Accomodations for stackable expansion boards

[rickyjames]

I strongly urge that accommodations be retained for a stackable project board by gathering unused / important control lines as well as the power lines to through-hole header layouts. There are several headers in the Alpha boards; whether or not they stay in on a size-shrink final version is TBD, I think.

Remember what made the original IBM PC a legendary hit: the ability to plug in third party add-on cards into its ISA bus, a concept still in use 30 years later in the x86 machines. To make the Rpi an equally legendary hit, you must allow for hardware expandability with stackable boards that fit its form factor.

Currently you have some headers on your Alpha board but I seem to recall reading that they were going to be removed in the size-shrink final version after using them as test points during debug. I sure hope not. Those headers are the key to expandability.

I would propose that not only should you have a through-hole header layout, but that you do not populate it on the final board. Several reasons. First, it saves cost for an additional component that most users won't utilize anyway. When you DO have a tinkerer come along, they will then have the option of soldering the header onto EITHER SIDE OF THE BOARD. Having the header as you have it in the Alpha board is actually very space inefficient - it is on the component side, so using it as an expansion path requires a LONG header and board spacers that gets the expansion board above the components. Better to put it on the smooth "bottom" side where only a small header is required to get "above" the SD card slot mechanism and only a few washers would be needed to support the board.

Alternately, if you DO populate the header, I recommend making sure that it is sized so that a mating connector on another board can fit into it and still clear the components on the Rpi. Preferably you should have a specific through-hole connector in mind for such a purpose that you have verified the fit and component clearance on, and be able to provide its part number thru DigiKey or some other vendor.

Bottom line, including an empty through-hole header will provide a starting point that will jumpstart a BUNCH of add-on board projects. Surface mount may be the key to affordability and size-shrink, but through hole is the key to getting tinkerers involved in actually making a variety of custom add-on boards.

IMHO, an easy route for a stackable project board for the Rpi is a custom PCB from ExpressPCB. I have used them for many projects and they are AWESOME. They do three copies of a custom double sided 3.8 by 2.5 inch fixed-size PCB (96mm by 63.5 mm) for only $51 (around 30 GBP) with 24 hour turnaround. This size is exactly in the ballpark for the size a Rpi stackable expansion board needs to be. Plus ExpressPC has free downloadable layout software that does everything you could want it to do, another plus for open-source sharing of stackable expansion board ideas that could start from a universal mounting hole / header hole template. They ship international to the UK and elsewhere. Highest recommendation.

http://www.expresspcb.com/expr.....iboard.htm

[ErvKosch]

Remember what made the original IBM PC a legendary hit: the ability to plug in third party add-on cards into its ISA bus, a concept still in use 30 years later in the x86 machines.

There will be a USB port on the unit. I know a stack-able module would be great (and cool) but there is a decent add on system already in place. Also there is going to be a serial-like debug port on the board as well.

[rickyjames]

USB is not a expansion path - it is a data bus, a hardware interface standard to facilitate data transfer at a minimum of 1.5 million bits per second. You've got to have a microprocessor to access its data stream, plus a data cable that snakes around a meter or two. This is totally overkill for something that may only need a few hi-low control signals, and could be packaged in a unified case that was only a centimeter thicker than one housing the footprint of the Rpi board alone.

Simplicity of design is a key design goal in engineering. Stackable custom expansion boards follows that concept.

Sigh. It seems that I am old. Boys and girls, once upon a time there was these things called 74LS series TTL digital logic chips...they were like legos for geeks....and you could do magic with them.

http://en.wikipedia.org/wiki/74LS

[Lob0426]

If you look at the back of the board you will see that the 1.27mm GPIO headers are through hole type. In another thread the person who is one of the board designers talked about going to larger strips. He also stated that the current headers could be removed (desoldered). A search of topics under GPIO may find you this information.

[rickyjames]

There would be great value in going from a 1.27 to a 2.54 mm header pin spacing because the latter is the standard protoboard spacing. Since the Rpi is primarily an educational tool, making it compatable with human-scale prototyping components would be highly desirable.

[Gert van Loo]

Board space!
If possible we will try to use 2.54mm (0.1") holes. If not next choice is 2mm. What I have no information about yet is the cost of keeping the holes open. That requires masking tape to put on each board before wave soldering (if we do wave soldering) or all through hole connectors need to be post soldered (laser/manual). We have an assembly expert on the team who can tell me. We are so busy with the next generation that we have not had time to answer all our own questions yet.

