Regarding low latency trace data transfer I tried higher baud rates than the default and they are honored, however, I did not measure lower transfer times. Of course, can lower number of points, or play with rbw.
How about the binary format for the scanraw data transfer. Currently takes 3 bytes per data point, one byte fixed with 'x', then the LSB and MSB.
raw_data = struct.unpack( '<' + 'xH'*N_points, raw_data[:-5] ) # ignore trailing '}ch> '
Since the entire binary blob is encased in '{' and '}' characters, why not drop the 'x' and have just an even number of LSB and MSB in between the brackets. This would yield a saving of about 1/3 of the total binary blob transfer on every trace and therefore lower the latency!
If backwards compatibility must not be broken just add a scanraw2 command or similar.
Already getting about 300 ms for 450 points plus delay of 100 ms to "encourage" parallelism, but would be nice to shave that 300 ms into a 200 ms.
Anyways, thanks for a great device and overall open source programmability!
VSS
Regarding low latency trace data transfer I tried higher baud rates than the default and they are honored, however, I did not measure lower transfer times. Of course, can lower number of points, or play with rbw.
How about the binary format for the scanraw data transfer. Currently takes 3 bytes per data point, one byte fixed with 'x', then the LSB and MSB.
raw_data = struct.unpack( '<' + 'xH'*N_points, raw_data[:-5] ) # ignore trailing '}ch> '
Since the entire binary blob is encased in '{' and '}' characters, why not drop the 'x' and have just an even number of LSB and MSB in between the brackets. This would yield a saving of about 1/3 of the total binary blob transfer on every trace and therefore lower the latency!
If backwards compatibility must not be broken just add a scanraw2 command or similar.
Already getting about 300 ms for 450 points plus delay of 100 ms to "encourage" parallelism, but would be nice to shave that 300 ms into a 200 ms.
Anyways, thanks for a great device and overall open source programmability!
VSS