The main breakthrough today was the data format. We plan to use the VME 64 standard or some close variation of it, for example, vme 64x. So, the data format will consist of 64bit words. Each word can be decifered as follows. The MSB(63rd bit) will signify which of two possible types the word is, it can be either a data word or a special word. If it is a data word, then the next 15 bits will be a time word. The rest of the 64bit word will be 6 bytes, each which is a consecutive ADC sample. The time word will corespond to one of the ADC bytes, and the rest will either preceed of follow by some number of clock cycles. So, if the time word corresponds with the nth ADC, then the nth+1 ADC has the time of the Time word plus one clock cycle, and the nth+2 is 2 clock cycles and so on. If the 64bit word is of the special type, then the that means it is either a start word, rollover word, or a stop word. In this case, the 15 bits that follow the special bit will agian be a time word. That time will be the time for the whole word. The next 8 bits will be an identifying word that will specify whether it is a start, rollover of stop word. The next 16 bits will be the Channel ID word so that we can tell which part of the detector this pulse came from. The next 3 bytes are not used(yet). The start word will come from an external signal that is 2 microsecond after T0. The Stop word will come from another external signal that tell us when our data taking period if over, most likely about 15us after T0. We will get a rolloverword whenever the wfd's 15bit clock counter runs out of digits. With a 500Mhz clock this will happen every 65.5 us.

Since we have many channels to read out and not every one of them will see an electron in any given fill, we will be doing a sparse data readout with VME block chain transfers. To help make this more efficiant, we will insist that a channel have more than n 64bit words stored in memory to be read out. At the moment we have chosen to make n equal to 3 because every channel will atleast have 2 words, a start and a stop, and possible a 3rd word from a rollover. We will do this by making the wfd reset its memory pointer if it has less than 4 words in its memory when it gets a stop signal. A typical signal from an electron will be about 45ns long. This is about 23 samples , and since our data will always come on groups of 6, the typical pulse will be about 24 samples long or 4 64bit words. These words along with the Start, Stop , and possibly a rollover will give 6 or 7 Words in a channel that got a hit. So, we should be in no danger of throwing out data by thowing out channels that have less than 4 words.

More to Come