Have you considered how the noise from a switching supply will affect the accuracy of the pressure transducer readings?
Yest, that too.
I'm using something like that and doing an average of ten readings with short delays. Do you think I can do something else? Test different delays before starting measurements...?
Well consider this:
Your sensor has a resolution of 0.1% of full scale. If you use a 250Ω resistor to do the 4-20mA conversion, the the voltage range is 1-5V or a 4V span, so full scale = 4V. 0.1% of 4V is only 4mV. You would want your noise to be much less than that.
Can you filter it in software? Maybe, depends on the nature of the noise.
My point is, a switching regulator will be much noisier than a battery. Will it affect your measurements? That is up to you to determine.
Good point @jim-p . I had some noise before that I managed to reduce a lot by adding capacitors. Well I think it was that. I can do more tests and try to explore the software solution.
Concerning the TPL5110, I just ordered it. But I don't think it will start the ESP32 when power is turned ON. Each time I upload my program, I need to manually press the Reset button to start the program. Do you know what should I modify in my schematic so that it starts automatically without pressing the button? (check the schematic below)
Thank you
Edit: Actually, I found that:
microcontroller - Power on ESP32 without pressing RST button after applying power (Custom PCB) - Electrical Engineering Stack Exchange
Could it just be my R1 / C3 components? Do you think 1uF would solve my problem?
The 3.3V supply must be above VDD3P3_RTC min and stable before EN is bought high. You have 1022uF on Vdd, that will make it very slow. Espressif recommends 10K and 1uF for EN but that depends on 3.3V being above VDD3P3_RTC min and EN being low
See the ESP32 datasheet for timing parameters
Why 1000uF?
Can't remember where I saw that. I read that long time ago in a forum to have capacitors at different values. I thought it was safe but I didn't think about that for VDD and EN. Still learning. You think the 1000uF is not necessary / recommended?
So in summary, changing C3 for 1uF and changing C1 (1000uF) for something else or just remove it?
Is it possible to control this CHIP_PU in the sofware with some delays to wait for 3.3V being above VDD3P3? (maybe stupid question)
Thanks
You may have other (unknown) stuff on the PCB that may need a capacitor but 1000uF for the ESP is way to much. 100uF should be good enough
There are microprocessor supervisory ICs that will hold the reset (EN) pin low until Vdd get to some voltage and then will release reset after some delay. Search Mouser/Digi-Key. They are usually tiny 3 or 5 pin ICs
Couldn't find anything after a quick search. I will try again later.
I think I need to start from scratch with a completely bare ESP32 and play with capacitors for my future circuits. I underestimate their importance.
Not unless you have experience with RF design. Stick with one of the modules.
Ah okay, I see. Thanks for this advice.
With such a drastic difference between awake and sleep time, your high input voltage, and the current requirements of an ESP32, I don't think a single regulator will do well for you. I would look into adding a second MCU to act as a low power sleep timer and supervisor.
Use a low Iq linear regulator (like the MCP1702-3302E previously recommended) to power the supervisor, and find a switching regulator with an enable pin so the supervisor can turn it on and off. A good regulator might also have a Power Good pin that indicates when it is fully up and working. If not, the supervisor could monitor the voltage with an ADC.
This being the Arduino forum, a good recommendation for the supervisor might be the ATMega328P (the chip the Uno R3 uses). It has a surprisingly useful feature called Power-save which lets Timer2 be clocked off a 32kHz watch crystal and draw less than 1 uA in that power mode.
You can use Power-save to time the sleep cycle, with the switching regulator held off (leaving the ESP and sensor unpowered). When it's time for a reading, the supervisor turns on the switching regulator, holds the ESP in reset, monitors the output voltage until it's good, releases reset on the ESP so it can start up and read the sensor, then waits for a signal back from the ESP (like a GPIO pin toggling) that says the reading is done. The supervisor can then reverse the sequence to power down and wait for the next measurement time.
This way, the high power section of the circuit (the switching regulator, ESP32, and sensor) is only on when it needs to be, while the part that's always on (the supervisor) can be made as low power as possible so that it only sips a couple of microamps.
Hey @Jiggy-Ninja, thanks but I think that's pretty much the same principle as this TPL5110, Jim-p was suggesting? I heard about this ATMega328P of course but never tried to used it. Maybe I should have a try too.
Thanks again Jiggy
In a sense it's similar, but there are pros and cons to each.
For the TPL5110, it's a purpose built chip that does the job it's designed for with no hassle. It's also able to optimize it's parameters far beyond what a more general chip is capable of getting (50 nA running current is an eye-catching spec). The downside is, it only does the job it was designed for. If you need it to do even slightly more than that, you need to hunt around for a new chip to replace it with or add to it. It's time interval is fixed in hardware with a resistor and will be hard to adjust if you need to make changes on the fly. It's also only 1% accurate, much worse than the time accuracy a crystal-based oscillator would get.
Using a microcontroller as the supervisor has the downside of up-front complexity, but it has the massive advantage of flexibility. It can use a crystal oscillator for more accurate timekeeping, intervals can be changed with just a software update, and it's greater capabilities make it able to handle more complicated power sequencing if you need it. It'll draw much more current than the dedicated time chip, but I doubt a 3S LiPo pack is going to care about a couple of uA.
Very clear explanation @Jiggy-Ninja. I still need to test both and decide what's best for my projects. I always use 2h intervals between each measurement, and I think I would prefer the TPL5110. But I'm curious too and can't resist to try the ATMega328P (I have one in one of my boxes).
Thanks for your time.
By the way, I've started watching videos and reading about the ATMega, but still didn't find any tutos about how to program the ATMega to just turn on my circuit containing the ESP32. Would you have a ref to share please?
Thanks
I doubt you'll find any tutorial for this specific application, since I thought of while reading your post. You basically just have to do the design the old-fashioned way, figuring out what you want each part to do and figuring out how to make it do it.
For the ATMega328P, you'll need the datasheet (which you can download from Microchip's website). This will tell you how to set up all the different peripherals and other parts of the controller that you want to use. In particular, you'll want to look for the Power Reduction Register to turn off all of the peripherals you won't be using to save as much power as you want.
Then look at what you need for Power-save sleep mode. This lets you clock Timer2 off of a different crystal (32kHz watch crystal) so it can stay running even when the rest of the chip is off, and can be made to wake the chip up after the counter finishes. You can use this to time your long 2 hour sleeps with minimum power usage.
Now for the hardware, it's pretty easy. You'll want 2 pins connected to the watch crystal for Timer2. You can configure the chip to run off it's internal 8 MHz oscillator when awake, so you don't need the extra 16 MHz CPU crystal.
For connections to the other devices, at minimum you'll need two. You need to control the regulator so one output pin will go to either the Enable input of the regulator or to a MOSFET power switch to disconnect the input from the regulator when you want to power down. An input will need to be connected to an output pin from the ESP.
Program the supervisor (the ATMega328P) to sleep in low power mode until a time for a sensor reading comes. Then you turn on the high power regulator (which will turn on the ESP and sensor) and keep reading the input coming from the ESP. Then, once the ESP's done everything it needs to do (read the sensor, saved it, transmitted it, whatever) it toggles that pin to let the supervisor know it's done. Then the supervisor can cut the power and go back to sleep to wait for the next measurement cycle.
That's a pretty general description, and there might be a lot of details you have to work out. But that's the fun of doing this stuff as a hobby. There should be lots of tutorials for using the ATMega328P off of it's development board (search for breadboard Arduino, you'll find tons). You can use those and guidance from the datasheet to start learning how to set it up exactly how you want to.