since I bought PPK2 my devices (ESP32S2) reached another level. My recent result is: temperature/humidity/light/PIR motion sensor that is sleeping and reporting every 3min or when motion happens powered by 1000mAh battery, after 380 days is still at 50% of the LiPo - I would NEVER reach this without PPK2 and spending some hours on proper design. Super nice video as usual
Awesome video, Robert! I'd like to add that rechargeable batteries have a noticeable self-discharge, so for IoT, primary batteries are better if you can afford them. But, if the battery is small, or the consumption peaks are high, it's also important to check the evolution of the internal resistance of the battery, because it will not be able to deliver the same amount of current towards the end of the charge (good decoupling capacitors help a lot). Also, the capacity curves are not the same for continuous consumption and for pulsed consumption: that factor also affects the duration of the battery. Last but not least: temperature also affects 1) consumption of the device, 2) battery capacity. Cheers!
Hi Robert, thanks for the video. As a lot of people says, the Nordic Power Profiler Kit II is a very good and cheaper option. I've using it for years and I've done a lot of battery-powered devices, resulting a very nice calculation of battery life. Absolutely recommended.
51:50 where you considered the batter self discharge current. For Li-ion rechargeable batteries, the self discharge current will be much higher. Thats why primary batteries are used in the long battery life product's.
Robert, your words at 1:02:10 came straight from my heart 🤩 I actually couldn't resist and bought the JouleScope 220 when I saw this excellent video. It will be a good improvement over my beloved LowPowerLab CurrentRanger. Especially because it uses an isolated USB port to transfer the data. Right now I am using a scope on an isolation transformer to measure current with high time resolution. But that's all tinkering compared to what's in store for me 😁
Here is my setup: 1)if you are using arduino ide export your code to Microchip studio or VScode depending on board that you are using check arduino core code and delete all code that turn on all unused peripheral like ADC, Timers etc. 2)If you are using Atmega base controller don't use standart arduino input output functions it’s spends milliseconds instead of hundreds of microseconds 3)Check current of all peripheral on your board (RS485 convectors, Step up, Step down convectors, GSM, Lora, Radio transmitters etc), if power consumption bigger that 1uA, add mosfet to turn on, off that device, if this is powerfull device or you need capacitors after mosfet, don't forget to add capacitor before, so you controller will not reset after you turn on your mosfet. 4)Quartz Crystal resonator is another device that take to much current, avoid to use it, if it's possible if you need RTC(Real Time Clock), use internal RC oscillator(a lot microcontrollers have it), and recalibrate it every time you success send data to server by time taken from server, 5 minutes per 10 days it's precession that I archive, if you need more accuracy use low power clock crystals(be careful transconductance of amplifier inside mcu must be at least 3 times bigger that crystal’s), low transconductance, high ESR of crystal means low reliability of crystal, noise, humidity can stop it. Modern mcu like Atmega328PB(not Atmega328P),STM32 have special functions that called up when this happens 5)You can use cheap low power mcu like atmega8, to turn off, on through mosfet mcu like ARM cortex A, atmega8 is enough to do a lot of simple task 6)Turn off all your perif like usart, spi etc., switch pin states of usart, spi etc. pins to high impedance without pullups, pulldowns before going to sleep. 7)If your battery don't provide enough current add big capacitor 8)Check what happened if it's not enough power during transmition, most GSM controllers for example simply resets it selves. 9)decrease time of sends and check how many sends of data your controller could provide before discharge battery, you can use lion rechargeable battery like samsung 18650 for that. 10)increase time of sends, and check current between sends by using your multimeter, or you can also use big capacitor instead of battery, connect it parallel to battery and after send data, disconnect battery. 11)Using previous information, you can calculate battery time of your device. 12)There are different types of battery like Lithium Thionyl Chloride high power type, Lithium Thionyl Chloride high energy type, rechargeable Lithium ion batteries. There are always been trade between is battery rechargeable, can it provide big currents, capacitance. So for example if your device need 10 times low current that battery could provide, you can try to find another battery with same size and price but lower current and bigger capacity. Buy battery that could provide 2-3 times lower current that you need and add capacitor, very often solution for GSM modules.
Hello, I would like to ask you a question. Regarding high-speed differential pairs, can we refer to the power layer? Will it have a bad impact? Thank you.
