Reset-after-Capture delay in Timer: a lost thread

[waclawek.jan]

In the past, we've discussed the fact that in STM32 TIM, there's a delay between Capture and Reset through the derived from the same signal than the Capture. This delay causes period measured using the method called "PWM Input" in RM, to be a couple of timer clock periods shorter than is the real period.

@Tesla DeLorean  experimentally characterized this delay and summed up his findings in a thread, which unfortunately got damaged twice - first at transition from Jive to Salesforce, as witnessed by the wildly varying "Edited by ST Community" timestamps during the first restoration attempt; and then again at the transition to Khoros. As it cannot be restored, @Lina DABASINSKAITE  and the crew kindly provided me with a pdf which contains probably screenshots from some internal database tool. Unfortunately, the original post together with the title has been already lost. I further removed personal information and some uninteresting footers:

So, the result is, the delay in question is two timer clock cycles.

This of course can be used as a compensation, but it's an experimentally determined value, with no guarantees for this being a constant, and no confirmation by ST. Indeed, ST does not even mention the existence of this delay; it surely would not look nice in the "PWM input" chapter.

So, as the second experiment shows, a sure method is to capture successive edges of the same polarity in a free running timer (i.e. without the slave-mode Reset) and calculate deltas "manually".

JW

[Tesla DeLorean]

@waclawek.jan thanks for digging that out of system.

As I recollect this was a synchronous test, using a u-blox LEA-M8F to provide a clock (30.72 MHz) for the system and a related 1PPS (100 ms 1 Hz)

[BarryWhit]

> ST must still have a pre-migration database backup. But I doubt they would spend the time excavating the information out, and they also can't provide access to outsiders due to sensitive information therein (or I could probably do it).

The information above was provided by ST, presumably by digging it up from the Salesforce backup. They may not have have a working Jive backup at this point, the Jive to Salesforce transition was messy, IIRC at that point Jive was already taken over by some competition or something similar, so basically defunct.

I too have some threads saved (unfortunately, not this particular one), as html/rendered, but that's hard, too: the "modern" internet world with its overjavascriptoiding makes html backup almost impossible; and as a bonus, the regular FF revamps ate up the easy possibility of making MHT/RFC 2557 backup/restores.

JW