Espressif ESP32 Dual Core SoC Features Faster WiFi, Bluetooth 4.0 LE, and More Peripherals

Espressif teased us about a successor to ESP8266 a few months ago that would support both WiFi and Bluetooth Low Energy, and John Lee, working for Espressif Systems, has now sent a letter to ESP8266 developers announcing the new wireless SoC with two Tensilica L108 cores and called ESP32.

ESP32
Illustration Only, Not the actual ESP32 SoC

Espressif ESP32 key improvements over ESP8266:

  1. Faster WiFi – Wifi has been upgraded to support HT40 speed (144.4 Mbps) and has a new RF architecture to simplify the application schematics
  2. Bluetooth Low Energy and Classic
  3. Dual core processor – 2x Tensilica L108 processors clocked at up to 160 MHz
  4. Low Power Mode Improvements – ADC conversions, level thresholds, etc.. can now be performed in deep sleep
  5. Peripherals – Capacitive touch, ADCs, DACs, I2C. UART, SPI, SDIO, I2S, RMII, PMW, etc… but no USB.
  6. More RAM – ~400 KB on-chip RAM
  7. Security – Hardware accelerated AES and SSL, and more undisclosed improvements.
  8. Simplified APIs – Not many details provided here, except WiFi APIs will be simplified, yet keep good flexibility and control.

One of my contact also informed me that there were two PGA (Programmable Gain Amplifier) blocks connected to two ADC blocks in the chip, which could be used for power metering with one PGA/ADC block used for voltage measurement, and one PGA/ADC block for current measurement. So ESP32 would offer a one chip solution for smart sockets. [Update: A few more information gathered from twitter: Close to 40 GPIOs, IPV4 and IPv6 support, [email protected] is where you want to write to apply for a board, QFN package, not much more expensive than ESP8266, …]

The letter promises that beta testing will start soon with about 200 applications boards to be sent out, and a bounty program will be launched for people who can successfully compromise ESP32 security. The CEO also made a call to engineers who are interested in working with the company in Shanghai.

Support CNX Software - Donate via PayPal or become a Patron on Patreon

50
Leave a Reply

avatar
50 Comment threads
0 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
16 Comment authors
Dr. Nissim ZurBig LESP32 Module Schematics and Board FilesNewracom NRC6101 802.11ah WiFi SoC Promises up to 1 km Range for Smart Grid and AutomationPaul Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
Johny007
Guest
Johny007

Looking forward to this one, because when I got ESP8266, it took me just few moments before I ran into low memory problems.

Member

@Johny007
What did you run on it ?

Johny007
Guest
Johny007

@Marius Cirsta
I was trying to make a remote controller for Xiaomi Yi camera. It involves using display, JSON decoding library. I forgot to mention I used LUA, because it already had all the libs I need and the language was simple to learn. I hope I will find some time to finishit it soon 🙂

zoobab
Guest

@Johny007
Switch from LUA to C?

aalku
Guest
aalku

@zoobab You can use C in ESP8266 too, even Arduino IDE.

Member

@Johny007
Your problem is that you’re using LUA. I know some folks have promoted the ESP8266 for using LUA and Python and these sort of scripting langauges but if you’re serious about embedded programming you should ( read must ) use something that can compile to native code like C or C++.
If you want to run LUA you should probably get a board that runs Linux but it will use more power and you’ll also probably need an USB wifi dongle. It’s also going to cost a bit more but shouldn’t cost more than $30 and that’s not going to give any out of memory errors.

zoobab
Guest

@Marius Cirsta
I stopped dreaming about an interpreted language for embedded system when my WRT54G got a new version of openwrt that was taking 100% CPU when the LUA web interface was eating all the resources.

Ali
Guest
Ali

Strangely, the image looks like that of an ESP8266 with part of the 8 and all of the 66 covered. The space between the ESP and 3 also is strange

Ali
Guest
Ali

Never mind, just noticed the tag says “Not the actual ESP32 SoC” @Ali

Abdullbasit
Guest
Abdullbasit

The picture shows a 32-pin package so how do we expect “close to 40” I/O pins ?
Unless the picture is fake.

