The ESP32… On A Chip

The new hotness in microcontrollers is the ESP32. This chip, developed by Espressif, is the follow-on to the very popular ESP8266, the cheap, low-power, very capable WiFi-enabled microcontroller that came on the scene a few years ago. The ESP32 is another beast entirely with two powerful cores, WiFi and Bluetooth, and peripherals galore. You can even put an NES emulator in there.

While the ESP32 is significantly more powerful, it has for now been contained in modules. What would really be cool is a single chip loaded up with integrated flash, filter caps, a clock, all on a 7x7mm QFN package. Meet the ESP32-Pico-D4 (PDF). It is, effectively, an ESP32 on a chip. It’s just the ticket if you’re trying to cram wireless, fast microcontroller wizardry into a small package.

At its heart, the ESP32-Pico is your normal ESP32 module with a Tensilica dual-core LX6 microcoprocessor, 448 kB of ROM, 520 kB of SRAM,  4 MB of Flash (it can support up to 16 MB), Wireless with 802.11 b/g/n and Bluetooth 4.2, and a cornucopia of peripherals that include an SD card, UART, SPI, SDIO, LED and motor PWM, I2S, I2C, cap touch sensors, and a Hall effect sensor. It’s quite literally everything you could ever want in a microcontroller.

Disregarding the just barely hand-solderable package and the need for a PCB antenna, the ESP32-Pico requires very few support components. Really, the only thing going on in the reference schematic is a bunch of bypass caps. This is, by far, the easiest and smallest method to put WiFi, Bluetooth, and a powerful microcontroller in a project. It will surely be a very, very popular chip for hobbyist electronics for years to come. Of course, it will be even more popular if Espressif also manages to put this chip in a QFP package in addition to the QFN.

Unfortunately, apart from the PDF released by Espressif, the details on the EPS32-on-a-chip are sparse. We don’t know when we’ll be able to get our grubby hands on a tray, tube, or reel of these chips. That said, there’s enough information here to start designing a breakout board. Have at it — we’d love to see what the community comes up with.

Shout out to [Dave] for the tip.

46 thoughts on “The ESP32… On A Chip

  1. ” It’s just the ticket if you’re trying to cram wireless, fast microcontroller wizardry into a small package.”

    You mean like hacking an existing graphing calculator to give it a network connection? Mahahahaha

    Yup! Now imagine what you can do if you dead-bug a HD camera chip on it, the battery or power supply would be the largest part of of your finished device. Or you could put the camera into your calculator too, now why on earth would you want to hide a camera in one though? …. ;-)

    1. Wow. I toyed with this idea in my head a few years ago but the technology just wasn’t there yet. It’s amazing that all of a sudden it would be easy. I would never use it to cheat of course, but it would have been fun to build a proof of concept.

      1. Even the antenna can be far smaller than most people would believe.

        Fractus, developer of fractal antenna technology, has launched of its smallest antenna for the ISM 2.4 GHz band. The 3.7mm by 2mm Micro Reach Xtend antenna is the size of a single grain of rice, providing device designers with more available space to enable new multimedia applications or reducing the space needed within devices.
        The miniature antenna was designed using Fractus’ patented fractal antenna technologies and developed especially for Bluetooth headsets and mobile handsets. It supports all standards working at the 2.4 GHz ISM band, including Wi-Fi, Bluetooth, and Zigbee.

        1. uh… “smaller than one would expect” is this said “one” assumed to be acquainted with wavelengths relation to frequency and impact on antenna theory? are we assuming this person knows what a quarter wave is? aka… do you think this is a youtube comment section?…know your audience shoulda just read “check this antena tech, they shrunk shrunk!and is not even cold out :D ”

          even if someone doesnt know the radio theories above… how many HoD readers do you wager have never opened a laptop panel and seen the wifi antennas contained with in? wasnt the scope things you could fit in a graphing calc?

          /end neurotic rant

    2. I dunno, why would you want a camera? If you’re cheating in exams, surely a keyboard and screen are good enough, doesn’t the keyboard support an alpha mode by design? Of course you could hack an alpha mode in anyway, but this means that the markings are already on the keys and case.

