Replacement for startKIT, now that is obsolete

[tggzzz]

When I've wanted a small simple cheap devkit for a low-volume embedded application, I've used the startKIT. Now that has become obsolete, I'd like to find an alternative - or use I'll use Arduinos, ugh!

For FPGAs, there are many small and cheap boards containing the FPGA, some memory, a mechanism for downloading the bitmap (often via JTAG), a voltage regulator chip, and connectors/headers for the i/o pins. The latter are connected to my application-specific i/o on a separate board. I haven't spotted any similar "daughterboards" / "carrierboards" / "mezzanine boards" for xCORE devices.

What startKIT features do I need:

• USB connectivity to a host process
• at least 8 cores dedicated to my application
• for development, tight integration with the IDE including program download and debugging (preferably via USB)
• similar number of uncommitted general purpose i/o lines, accessed via one or two general purpose connectors/headers (probably 0.1" pitch including a sufficient number of grounds!)
• voltage regulator, to convert the USB's 5V to 3.3V
• small and cheap

What startKIT features don't I need:

• compatibility with startKIT's connectors
• analogue i/o; that's part of my application specific i/o on my board
• LEDs and switches, for the same reason
• input protection, except for the USB connection

So, has anybody spotted any such daughter/mezzanine boards, or similar?

If not, are there any pointers to what I would have to develop? Assume I can design PCBs, but do not have XMOS-specific knowledge of how to integrate xCORE chips with USB with the IDE.

Thanks.

[eez-open]

I'm absolute beginner and started to research XMOS recently. As an exercise I set as first goal making an "evaluation board" (open source, of course) that could be used for testing XMOS platform and can be soldered at home. I hope that during the process of designing it I'll learn a lot about xCORE. It's almost in line with your wishlist with exception of many GPIO (only two is currently exposed). Another missing thing is integration with IDE over USB. But, it has place for adding Ethernet PHY, LCD, SDRAM and few SPI peripherals. I planned to use xTAG that is fairly priced and with full implementation of XSYS connector (i.e. with 2-wire xlink 0) it should be possible to use it for realtime debugging. Do you thing that such functionality is possible to manage over USB without xTAG?

[tggzzz]

Thanks for those pointers - I'll monitor it to see how you progress, and to learn anything I can from your experience!

Two benefits that I like about the startKIT are:

• its simplicity - very few components which mean it can be small and cheap. It also has the advantage/disadvantage that the i/o is "uncommitted" so I can/must add whatever is necessary for the specific project at hand
• the USB interface and programming model, which is ideal for connecting to a host computer for normal operation (i.e. not just programming)

But I haven't yet assessed whether anyone else has duplicated those benefits in the form of a simple board.

[eez-open]

Maybe I can manage later to make a "fork" that comes with more "uncommitted" pins. Currently it's up to user to decide which of added SPI peripherals to solder on board or not. Regarding the USB interface and programming model, what I found so far startKIT was a special kind of the "beast". It looks to me that its one tile is used to provide functionality of xTAG (that is in fact a one tile MCU with USB!). For more info and better understanding I believe that we need here some (official) inputs from XMOS team/support.

[Gothmag]

The startkit wasn't that special. It was a 16core mcu with 1 core locked out for USB and had a bootloader for programming afaik. I've made a couple custom xmos boards now and I did look into adding the usb only functionality but what I found was that it wasn't cost effective for the quantity you'd make. Better off to just include the xtag and header. Maybe someone else has had a different experience.

The software side is also more complicated than I'd like. The only sensical design I came up with was adding a second xmos cpu to duplicate jtag functionality. Other than that you could add a small USB capable mcu that can just program the quad flash on board or inside mcu but you completely lose ability to do anything other than program it.

[eez-open]

Instead of using other MCU (and have additional maintenance issue with its own firmware) what's about USB-to-SPI bridge such as MCP2210? Perhaps it's also MCU too, but comes already programmed to perform just bridging.

[Gothmag]

You could but the xmos requires quad spi. So you'd need to be able to program quad, or swap to spi, then swap back to quad on the flash. I'm not sure either is feasible with that part

[eez-open]

I'm not so sure, otherwise why SPI master and slave modes are supported as boot options?

[tggzzz]

The startkit wasn't that special. It was a 16core mcu with 1 core locked out for USB and had a bootloader for programming afaik. I've made a couple custom xmos boards now and I did look into adding the usb only functionality but what I found was that it wasn't cost effective for the quantity you'd make. Better off to just include the xtag and header. Maybe someone else has had a different experience.

The software side is also more complicated than I'd like. The only sensical design I came up with was adding a second xmos cpu to duplicate jtag functionality. Other than that you could add a small USB capable mcu that can just program the quad flash on board or inside mcu but you completely lose ability to do anything other than program it.

What is "special" about the startKIT (and the Arduino) is the combination of low-cost hardware, plus IDE development environment, plus runtime comms over USB to a process in the host PC.

It shouldn't be too difficult to pick a 2-tile xCORE device with USB comms and put it on a board. OTOH, I don't know the software that is available nor how easy it is to integrate all software components - so that is relatively risky.

[Gothmag]

Well that's the thing you would have to have the cost subsidized or build in massive quantities that likely wouldn't sell through, maybe both. The arduino stuff is actually a good example because the genuine boards can be made for significantly less but you don't see that in the price unless you buy iffy clones where total cost can be near mcu cost alone.

The startkit software is sort of available in bits and pieces but not entirely and I don't know if there is a plan to release that now that it is EOL. The software is doable but I can't find the docs right now on what was found to be involved. I decided it was not worth pursuing because the profit wasn't there and I don't think the market is either. Since the startkit was released in 2013 and the xs2 processors are out your best bet for low cost is that xmos decides to try again.