TC
Guest
TC

@Abdullbasit
it’s not a real picture

Member

cnxsoft :
@Abdullbasit
That’s why I wrote “Not the actual ESP32 SoC”.

Lol, next time just pot a big question mark 😀

Member

I wonder if the low-power improvements include being able to continue execution after a sleep instead of essentially resetting after a deep sleep? Maybe some extra sleep levels beside “deep”?

And what’s the purpose of having two processors? Could one be dedicated to communications (Wifi/BLE), while the other can focus on I/O? Do they share memory?

Having more GPIO would be good too.

Paul
Guest
Paul

> instead of essentially resetting after a deep sleep?

No, any chip which wants to get as low as possible in power consumption effectively powers 99% of itself off and then gets reset.

Also, mind that this new chip will be more power hungry than ESP8266 – you just can’t have so much stuff inside and have it eat less. Mere leakage current will be much higher.

Paul
Guest
Paul

> I stopped dreaming about an interpreted language for embedded system when my WRT54G got a new version of openwrt that was taking 100% CPU when the LUA web interface was eating all the resources.

That’s because it’s Lua. I have a router with 128Mb RAM in front of me, and on it, OpenWRT’s LuCI is as slow as on 16Mb router. Because it’s Lua. And not, Lua by itself is not *that* slow. It’s just any software needs some optimization after being written. And Lua is pretty ugly language that nobody wants to deal with more than barely needed – write something, it somehow works, hands off, because even looking 2nd time at it is ugly. Otherwise, everyone loves Lua.

Johny007
Guest
Johny007

@Marius Cirsta
Yes, I will try to rewrite it for C. I wanted to put it together ASAP, parts worked, but when I tried to run both. There was not enough memory.

Andrew
Guest
Andrew

Wow, this will be great. Having BT4.0 and wifi in a single chip is going to make some very flexible modules, and might be a signifigant enabling factor to this whole “IoT” thing to actually happen. Now, bluetooth communications can be bridged to wifi, and thus out to the wide area internet, or as people call it these days, “the clouds”.

Although, really, they need a 15.4 radio in there. If they already have a 2.4ghz baseband, I wonder how hard it would be to get 15.4 operation in there as well.

Deets
Guest
Deets

@Andrew

Yep if was configurable for 802.15.4 or 802.15.1 like cc2650 then it would really crush everything.

Member

@Deets
It has 802.15.1 – that is Bluetooth classic

802.15.4 is Zigbee which is dying, and Google Thread which isn’t deployed anywhere.
Zwave is proprietary and only made by a single company.

Member

CSR supports a BLE mesh, I wonder if Espressif will?
http://www.csr.com/products/csrmesh-development-kit
I read somewhere standard meshing was going to added to the next rev of BLE.

With a BLE mesh you don’t really need 802.15.4 anymore.

ESP32 is a killer chip for making smart outlets. It can blanket the room with wifi and BLE. Plus do power consumption monitoring on the outlet. All for a BOM well under $10. BLE is excellent for making temp sensors, PIR, contact, etc.

Deets
Guest
Deets

@Jon Smirl
Yes I am aware. But if it could do wifi plus configurable BLE or 802.15.4 that would be great. ZigBee is not dying especially since ZLL is the defacto smart bulb stack. And there are plenty of other 802.15.4 based stacks.

If it could do all 3 like RS9113 at a fraction of the price that would rock too.

Paul
Guest
Paul

@Jon Smirl

> ESP32 is a killer chip for making smart outlets.

Which brings a question I wanted to ask for a long time: did you see any consumer off-the-shelf product using esp8266? I never saw any, even though I watch tear-down scene and crack stuff open myself from time to time. For example, last smart power socket I cracked used RTL8189EM – pretty old WiFi MCU nobody heard about before, a refreshing change from 88MC200, which appears to be a market leader. No esp8266 anywhere.

