Academia.eduAcademia.edu

Figure 2 – uploaded by Saeed Farsi

See full PDFdownloadDownload figure
TABLE II. PERFORMANCE OF CMPM AND MPM FOR MEASURED WCDMA SIGNAL RECHING THE 2 DB COMPRESSION POINT in which $(n) and y(n) are the estimated and the measured output, respectively.

Table 2 II. PERFORMANCE OF CMPM AND MPM FOR MEASURED WCDMA SIGNAL RECHING THE 2 DB COMPRESSION POINT in which $(n) and y(n) are the estimated and the measured output, respectively.

Related Figures (7)

As can be clearly seen, the number of coefficients in the generalized formulation has been increased. This rise leads to a higher computational load. This paper proposes a new model, which has the simplicity of the memory polynomial formulation, but partly benefits from the cross terms, like in GMPM. The idea is to use the magnitude of the present input sample to compute the nonlinear weights of the memory terms. This modification is originated from the fact that the system has memory, and its recent samples are linked together. Evaluating the validity of this assumption is the subject of this paper. The suggested new model is written as
As can be clearly seen, the number of coefficients in the generalized formulation has been increased. This rise leads to a higher computational load. This paper proposes a new model, which has the simplicity of the memory polynomial formulation, but partly benefits from the cross terms, like in GMPM. The idea is to use the magnitude of the present input sample to compute the nonlinear weights of the memory terms. This modification is originated from the fact that the system has memory, and its recent samples are linked together. Evaluating the validity of this assumption is the subject of this paper. The suggested new model is written as
Figure 1. Standard architecture of MPM, b) proposed architecture for MPM and c) proposed architecture for CMPM. architecture in Fig. 1b is used. This is mainly due to the higher number of complex-complex multiplications in Fig. la. It is worth mentioning that the propagation delay difference due to the branch in Fig. 1b connecting the input to the last multiplier is not a critical issue and can be compensated, if required, by inserting a delay block in that branch.
Figure 1. Standard architecture of MPM, b) proposed architecture for MPM and c) proposed architecture for CMPM. architecture in Fig. 1b is used. This is mainly due to the higher number of complex-complex multiplications in Fig. la. It is worth mentioning that the propagation delay difference due to the branch in Fig. 1b connecting the input to the last multiplier is not a critical issue and can be compensated, if required, by inserting a delay block in that branch.
LE I. NUMBER OF FLOPS FOR DIFFERENT OPERATIONS [6]
LE I. NUMBER OF FLOPS FOR DIFFERENT OPERATIONS [6]
The assessment of the performance can be made easier with table II. The criteria in table II for assessment is the normalized mean square error (NMSE), defined by Figure 3. Modeling performances of MPM and CMPM for Amp #1 with a measured 1x WCDMA signal reaching the 2 dB compression point of the PDITTIT.
The assessment of the performance can be made easier with table II. The criteria in table II for assessment is the normalized mean square error (NMSE), defined by Figure 3. Modeling performances of MPM and CMPM for Amp #1 with a measured 1x WCDMA signal reaching the 2 dB compression point of the PDITTIT.
Figure 2. Measurement set-up for modeling.
Figure 2. Measurement set-up for modeling.
Figure 4. Modeling performances of MPM and CMPM for Freescale Amp #2 with a measured 1xWCDMA reaching the 2 dB compression point of the NMIIT
Figure 4. Modeling performances of MPM and CMPM for Freescale Amp #2 with a measured 1xWCDMA reaching the 2 dB compression point of the NMIIT
Figure 5. Modeling performances of MPM and CMPM for Freescale Amp #2 with a measured 1xWCDMA reaching the 2.5 dB compression point of the DUT.
Figure 5. Modeling performances of MPM and CMPM for Freescale Amp #2 with a measured 1xWCDMA reaching the 2.5 dB compression point of the DUT.
Related topics:
Electronic EngineeringComputer ScienceCode Division Multiple AccessElectronic Circuit
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved