[quote=āgodefroiā]The problem here is that in this space youāre competing with things like Raspberry Pi, which is much cheaper, vastly more capable, and has all sorts of things like USB (host and client), ethernet, video and audio output, and SD built in. While NETMF is easy to develop for (for us .NET developers, anyway), it still has a stiff uphill battle.
[/quote]
How would you take a Pi based prototype and turn it into a product?
What would such a solution cost to start off (without your specific BOM)? The CPU, memory, and all the other associated support parts.
From the Pi forum regarding the BCM2835:
"You canāt order small quantites of this chip, Broadcom dosnāt operate like TI, ST Micro or Atmel. These chips were designed to be placed into cell phones by OEMs. "
āI did a quick search at Digi-Key and you can buy TI OMAP and Freescale MX31 in single units: they cost about as much as a RasPiā
I donāt want to take anything away from the success of the Pi but it is hardly the kind of platform I would want to deploy in my product in large quantities. The Pi is great for what it was designed for, education, hacker, and small one-off projects. With no internal Flash and very minimal I/O its not very flexible as a control platform. The BeagleBone is much more suited for an embedded Linux platform than the Pi.
Comparing Pi/Beagle to the NETMF is an apples/oranges comparison. The Pi/Beagle consume more power, have a larger software load and probably cost more to actually turn into a product. (as pointed out by @ mhectorgato). The Pi is heavily subsidised by Broadcom and the BeagleBone by TI. To actually embed their capabilities on your own board would be much more than the $35 (Pi) or $45 (BB). (In its defense, the BB could be used on a custom board far easier than the Pi.)
Embedded Linux has its place. But it is a lot of over kill for most embedded applications. It is by design a full blown general purpose OS, just shrunk down to fit in limited memory devices.
Bare metal 8bit microcontrollers, Arduino, NETMF and Linux all have their places in the IoT world. If youāre a hacker or solving a simple one-off problem with limited distribution your requirements are much different than shipping thousands of devices all over the world.
It is far more helpful if we as a community of developers help each other solve our given challenges with the best tools available; rather than bicker about which platform is best when we have no constraints, application or parameters to constrain the discussion. We have at our disposal more capability to create new and great products than ever before in the history of technology. Lets challenge and help each other to build great things.
@ mhectorgato - I see your point here. Iām new to this microcontroller game, thatās why all these questions gets asked I tought maybe the RPi had a chance because of all the hype it have gottenā¦
As for instant on. Iām trying to make some automotive stuff, that canāt draw power when the car is parked for a while. Noone would like a dead battery when they get back from a one week vacation. Maybe Iām looking at this all wrong. but i guess i learn more as I thinker with this stuff
[quote=āskellerā]
Comparing Pi/Beagle to the NETMF is an apples/oranges comparison. The Pi/Beagle consume more power, have a larger software load and probably cost more to actually turn into a product. (as pointed out by @ mhectorgato). The Pi is heavily subsidised by Broadcom and the BeagleBone by TI.[/quote]
Weāve been over this. I canāt speak for the BeagleBone, but the Pi isnāt subsidized by anyone. There is no evidence to the contrary, and the foundation has repeated that.
I doubt you could put together a BOM for a NETMF solution with the cababilities of the Pi for less than the Pi, even if you didnāt include manufacturing costs. Remember to include things like an Ethernet jack and PHY, SD holder, USB ports, power regulation, RAM, flash, etc.
Also remember that only the Cerberus around here is using internal flash and RAM. The Hydra, G120, and G400 are all using external flash and RAM.
Further, because of NETMFās limitations, lots of interesting IO can only be done by external boards, like the DL40 and PulseInOut, MaxO, IO60P16 modules, all of which could be used just as well with the Pi.
@ Architect - If cheaper means coding C in the browser, then Iām all for using a few more bucks to get back into a familiar IDE and productive language/framework and having fun
[quote=ādanibjorā]
@ mhectorgato - I see your point here. Iām new to this microcontroller game, thatās why all these questions gets asked I tought maybe the RPi had a chance because of all the hype it have gottenā¦[/quote]
Thereās nothing wrong/bad about Pi, itās just a different environment (hardware/software/conceptually) all around. It offers capability and performance that most, if not all, .NETMF boards canāt do at a very attractive price point.
