Logging Rate Root Cause Analysis

[niekky]

It was last diagnosed that the firmware has some unnecessary delay blocks in it which cause the logging rate to be slow.

Task:

• Run data collection experiment with the test cart, collect live data and data stored in flash logging rate.
• Run experiment with no instruments and with instruments to see if it's a hardware issue.

[ETSells]

Request profiling the following performance characteristics using start_time = HAL_GetTick(); tDelta = HAL_GetTick() - start_time;:

• tDelta over sensor_dump
• tDelta over each sensor type (all of the PTs, all of the ADCs, all of the TCs)
• tDelta over each individual sensor (PT 1, PT 2)
• tDelta for flash logging

[ETSells]

also note @niekky: plenty of our recent firmware optimizations are not available yet in mod, so i recommend prioritizing getting mod up to date.

for example, if we remove the delay call here, we'll see noticeable speed improvements

[ETSells]

@niekky @Bigmouseears With the flash delay removed in a rough test, I'm seeing ~2000 hz. This doesn't account for everything (maybe sensor dump takes longer, maybe there'll be some interrupts and data transfers that slow down the main loop), but ideally it's basically an instant speedup. We'll see.

Before:

After:

[niekky]

Interesting, did we have mod updated yesterday and the result wasn't much better? How did you do the rough test?

[ETSells]

Rough test was with GDB over USB, no RS485. I literally just ran the data logging loop. We supposedly did have mod updated yesterday with little new results, but idk if i trust it 100%.

[niekky]

Yeah so application with RS485 indeed does run slower than with USB. I'll have to look into it myself and let you know my theories.

[ETSells]

it'd make sense to be a little slower, but the degree to which it is doesn't make sense to me since this isn't blocking I/O.