[mgb]

Do you want to connect directly to these pins given the lack of protection on them?

If you are going to need a daughter/breakout board with esd protection, source resistors etc then is there a problem making this connect by a 1.27mm or a z-cable or similar?

[ErvKosch]

Your points are valid. I guess two reasons not to do this:
1. Cost
2. This is suppose to be a mini PC.

Its great for embedded projects but the projects focus is on creating a basic computer. I suppose you could throw in a #3 - This is Version 1.0.

The design and everything will pretty much be open. Someone may take this design and run with it into an embedded project like what you're talking about.

[mgb]

I was thinking that anyone who wanted to make an add-on and keep the form factor for some embedded project is going to need to use SMD so the 1.27mm GPIO isn't a problem.

For education use - any protoboard is going to need a dedicated area of protection components (unlike an arduino). If these daughter boards are going to be manufactured by someone then adding a "difficult to deal with" connector to the raspberry isn't a deal killer.

ps. I can finally admit that it was me that burned out 2 of the school's BBC micros by connecting motors to the user port - and not understanding about current limits!

[rickyjames]

For anybody that's curious about what it would take for direct "embedded PC" hookup to the ARM processor used by Rpi, below is a link to the 750+ page Technical Reference Manual.

http://infocenter.arm.com/help.....p7_trm.pdf

There is a "debug test access port" that connects to the ARM core, but repurposing some of its capabilities for embedded control is obviously a non-trivial exercise. Still, if header access to this debug port is provided on the board, who knows what could eventually be dreamed up as an add-on to Rpi? It would be, er, a very educational exercise to try...

@mgb - I think you are quite correct about needing protective circuitry on any daughterboard to protect the integrity of the Rpi signals, and that probably would be SMT by the time you made it all fit on a same-footprint board. LOL about blowing out your school micros during your interface training. I can relate. Nothing like a big pop and flash to teach you a lesson you'll never forget - you must never let the magic smoke out of the component.

@ Gert: Dude, you guys have done an outstanding job on hardware development, I am standing here clapping my hands in your honor. Your circuit board layout for the Alpha boards is a joy to behold - double SMT with wave solder thru hole. It's easy for me to stand here and babble on about "gimme headers", as a board layout geek myself I understand just how very hard it would to make it happen on your final size-shrink version when it's not even on the list of primary requirements. No matter what makes the final cut, you and your team have done an outstanding job.

Enlarge these JPGs to their max size and behold true art:

http://www.raspberrypi.org/wp-.....CF1808.jpg

http://www.raspberrypi.org/wp-.....CF1809.jpg

J1: DC Jack (6-20v input provisionally)
J2: UART serial console (debug)
J3: SD/MMC/SDIO memory card slot (underside)
J4: HDMI connector providing HDMI 1.3a out
J5 or J9: GPU JTAG (ARM11 pinout; no-fit on production boards)
J6: Audio connector: 3.5mm stereo jack
J7: Composite Video connector: RCA
J8: Either 1x USB 2.0 (Model A)
or LAN9512 providing 10/100Mb Ethernet and 2x USB 2.0 (J10: Model B)
J10: 10/100Mb RJ45 Ethernet jack
J11, J14: 1.27mm header providing ~16 GPIOs at 3v3, I2C and SPI interfaces and ARM JTAG.
J12: 1.27mm header providing DSI interface
J13: 1.27mm header providing MIPI CSI-2 interface (camera)

[fordp]

I am interested in this topic. I would like to see a 2.54mm pitch GPIO section, as stated above unpopulated. I understand you do not have much space in which case limit it to a smaller number of connections. Do however bring out ASYNC Serial, SPI, I2C and any spare USB's as well as GPIO's (as well as power of course).

The cost to leave them unsoldered in a volume product is zero as you simply sit the board in a carrier when in the flow solder process that masks that section off. The carrier can do other things that increase yields.

This will then make the board useful for robot projects and countless hobby electronic projects as well as the core software development one.

Great project guys/ I cut my teeth on Apple II, ZX81, Spectrum and BBC Micro and as a result has had a long career in electronics so far.