Hi,
if you would want to measure the instant in time you do something() in software with a scope or logic analyser (by toggling a pin once) on a typical x86 PC (e.g. pci, serial, lpt available, ok maybe not THAT typical anymore ;) running linux, what kind of I/O would you choose? latency should be minimal, but without major hardware hacks (no soldering to CPU pins or similar please :)
Hi Stefan,
On Wed, Dec 5, 2012 at 6:55 PM, Stefan Tauner stefan.tauner@student.tuwien.ac.at wrote:
Hi,
if you would want to measure the instant in time you do something() in software with a scope or logic analyser (by toggling a pin once) on a typical x86 PC (e.g. pci, serial, lpt available, ok maybe not THAT typical anymore ;) running linux, what kind of I/O would you choose? latency should be minimal, but without major hardware hacks (no soldering to CPU pins or similar please :)
I did something similar for quantifying latency and jitter for a control system (like a PID loop). What is your margin of error? If you can only tolerate an error of about 1 nanosecond, you may have to take great pains to make sure you have an accurate clock. Some physicists had to do that to resolve their questions: http://en.wikipedia.org/wiki/Faster-than-light_neutrino_anomaly
You could read from a PCI or PCI-e device and a scope or logic analyzer on the bus would know the instant that happened.
Hope that helps, David
-
David Hubbard -
Stefan Tauner