Robert .. would you like to make a video about - INVERTERS - I have a project, in it is an inverter 12/220 volt .... is the inverter simply a transformer ? .... thanks for even short practice explanation 🎉
Nice research, but I might say that the calculating of dc-dc efficiency is inaccurate and depends on the duration of deep sleep (the more time in sleep - less efficiency of the dc-dc). Also the quiescent current of dc-dc is forgotten. So, the efficiency of LDO can be better with the the bigger periods of transmitting data. One more thing about 10% of the capacity of the chosen battery, after dropping to 3.5V it will drop dramatically to 3V sooner than expected (because of non-linearity of the discharge curve), so, choosing a dc-dc instead of an LDO to get more capacity is not so sufficient. The last thing, in low power devices with all optimizations the quiescent current of the chosen IC (LDO or DC-DC) becomes more sufficient
Hi thanks for the feedback! DcDc efficiency losses are calculated on the average load current during the lifetime of the device. In the example that is around 5.3mA and it is where the DCDC efficiency is extracted. The DCDC efficiency reported in the data sheet will also take in account quiescent current , it is POUT/PIN (generally).
@@andrealongobardi2084 hi! Exactly, but the average current drops with bigger periods of wakings, so, the efficiency drops with lower current. For instance, with mcp1700t-3302e/tt there will be less power hungry with 1.6 uA of the quiescent current. I might add that it cannot handle 500 mA for a long time, but for your use case it will be able to and won't be overheated (or add extra capacitance). Just my point of view
video explains how we went from 8 days to over 300 days: 1) measure current consumption with proper equipment 2) use sleep mode 3) BUCK/BOOST is better than LDO 4) Optimize your firmware
@@raguavivathat is typical. I assume you are new to this kind of stuff. Content creators would use higher quality images that might not be the actual. However the same concepts apply.
@@raguavivano, you're being an overly literal sperg. I would like to see more details on how to optimize circuits and code, but he isn't "lying". It's not as in depth as it might be, but it's exactly the information that popped into my head and what I've done before on the job.
These interviews are gold, it is a shame that they don't get that many views, they are a really nice introduction to the subjects.
since I bought PPK2 my devices (ESP32S2) reached another level. My recent result is: temperature/humidity/light/PIR motion sensor that is sleeping and reporting every 3min or when motion happens powered by 1000mAh battery, after 380 days is still at 50% of the LiPo - I would NEVER reach this without PPK2 and spending some hours on proper design. Super nice video as usual
Awesome video, Robert! I'd like to add that rechargeable batteries have a noticeable self-discharge, so for IoT, primary batteries are better if you can afford them. But, if the battery is small, or the consumption peaks are high, it's also important to check the evolution of the internal resistance of the battery, because it will not be able to deliver the same amount of current towards the end of the charge (good decoupling capacitors help a lot). Also, the capacity curves are not the same for continuous consumption and for pulsed consumption: that factor also affects the duration of the battery. Last but not least: temperature also affects 1) consumption of the device, 2) battery capacity. Cheers!
Awesome, listened to Andrea on the amp hour podcast. And now this video with more technical information, Thanks Robert and Andrea!
Hi Robert, thanks for the video. As a lot of people says, the Nordic Power Profiler Kit II is a very good and cheaper option. I've using it for years and I've done a lot of battery-powered devices, resulting a very nice calculation of battery life. Absolutely recommended.
Very comprehensive and full of valuable content! Thank you for sharing.
51:50 where you considered the batter self discharge current. For Li-ion rechargeable batteries, the self discharge current will be much higher. Thats why primary batteries are used in the long battery life product's.
Robert, I don't know if you even need any encouragement to continue with what you're doing, but your content is amazing.
Interview videos don't get many views, but I really like making them. It is very nice to hear that some other people like them too.
What an interesting and well presented topic ! Thank you Andrea and Robert.
Robert,
your words at 1:02:10 came straight from my heart 🤩
I actually couldn't resist and bought the JouleScope 220 when I saw this excellent video. It will be a good improvement over my beloved LowPowerLab CurrentRanger. Especially because it uses an isolated USB port to transfer the data. Right now I am using a scope on an isolation transformer to measure current with high time resolution. But that's all tinkering compared to what's in store for me 😁
Excellent discussion guys! Very informative
Here is my setup:
1)if you are using arduino ide export your code to Microchip studio or VScode depending
on board that you are using check arduino core code and delete all code that turn on all unused peripheral like ADC, Timers etc.
2)If you are using Atmega base controller don't use standart arduino input output functions it’s spends milliseconds instead of hundreds of microseconds
3)Check current of all peripheral on your board (RS485 convectors, Step up, Step down convectors, GSM, Lora, Radio transmitters etc),
if power consumption bigger that 1uA, add mosfet to turn on, off that device, if this is powerfull device or you need capacitors after mosfet,
don't forget to add capacitor before, so you controller will not reset after you turn on your mosfet.
