In any communication system, such as in OFDM signaling systems, interfacing between digital and analog domain is a required function within the system. This thesis focuses on such analog interfacing factors, specifically the impact it has on the OFDM baseband signal in typical hardware systems. To investigate these factors, hardware experiments are performed then emulated in an elaborate software model to verify contribution from DC offset error, gain error, I/Q amplitude imbalance, and fixed DAC /ADC resolution. To perform the hardware experiments in this study, a Texas Instruments (TI) TMS320C6711 DSK development kit, equipped with a C6711 floating point processor was studied, programmed, and configured to perform OFDM modulation needed. Further investigations use software simulations to isolate each analog factor and combine them to characterize its effect on performance. In addition, the simulations also examine, ADC saturation, and relative I/Q delay. The performance of received signal at the demodulator is measured by MSE of the waveform (M SEwaVe), M SE of the constellation (MSEcst), signal to noise ratio normalized to a bit (SIRbit).
The results of the experiments and simulations reveal the sensitivity of demodulation performance and magnitude of impact each case has on OFDM signaling. In simulations, the maximum DC offset distortion permitted in the most sensitive modulation scheme probed is 0.15% for the case of 256 QAM, and ADC saturation due to DC offset has no obvious impacts for the values probed. Uncompensated gain error from 0.5% to 1% in received samples for 256QAM, results in SIRbit decrease 6dB. Similar to the effect of gain error, SIRbit degrades in the same manner for I/Q amplitude imbalances. In situations where resolution is less in the receiver relative to the transmitter, the SIRbit performance degrades by 6dB for every bit decremented. Up to 10% I/Q delay caused by filtering results in approximately 0.5dB of variance in SIRbit for all constellation sizes investigated. However, all these numerical figures are for ‘isolated’ cases, when factors are combined the results cannot be predicted by superposition of the ‘isolation’ data cases.