lgt8fx
lgt8fx copied to clipboard
dbuezas
How to use sleep modes on LGT8F328P ?
Hi !
So I moved from Arduino Nano to the LGT8F328P and I feel a little bit lost. The LowPower.h library I was using is not working anymore.
I'm looking for another way to reduce consumption for a certain period of time (2 seconds). I already have the LGT file installed in my hardware arduino location.
Thanks for your help !
I too have problems with the power consumption. The lowest I can get is 170µA, nowhere near the 1µ[email protected] that is mention on the manual or that I can achieve with a atmega328.
I think in this lgt8fx package the sleep modes controlling software elements and examples are unmodified and not tested.
To set the LGT8F328P to (deep) sleep mode (and wake up settings) is much more complicated than in case of ATmega 328P because of -LGT8F328 has more (deep) sleep control registers ( PRR1, DPS2R, IOCWK and more bits in MCUCR and MCUSR ) -it has more wake up modes/sources ( IO level change, watchdog, ADC, etc. ) -has more circuit design restriction ( eg. need pull-up or pull-down resistors connected to some IO-s to achieve lower power consumption) -need to power down some more internal circuits than ATmega ( eg. low voltage detector (LVD) or internal voltage source circuits and more )
You can learn from databooks and untested examples :
https://github.com/dbuezas/lgt8fx/tree/master/lgt8f/libraries/PMU
https://github.com/dbuezas/lgt8fx/tree/master/lgt8f/libraries/WDT
https://github.com/dbuezas/lgt8fx/blob/master/docs/LGT8FX8P_databook_v1.0.5-English.pdf
https://github.com/dbuezas/lgt8fx/blob/master/docs/Porting_from_LGT8FX8D_to_LGT8FX8P_v1.0.0.pdf
( the lgt8f328p_spec.h is integrated into this package so you don't have to include it)
Hi everyone,
Well 170μA seems already low hahaha, with the 5v pin I obtained 1mA for a project with the purple LGT8F328P board, I was happy about it. But I tried the same with the green board, and I obtained 25mA, impossible to get lower values. I used destroyed led technique, no 5v regulation, and deep sleep mode code. The green and purple boards seems to have a different PCB layout...
For the deepsleep mode I directly use the internal sleep register as the lowpower library is not compatible with this boards.
Le mar. 5 juil. 2022 à 07:29, LaZsolt @.***> a écrit :
I think in this lgt8fx package the sleep modes controlling software elements and examples are unmodified and not tested.
To set the LGT8F328P to (deep) sleep mode (and wake up settings) is much more complicated than in case of ATmega 328P because of -LGT8F328 has more (deep) sleep control registers ( PRR1, DPS2R, IOCWK and more bits in MCUCR and MCUSR ) -it has more wake up modes/sources ( IO level change, watchdog, ADC, etc. ) -has more circuit design restriction ( eg. need pull-up or pull-down resistors connected to some IO-s to achieve lower power consumption) -need to power down some more internal circuits than ATmega ( eg. low voltage detector (LVD) or internal voltage source circuits and more )
You can learn from databooks and untested examples : https://github.com/dbuezas/lgt8fx/tree/master/lgt8f/libraries/PMU https://github.com/dbuezas/lgt8fx/tree/master/lgt8f/libraries/WDT
https://github.com/dbuezas/lgt8fx/blob/master/docs/LGT8FX8P_databook_v1.0.5-English.pdf
https://github.com/dbuezas/lgt8fx/blob/master/docs/Porting_from_LGT8FX8D_to_LGT8FX8P_v1.0.0.pdf ( the lgt8f328p_spec.h is integrated into this package so you don't have to include it)
— Reply to this email directly, view it on GitHub https://github.com/dbuezas/lgt8fx/issues/202#issuecomment-1174630070, or unsubscribe https://github.com/notifications/unsubscribe-auth/AYJPAGV2OZ44IVY2NBUGGIDVSPB4LANCNFSM5RC6AS7A . You are receiving this because you authored the thread.Message ID: @.***>
The purple and the green boards has USB to serial chip CH340 or HT42B534 ( and AMS1117 LDO regulator ) which has some current load.
I think in this lgt8fx package the sleep modes controlling software elements and examples are unmodified and not tested.
