We describe a time‐to‐digital converter able to measure intervals as great as 100 ns with a resolution of 4 ps rms. It achieves this large dynamic range by simultaneously sampling four sinusoidal wave forms (sine and cosine waves at 200 and 6.25 MHz) derived from a single quartz oscillator. Twelve‐bit analog‐to‐digital conversion of the 200 MHz waves yields the high time resolution. Eight‐bit conversion of the 6.25 MHz samples removes the cycle ambiguity of the 200 MHz data. The digital words are pipelined in a fully parallel data flow architecture. A first‐in first‐out stage in the pipeline derandomizes the random event arrival times. A subsequent stage in the pipeline uses an arctangent function to convert the sine and cosine pairs into linearized measures of event time. These are subtracted to yield start–stop time interval sizes for individual photoevents. The minimum start–stop interval is 50 ns, set primarily by the cycle time. Because the same processing is employed for the start and stop events, a large class of potential error and drift phenomena are eliminated. The digitizer provides an accurate way to decode the outputs of delay line detectors, offering high event throughput and extremely good long‐term timing accuracy. As a side benefit, the pipeline data flow architecture permits simple breadboarding and low‐throughput testing of the system stages with the arctangent work implemented in a personal computer. This arrangement is also very convenient for logging diagnostic evaluation data. The same front‐end and data flow architecture is directly applicable to very high‐speed applications where the event processing is implemented in a digital signal processor.