If you need an industrial-strength IoT product, you need an industrial-strength WiFi chipset. For our own household hacks, we’re totally happy with the ESP8266 chip. But if you need to connect to the big, scary Internet you’ll probably want state-of-the-art encryption. In particular, Amazon insists on TLS 1.2 for their Web Services (AWS), and we don’t know how to get that working on the ESP.
[Anuj] designed a breakout board called the knit which includes a Marvell MW300 WiFi SOC. This chip has an onboard ARM Cortex M4F running at 200 MHz, which means you’ve got a lot of everything to play with: flash memory, RAM, a floating-point unit, you name it. And Marvell’s got an SDK for using AWS that includes things like an operating system and peripheral support and other niceties. TLS 1.2 is included.
Best of all, a MW300 breakout is reasonably affordable (though more expensive than the mass-produced ESP8266 modules, naturally) and it’s an entirely open design. [Anuj] also seems to be setting up for a production run, if you don’t feel like making it yourself.
The MW300 is in all sorts of commercial IoT designs, and it’s a battle-tested go-to for interfacing with “the cloud” securely. The only hobbyist-friendly board that’s similar is the Adafruit WICED WiFi Feather, but it’s more expensive, less powerful, and out of stock at the moment, which just shows the demand for something like this.
Of course, if you need more integrated peripherals, you could just hack up a “Hello Barbie” toy which has the same chip as well as sweet audio codecs and a nice fat flash ROM.
We think it’s neat that [Anuj] would make and test a breakout for this powerful little WiFi SOC. We don’t need one for our projects right now — we’re running in entirely insecure mode — but it’s good to know what your options are. (We’re also looking into esp-open-rtos for the ESP8266 — we know they’ve been working on TLS 1.2 encryption, but we don’t know what their status is at the moment. Anyone?)
Yes the Adafruit Feather is good, also the Arduino.cc MKR1000 you haven’t mentioned. If you really want to compare items, you should consider a table of features+price ala http://21stdigitalhome.blogspot.com/2015/06/comparing-modern-microcontroller-wifi.html. Oh and never discount the toolchain ease of use, if you need to be a PhD in EE, it doesn’t matter if it’s cheaper, it costs more in the long run.
Can’t agree more. Don’t want anyone jumping through hoops to get a basic blink.
The officially supported toolchain is gcc-arm, and openocd will be the defacto debugging tool.
Plus support for Eclipse IDE
Why is it industrial strong? Explain please, because so far the ESP is more mature and more robust…
As it says: “In particular, Amazon insists on TLS 1.2 for their Web Services (AWS), and we don’t know how to get that working on the ESP.”
So nobody has ever tried to build MatrixSSL, wolfSSL, or mbed TLS for the ESP?
esp-open-rtos supports TLS v1.2 through mbed TLS.
I am not sure if ESP* is being used in any commercial products, but this is. Case in point the Hello Barbie. Mostly because of the security features.
Of course, the community behind ESP is crazy awesome. The tools, examples, support for different environments, etc. are the real value for ESP.
And like Elliot said, a better comparison is with the Feather Wiced.
> I am not sure if ESP* is being used in any commercial products,
There’s this one, arguably it’s not very well designed.
http://hackaday.com/2016/02/06/cheap-wifi-outlets-reflashed-found-to-use-esp8266/
>Why is it industrial strong? Explain please, because so far the ESP is more mature
The Marvell platform was developed before the ESP came around. It just hasn’t been available to people that don’t want to order 10,000 units before now. The SDK is much better than all of the half baked crap out there for the ESP.
Better for 20 year veteran engineers. I can get a kid up to speed on an ESP in less than 2 hours with a demo project finished. It takes that long to set up the dev chains and IDE for the Marvell.
>I can get a kid up to speed on an ESP in less than 2 hours with a demo project finished.
Has nothing to do with how mature the product is.
We are doing a small production run https://makerville.io/knit/early-access/ ! Ships in July
Do sign up !
I totally agree.
That’s why all the supported tools are going to be FOSS projects
gcc-arm for the toolchain, openocd for debugging.
I personally use Eclipse IDE along with these things
BTW, me and Rohit aka indiantinker have created a project on HaD.io too
https://hackaday.io/project/10082-makerville-knit
The Marvell MW300 looks really interesting given how many GPIO-pins and buses it offers and the very generous amount of RAM when compared to the ESP8266. I’d totally be interested in a devkit, but a quick look at the specsheet seems to indicate the SoC supports a lot more pins than this board brings out — why limit the number of pins to fewer than the SoC actually supports?
