Small ESP32 Module for Easier Custom Design

Unhappy with the ESP32-WROOM module, I designed an ESP32 D2WD module to ease custom board assembly and enable small, wearable ESP32 devices

Similar projects worth following
My journey to develop a suitable small modular ESP32 based on the new ESP32 D2WD IC for use in portable and wearable projects. Let the fun begin!

The ESP32 is a promising MCU with concurrent wifi/BLE, two 240 MHz CPUs, a floating point unit, an RTC, and lots more GPIOs than the also-very-useful ESP8266/85. I prefer to design my own circuits rather than use modules but sometimes a well-designed module can speed the prototype process and ease assembly work in custom boards. The commonly used ESP32-WROOM-32 module is one I have used in this way, but I found its size rather gigantic and limiting. I was only able to expose a few of the GPIOs without widening this board even further:

Sensors on the back turn this into a motion sensing utility board intended for pocket wear but the size of the WROOM module plus pads (20 x 26 mm) makes very small board design prohibitive.

The WROOM module is this large partly because of the external SPI flash used for both data storage (via SPIFFS, for example) and program flash. In addition, the WROOM uses the standard 6 mm x 6 mm ESP32 QFN48 package. The WROOM uses connector pads intended to solder to the sides of the module instead of below like the Rigado BMD-350, for example. Lastly, the WROOM is designed with a rather large pcb antenna which takes up a large part of the module length.

With the advent of the ESP32 D2WD, this can now change. The ESP32 D2WD is embedded with 2 Mbytes SPI flash, so no need for an external flash chip. The ESP32 D2WD is a 5 mm x 5 mm QFN48 package, 69% as large as the standard 6 mm x 6 mm QFN48 ESP32. The combination allows the creation of much smaller ESP32 board and module designs.

First of all, just the design of a development board using the ESP32 D2WD requires a smaller board area (~12.7 mm x 12.7 mm) for the ESP-specific circuit.

Upside is smaller ESP32 IC size and no external flash. The downsides include two fewer GPIOs (losing GPIO16 and 17) and much harder hand assembly. The pitch of the pins is a terrifying 0.35 mm, difficult for all but the most expert solder paste slingers.

In order to ease the burden of this difficult assembly I designed a small module based on the circuit in the above board. The idea is to have the module manufactured in China at a reasonable cost, and then make use of the module with a much easier soldering process.

I first captured the ESP32 D2WD circuit onto a 12.7 x 12.7 4-layer pcb, added three mm in length for a chip antenna and designed a reasonable routing of the 24 available GPIOs and 8 power pins to back pads on the board periphery to make mounting using (low-temperature) solder paste and a hot plate/hot-air gun very easy. This is following the soldering strategy Rigado uses for its nRF52 modules like the BMD-350 which I use a lot and find the soldering of the module quick, easy, and reliable. I want the same experience for the ESP32.

I redesigned the development board into a daughter board to accept the module and I am in the middle of assembling everything and testing the concept.

Tomorrow's task is to assemble one or more of the ESP32 modules and then solder them onto the daughter board. The required pullups on nRST, GPIO0, and GPIO12 (the embedded flash runs at 1.8 V, which is why we lose two GPIOs) as well as the pull down on VDD_SDIO are already on the module, so there are fewer passives required on the daughter board. I included the 32.768 kHz (ABS05) crystal and passives on the module since eventually I would like to be able to make use of the RTC clock and sub-processor. I left off the 270 pF capacitors connecting pins 38/39 and 36/37 on the module, of course, but also on the daughter board, I can't remember why.  It is easy to remove the caps if these GPIOs are needed for inputs. Maybe with two fewer GPIOs I was thinking touch pins are less useful? I added a 0603 10 uF bulk capacitor on the VCC input to the module for good measure.

There is no guarantee this is going to work, or is going to work well. I have never used this particular type of chip antenna before but I find it hard to believe...

Read more »

ESP32 development board module acceptor

x-zip-compressed - 248.49 kB - 03/04/2018 at 02:04


ESP32 module design v.01a

x-zip-compressed - 83.36 kB - 03/04/2018 at 02:03


  • First Assembly of ESP-PICO-D4 Development Board

    Kris Winer03/20/2018 at 01:48 2 comments


    I got both the ten ESP32-PICO-D4 ICs I ordered from Electrodragon and the three pcbs I designed for them from OSH Park today. So I immediately went to work assembling the components to make my first ESP32-PICO-D4 development board.

