ESP32 projects: When do you move from breadboard to a custom PCB?

[rachana062]

Hi everyone,

I’ve been working through an ESP32 project using ESP-IDF in VS Code, starting with a simple LED blink application on GPIO18.

While exploring embedded development and PCB design resources from PCBCool, I started wondering at what point it makes sense to move an ESP32 project from a breadboard to a custom PCB.

For a basic prototype, the development board is convenient, but once you start adding sensors, communication interfaces, and other peripherals, wiring can get messy.

For those who have developed ESP32-based products, what usually determines when you design a custom PCB? Is it mainly the number of peripherals, reliability, noise considerations, or the need to reduce the overall size?

Would be interested to hear how you approach this in your own projects.

[nagelc]

I haven't done any ESP32 projects, but have used arduio mini, NRF5, and some other modules. I think the logic is the same. For me, it was 3 things: getting tired of dealing with wires; wanting to get the size down; and wanting enough of a sensor design. I figured out how to get a temperature sensor reliably working, and wanted one in every room.
I started from knowing almost nothing, eventually landed on KiCAD for pcb design, and designed a board with a radio, dallas sensor, and wired connection to a 2 AA battery pack -- pretty much the basic MySensors Design.
Haven't used PCBCool, but there several manufacturers that will give you a cheap price on a prototype run of 3 to 10 boards. They usually have a size limit for that price, so that also forces you into a given size.
I made some mistakes, but for the small prototype price, I could throw away my mistakes and go buy an updated board.
Another thing to consider is lead time. This is a hobby for me, so I can afford to wait for the cheapest delivery, but it can be a bit of a wait depending on where the boards are coming from.

[OldSurferDude]

I generally agree with @nagelc. The nano-computer being used is irrelevant. The quantity of devices being built is the biggest factor for me. This is what I did when I made ten of my "Universal Light Switch".

I accidentally did something that turned out to be quite the boon. This project uses and ESP-12F (ESP8266) It has 8 holes on each side of the board. My the distance between the two rows of holes on the board were a smidge narrower than that of the ESP-12F. Instead of soldering the ESP-12F directly to the board, I put pins in the board. This effectively created a low-insertion-force connector between the board and the ESP-12F. Thus I could easily remove the module for reprogramming if OTA failed. I had brought out Tx, Rx GND, GPIO0 and RST for this purpose, but this was a lot easier in the long run. On the downside, I found that if you remove and replace the same board several times that the contact material can be worn off. At this point, I solder the module to the pins.

Second would be how much experimenting I would be doing. For example, I was trying to use ultra-sonics to measure the amount of water in a barrel. I went through a number of sensors, so on the PCB I brought out I2C, SPI, and some I/O pins. (Ultra sonic didn't work for me and I ended up with a float gauge) Also bring out points where noise filter capacitors would be helpful.

For the 1-off, (Low Cost Energy Meter LCEM), I used an Elegoo PCB breadboard. I soldered wirewrap wire between points and super-glued the wire to the board.

I usually use a solderless breadboard first for making sure the core circuitry works (processor, power supply, etc.) I'm not a fan of these because the "spring-loading" wears out pretty quick and you'll tear your hair out trying to fix what was working yesterday (I'm bald).

On a tangent, another issue is the choice of the nano computer. I find that for basic sensors (on/off, relay, temperature, etc.) an Arduino Nano with an nRF14 radio talking to MySensors gateway to be cost effective and robust. An Arduino Nano with no radio as a gateway can provide I/O to a PC. The ESP devices are great, having built-in WiFi and easy OTA. My experience shows two downsides, though. With their sensitive 3.3V I/O, they can't take much abuse. My Asus router can only handle about 25 WiFi connections (devices losing connection and cannot reconnect until another device loses connection), so a lot of devices could be problematic. My LCEM uses an Arduino Nano.

Home Assistant (HA) has an offshoot called ESPHome. My experience with ESPHome, and HA in general, is that the learning curve is quite steep and the implementation is not as straight forward as MySensors. I do find it a great platform for collecting and analyzing data. There is an HA "MySensors Integration" to which MySensors devices can send data.

Please share with the community the your experience as you move forward with your project. What projects are you thinking about doing?

[ejlane]

I'm in a similar boat.

If the board is small/simple enough and I'm only doing 1 or 2, then yeah, maybe I'd just do it with point to point wire and hot glue. But these days for anything complicated I'll just throw together a PCB. Yeah, it takes a week or two to arrive, but I'm usually busy getting parts, other parts of the project, or just life in general, so I plan around it and it's not an issue.

These days the PCBs are so fast and cheap from China that it's almost never a question except for the simplest cases. It just saves SO much hands-on time to do. I ended up getting a bunch over the years from JLCPCB, and have always been happy with them. Now I don't even look at other shops, and just go to them each time.

I've had very good luck with the ESPHome devices. I had issues with Mysensors, and needing things that were in the alpha code that never got released. I guess these days if you have Claude or ChatGPT go through it, maybe it wouldn't be so painful to find and fix the error. I just didn't have the time, and ESPHome, while it had a learning curve, already had support for the devices that I wanted to use.

My wifi at home is rock solid, and can handle I don't know how many hundred devices, so that does help that the ESP devices don't have any incremental cost in that regard for me. As far as the 3.3V, that hasn't seemed to be much of an issue for years for me - most newer devices seem headed that way or even lower voltage once you get some of the higher performance chips.

Anyway, not trying to argue that anyone is wrong, this is just my experience of it. :)