Paul
Guest
Paul

@cnxsoft

Thanks for the hint. Link you gave doesn’t show a picture for me, but clicking around, I arrive atcomment image . But I can’t: 1) relate esp8266 and module on that picture (marked KK3000); 2) relaibly relate that teardown and that product on aliexpress. KK3000 is module from KanKun, FCC info here: https://fccid.io/2ACJ2KK3000 . In the public docs, nothing contradicts it being an esp8266, though any confirmation would be in non-public parts.

Still, good find, and given upcoming 11.11 trashdump on aliexpress, may be a good toy for folks with a hammer ;-).

Paul
Guest
Paul

Ok, there’re blurred internal photos: https://fccid.io/pdf.php?id=2660264 . Indeed, that would roughly match what’s on eps8266 board, except: logo on the main chip isn’t Espressif; there’re too many discrete components for a typical esp8266 board.

Paul
Guest
Paul

At least it’s still QFN32.

Paul
Guest
Paul

Ok, flaming logo on a chip is Konke (former Kankun) logo: http://ikonke.com/ . Espressif really should provide service “Get a esp8266 chip with your own picture on it!”.

Member

@Paul
Actually there were a lot products based on the ESP8266, but it had the ESP8089 label on it (they are the same die). A large number (many, many millions) of Chinese tablets and STBs have shipped using the ESP8089. The focus was mainly on the tablet market. ESP32 is going to be the same way, now those tablets get 2.4Ghz wifi and BLE support.

Standalone devices are a secondary market. Most Chinese manufacturers need turnkey reference designs to be able to put something into production. Then they make slight modifications to those reference designs. Last time there weren’t any reference designs expect the WROOM module (which became the ESP12). I believe there will be more reference designs this time. Hopefully more reference designs will turn into more retail products.

ESP8266 is an optimal chip for LED lighting since it is rated up to 120C. Too bad no one has put it into a light bulb. It is also an excellent chip for making in-wall dimmer switches. ESP32 does both of those and adds excellent smart outlet support. All three of these are high volume opportunities. I don’t know anyone pursuing these, which is a shame since the Espressif solution is the lowest cost one in the market.

I keep seeing things built around Broadcom Wiced. That solution is 10x the cost of the Espressif one with about the same functions. So it is a marketing victory, not an economic one. For example the Amazon Dash button. It costs about $10 to make it on Wiced, same button using ESP8266 can be made for under $2. Amazon is aware of the ESP8266 solution, but they ship Wiced. Since they sell for $5, they are losing $7-8 a unit after shipping. So is Broadcom selling Wiced at a loss to Amazon for $1, a price no one else can get? Nobody knows.

Paul
Guest
Paul

@cnxsoft

You must be kidding me, there’re e.g. MT6260 smartwatches which have “CSR German chip” in advertising. Closer to Espressif, they initially released spec (and SDK headers!) with pin pull-down support, and then retracted it. Their main SDK header is called “user_interface.h”, even though it has nothing to do with user interface, and inside those headers, there’s stuff like “buad” (instead of “baud”), so there’s lot of misinformation in Chinese stuff (from Western point of view). Ah, and straight in this same department – I don’t know why you think Mr. John Lee is CEO, according to the twitter account he is “i am love programming” and not “I’m CEO, bitch”. I know another guy as Espressif CEO, his English is definitely not “i am love programming”, and his Linkedin account still says he’s a CEO.

Paul
Guest
Paul

@Jon Smirl

> Actually there were a lot products based on the ESP8266, but it had the ESP8089 label on it

That’s completely different market indeed, not IoT at all.

> Most Chinese manufacturers need turnkey reference designs to be able to put something into production.

We perfectly know that there’re whole bunch of such reference designs, ESP12 is 12 for a reason.

> I don’t know anyone pursuing these, which is a shame since the Espressif solution is the lowest cost one in the market.

That answers my question.

> Amazon is aware of the ESP8266 solution, but they ship Wiced.