      Unless you want to do sign language, and swap the screen for a TFT, and hope nobody notices. Which they certainly would. A mic / earpiece would be better, but obviously noticable. Kbd + screen seems the way to go. All you need them is to squash in the ESP for comms with your cheating friend outside. Of if you wanna get really clever, and maybe strap some more RAM in there, port Linux, and have access to Google and

      Actually that would be super-clever. Maybe have a gateway on your PC at home though to format screens better for a lo-res Texas Instruments mono LCD. Assuming the last WAP portal is dead. Shame if it is actually cos that’d be easy to develop a browser for, a WAP web would be nice for calculator networking in general. I think somebody ported TCP / IP to the TI calcs.

      Then there’s intra-classrom IRC chat, but that’s just asking to be spoiled by some blabbermouth, or somebody getting it wrong and asking out loud “how do you get to the bookmarks?”.

      1. Given the camera, I would assume a remote person (team?) would be receiving images of the test/exam and reporting answers back over a simple text stream. Would work particularly well for multiple-guess exams!

    3. this might show my age…but i know for a fact that the ti89 could be transfered to the ti83 shell and all that was required was shaving down a few plastic tabs on back of the case. …if you have already used the 89 for a few years the keys not matching arnt troublesome, dont think just press and u get what you want assuming any relevant athorities are expecting you to prove you are detail oriented with their usual head in the clouds aproach of teaching… you should be good to go. tldr: it all comes down to the odds a prof notices the gui on a 3 inch grayscale screen from several feet away.

      wifi would be great for some chem or physics stuff, but for math having an 89 with its symbolic manipulation is probably as good as wifi any day. at least if you are the fucks order rules up type with a never caught math disability.

  2. I don’t know if Espressif or Sprite_tm (sorry, dude, I don’t give up) is listening or anything like that… but a 4g version of the ESP chip would be freakin’ amazing… ;) specifically what I’m picturing is an as-cheap-as-humanly-possible microcontroller that can do 4g voice and data communications. Imagine if you could buy a dumbphone at Dollar Tree (or Poundland, for those in the UK, sorry I don’t know what the equivalent 1€ chain or chains would be for the rest of Europe…)? Not a *toy* phone — an actual, working mobile/cell phone! That’s what *I* would do with it… a cheap, super-basic phone for anyone, even those who can’t afford to pay $15/mo for three decades for a proper smartphone.

    1. You could however get 100 pieces of your own OEM cellphone made for $12,– each… With touch and Android and color display. There are also developer boards out there, but the toolchain is way sketchy ;-)
      They even come in racecar-designs! Based on some obscure China platform made from all sorts of ICs custom-bonded inside a larger package. CCC had a talk on that last year or the year before.

      1. There are also $5 dumbphones on eBay, but they’re all 2g. I want to see a product that’s both more compatible with modern networks, and less expensive to manufacture at the same time. If Espressif can do this, they /should/ be able to do that as well.

        1. Got any links or example brands?
          I need a few 850MHz GMS handsets for a project and in general I’d buy a few 5usd cellphones just for tearing them to pieces to see how it was possible to make them so cheap.

          Just imagine if some folks manage to reverse it like the osmocom folks did with TI Calypso and created osmocomBB.

          1. You can get watch-phones for $7 according to some kid on Youtube. Mine was 17 quid, same (GT-08) model. I’ve seen them listed for $7 but paid 17 cos I figured it was cheap enough, and stood a better chance of being a working one. Takes a SIM card, is a watch. Takes SD-card too, and has GPRS as well as GSM. Also plays MP3, but not video (which is a shame, since I’m sure it’s an ARM chip in there, surely some custom low-quality low-overhead codec would do the trick? ). Also has Bluetooth, both as host and client (audio and SPP / GTMM) and a touch screen. And a strap. Colour LCD about 128×128 maybe.

      2. If you’re talking the Gongkai phones Bunnie wrote about, they aren’t smartphones, and they’re limited to 2G. Often using Mediatek’s older chipsets, and popular in phone-watches. You do get GPRS though, 2.5G.

        Or are these really smartphones, full-size screen and all? Surely this would be worth any small business buying 100 and selling them for, say, 2 or 3x the buying price? For a smart phone that undercuts the market rate by half. If these really are full-screen cap-sensing smart phones, plz supply a link.

  3. If only all features of the ESP32 would work.
    And don’t get me started on the Arduino side…
    But yeah, this will be the hot fuzz for the next years – once you can use it without going crazy.

    1. Its just all a matter of reading the documentation and experimentation. They have gicen a mem map so you can implement functonality before the SDK matures fully. Plus their documentation is quite good I have to admit especially on the SDK. They are continually improving though. Been on the ESP32 for over 6 months and the guys at expressif were kind enough to even send a direct quote for their WROOM32 modules for a product we were making with no MOQ. I am sold.

