I have enough trouble programming one at a time lol! 
1 Like
Youâre not deluded Justin; go for it.
Not via CAN, but via Serial link. Each ”C is connected to 2 others via two (2) of the numerous UARTs. Itâs more like a quad intel quickpath setup. Each message is sent in a frame, where the frame identifier has a from and to address. The addresses range from zero (0) (which means all) to 4. When a message is sent, itâs always from a ”C to either a single other, or all. If you receive a message that isnât addressed to you, then you forward it on the interface that it did not arrive from. If it is is addressed to you (e.g. 3) then you deal with it and reply to sender if necessary. If it is addressed to all (i.e. zero) then you deal with it, reply if needed, and forward the message.
1 Like
Not too awfully different from a design Iâve got whiteboarded; except I wasnât looking to make multiple MCUâs work together, but rather, to have redundant N+1 failover of the MCUâs.
If one MCU hangs or fails, another takes over operations.
(For a decade now Iâve built highly available supercompute architecture, as I venture in to Industrial control, I want to apply many of the same hardware failover concepts to embedded architecture)
What constitutes the failure of hanging of a system? It seem that youâre looking at the building the same type of solution thatâs used in Avonics Computers, or in the Space Shuttle. A prime number of computers receive the same inputs and calculates the outputs. An electoral system check that the outputs are the same. If they are not, then majority rules, and the one(s) that did not have the âcorrectâ output getâs reset.
Sort of. Space destined computers are in a redundant array where outputs are compared because radiation can flip bits around. For every operation results are compared and majority wins (it is exceedingly unlikely as you add peers that radiation would cause the same errors in RAM or CPU operation).
What I need isnât that elaborate. I just need to determine âmodule A is no longer responding or working, Module B takes over.â In the world of clustered computing, you have high availability through the cluster. Every X period of time all cluster nodes communicate over a redundant pathed cluster network. If one fails to check in or participate for Y cycles, and there is no path found working to that node, it is evicted. A new node is elected to take over itâs role.
Now thatâs a bit overkill for what I need; I wonât be failing over file shares, virtual machines, or the like. Just compute operations. So if microcontroller #x fails to check in, microcontroller #y starts taking over. Microcontroller #x is reset (physically) by the remaining nodes to attempt to bring it back online in a working state.
To achieve same state, all IOâs would have to be linked and all MCUâs would have to be doing the same exact operations; except âpassiveâ microcontrollers would not be setting IO states, driving DAC output, PWM output, etc. Theyâd still be doing logic, but the lines of code which set state would be skipped until they are told they are âactiveâ.
In this scenario all MCUâs are doing the exact same things all the time; but X number of them are passive - not driving any logic high or low, or whatever.
If the active MCU fails (goes in to garbage collection for 15 minutes, or just up and quits), the remaining MCUâs all have the correct logic state because theyâve been doing the same stuff as the primary MCU was doing; just without triggering logic. So when one is elected to be active, it would trigger a subroutine that sets state appropriately on all IOâs, DACâs, PWM, and so on, based on what itâs current state should be.
Thus if MCU #1 fails and MCU#2 takes over, relays, inputs, ADC, PWM, etc are all set to what they should be immediately, restoring the system to working state.
Make sense?
1 Like
Sounds like you need an analog multiplexer for each output. with an external clock that counts from 0 to 4. I think you would also need a common going to each of the ”Cs. No need to change the code (as that will change the timings).
1 Like
v1.1 Production boards ready
4 Likes
How you did using .NET MICRO 4.4 at VS2017 ???
Can you provide a way - please how you did it.
@valon_hoti_gmail_com - Iâm thinking that you just have to edit the VSIX to allow it to be installed in VS2017, but I am waiting for @Justin for confirmation.
Not me as i dont play well with Arduino 
1 Like
How did I not know there was NETMF 4.4?? Been using 4.3 this entire time
itâs been released but is not supported on any commercial scale hardware platform, certainly not GHI. There were some bug fixes and no new capabilities, so GHI have not put any effort into it - you can see TinyCLR is their focus. You donât need to worry about 4.4.
Right onâŠI am definitely going to have to get me of Justinâs chips to play with Tiny CLR on at homeâŠthis is the only embedded project I will be getting paid to do but I definitely want to keep developing my skills / its a shit load of fun.
1 Like