Firmware feature lists

Here are some utility-mode features that I think will still be useful:

detect/report sync error if BC detected on input cable does not match BC inferred from CDF backplane signals
output correct EE word (e.g. even if input BC or L2B is wrong): useful in case in which SVT is playing spy buffer fake data with no knowledge of Trigger Supervisor's true BC, L2B; we need to be able to send the right EE word to L2 and DAQ
fix LEDs at U of C (make sure all LEDs work as they should on a typical SVT board; there is at least one schematic bug on the connection of one of the four LEDs per channel to the 74HCT123)

Here are some utility-mode features that I think are already implemented (i.e. were checked off on an old to-do list):

512 words times 4 L2 buffers for VME readout of SVT cable data (one known bug: about 0.1% of time the last word of data does not read out correctly over VME, though it is stored correctly in memory; work-around is to read one extra word)
ability to write directly from VME to output connector
ability to read from input FIFO to VME
event-by-event timing info for each L2 buffer
option to suppress Data Strobe to downstream board (e.g. so that we send nothing to L2 during standalone tests)
option to ignore hold signal
internal state machine can wait for backplane L1A (e.g. for back pressure) or not, according to VME-programmable register
internal state machine can look for SVT input data or not (this would provide XTRPemu functionality, which we may not need anymore)
clear counters before seeing first SVT word
increment Level 1 Accept counts properly
512-word RAM (instead of 1K)
VME FIFO block-read option
Test Mode (mode in which data processing is stopped)
read channel ID via VME bits
2K words of spy buffer
for Marco's monitoring program: use spare SRAM space (1.2KB) for circular buffer of last 1K 10-bit processing times in microsecond increments
fix FRAM at U of C (the power was not connected on the artwork)

Here are some utility-mode features that are probably not necessary, but I include them because I once thought they might be useful, or they would emulate some functionality that we currently have in crate b0svt07 (but may not need)

option to pump OSPY data into output cable on each event, which would replace the current "fake track data" option that we have kludged together with Hit Finders and Mergers
option to pump FRAM data out to output cable on each event?

Here are some processor-mode features that I think are needed (and are now all done):

do multiply/add in logic for beam spot instead of using big lookup table
use FRAM for sin/cos table
subtract double-buffered beam spot (allows perfectly emulatable beam subtraction)
also use FRAM for phi linearization
suppress duplicate tracks (done)
expand duplicate filtering to include hit quality bits (in addition to chisq bits) in its decision: this may be useful in 4/5 mode

Here are some processor-mode features that we can probably skip, especially if utility mode has these features:

L1A back pressure
in/out hold programmable