To set the LGT8F328P to (deep) sleep mode (and wake up settings) is much more complicated than in case of ATmega 328P because of -LGT8F328 has more (deep) sleep control registers ( PRR1, DPS2R, IOCWK and more bits in MCUCR and MCUSR ) -it has more wake up modes/sources ( IO level change, watchdog, ADC, etc. ) -has more circuit design restriction ( eg. need pull-up or pull-down resistors connected to some IO-s to achieve lower power consumption) -need to power down some more internal circuits than ATmega ( eg. low voltage detector (LVD) or internal voltage source circuits and more )
You can learn from databooks and untested examples : https://github.com/dbuezas/lgt8fx/tree/master/lgt8f/libraries/PMU https://github.com/dbuezas/lgt8fx/tree/master/lgt8f/libraries/WDT https://github.com/dbuezas/lgt8fx/blob/master/docs/LGT8FX8P_databook_v1.0.5-English.pdf https://github.com/dbuezas/lgt8fx/blob/master/docs/Porting_from_LGT8FX8D_to_LGT8FX8P_v1.0.0.pdf ( the
lgt8f328p_spec.his integrated into this package so you don't have to include it)
Has anyone achieved the 1µA in standby mode? (This is easily achieved on a a atmega328p)
Running the same code
Are any of the interested parties still involved in this or has this issue passed from usefullness?
@alxgarza
Has anyone achieved the 1µA in standby mode?
I tested and reached current lower than 2 µA in deep-sleep mode. (Measured with a 3 and 1/2 digits METEX multimeter.) I was used the SSOP20 Mini style board for testing because only three components had to be removed. The voltage regulator, and 2 LED's resistors. (The LQFP32 Mini style board can be just as good.) At the first test I didn't remove the builtin LED's resistor so it turned out the pull-up resistor switched on when the deep-sleep mode started. After this I removed this resistor and connected an external LED with a 10k resistor between D13 Arduino pin and VCC, because when pull-up resistors switching on this LED won't draw any current.
// LGT8F328P sleep test DPS2
#include <avr\sleep.h>
// Sleep mode 111 binary
#define SLEEP_MODE_DPS2 SLEEP_MODE_EXT_STANDBY
void setup() {
pinMode(LED_BUILTIN, OUTPUT);
digitalWrite(LED_BUILTIN,HIGH);
delay(8000); //wait 8 sec for current measure before sleep
digitalWrite(LED_BUILTIN,LOW);
delay(500); //wait more 1/2 sec before sleep
cli();
IOCWK = 0; // port D level change wake up. 0 = No wake up when level changes
set_sleep_mode(SLEEP_MODE_DPS2);
sleep_enable();
DPS2R = 1<<DPS2E; // DPS2 enable (bit 3), LPRC timer disable (bit 2)
sleep_cpu();
}
void loop() {}
@alxgarza
Has anyone achieved the 1µA in standby mode?
I tested and reached current lower than 2 µA in deep-sleep mode. (Measured with a 3 and 1/2 digits METEX multimeter.)
I replicated your [code] (https://github.com/dbuezas/lgt8fx/issues/202#issuecomment-1419681782 )and you are a genius. I got 0.67uA (670nA)!
@alxgarza you are more than welcome to check other [see bottom] discussions related to power consumption. I think it is quite easy to reach proper power saving to your needs:
~16uA: DPS0/1 sleep with SRAM retention and with WDT for wake up. ~6-7 uA: DPS0/1 sleep with SRAM retention but w/o WDT. ~5 uA: DPS2 sleep w/o SRAM retention but with LPRC for wake up. ~0.6uA(600nA): DPS2 sleep w/o SRAM retention, w/o LPRC for wake up, only setting PortD as wake up pins.
Those are measured with Vcc = 3.3V. If you are willing to take the privilege that lgt8f328p can run at full speed at Vcc=1.8V, then these current can be significantly reduced.
~1.1 uA: DPS2 sleep w/o SRAM retention but with LPRC for wake up. ~0.2uA(220nA): DPS2 sleep w/o SRAM retention, w/o LPRC for wake up, only setting PortD as wake up pins.
More power consumption measurements on DPS2 mode - with or without periodic wake up
@alxgarza you are more than welcome to check other [see bottom] discussions related to power consumption. I think it is quite easy to reach proper power saving to your needs:
~16uA: DPS0/1 sleep with SRAM retention and with WDT for wake up. ~6-7 uA: DPS0/1 sleep with SRAM retention but w/o WDT. ~5 uA: DPS2 sleep w/o SRAM retention but with LPRC for wake up. ~0.6uA(600nA): DPS2 sleep w/o SRAM retention, w/o LPRC for wake up, only setting PortD as wake up pins.
Those are measured with Vcc = 3.3V. If you are willing to take the privilege that lgt8f328p can run at full speed at Vcc=1.8V, then these current can be significantly reduced.
