Quote
AndrewBCN
The big hurdle I see for this new project is that basically it offers almost the same feature set as a Duet board (except for the fact that the Duet includes 4 Allegro A4982 steppers drivers soldered in), and yet one of its objectives is to reach a manufacturing cost similar to that of an Arduino Mega 2560 + RAMPS combo, something I would deem is very difficult if not impossible.
The parts cost for the board I propose is significantly less than the parts cost for an Arduino and a RAMPS, if purchased in the same quantities. For example, the ATSAM3S4B is less than half the price of the ATMEGA2560, and it does not need the USB-to-serial converter. So if it were taken up by the same Chinese manufacturers, then it should eventually cost less than the $35 you quote, including the accessories. Of course, if the Chinese clone manufacturers started making the Duet, then its price would come down too.
The board I propose will also cost less to manufacture than a Duet, mostly by using a microcontroller with fewer pins and without on-chip Ethernet, leaving off the other Ethernet bits, leaving off the digipot chip that provides individual control of stepper motor currents, and using a 2-layer PCB instead of 4 layer (which is only possible because the stepper drivers are off the board). My planned selling price for the first batch would be about £35 + shipping + tax, which equates to about $54 - so less than half the price of the Duet (but you would still need to add your choice of stepper drivers).
Quote
AndrewBCN
Wouldn't it be a lot faster to develop, simpler, cheaper and more effective to just move the heavy duty processing to the Linux box and leave only the nitty-gritty realtime pulse generation and sensor reading to the Mega 2560?
The mega2560 is not a serious contender for new designs on account of its costing more than twice as much as a more capable 32-bit chip. I'd be reluctant to move anything to a Linux box, because I like the fact that both my Duet-powered printers are ready to be controlled within a second of being powered up. But it would be nice to have a multicore chip so that the hard real-time functions could be separated from the TCP/IP stack etc.