This is the traditional vector-signal generation architecture. Frequency hopping may be implemented in two ways: by synthesizing a baseband IQ pair with the required frequency shift for each symbol, or by changing the LO frequency at the IQ modulator. Practical implementations for the baseband generation of the hopping signal require dual-channel AWGs with sample rates around 2 Gsps and analog bandwidths in the 1 GHz range. The implementation of frequency hopping by controlling the carrier frequency at the IQ modulator requires the capability of hopping more than 1 GHz in less than 70 ns.

Current implementations, given their limitations in sample rate and hopping speed, are limited to the generation of non-hopping signals. As two independent signal paths are used for I and Q baseband components, their alignment is extremely critical to obtain satisfactory results. Careful and long calibration procedures requiring additional high-performance analysis equipment are necessary and may have to be carried out frequently (due to the associated thermal and time-delay drifts).

In this method, a single-channel AWG is used to generate a UWB signal to feed an upconverter covering the required frequency range. Practical requirements for the AWG depend on the implementation of the hopping frequency operation. A sampling speed of 1.5 Gsps is the minimum requirement to generate a non-hopping signal. Generating a hopping signal would require twice rate (greater than 3.2 Gsps). Upconverters used in such a system would require a minimum of 750 MHz or 2 GHz upconversion bandwidth for non-hopping and hopping signals, respectively.

Although this method also requires careful magnitude and phase calibration procedures to reach the highest levels of modulation and spectral accuracy, its requirements are much less demanding, as the I and Q components are by definition aligned, and share the same signal path. The main limitation of this strategy is handling the signal images that will show up in the spectrum. This effect may be minimized by using an analog bandpass filter covering the target band. The amplitude and group-delay distortions introduced by the bandpass filter may be compensated as part of the calibration procedure.

In this arrangement, a single-channel AWG generates the UWB signal directly at the final frequency. The speed and the analog bandwidth requirements for the AWG depend mainly on the specific band groups to be covered, and not on the hopping nature of the final signal. For band group 1 (max. frequency 4.752 MHz) a minimum of 10 Gsps sampling rate and 5 GHz analog bandwidth are necessary. Band group 2 requires 15 Gsps sampling speed and 7 GHz analog bandwidth.

The Tektronix AWG7102 is capable of generating 5.8 GHz bandwidth waveforms at 20 Gsps, so it is possible to generate hopping signals in band group 1 with sufficient performance margin. Direct RF synthesis requirements for calibration are low. Controlled thermal behavior, low drift, and the elimination of additional external equipment allow this setup to maintain an acceptable signal quality for extended time periods, using only factory-level calibration.

Experimental data has been gathered using the test setup shown in Figure 3, which is an implementation of the direct RF synthesis architecture. All the WiMedia signals were generated using a Tektronix AWG. The AWG7000 series AWG has an extremely high sample rate, bandwidth and signal fidelity. For example, the unit features sample rates from 5 Gsps to 20 Gsps (10 bits), together with one or two output channels. The instruments also run on open Windows (Windows XP), enabling connectivity with peripherals, and compatibility with third-party software.

All tests were performed using a high-bandwidth digital sampling oscilloscope (DSO) with a sampling rate of 40 Gsps and a 15 GHz bandwidth, a 64 megasample record length, and UWB-analysis capability. Amplitude and phase distortions introduced by the oscilloscope are extremely low due to built-in real-time, DSP-based compensation techniques, and time-domain calibration procedures performed during the manufacturing process. The oscilloscope's high accuracy and traceability, therefore, makes it suitable for waveform-generator calibration procedures.

Basic experimental results are grouped by EVM data sets gathered for different bit rates and modulation schemes (QPSK and DCM). All data generated by this setup suggest that direct-RF synthesis would yield an EVM of around -30 dB better than the -19.5 dB EVM specified in the standard, as reflected in the data in Table 3. This setup also has the ability to frequency hop without the necessity of an external frequency hopper, due to the high sampling rate and bandwidth of the AWG.