#include "mbed.h"

Timer timer;
AnalogIn adc(A0);

int main()
{
    float a = 0.0;
    unsigned short b = 0;

    printf("\nTesting floating point A/D performance...\n\n");

    timer.start();
    for (int i = 0; i < 1000000; i++) {
        a = adc.read();
    }
    timer.stop();
    printf("\tResult (%f): %fus\n", a, timer.read_us() / 1000000.0);
    timer.reset();

    printf("\nTesting integer A/D performance...\n\n");

    timer.start();
    for (int i = 0; i < 1000000; i++) {
        b = adc.read_u16();
    }
    timer.stop();
    printf("\tResult (0x%X): %fus\n", b, timer.read_us() / 1000000.0);
    timer.reset();
}

I just benchmarked the AnalogIn performance on several of my microcontrollers, and the LPCXpresso11U68 board is extremely slow! I measured 176µs for a floating point reading, and 173µs for an 16-bit integer reading. By comparison, the LPC11U24 (which has basically the same peripherals as the LPC11U68) only takes 18µs for a floating point reading, and 14µs for an 16-bit integer reading. Something is clearly amiss here... I've opened an [[https://github.com/mbedmicro/mbed/issues/345|issue over at GitHub]], and I've attached my benchmark code and findings in case anyone is interested: <<code main.cpp>> #include "mbed.h" Timer timer; AnalogIn adc(A0); int main() { float a = 0.0; unsigned short b = 0; printf("\nTesting floating point A/D performance...\n\n"); timer.start(); for (int i = 0; i < 1000000; i++) { a = adc.read(); } timer.stop(); printf("\tResult (%f): %fus\n", a, timer.read_us() / 1000000.0); timer.reset(); printf("\nTesting integer A/D performance...\n\n"); timer.start(); for (int i = 0; i < 1000000; i++) { b = adc.read_u16(); } timer.stop(); printf("\tResult (0x%X): %fus\n", b, timer.read_us() / 1000000.0); timer.reset(); } <</code>> \\ {{/media/uploads/neilt6/adc_benchmark_table.jpg}} {{/media/uploads/neilt6/adc_benchmark_chart.jpg}}

Testing floating point A/D performance...

        Result (0.263492): 175.767955us

Testing integer A/D performance...

        Result (0x4314): 172.541668us

Testing floating point A/D performance...

        Result (0.271306): 14.500005us

Testing integer A/D performance...

        Result (0x4574): 10.791682us

Hi, Thanks for your detail report. I have checked the LPC11U68 library source code and find the sampling clock is an issue. Current mbed library for the LPC11U68 use 500kHz sampling clock. [CLKDIV = 100] <<code>> Testing floating point A/D performance... Result (0.263492): 175.767955us Testing integer A/D performance... Result (0x4314): 172.541668us <</code>> I locally changed this value to be 1. i.e. 50MHz sampling clock. [CLKDIV = 1] <<code>> Testing floating point A/D performance... Result (0.271306): 14.500005us Testing integer A/D performance... Result (0x4574): 10.791682us <</code>> I will make a pull request for this when I finish other testing. Regards, Toyo

Cool thanks! Also, holy crap, that's as fast as the [[/platforms/LPCXpresso1549/|LPCXpresso1549]]!

Btw it depends on the settings also for the other targets. The KL25 should be able to do that also, but currently it is set for 16-bit mode, and iirc also uses some hardware averaging.

33µs on the KL25Z seems plenty fast enough to me. I think most people would prefer accuracy over raw speed (within reason of course).