4)Quartz Crystal resonator is another device that take to much current, avoid to use it, if it's possible if you need RTC(Real Time Clock), use
internal RC oscillator(a lot microcontrollers have it), and recalibrate it every time you success send data to server by time taken from server,
5 minutes per 10 days it's precession that I archive, if you need more accuracy use low power clock crystals(be careful transconductance of amplifier inside mcu must be at least 3 times bigger that crystal’s),
low transconductance, high ESR of crystal means low reliability of crystal, noise, humidity can stop it. Modern mcu like Atmega328PB(not Atmega328P),STM32 have special functions that called up when this happens
5)You can use cheap low power mcu like atmega8, to turn off, on through mosfet mcu like ARM cortex A, atmega8 is enough to do a lot of simple task
6)Turn off all your perif like usart, spi etc., switch pin states of usart, spi etc. pins to high impedance without pullups, pulldowns before going to sleep.
7)If your battery don't provide enough current add big capacitor
8)Check what happened if it's not enough power during transmition, most GSM controllers for example simply resets it selves.
9)decrease time of sends and check how many sends of data your controller could provide before discharge battery, you can use lion rechargeable
battery like samsung 18650 for that.
10)increase time of sends, and check current between sends by using your multimeter, or you can also use big capacitor instead of battery, connect it parallel to battery and after send data, disconnect battery.
11)Using previous information, you can calculate battery time of your device.
12)There are different types of battery like Lithium Thionyl Chloride high power type, Lithium Thionyl Chloride high energy type, rechargeable Lithium ion batteries. There are always been trade between is battery rechargeable, can it provide big currents, capacitance. So for example if your device need 10 times low current that battery could provide, you can try to find another battery with same size and price but lower current and bigger capacity. Buy battery that could provide 2-3 times lower current that you need and add capacitor, very often solution for GSM modules.
Thank you for sharing this, really good tips.
Would love to see the numbers for bluetooth
Hello, I would like to ask you a question. Regarding high-speed differential pairs, can we refer to the power layer? Will it have a bad impact? Thank you.
Robert .. would you like to make a video about - INVERTERS -
I have a project, in it is an inverter 12/220 volt .... is the inverter simply a transformer ? .... thanks for even short practice explanation 🎉
How he is calculating days by taking sampling time and clock frequency ?? , I can't understand
Nice research, but I might say that the calculating of dc-dc efficiency is inaccurate and depends on the duration of deep sleep (the more time in sleep - less efficiency of the dc-dc). Also the quiescent current of dc-dc is forgotten. So, the efficiency of LDO can be better with the the bigger periods of transmitting data. One more thing about 10% of the capacity of the chosen battery, after dropping to 3.5V it will drop dramatically to 3V sooner than expected (because of non-linearity of the discharge curve), so, choosing a dc-dc instead of an LDO to get more capacity is not so sufficient. The last thing, in low power devices with all optimizations the quiescent current of the chosen IC (LDO or DC-DC) becomes more sufficient
Hi thanks for the feedback! DcDc efficiency losses are calculated on the average load current during the lifetime of the device. In the example that is around 5.3mA and it is where the DCDC efficiency is extracted. The DCDC efficiency reported in the data sheet will also take in account quiescent current , it is POUT/PIN (generally).
@@andrealongobardi2084 hi! Exactly, but the average current drops with bigger periods of wakings, so, the efficiency drops with lower current. For instance, with mcp1700t-3302e/tt there will be less power hungry with 1.6 uA of the quiescent current. I might add that it cannot handle 500 mA for a long time, but for your use case it will be able to and won't be overheated (or add extra capacitance). Just my point of view
Or instead of buying an expensive ass device you could just measure the current with an arduino. Log it to an sd card and do the maths in excel...
Not accurate, not fast, not high bandwidth. Accuracy and precision are key here
@@lolsypussy whats the min req?
Dislike for using a click-and-bait thumbnail, you just show a power profiler but do not talk about what the image promises.
video explains how we went from 8 days to over 300 days: 1) measure current consumption with proper equipment 2) use sleep mode 3) BUCK/BOOST is better than LDO 4) Optimize your firmware
@@RobertFeranec the thumbnail shows a specific board and a 1.5volt battery, then you talk about something else, this is dishonest.
@@raguavivathat is typical. I assume you are new to this kind of stuff. Content creators would use higher quality images that might not be the actual. However the same concepts apply.
@@lolsypussy If many people do something wrong, it doesn't become right. Chances are this video was an advertisement.
@@raguavivano, you're being an overly literal sperg. I would like to see more details on how to optimize circuits and code, but he isn't "lying". It's not as in depth as it might be, but it's exactly the information that popped into my head and what I've done before on the job.