But I agree that thereās too much hype about it. People using it to turn lights on/off via a network connection and are acting like itās something revolutionary.
The definition of āproductiveā, āfamiliar IDEā and āhaving funā vary from person to person greatly.
I agree 100%. But then, I also think the G400 is just a slow Pi without most of the useful peripherals, and with more I/O options, for a way higher price.
I bring up the Pi because I think that microcontrollers should run on the small, cheap, low-power end of the spectrum, where they were designed to run. Once youāve gone past the cost of a Pi (by an order of magnitude, even) without improving on the capabilities, then, in my book, youāre squarely out of microcontroller territory.
Iād like to see NETMF return to its roots, where it can shine, instead of seeing it compete (poorly) in a niche it can never own. If it did, it could take over the world. If not, itāll always be an also-ran. There were NETCF devices with less powerful hardware than the G400, and NETCF on something like the G400 would easily outperform NETMF (unless, of course, youāre only measuring boot-up time).
Yes, you cannot securely communicate over https with the Cerb if you cannot control the web service you communicate to. Such as a public web service. I host a web service on the Cerb and encrypt only WAN messages. But if you are communicating from the Cerb to a service in the WAN and you have no control over that service, youāre right it, there is no way to communicate unless you add https support. Having said that, I donāt see why https could not be implemented within the hardware constraints of the Cerb.
However, with a really good SSL stack designed for small embedded systems, it should be feasible. But even if you license such a stack (no Apache 2.0ā¦), Iād expect a [em]lot[/em] of work for the integration into NETMF, and for optimizations, until it works reliably (!) and you bring down opening an SSL connection from, say, two minutes down to something like ten seconds. I shudder when I think of how much effort and time it took to make even only TCP run reliably with NETMF 4.3.
Of course, Iād only use SSL where I donāt have the other endpoint under my own control. SSL is a camel through the eye of a needle, as far as inexpensive IoT devices are concerned. Not a good fit at all. But often the other endpoint is beyond your control and you have no choice.
So if we want to take over the world, with NETMF focused on the microcontroller ānicheā (which unit-wise is far larger than the more powerful Pi-class microprocessors), then someone needs to accept this challenge, and deliver on itā¦ And with the additional 64 KB of RAM in the newer STM32F429 chips, it just got a bit easierā¦
AES is supported in some microcontrollers through additional hardware, so this is a valid choice. Such hardware support is available in some of the STM32 microcontrollers as well. A Cerberus variant with such a microcontroller may not even require a layout change.
Iād prefer something thatās less expensive than AES in terms of processing resources, yet with better encryption than SSL (which is compromised by the NSA, it seems). Actually we are currently investigating this issue, but itās too early to tell what will come out of that feasibility study.
Hi All,
Iām totally new in gadgeteer prototyping, Iām a senior C# developer experimented in building IA, real time systems, Image analysis,ā¦
I so consider Iām mastering .NET & C# and a zero in electronic.
Iām new in gadgeteer, I started it because by chance, I found an article which let me thinking that Iāll be able to realize an project I imagine years ago (nearly 10 years).
In fact all is not perfect in gadgeteer, but we have a quality support. For my own, with an absolutely empty knowledge in general electronic, I was able to build a drone control board.
And my drone is flying like hell, without nobody piloting it, only through Gadgeteer components, nothing else.
So I would say that Gadgeteer is a fantastic technology which let you quickly prototyping lots of things.
If I had to build my drone from scratchā¦ I wouldnāt have completed at all, and certainly miss my target market.
So, we can say what we want, expensive, unstable, slowā¦ the fact is that my drone is totally stable (I let it run for more than 48 hours with permanent I/O), totally initialized in 2 sec, close to real time transaction, and for some small money (less than 500$ in my case) a huge profitability (more than 100x).
So I would thanks Microsoft for that technology and GHI for the components and the support and hope it will run for a long time !