      1. I’m looking for Espressif to release their agnostic WiFi support they promised, currently only FreeRTOS is supported although Espressif promoted the port of NuttX for ESP32. The basic port is working, but without the support for WiFi, Bluetooth, CAN, SDCard, etc it is better use use a MCU with BCM43362 that works fine with NuttX (i.e. Particle Photon Board)

  4. Oh Boy.
    There are so many things that could be done today that were not possible to conceive 20 years ago. And you think the public is maybe 10 years behind what they can actually do.
    I am so amazed at what is out there. And the things we can do at home now.
    Can you imagine what is out there that we don’t know about.
    And this is what a china company has done.
    I wonder what it is going to be like 5 years from now. SO SO mind blowing.
    And i am still working on using up my esp8266 stock up. And now I have to get more then 2 esp32.

    1. Military or TLA technology rarely exceeds what is possible in the private sector. What they DO have is the ability to spend way more on something that wouldn’t ordinarily be a viable product.

      1. Depends on the kind of tech. Some tech there is little or no market for in the private sector – either because it’s illegal, or because you simply do not need those kinds of tech if you aren’t waging a war. For example I’m sure the military has way better missile guidance tech than what you can buy from any private company, if you can even buy that at all.

    1. This is most of the stuff under the metal can in the Adafruit boards, in a single QFN chip that is about the size of a grain of rice. Totally different product, this is more of targeted at people designing complete integrated products, whereas the Huzzah32 is a developer board that breaks out the pins of the ESP32 and has power and stuff.

  5. The module is still probably a better option if you plan to manufacture and sell more than 1 and less than say, 100k of your IOT product. I’ve never used an ESP module in particular, but if the module is FCC, WiFi, and Bluetooth certified, then it makes a lot more financial sense to leverage those certifications as much as possible. Antenna tuning also requires significant expertise and specialized tools (VNA, spectrum analyzer, etc), so for a consumer device, pre-certified modules are a no-brainer for mid-level production volumes – especially at the pricepoint of the ESP modules.

    1. Hmmm, the datasheet claims “Given that no other peripheral components are involved, module welding and testing is not required either.” I assume by testing, they mean certification? The datasheet doesn’t specifically mention FCC, Wifi, or BLE / Bluetooth certifications. Note that for all BLE / Bluetooth devices, the Bluetooth SIG requires qualification / listing, even for a pre-qualified modules. I clicked the “Download certificates for Espressif products from here.” link in the datasheet, but saw nothing there for the ESP32-PICO-D4.

  6. Why would they use QFP, if they already have QFN?
    For many manufacturers, it’s the other way around. They have QFP, their customers want QFN, but they’re struggling to get the QFN packages qualified.

    1. Mostly to appeal even more to the maker market. QFP is a lot less intimidating for people without well equipped workshops. Anyone with patience and a half decent soldering iron can do QFP. QFN is a pain without an oven or hot air.

      On the other hand, size concious designers will appreciate the QFN package.

  7. Hi
    I have been hacking on the ESP32 since the early release – Best micro processor experience that i ever had. Choose Arduino or ESP-SDK for simple or real time OS. Basic wireless lan stuff is up after 30 min hacking and you just keep adding stuff that you normally would require a bigger system to like the eg Raspberry pi
    To see Espressif has a path to smaller devices is so nice. Module produces will build breakout with this so we do not have to – that has been the model also with the esp8266.
    I do integrated audio class D stuff day time job – and has found the EPS32 to be the sweep spot on price, processing power and hardware interface to cover the next wave of audio streaming applications ranging Hi-fi to Alexa/Echo like application. But lets keep that as a secrete – not a lot have seen what is coming.
    Regards Jakobsen

      1. I have not seen that setup yet – but the new module and revision has plenty of sram for buffing and an audio pll that will do for higher end audio. But it very much depend how you device you high end audio and want to balance your full audio path – eg trying to run the audio stream through a 5$ esp32 and a 5$ integrated class D amp over high end speaker cable at 20$ / meter and 200$++ speakers mite still be high end audio audio if you box it up to look like million. /Jakobsen

    1. I think CAN got dropped from the datasheet of the ESP32 as well, but that’s just because we don’t have a driver for it in the SDK (yet). The peripheral is still there and the third-party driver found on GitHub works well, as I heard.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.