Thoughts on computing device design:
In looking at the RaspberryPi and the Panda Board I have seen where there may be a different way to build “similar” devices.
The problem with all of these small devices is board size versus space for connectors and components. Another consideration is part of this is the cost. Larger board and more “layers” equals higher overall production cost.
Current theory is to build it all onto a single board to create a SBC (Single Board Computer). This complicates the design as you are trying to fit a ton of components and connectors on to the same board. You end up trying to roll all of the features someone MIGHT want onto the core board. If the device is for a specific use this is the probably the cheapest way to design it.
Another way is to design a core device, then add-ons that give a broader use of the device. A good example of this is the arduino and its “sheilds”. In general the Arduino style system cannot really be called cheap, at least when it may take several sheilds to get an end design.
So if you want to design a less expensive SBC system I am proposing a “hybrid” design theory. Designing a limited scope SBC that has multiple, limited design roles.
To start I will base this design on the Raspberry Pi model types. It has two seperate design roles. The A model, which is a single USB port device versus the B model, a dual USB plus Ethernet Device. The core “board” has exactly the same components, SoC, video connectors, sound and power. The difference is the USB and the Ethernet connections.
So to hybridize the design we are going to remove the USB and Ethernet connectors and compononents from the “core” board. These are the components that change between the two models. Later we will look at some other components that may be moved off the “core” board for additional space and cost savings.
The A model really only needs a Connector to have its USB. There are some other components there also that will remain on the core for uniformity, such as the resistors on the data lines (R36/R37 on a Raspberry Pi). A 4 pin header to a dongle would be the only part needed to make the core board into an A model. The area that is currently unpopulated on an A model does cause additional cost (PCB size). This includes the unused traces, additional holes and vias that are not needed on an A model.
The B model needs a Hub. This “hub” is (the LAN9512 “chip”on Raspberry Pi, LAN9514 on a Panda Board). It provides 2 USB and an Ethernet ports. The Ethernet connector, USB connectors and the LAN9512 chip comprise approximately 20% of the board space. This does not take into account the “traces” that are needed to connect these components to the SoC. By removing these components from the core board it should be possible to reduce the cost of the base model, the A model. The 4 pin header would lead into an off board hub/Ethernet. Of course you could also have just a USB hub without the Ethernet with a different “breakout board”. The core board would again have less vias and traces so lower expense in design and manufacture.
Additional design considerations:
In either of those use cases you would be powering the single USB or hub with the same power as is being supplied to the board. This means that there would still be a limit as to how much could be powered through the available USB ports. An alternate proposal would be to move the power system off of the core board. You would remove even more components from the core board. The PSU could be sized for the model/use case it is to be used with. In the case of a powered hub you would not now need an additional PSU. In this design you would supply power to the breakout board which would in turn power the core board. Again reducing the size of the core board. This means you have to build breakout boards, for anything but the core board. The breakout board should be cheaper for its size than the same area of the core board. It would most likely be a 4 layer rather than a 6 layer board. The 4 pin header would still be the way to connect the breakout boards to the core board. If the 4 pin header was placed close to or in the GPIO header strip, it would make custom breakout/daughter boards much easier to design.
Choosing the style of board you need would be easier, the core board is the same. The breakout boards are different. The Models possible from the core board would only be limited by what could be connected to that 4 pin header, and the GPIO. Power limitations would be less of a consideration in this hybrid design. The core board will always have the same power usage range, the breakout boards will differ in power usage. Power protection would only have to provided to the header. Any additional protection would be incorporated into the Breakout boards (fuse). The core board plus 1 of the 2 basic breakout/daughter boards would be the basic sales model. There would be additional market opportunities in other board models.
Single USB (A Model).
dual USB with Ethernet (B Model).
4 port USB hub.
4 port USB hub plus Ethernet.
7 port USB hub.
Ethernet plus USB controlled relays for home automation.
The real question is would seperating a design into multiple boards add price? Would retailers try to add in additional profit margin on multiple board designs, even on the base models, or would a hybrid design be able to hold a price point? Would it make retailers happier as there would be an avenue to make higher profit breakout boards in addition to the base model breakout boards? Would this reduce the overall price of ownership?
A secondary question is whether moving the power input to a breakout board would make it harder to sell the core boards as individual products?
Users would be able to have a device that is easier to setup for their specific use.
Able to put connectors on fewer sides of the PCB.
Case design would be more complicated.
More pieces to the design.
Not as straight forward for new users.