Thanks for your input!

Another easy way to reduce consumption, what I found out today, is to reduce the core voltage. I'm using the Pico at 50 MHz and can go as low as 0.90V and it saves me about 2mA compared to stock, which is pretty sweet.

I'm using the Pico SDK'svreg_set_voltage() command which is very easy to use.

50MHz will cause it freeze when using 2 cores. Not test for 1 core codes.

Random kinda-off-topic thing I just noticed: I think clk_adc runs even without doing any analogRead().
At the top of void setup():

    Serial.begin();while (!Serial) {};delay(1000);    Serial.println(clock_get_hz(clk_adc));    Serial.print("frequency_count_khz(CLOCKS_FC0_SRC_VALUE_CLK_ADC):");    Serial.println(frequency_count_khz(CLOCKS_FC0_SRC_VALUE_CLK_ADC));clock_stop(clk_adc);    Serial.println(clock_get_hz(clk_adc));    Serial.print("frequency_count_khz(CLOCKS_FC0_SRC_VALUE_CLK_ADC):");    Serial.println(frequency_count_khz(CLOCKS_FC0_SRC_VALUE_CLK_ADC));

output:

48000000frequency_count_khz(CLOCKS_FC0_SRC_VALUE_CLK_ADC): 480000frequency_count_khz(CLOCKS_FC0_SRC_VALUE_CLK_ADC): 0

50MHz will cause it freeze when using 2 cores. Not test for 1 core codes.

Works for me, even down to about 20 MHz. Although, at this point the USB connection drops out, but the Pico still works otherwise. With 50 MHz, even the serial USB connections works like normal for me.

Random kinda-off-topic thing I just noticed: I think clk_adc runs even without doing any analogRead().

Interesting! 👍🏻

I see, but does this compile with the current core? Because AFAIK you have to build some extra dependencies into to core to get it to run.

@lyusupov I use your code and run like this:

            Serial.println("start sleep.");            sleep_run_from_xosc();            datetime_t alarm = { 0 };            Serial.println(" wait alarm.");            sleep_goto_sleep_until(&alarm, NULL); /* TBD */            Serial.println(" awake");            // back from dormant state            rosc_enable();            clocks_init();            rp2040.restart();

Then the output stuck at "start sleep.", and never back from sleep... Any clue?

Just add a libray.json to it and add symlink to the dir will works in platformio.

Which header files have to be included for using set_sys_clock_khz()?

You only needpico/stdlib.h for that.

Sure, here is the relevant code section. It's all the tricks described above. Note that I get to 3ma sleep onmy custom board, the Pocket Integrator, which runs at 1.8v when awake. Other boards have other parts, drains, needs, etc. (Also, my ammeter is something I got cheap off of Amazon. It says 3ma, but I've never actually calibrated it.)