~1.1 uA: DPS2 sleep w/o SRAM retention but with LPRC for wake up. ~0.2uA(220nA): DPS2 sleep w/o SRAM retention, w/o LPRC for wake up, only setting PortD as wake up pins.
More power consumption measurements on DPS2 mode - with or without periodic wake up
Has anyone tried to put it to sleep with a 32.768LHz crystal? Waking up every second.
No I haven't try that since running a crystal usually increases sleep power consumption, unless you would like to sacrifice some power for accurate timing during sleep. Frankly I would like to use a dedicated RTC IC for that purpose.
No I haven't try that since running a crystal usually increases sleep power consumption, unless you would like to sacrifice some power for accurate timing during sleep. Frankly I would like to use a dedicated RTC IC for that purpose.
On a ATMEGA328p, I have successfully achieved < 10µA in sleep mode running the RTC with the Asynchronous Timer and a 32.768KHz crystal. Running with internal clock @ 8Mhz, 3V and waking up every second for housekeeping. This is on Power-save mode. In reality it consumes ~1µA in sleep mode, wakes up just to add a second. So in average power consumption is about ~5µA.
https://ww1.microchip.com/downloads/en/Appnotes/Atmel-1259-Real-Time-Clock-RTC-Using-the-Asynchronous-Timer_AP-Note_AVR134.pdf
But on the LGT8F328P I can't get below ~135µA. With the same setup.
One of the differences between ATmega and LGT8F328P is how to handle of the internal analog devices during sleep. LGT8F328P in the most of the sleep modes not disable these analog devices except DPS2 sleep mode. These devices may draw some current if they were in use before sleep: ADC, two Analog Comparators, Low Voltage Detector (LVD is equivalent of BOD), DAC, Internal Voltage Reference, Differential Amplifier, Voltage Division circuit
It is also advisable to disable the two internal oscillators in that case when the external 32 kHz crystal used during sleep.
There is also an option to turn off the FLASH during sleep. This should be set (6th bit of MCUCR) before going to sleep similar to the sleep enable bit in SMCR
In the LGT8Fx databook there is no mention of how the SWD port behaves during sleep. So it might be a good idea to disable during sleep.
And there is a completely undocumented LDO. I don't know what amount of current it consumes. I found the disabling code in the PMU library and in this doc Porting_from_LGT8FX8D_to_LGT8FX8P_v1.0.0.pdf
And the Timer 3. It could be disabled by writing the appropriate bit of PRR1 register.
@LaZsolt , what a nice document you found :)
I feel this side information quite important: (Deepl translate)
In DPS2 mode, all IOs have a weak internal pull-up resistor (~20K~40K) automatically turned on, which needs to be taken into account when designing the system!
And there is a completely undocumented LDO. I don't know what amount of current it consumes. I found the disabling code in the PMU library and in this doc Porting_from_LGT8FX8D_to_LGT8FX8P_v1.0.0.pdf
I found that LDO shutdown code also from PMU lib and that doc you mentioned. And I tried to test and reverse engineer:
Test conditions: Clean circuit. No LDO, LED, No pin attached to anything except for VCC pin and GND pin connected to joulescope and RST pin pulled up. Sleep mode: DPS0 or DPS1 (doesn't matter) WDT: disabled
Key code section
DDRB = 0; DDRC = 0; DDRD = 0; PORTB = 0xff; PORTC = 0xff; PORTD = 0xff; //Stop floating pins
ADCSRA = 0; //I always need to turn off ADC explicitly to get good power saving even if I didn't turn it on.
//Enable WCE bit so that LDOCR can be written within the next 4 clock cycles.
LDOCR = 0x80;
//LDOCR bits: WCE na na na PDEN VSEL2 VSEL1 VSEL0
LDOCR = 0b010; //0b0010 = 0b0011=0b0111=0b0001 7uA, 0b1xxx 42uA
Therefore, VSEL2 VSEL1 VSEL0 doesn't matter. I suspect these bits might be LDO trimming register or output voltage register. However, if you disable PDEN (I guess it is Power Down ENable register) the LDO will be disabled in power down mode (DPS0 and DPS1). But I am not sure if LDO supplies the power to SRAM so disabling it means you lost SRAM data?
I will do more test.
----------Test result------------ SRAM is retained even if LDO is turned off. So the LDO only ouput Vcore, not SRAM.
@sullivanzheng
LDOCR = 0b010; //0b0010 = 0b0011=0b0111=0b0001 7uA
This is pretty low consumption with SRAM retention. When you measured this value was the FLASH power and SWD interface shut down?