    The assembly of the LDO, battery charger, USB, FTDI chip, etc went as usual as I have done this dozens of times so nothing interesting there. The assembly of the PICO IC was super easy, no terrifying 0.35-inch pitch IC like the ESP32-D2WD. In fact, the PICO is easier to solder well than any of the ESP32 ICs I have tried. More like soldering the ESP8285/66. And with most of the components embedded in the PICO putting the entire board together took less than an hour, where usually just the ESP32 part takes this long.

    The board worked first time out. I was able to flash a program using the Arduino IDE using ESP32 dev module, DIO, 80 MHz flash speed, 921600 UART speed, and 2 MByte flash selections (I got the flash size wrong but it didn't seem to matter). I got the led blinking, serial data (battery voltage and chip temperature) streaming to serial. I need to test the pin functionality, etc but so far so good!

    I tested the wifi and was seeing up to four networks including my home network, an extender, a laptop and my neighbor's wifi network. The Antenova performance seemed to be about as good as the pcb antenna on the ESP32-WROOM module and slightly poorer than a monopole copper wire but it is hard to tell with RSSI alone and I would like to find some convenient way to test and optimize the antenna performance short of buying a network vector analyzer.

    Here is what the board ended up looking like:

    Overall I am so pleased with the performance and ease of assembly and low cost of the ESP32-PICO-D4 that I am abandoning any attempt to create a module of my own since Espressif has done such a wonderful job of it. Kudos Espressif!

  • Skunked!

    Kris Winer03/09/2018 at 02:13 0 comments


    While trolling the Hackchat forum I happened to come across a mention of the ESP32-PICO-D4 which I had never heard of. Well, it turns out this is a SoC (System-on-a-Chip) like the ESP32-D2WD with 4 MBytes of embedded flash instead of the 2 MBytes of embedded flash of the D2WD. Furthermore, even though the PICO is 7 mm x 7 mm (2x larger in area than the 5 mm x 5 mm D2WD) it contains the 40 MHz crystal, and almost all of the passives required to make it work. Plus, as a QFN48 package with 0.5 mm pitch it is a whole lot easier to assemble than the D2WD, which still gives me nightmares.

    So here is the easy-to-use, compact module I was trying to make already available for about $5. Not much point fooling around trying to make my own module then when Espressif has already done it and better than I ever could.

    So, I bought ten of the ESP32-PICO-D4 from Electrodragon, replaced my module with the ESP32-PICO-D4 and submitted the new development board to OSH Park. I expect to get the pcbs and ICs  at about the same time and have a working board by the end of March. Here is what it looks like unpopulated:

    I chose to add a 32.768 kHz crystal for the RTC, but this is not required. I would have had the 0603 10 uF capacitor anyway for the board using my own module and the extra 100 nF capacitor is in the reference schematic. As is a pi matching circuit to tune the RF to optimize the coupling to the Antenova antenna. But I expect either a small capacitor or 0 Ohm jumper will be enough to connect the LNA-IN to the antenna. So really, almost no external components are really needed with the PICO solution. Sweet!

    I kept the FTDI USB-to-UART connector and battery charger and buttons, etc that I use for all of my ESP32 development boards.

    So this could be nearly the perfect solution for rapid customization of hardware designs using the ESP32. And I like the fact that the PICO has 2 Mbytes more flash and allows use of GPIO16/17 (since the PICO flash operates in the 3.3 V domain) unlike the D2WD IC.

    I will report on the board assembly and performance in a couple of weeks. In the meantime, I am suspending development of my own module. No point now really...

  • Wifi Antenna Performance

    Kris Winer03/04/2018 at 19:42 0 comments


    I ran a simple wifi scan program as a test of the antenna performance and found the chip antenna quite disappointing. For the 1.2-inch wire antenna, I found the two wifi hotspots in my house (router and extender) at RSSI of -27 and -59 dB, on average, and about half the time I also detected the neighbor's wifi at -93 dB. With the ESP-WROOM dev board and its pcb antenna I did a little worse, finding the two in-house stations at -30 and -67 dB and once I also found the neighbors at -91 dB. Worst of the three was the chip antenna, which found both in-house hotspots at -38 and -71 dB and never saw the neighbors wifi. This is only a semi-quantitative result but it is clear that either the chip antenna is improperly designed (I guessed, after all) or the matching circuit is way off. I think this means short of a network vector analysis and pi match tuning, I am going back to the wire antenna solution...

