I have an Stimer capture ISR that is simple and quick. I toggle an IO pin in the ISR. Capture input is driven with a 5KHz square wave. With the scoped hooked up I can see once in a while I miss a capture IRQ, I tracked it down to coinciding with an SPI transmission. I am using Artemis ATP with Arduino IDE with latest Apallo3 2.2.1 library. I am running 2 threads with the mbed rtos. One is a 5ms thread that does the SPI master stuff, the other is a 50ms thread doing I2C and serial stuff. It seems the SPI transfer func is responsible, I have eliminated other things like beginTransaction and whatnot. I have change the capture ISR priority from 0 to 7 with no change. Tried many things, been looking inside the mbed lib and not seeing anything with am_hal_iom not seeming to touch timer ISR regs… Any ideas? I am an embedded controls engineer for 25 years so I am not a newbie.
SPI in the Apoll0 library 2.2.1 is NOT using interrupts. It is using the hal-function am_hal_iom_spi_blocking_fullduplex(). The IOM interrupts are disabled and the data is gathered by polling the IOM-FIFO in a while() loop.
From other projects I did, I know that MBED brings a HUGE overhead. Many of the time-critical programs are not working (stable) on V2.x.x. like Software Serial(https://github.com/paulvha/apollo3/tree … wareSerial) , OneWire(https://github.com/paulvha/apollo3/tree … re_on_uart ) You could try with v1.x.x, but it does not support threads.
potential solution could be:
you must Ensure that SPI and capture ISR do not conflict in resource usage.
Investigate if SPI transmission delays coincide with missed capture IRQ.
Thanks all for the suggestions, I did track it down to a slow fall time. For whatever reason, the IRQ didn’t happen during SPI transmission, but by putting a schmitt trigger on the input to drastically increase the slew-rate, I now have no missing pulses… Go figure.