YAES - Yet Another Embedded System? Crowfunding advice needed!

[s.voulaz]

Hello everybody,
I'm going live to pick some suggestions from you fellow colleagues of EEVblog. As a long-time passionate electronic designer and software programmer, I'm mostly focused on the design of digital embedded systems - from tiny 8-bit bugs to state of the art SoCs. Some years ago I leaded the design of a modular embedded platform (U-MoBo), aimed to become a workhorse for labs (including schools) and an handy way to create decent prototypes or even support low numbered series. Long story short, most of the people jumped off the train and I kept holding the ball - the idea was reasonably good, but we didn't share the intentions.

It's a pity that the project never found its way to the market, since it was generally appreciated from a technical point of view and it was used as a base to develop actual products. For a certain period of time Arrow (Italy) backed the project, using it as a mean to promote its linecard. The big obstacle was definitely the cost and the lack of a community, since we were unable to give it the right spin and the numbers were too low to make it interesting also from a commercial point of view.

I would like to open a discussion here to understand if it'd be worth to try the way of crowfunding - Kickstarter, IndieGoGo and Crowd Supply being the primary options. To share an idea of what it's all about, have a look here: https://tinyurl.com/y2g7wqg6 - it's a presentation originally used to promote the design. It's all about a carrier board with a SoM, with additional expansion modules (ExM) and a display. What makes is "different" is the way it is engineered and the high modularity. And yes, it is intended to be fully open source (CC BY) - including OrCAD schematics and PCBs.

The project is in an advanced stage of development, with full working samples and quite a good documentation. A final hardware review is needed and I'm currently designing a new SoM based on NXP's i.MX8M mini. So it's not just vaporware. Back in 2016 some kits have even been released to actual customers during workshops, while others have been used to build POCs. Besides, the i.MX6 SoM has been actually used in a couple of real products (but at that time the SoM design was not Open Source).

Now for the sore points.

Software: I'm not a linux/yocto expert and my guru friend once following this activity is now working full time for one of the big players in the SoM market. For a series of reasons we've been unable to build a community around this project, so the git repo and the wiki are just dead lost in hyperspace. Of course the project needs someone taking care of this, I'm still trying to find a good way to issue an effective call for duty.

Cost: this is definitely not a 10$ gadget and it's not (yet) made in China. A single Hirose 80-ways connector costs more than 1$ USD (for quantities) - and there are 22 in a fully populated kit. Multi-layered, impedance controlled PCBs are not the cheapest ones and the assembly of a SoM with many tiny components (down to 0201) does not come for free. But the biggest issue is about quantities. Samples can of course come out of my pockets (as they actually had), but even a pre-series of 100 pieces is simply prohibitive. Things start to be viable talking of "one thousand", but according to some (outdated) estimates the production cost for "1K of i.MX53 SoM + full size BaseBoard" was around 170€ - excluding the display and any expansion module - no WiFi/Bt, no HDMI output (or whatever), just bare metal. That means that the final cost of a complete kit could easily reache five hundreds bucks... still interested?

OK, so far so good. That's a starting point. Thanks to everyone for reading - I look forward to your comments to open a fair and useful discussion. Comments and suggestions (including that of leaving for a vacation) are more than welcome!

Good work!
Stefano

[nctnico]

I'm working on a commercial iMX8 project myself. I recommend to run Debian instead of using Yocto. With Debian you only need to care about getting the right kernel and right drivers together. The rest of the software is taken care of by Debian. I'm using the iMX8EVK as a development platform for the software (I run Eclipse on it!) so no more endless cross-compiling.

I'm not sure a SOM is very usefull because you'll need an expensive board anyway with controlled impedance traces. What could be interesting is to have a micro-ATX style board so you can use standard PC hardware. A PCIexpress bridge chip can be used to get extra PCI express slots. And the micro-ATX form factor doesn't rule out space for Arduino / RPI shields.

[ebastler]

Technical design considerations aside, I think you are in a difficult position here. This kit will be expensive -- you want some margin on it too, so will probably have to ask for $500 or even more. Your main sales pitch seems to be that it is scaleable and future-proof. But will people be willing to bank on a small, unknown startup when they are looking for a "future-proof" solution?

I think this particular project would benefit a lot from a bigger player (like Arrow) backing it. It would not only buy you scale via a broader distribution channel, but it will also buy you credibility in the market, and that seems important for this type of product.