  • First Module Assembly...Success!

    Kris Winer03/04/2018 at 01:48 0 comments


    Phase 2 is assembling the module onto the acceptor board. Now this process is straightforward and I expect success in the sense that I will be able to actually solder it uneventfully. Here it is:

    To my delight, I was able to flash the same blink program and get it to work the first time out. The battery voltage monitor on pin 34 seems to be reading too high, but the led blinks and I am reading the ESP32 chip temperature. Now to do some more testing in the next few days to compare the performance between the integrated board and the modular one. But overall, I declare victory!

    Now if I choose to go ahead and have some modules made there are a couple of changes I would make. One is I would use an 0.8-mm-thick pcb to reduce the overall height of the module and make reflow easier. I would also insist that the module have straight edges instead of the somewhat ugly mousebite remnants that come with the OSH Park boards. I would add more ground vias to the ESP32 thermal pad to improve the heat sink and reduce the inductance between the board and IC. I would probably also add a thermal pad onto the acceptor board like the WROOM module for the same purpose.

    I am not sure about the metal can. There are design houses that can make custom RF enclosures but I am not sure I want to incur the expense. Perhaps I will make one more design iteration to add the vias and thermal pad before I declare this modular design good enough. Some testing of the peripherals is in order as well while I wait for the design revision.

    Somewhat to my surprise, it looks like this idea not only has merit but is cursed with working the very first time!  Maybe I should quit while I am ahead...

  • First Module Assembly

    Kris Winer03/04/2018 at 00:49 0 comments


    Just completed Phase 1 of the ESP32 module assembly. The strategy in a nut shell is (Phase 1) to assemble an integrated development board using the ESP32 D2WD and then a module with same, then (Phase 2) solder the module to the acceptor board and (Phase 3) compare performance of the integrated board to the modular board.

    Here is the collection of boards at the end of Phase 1:

    Initial assembly of the two ESP32 ICs and surrounding circuit was reasonably quick but it took me a long time to cure the last of the solder bridges. This 0.35-mm-pitch is tough!

    I verified I could flash a blink program (ESP32 Dev Module, QIO, 80 MHz flash speed, 2 MByte flash, 921600 upload speed) onto the integrated board using the Arduino IDE and the led works and I am getting battery voltage and ESP32 IC temperature on the serial monitor as requested. So at least the basic design seems sound and the integrated board works.

    I soldered a 1.2-inch copper antenna on the integrated board, which provides adequate reception. One of the things I want to test is the reception of the chip antenna relative to this monopole wire antenna solution.

    Here is a close up of the module before I assemble it onto the mother/acceptor board:

    Everything except the ESP32 D2WD itself is easy to assemble. I am using the 40 MHz HSE crystal in this design even though I had used the same 26 MHz crystal in my original ESP32 and ESP8285 development boards. The ESP32 works with the 26 MHz crystal but since the current version of the ESP32 reference design uses 40 MHz so will I.

    Next task is to solder the module onto the acceptor board:

    Next, Phase 2...

View all 5 project logs

Enjoy this project?



Xasin wrote 09/06/2018 at 18:11 point

I really like the possibility of a smaller yet equally powerful board. Of course, there's always the ESP-WROOM-02 featuring a much smaller size, but ... It just looses a bunch of GPIO.

Good luck! I wouldn't want to solder this myself (that chip size IS daunting), but if you get a few models working that'd be awesome!

EDIT: I didn't realize the ESP-WROOM-02 was only using the ESP8266!
That's a bummer.

  Are you sure? yes | no

Kris Winer wrote 09/08/2018 at 04:17 point

I abandoned this project since Espressif has done a better job with the ESP32 than I was doing (even though mine worked just fine). The ESP32-PICO-D4 6 mm x 6 mm IC only requires an external capacitor (actually, even this is optional), is super easy to solder (being a QFN), and only costs about $4 when you can find them. All of my custom ESP32 boards will use this from now on.

  Are you sure? yes | no

Xasin wrote 09/08/2018 at 09:09 point

Ohhh, now THAT is what I want to hear! 

Sorry your project didn't work out, but ... With such a chip, it's hard to say no. Espressif should really learn how to advertise though, because up until now I had no idea there was such an all-in-one chip available.
Thanks for the info!

  Are you sure? yes | no

AVR wrote 03/19/2018 at 16:44 point

nice work here! very similar to my Vento board variants!! I'll be looking at your work as I revamp my schematic.

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates