of -14 db below full scale and is therefore in the usable range. By applying the formula above, RLEVopt = Lucent Technologies Inc. 3 2 m

4 Example for a Gain Tracking Algorithm Required in a GSM Terminal Based on the Lucent Technologies Sceptre Platform Application Note Gain Tracking Algorithm The RX level of 128 samples from the burst processed by the digital receiver RXLEV is available at the memory location rx_rss. For the proposed algorithm, this value should be evaluated and reported in db relative to full scale at input of the baseband AD converters of the CSP1088. This average level over burst n is referred to as RXAD(n) (negative value). The absolute RX level at the antenna input in dbm during burst n is referred to as RXANT(n). The appropriate Software(SW) has an internal table for mapping desired full scale gain (GD) to gain control word (GCW, control word for CSP1088/ W2020/ LNA gain setting). RXANT(n) + GD(n) = RXAD(n) GD GCW(GD) (GD(n) positive value) GD should be considered as the sum of the variable gain (W2020, CSP1088, LNA) and the constant gain of the receiver GRF. The SW maintains a pipeline of GD(n) values such that when an RXAD value is reported, RXANT may be derived: RXANT(n) = RXAD(n) GD(n) There is an optimum RXAD = RXAD(opt) With the assumptions made in the previous chapter, this value is: RXAD(opt) = 12 db. The gain control algorithm, RXANT = f(rxant(n)), has to predict a value of RXANT. The gain will then be set as: GD = RXAD(opt) RXANT The algorithm, to be applied to all received bursts, is as follows: IF (burst n not subject to DTX) IF (RXAD(n) > RXAD(opt)) RXANT = RXANT + RXAD(n) RXAD(opt) ELSE RXANT = RXANT MIN(RXAD(opt) RXAD(n), GINC_NO_DTX) ELSE (burst subject to DTX) IF (good burst) IF (RXAD(n) > RXAD(opt)) RXANT = RXANT + RXAD(n) RXAD(opt) ELSE RXANT = RXANT MIN(RXAD(opt) RXAD(n), GINC _DTX) RX AD(n) = 10 * log RXLEV (n) 4 Lucent Technologies Inc.

5 Application Note Example for a Gain Tracking Algorithm Required in a GSM Terminal Based on the Lucent Technologies Sceptre Platform Gain Tracking Algorithm (continued) In general, increases in the detected level are always tracked immediately, decreases in the level are tracked slowly (the increase in gain per TDMA frame is limited). Bursts that are always sent by the BS always affect the AGC as described. Bursts that may be subject to DTX, and so are not necessarily sent by the BS, only affect the AGC if the burst is good, i.e. there is some evidence that the burst was sent. The criteria for good bursts should be based on the midamble_detect_flag in the rx_status_word out of the digital receiver. GINC is dependent on the type of bursts. It determines the speed with which the variations in the signal level are tracked. On channels which are not subject to DTX (always transmitted by the BS) GINC_NO_DTX should have a higher value then GINC_DTX. This makes sure, that in case the signal drops and the digital receiver can not find the midamble, the signal can be tracked. It is recommended to set: GINC_NO_DTX = 2 db, GINC_DTX = 1 db. On BCCH, PCH, CCCH, SDCCH channel configurations, all bursts are NOT subject to DTX. On TCH/FS, SACCH and TCH bursts (SID) are not subject to DTX. All other bursts are subject to DTX. Please note that the LNA should be enabled as late as possible in case the W2020 is in receive mode whilst the CSP1088 is performing the dc calibration. This reduces the distortion caused by the receive signal. On TCH/FS channels with hopping configuration where MA includes the BCCH frequency, a second RXANT has to be maintained for samples taken on the BCCH frequency. (This is because the BS may use power control on the downlink, but may not use power control on the BCCH frequency). Note that bursts on the BCCH frequency are NEVER subject to DTX. The constant gain of the receiver GRX may vary from board to board, over receive band as well as over temperature. It is recommended to measure GRX during production and to store the appropriate values in EEPROM. Implementation Aspects As mentioned before, the appropriate Software(SW) has an internal table for mapping desired full scale gain (GD) to gain control word (GCW, control word for CSP1088/ W2020/ LNA gain setting). There is no necessity in a practical realization to assign one GCW to each possible GD. A quantization Q should be introduced such that GD(n)' = GD(n) / Q. The gain will then only be updated if RXAD(opt) RXAD(n) Q. This also results in a shorter mapping table (GD' GCW(GD')) to be maintained by the Software. It is recommended to set the quantization to Q = 2. To avoid overloading in the receive path, the gain of the W2020 should be set to the minimum possible value. This means, depending on the receive signal strength, the CSP1088 gain should be set to its maximum value of 18 db. Lucent Technologies Inc. 5

6 For additional information, contact your Microelectronics Group Account Manager or the following: INTERNET: U.S.A.: Microelectronics Group, Lucent Technologies Inc., 555 Union Boulevard, Room 30L-15P-BA, Allentown, PA 18103, , FAX (In CANADA: , FAX ), ASIA PACIFIC: Microelectronics Group, Lucent Technologies Singapore Pte. Ltd., 77 Science Park Drive, #03-18 Cintech III, Singapore Tel. (65) , FAX (65) JAPAN: Microelectronics Group, Lucent Technologies Japan Ltd., 7-18, Higashi-Gotanda 2-chome, Shinagawa-ku, Tokyo 141, Japan Tel. (81) , FAX (81) For data requests in Europe: MICROELECTRONICS GROUP DATALINE: Tel. (44) , FAX (44) For technical inquiries in Europe: CENTRAL EUROPE: (49) (Munich), NORTHERN EUROPE: (44) (Bracknell UK), FRANCE: (33) (Paris), SOUTHERN EUROPE: (39) (Milan) or (34) (Madrid) Lucent Technologies Inc. reserves the right to make changes to the product(s) or information contained herein without notice. No liability is assumed as a result of their use or application. No rights under any patent accompany the sale of any such product(s) or information. Copyright 1996 Lucent Technologies Inc. All Rights Reserved Printed in U.S.A. AP97-004WDSP