A 32-bit pipeline accumulator with carry ripple topology is implemented for direct digital frequency synthesizer. To increase the throughout while hold down the area and power consumption, a method to reduce the number of the pre-skewing registers is proposed. The number is reduced to 29% of a conventional pipeline accumulator. The propagation delay versus bias current of the adder circuit with different size transistors is investigated. We analyze the delay by employing the open circuit time constant method. Compared to the simulation results, the maximum error is less than 8%. A method to optimum the design of the adder based on the propagation delay is discussed. The clock traces for the 32-bit adder are heavily loaded, as there are 40 registers being connected to them. Moreover, the differential clock traces, which are much longer than the critical length, should be treated as transmission lines. Thus a clock distribution method and a termination scheme are proposed to get high quality and low skew clock signals. A multiple -type termination scheme is proposed to match the transmission line impedance. The 32-bit accumulator was measured to work functionally at 5.3 GHz. Direct digital frequency synthesizer (DDFS) is able to generate sinusoids with sub-hertz resolution, good spectral purity, fast frequency switching and phase continuity on switching. For the next generation radar and communication systems, DDFS operating at GHz-range clock is required. However, the operating speed of DDFS is limited by the phase accumulator (PA), which consumes large area and power. Many architectures and designs have been reported in the literature for the PA in DDFS, such as the carry-ripple adder (RCA) [1], the carry look ahead adder (CLA) [2] and the pipelined adder [3,4]. Though faster than RCA, the significant fan-in and fan-out requirements of CLA architecture lead to lower clock rate as the bit width increases. The pipeline accumulator offers significant speed improvement compared to the CLA, which has large propagation delays before a valid sum is produced. The carry ripple pipeline accumulator is more practically suitable for GHz DDFS with 32-bit resolution in terms of compact layout and hardware implementation. A 32-bit pipeline accumulator with 8-stage pipeline is illustrated in Figure 1. A number of pre-skewing and deskewing registers are required to keep the input and output of the accumulator coherent. To minimize the area while achieving the desired operating speed, a new hardware efficient PA is proposed in this work. This method reduces the number of registers required in the pipeline without performance degradation. The optimum design in terms of power and propagation delay is investigated. The clock for the accumulator is critical for high speed operation. A clock distribution method and a termination scheme are proposed.