Ultra Fast Multiscalers / Questions and Answers

Concerning the fast P7889 FIFO_1, please correct us if we are wrong: The FIFO_1 can store STOP signals for 1024 time slices, each represented by a 64-bit word and covering 6.4ns: one bit for each 100ps time bin. Together they can cover 1024*6.4ns = 6.5µs

Your understanding is quite right.

Empty 6.4ns words are not stored in the FIFO_1. Thus, the total time covered depends on the STOP rate and is at least 1024*6.4ns = 6.5µs but most often far longer. Yes, there are parallel registers that record the corresponding TAG information and also a "slow" time (6.4ns resolution) with each 64bit event word. The processing time per STOP event (to convert a FIFO_1 word into a 64bit time word --> FIFO_2) depends on how many STOPs are in each 64bit event word. With only a single STOP in each 6.4ns word the processing-time/STOP is approximately 44.8ns (--> 22.3MHz). With 64 STOPs in a 6.4ns word it would be only 19.6ns (--> 51MHz). Thus, to empty a full FIFO_1 (with only one single STOP in each 6.4ns word) takes about 1024*45ns = 46µs. Or 1024*64*20ns = 1.3ms for a fully saturated FIFO_1 (not a condition to happen in reality). And, as soon as there is any data in the FIFO_1 the processing machine starts to empty it. This means, the total number of time covered by FIFO_1 might be much larger as it is already emptied while new STOPs arrive.

Ultra Fast Multiscalers