Fix PicoScope streaming issues #188
Conversation
drslebedev
commented
Dec 18, 2025
drslebedev
commented
- Avoid duplicate digitalOutSpan.publishTag() when samples are dropped
- Increase internal driver buffer size to gain performance for high sample rate
- Log actual sample rate in verbose mode
- Fix TimingMatcher alignment for negative deltas (std::floor) and use a single safeSamples threshold
drslebedev
force-pushed
the
fix_pico2
branch
from
d3066f7 to
996440c
Compare
| return std::unexpected{Error{"No channels configured"}}; | ||
| } | ||
| scope.data.resize(65536); // todo: figure out better buffer size strategy | ||
| scope.data.resize(16 * 65536); // todo: figure out better buffer size strategy |
There was a problem hiding this comment.
Don't hard-code this number; fix it, e.g., by making it depend on the sampling and polling rates, a configurable integer parameter, and a configurable minimum size.
There was a problem hiding this comment.
Updated! The buffer size now depends on the sample rate and the number of active channels. Polling rate isn’t considered yet, because the user can’t configure it, and the PicoScope block’s effective polling rate is simply the scheduler’s processBulk execution rate.
const std::size_t kBaseBuf = 16384;
const std::size_t mult = std::max<std::size_t>(1, static_cast<std::size_t>(std::ceil(freq / 1e6)));
scope.data.resize(activeChannels * mult * kBaseBuf);I didn’t expose the base buffer size as a user setting, because it is not intuitive and depends on the formula we use (and that formula may change in the future). What would likely be more intuitive for the user is the ability to configure the polling rate, but that isn’t supported right now.
| actualFreq = detail::convertTimeIntervalToSampleRate(timeInterval); | ||
| started = true; | ||
| if (scope.verbose) { | ||
| std::println("actualFreq:{}, timeInterval:{}, units:{}", actualFreq, timeInterval.interval, magic_enum::enum_name(timeInterval.unit)); |
There was a problem hiding this comment.
This print-out will kill any performance metric for polling rates > 100 Hz
There was a problem hiding this comment.
This message is printed only once at startup, so it shouldn’t affect performance. It’s also only printed when console_verbose is set to true.
- Avoid duplicate digitalOutSpan.publishTag() when samples are dropped - Increase internal driver buffer size to gain performance for high sample rate - Log actual sample rate in verbose mode - Fix TimingMatcher alignment for negative deltas (std::floor) and use a single safeSamples threshold Signed-off-by: drslebedev <dr.s.lebedev@gmail.com>
drslebedev
force-pushed
the
fix_pico2
branch
from
996440c to
506d967
Compare
|