// ... after shutting off LEDs, accelerometer, etc ...// idle core 1rp2040.idleOtherCore();// power down analog referencedigitalWrite(Aref_enable, LOW);// reduce clock speed for mega-savings!// 18mhz seems to be my slowest option set_sys_clock_pll(756000000, 7, 6); #ifdef SLEEP_USB// power down USBSerial.end();pll_deinit(pll_usb);#endif#ifdef LOW_VREG// reduce internal chip voltage?vreg_set_voltage(VREG_VOLTAGE_0_90 ); // 0_85 crashes, but 0_90 works & saves close to a milliamp#endif#ifdef SLEEP_FLASH// power down flash// (saves very little power; 10-40ua according to datasheet)#define FLASHBUFSIZ 4        // FYI these commands are specific to my flash chip; yours may be differentbyte flash_ibuf[FLASHBUFSIZ] = { 0xb9, 0, 0, 0 };  // power-down cmdbyte flash_obuf[FLASHBUFSIZ];flash_do_cmd(flash_ibuf, flash_obuf, FLASHBUFSIZ);#endif//////////////////// try to sleep --// until any button pressed,// or activity on wake or play pin.//playPinRose = false; // the playPin-watching ISR will re-true this if play LED lights during sleep.unsigned int napsTaken=0;bool anyButtonPressed = false;unsigned long totalNapTime_ms = awakeTimer_ms;        // FWIW I haven't figured out which sleep_goto_dormant_until_X() call in the pico sdk         // will wake up on activity from any one of a set of pins. That might save even more power.        // Instead, we just take 100ms naps & then test all the pins, over and over ..do {sleep_ms(SLEEPYTIME);// versus doing it this way -- less power in theory, but something keeps waking it up prematurely ...//absolute_time_t t = make_timeout_time_ms(SLEEPYTIME);//best_effort_wfe_or_timeout(t);napsTaken++;if (tempSettings.napWink) { // Wink the debug led every 4 seconds during sleepif (napsTaken % 40 == 1)leds[4].on();else if (napsTaken % 40 == 2)leds[4].off();}                 // pet the watchdogrp2040.wdt_reset(); readButtonsQuick();anyButtonPressed = btn1pressed || btn2pressed || btn3pressed || btn4pressed;} while ( !(// wake up if the play LED has lit, playPinRose // or if any button was pressed|| anyButtonPressed ));//////////////// wake back up:#ifdef LOW_VREGvreg_set_voltage(VREG_VOLTAGE_1_10);#endif#ifdef SLEEP_USBpll_init(pll_usb, 1, 1440000000, 6, 5); // return USB pll to 48mhztusb_init(); // re-init Adafruit TinyUSB library#endif#ifdef SLEEP_FLASHflash_ibuf[0] = 0xab;  // power-up cmdflash_do_cmd(flash_ibuf, flash_obuf, FLASHBUFSIZ);#endifset_sys_clock_khz(MCU_MHZ * 1000, true); #ifdef SLEEP_USBSerial.begin(115200);#endif// power up analog referencedigitalWrite(Aref_enable, HIGH);// wake core 1rp2040.resumeOtherCore();// then wake up LEDs, accelerometer, etc.

@myklemykle that's some nice work. Do you think it would make sense to include it in the core somehow?

Also, just in general, if you're disabling the flash clocks then you're also going to need to move the code into RAM (i.e.__not_in_flash_func()), no?

I'd be happy to contribute this to core in some format. However the way I'm doing the actual sleeping in the middle of this function -- sleeping until one of N pins is active -- might not be what most people need. So I'm not sure what options to put there instead.

As for the flash thing ... I agree that makes sense, but OTOH I never did move the sleep function into RAM and it still works. Perhaps the entire function is loaded into the XIP cache before it runs?

Also I have to admit that the sleep/wake of USB does not seem to me to be 100% solid. In my testing, if I sleep/wake USB more than 4 or 5 times in quick succession (like within a minute), the host OS (MacOS 11.7 in this case) seems to get grumpy and reject the device until it's unplugged/replugged. I'm not sure if this is an actual fault on the Pico side of if it's something caused by whatever Serial.end() and tusb_init() say to the host if you call them over and over. I'm not even sure that calling tusb_init() multiple times is supported by TinyUSB. So the USB sleep might want to be controlled by an option flag.

As for the flash thing ... I agree that makes sense, but OTOH I never did move the sleep function into RAM and it still works. Perhaps the entire function is loaded into the XIP cache before it runs?

The XIP cache wouldn't know what a function looks like, but it might have a large enough cache line/prefetch to keep things running in most cases with this code. I wouldn't like to count on that, though...a slightly different placement of the function might end up not making it into XIP before turning off the flash. Real pain in the butt to diagnose in the field after an OTA!

@dlee99 Did it work?

Yes, it did. Although I'm only getting down to 12mA, which could be from the previous power issues discussed before in this post and other posts or because I am using the Adafruit Feather rp2040 which may have other peripherals that are using power even when the RP2040 sleeps.

sleep_goto_dormant_until_pin(SELECT, 0, LOW); // wake up on active low.

Also, I ended up not use your library you made. But it is a great reference to see how it's implemented on the arduino-pico core.

Did you try any of the options suggested in#1544 (comment)?

Yes, I tried and only got 1mA difference. From 13mA to 12mA on my multimeter.

This is what I have in place before I put the controller to sleep.

set_sys_clock_khz(18 * 1000, false);
Serial.end();
vreg_set_voltage(VREG_VOLTAGE_0_90); // 0.85v 0_85 indeed does crash and cause the rp2040 to restart.
pll_deinit(pll_usb);

Interesting, nevertheless.

@dlee99 I guess the MCP73931/2 LiPoly charger and the NeoPixel could be the extra current consumers on that board. But the NP is supposed to be only about 1ma at idle, and the charger is supposed to be less than that.
Just curious: are you sleeping the second core?