mirror of
git://projects.qi-hardware.com/ben-blinkenlights.git
synced 2024-11-27 15:41:53 +02:00
09de2454ef
Third time lucky, I hope. -fno-tree-cselim is much more specific than disabling all optimization and results in a considerably less severe performance reduction (about 30-40% of -O0). |
||
---|---|---|
.. | ||
hw | ||
Makefile | ||
README | ||
ubb-la.c |
UBB pattern capture =================== ubb-la is the counterpart of ubb-patgen: it captures a pattern on DATx and prints on standard output what it has received. The output format is the same as the one used by ubb-patgen, i.e., hex digits with repetitions indicated by {number}. ubb-la is currently a proof of concept implementation and therefore only provides the most basic functions. Capture frequencies are the same as for ubb-patgen. The sample size is fixed at 8008 samples. Sample rate ----------- The frequency in MHz is selected with the option -f. For example, this would select 10.5 MHz: # ubb-la -f 10.5 bus 10.5 MHz controller 168 MHz Note that - unlike ubb-patgen - there are no SI prefixes or rounding modes to select. ubb-la simply picks the frequency closest to the one specified. A list of available frequencies can be obtained with # ubb-patgen -q The default sample rate is 1 MHz. Clock output ------------ The CLK line (CLK/TRIG) is usually configured as an input. If the option -C is present, ubb-la outputs the MMC bus clock instead. Trigger ------- Capture only starts after detection of a trigger. By default, ubb-la triggers when the CLK/TRIG line changes. The trigger pattern can be selected with the option -t pattern/mask, where ubb-la triggers when (pins & mask) == pattern For example, # ubb-la -t 2/3 would monitor the lines DAT0 and DAT1 and trigger when DAT0 is 0 while DAT1 is 1. In the pattern and mask, DAT0 has the value 1, DAT1 is 2, DAT2 is 4, DAT3 is 8, and CLK/TRIG is 16. ubb-la turns off interrupts while waiting for a trigger. The only ways to exit are either detection of a trigger or a button press on the Ben's keyboard. In the latter case, ubb-la exits immediately and does not capture any samples. When the trigger is present, it takes roughly 550 +/- 250 ns plus one sample time until the first sample is taken. Known bugs ---------- At higher sample rates (observed at 42 and 56 MHz; not observed at 24 MHz), the first sample may have an incorrect value and should be ignored. Interrupting ubb-la while a capture is in progress could cause memory corruption. The passive ubb-la circuit loads the inputs such that they will only work if there is a strong driver. In particular, signals held only by a pull-up or pull-down resistor are likely to be compromised. The respective resistive loads are: Signal To GND To 3.3 V (max) (max) ------------------------ DAT0 1.1 k 11 k DAT1-3 10 k 10 k TRIG 330 k 330 k