Well, with high-volume Western producers it’s all clear – 1-year old product is too fresh for them. I meant exactly Chinese shanzhai/gongkai – they are interested in quick kinda-innovation and low-cost BoM, and yet it seems to need a lot of preaching and beating to make them use something new too (why I call it kinda-innovating, there’s actually very little (relative) innovation in Chinese market, 95+% just copycatting each other).

Member

@Paul

I believe those other designs ESP1, ESP2, etc came from a university, not Espressif.

Paul
Guest
Paul

@Jon Smirl

Sure, they came from 3rd party. And do you say that a manufacturer won’t use a schematic/module unless it comes from a chip vendor? I’d say that availability of multitude of 3rd-party designs is a good sign – it means there’s/it’s possible to obtain docs enough to create such design, while not being a chip vendor iself.

Deets
Guest
Deets

@Jon Smirl

I don’t think it is 10x the cost for wiced. Wiced modules are available at retail for $8 single quantity and that includes the module manufacturer margin and the retailer margin. Plus the broadcom IP portfolio is strong whereas something like esp may have some risk as far as patent trolls and other IP holders.

Member

@Deets
One of the Espressif founders previously spent many years doing IP licensing. They are quite confident that they have the licenses they need. Of course nothing can protect you from trolls appearing out of the woodwork.

Broadcom is definitely spreading IP fud about Espressif. I personally received an earful of it at a trade show. If they had anything real they’d act on it instead of just spreading rumors.

rudi
Guest

hi
your pic is not the right 🙂

Think ESP32 is a QFN 48 7×7
with 35 GPIO… not 40 ..
http://esp32.de/head.JPG
best wishes
rudi 😉

rudi
Guest

@Paul
…QFN48 7×7

rudi
Guest

Paul :
At least it’s still QFN32.

ESP8632 “ESP32” is still a QFN48

Paul
Guest
Paul

For folks who didn’t get “hammer” reference above – here’s how ladies do it:comment image

trackback

[…] and the other used by the application software, similarly to what has been done for the upcoming Espressif ESP32. The main target applications include smart grid and extending “Wi-Fi” coverage, but […]

trackback

[…] teased about their ESP32 WiSoC with WiFi and Bluetooth LE a few weeks ago in a letter addressed to developers, but did not release […]

Big L
Guest
Big L

@Jon Smirl Zigbee is an excellent technology that just works, and is especially good for sensor networks, whilst BLE is a truly HORRIBLE technology and rarely works properly. Zigbee is most certainly not dying. I’m not sure where you get your facts from, but they couldn’t be more wrong. Also, z-wave is an alliance of 325 different companies all making z-wave products, so saying it’s “only one company” is pretty ridiculously incorrect also.

Either you’re extremely ill-informed, or you’ve got a barrow to push. Either way, it’s very irritating.

Member

@Big L
Zwave chips are manufactured by a single company Sigma Designs. Those 325 companies all get their chips from this single vendor. Something happens to Sigma Designs and it’s the end of Zwave. If Sigma doubles the prices of their chips, you’re paying it or abandoning your Zwave system.

Zigbee is dying because it is being taken over by 6lowpan both of which run on 802.15.4. Zigbee is also a mess because the various vendors making Zigbee devices tweak the standard so that a lot of the devices won’t interoperate with other vendor’s devices.

I would say BLE does not have enough deployment yet to really know how it will fair.

Dr. Nissim Zur
Guest

The Zigbee die from the 4 main reasons:
1. There is no smart phone support it. All phones support ble.
2. The need to add gateway Zigbee-WiFI to internet bridge add ridiculous more cost
3. More than Zigbee 7 to 15 nodes is not possible since the mesh is flooded and reaction to command is delayed 45 to 90 sec and more. See Philips HUE project.
4. Total BOM cost for IoT devices with BLE + mesh is lower.
CSRmesh is limited to fix profile and it deliver with close lib mesh code.
We developed BLE with mesh that bridge this limitation with also ble long range 200 meters. see our web site http://www.elinistech.com
[email protected]