The FLASH power can be disabled during sleep by the FPDEN bit of MCUCR. SWD interface can be disabled by the SWDD bit of MCUSR.
I think LDO (which can be disabled by PDEN bit of LDOCR register) is powering some part of the analog modules such as Internal Voltage Reference. So when LVD module (ATmega BOD equvalent) is enabled during sleep to detect low VCC then LDO must be not disabled. I will do some test to measure it. The output of the internal reference voltage will be put on AVREF pin if ADTM bit of ADCSRC is set. (测试模式，从 AVREF 端口上输出内部参考电压)
I think LDO (which can be disabled by PDEN bit of LDOCR register) is powering some part of the analog modules such as Internal Voltage Reference. I will do some test to measure it. The output of the internal reference voltage will be put on AVREF pin if ADTM bit of ADCSRC is set.
Not true. :disappointed:
- To place out on a pin the voltage of IVREF with setting the ADTM bit of ADCSR isn't working. But I could measure it on the DAC output pin.
- When LDO was powered down the IVREF voltage does not change.
How to use sleep modes on LGT8F328P ?
The anwer is the LGT LowPower library.
Just uploaded: https://github.com/LaZsolt/lgt8fx/tree/master/lgt8f/libraries/lgt_lowpower
It's under constuction, not tested, but I look forward to your questions.
@LaZsolt when you think it is good enough we can add it to the core libraries or at least link it in the readme and add an example to the core, wdyt?
I need one or two weeks for publish. The readme has some garbage in the middle and I would like to make some tests and make more examples (eg. to present how to turn off analog modules).
Do you have examples of all the chips/cores that you need?
On Tue, Mar 28, 2023 at 3:30 PM LaZsolt @.***> wrote:
I need one or two weeks for publish. The readme has some garbage in the middle and I would like to make some tests and more make examples (eg. to present how to turn off analog modules).
— Reply to this email directly, view it on GitHub https://github.com/dbuezas/lgt8fx/issues/202#issuecomment-1487485857, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACPEX7DRMTTLHLKBIVXA3G3W6M36VANCNFSM5RC6AS7A . You are receiving this because you commented.Message ID: @.***>
Do you have examples of all the chips/cores that you need?
Not really. I plan to convert the existing examples from LowPower and PMU ilbraries to LGT. After this I will create some new. I don't have a detailed idea of the the new examples at this moment.
PMU.h : lgt_LowPower.h 1:0
PMU is working, but my adapted LowPower not. The disassembed code looks perfect, but somehow not good.
The lgt_LowPower library started working finally. The difficulties were caused by faulty WDT operation.
Today I tested only the Idle sleep mode and created the first 2 examples. https://github.com/LaZsolt/lgt8fx/tree/master/lgt8f/libraries/lgt_lowpower/examples
I'm already very tired. I will continue the tests in the next few days.
Today I created and tested one more example. I have written many comments in the source. Aren't these many comments annoying? https://github.com/LaZsolt/lgt8fx/tree/master/lgt8f/libraries/lgt_lowpower/examples/adcNoiseReduction
I think explaining your rationale this way is excellent. People can understand better why you code this way and if someone understands your objective and has a smarter way of coding that, you may get good feedback.
Today I created and tested one more example. I have written many comments in the source. Aren't these many comments annoying? https://github.com/LaZsolt/lgt8fx/tree/master/lgt8f/libraries/lgt_lowpower/examples/adcNoiseReduction
No, absolutely not. They are very helpful and educational. Many times I've read my own code and wondered what the heck I was thinking. To have this insight into the thoughts of someone else is invaluable. Please be as verbose as you wish to be! Please don't feel that your comments are in any way a negative.
Thank you all for your feedback. :slightly_smiling_face:
It used to happen that there were twice as many comments as code in my sources. At the end the code was almost unreadable because lost in remarks.
Adapting the Low-Power library to LGT8F328 is much more complicated than I thought. I've been doing it for 3 months now, but there are still surprises.
I tested the library example of the wake-up capabilities of the external interrupt pins yesterday, but not worked. After reading the databook, it turned out that the external interrupt flag bit will not set while interrupts are disabled with cli. Interrupts must be disabled with cli due WDT error. This error causes the MCU to reboot when the sleep time expires if the interrupts not disabled with cli.
So when the sleep period is set to SLEEP_FOREVER the external interrupt pins could work if interrupts reenabled before sleep instruction... Well, then I started rewriting the affected parts of the source code. again