There are 2 ICs in the family, 88MW300 and 88MW302. The 302 has more pins as well as USB OTG support. It’s available as a dev kit on amazon.com
http://www.amazon.com/Globalscale-MW302-IoT-Starter-Powered/dp/B0168DLQHI/
The Knit on the other hand is based on the 88MW300 (for price reasons of course ;)
Plus the module itself, flash, boot pin, etc. take up some pins too. So we lost a few there.
Oh, right, I must have looked at the 88MW200 specsheet or something. Well, I suppose that makes things a tad clearer. I can’t seem to be able to find a proper specsheet for the 88MW300, how many of each I/O-interface does it offer — are there more than the one SPI-bus that’s already in use by Flash, and if there aren’t, can that same bus also be used for peripherals or does it being used by Flash preclude that?
We are doing a small production run https://makerville.io/knit/early-access/ ! Ships in July
Very nice.
I have a similar board I’m using for my own projects based on the older MC200 platform. The ESP stuff is really junk in comparison.
We are doing a small production run https://makerville.io/knit/early-access/ ! Ships in July
This board is going to cost $14.90, but for $12.90 you can already get a LinkIt Smart, running a full OpenWRT Linux distro, which is probably much more powerful.
http://www.seeedstudio.com/depot/LinkIt-Smart-7688-p-2573.html
The non-Duo Linkit Smart 7688 only has one SPI-bus and it’s already in use by the internal flash — I don’t know if you can use the SPI-bus on it at all for running your own peripherals when it’s used for the flash. I, personally, count that as a major negative point for it.
The Duo – version of it that combines the MPU with an Atmega MCU, similar to the Arduino Yun, is just as stupid a concept as the Yun is: they opted not to bring out any of the SoC’s own pins, only those of the MCU, even though the SoC has plenty of them already on its and you could perfectly well use both the SoC’s pins and the MCU’s pins for your projects. Marrying an MPU and MCU on the same board is a perfectly valid idea, but then deliberately not allowing people to use the pins of the MPU? No, that’s fucking ridiculous, especially when you’re already selling a board where those pins from the exact same god damn MPU are brought out!
‘especially when you’re already selling a board where those pins from the exact same god damn MPU are brought out!’
And that is exactly why they are not on the other board. Follow the money :-)
I don’t get your point. The board where the MPU’s pins are available is the cheaper one.
Really? Bizarre so.
Or you could just have all your “things” hand over their data to a “broker” runing on a Linux server (Orange Pi etc.) and have the broker deal with all the external services as well as acting as a firewall and monitoring system for the state of your little cluster of things. This also allows you to partition off the things from the rest of the LAN to protect it from intrusion via the WiFi side.
Esp32 coming soon
It has been “coming soon” for a good while now, and it’ll take even longer for any good devboards for it to be realized.
It has always been targeted for summer 2016. There will be modules and breakout boards right away, based on what was learned from esp8266.
I’ve given up on the ESP…I’ve built simple devices that run perfectly for a few weeks then go in fatal wdt infinite boot loops. Don’t know if it’s the ESP, the flash , crap Chinese assembly or whatever and I’m not going waste any time trying to salvage a cheap crappy part. I’m moving on to the feather and/or the mkr1000. I’d rather spend $50 on solid reliable part than send hours trying to fix a POS. I hope the Feather and mrk1000 turn out to be reliable and get the kind of support the ESP has
We are doing a small production run https://makerville.io/knit/early-access/ ! Ships in July
Hi Anuj – just thought I’d let you know that I tried to order one of your early access boards (I’m doing some testing for a project that will need a few hundred wifi chips and the 8266 suffers from verrryyy slow packet responsiveness..) and it wouldn’t let me order it with my Australia credit card – says ‘can not purchase with that card’.
Which is strange as it’s my main card from a big bank here, that I use all around the world both on the web and in person….
You tried the PayPal button right ?
One Buy button is for folks in India, and the other buy button is for everyone else.
https://www.paypal.com/in/cgi-bin/webscr?cmd=_flow&SESSION=ME2xscRZUwR0vbOIbcYbcPXs0LBzs5nwE1hr6XkeMZWqPGbmmkpXX0Tr6ZK&dispatch=50a222a57771920b6a3d7b606239e4d529b525e0b7e69bf0224adecfb0124e9b61f737ba21b0819848475f0da5465a2ea26eae033cbe3bda
Can you send me a mail at anuj@makerville.io ? If it still doesn’t work, I’ll drop a mail to PayPal support.