MultiSpork

Looks like I might have recruited a software collaborator with some serious Android chops. Watch this space.

Well, it looks like I won't have boards in hand until the 25th or 26th, which means I can't use them as a prop in my quarterfinals video. 

OTOH, I'm forging ahead on the documentation front. I just checked in a system description document that covers the overall intent and the hardware design. 

I realized that to make a really good impression on the quarterfinal video, I need to have some nice wireframes of the proposed app. I'm not any great shakes at that (and I'm still working on finding a software collaborator), but I think I can hack something out in time. 

Just sent the first prototype out to OSHPark. Since it's four layers, it's going to be cutting it very close to have it here in time for the quarterfinals video. I suppose I need to design and show off a case and some wireframes of what the app will look like.

I finished the routing! Off to OSHPark in the morning. Hopefully I'll have a PCB to include in my video for the 20th.

Now it's time to route the prototype MultiSpork board... and it's kind of a bear, with over five hundred air wires. Even as complex as the board is, it shouldn't be too troublesome to route it with four layers. 

At this point, I've added top and bottom digital and analog ground pours. This took the air wire count down to about four hundred. Since this board has four different power rails (+10V, -10V, +5V, +3V3) and two ground rails (digital and analog) this is just the beginning. I'll add a couple of internal pours to distribute the ground and the other power rails as needed, especially to the series protection diodes.

Once the various pours are in place, the next thing to route is the fastest & most path sensitive signals. In the case of this board, that's the 2.4 GHz signal path, the clock crystals for the MCU and the WiFi chip, and the USB signal lines. One thing that really helps with keeping everything organized is to make all of the traces on each layer more or less parallel. In the case of this board, the bottom layer runs crosswise, the top layer lengthwise, and the two inner layers alternating

Next up is the somewhat slower but widely distributed SPI bus. That bus goes to every major chip on the board as well as the SD Card slot, so it goes everywhere. It's also a fast SPI bus, in order to ship data back and forth from the MPU to the MAX11300 at the speed required. 

With all the fast stuff laid down, next step is to get the analog and the regulators laid out. None of that is particularly hard, since those groups of components are laid out to make this bit easy. Once those are in, hook up the power rails. 

The last bit is all the low-speed logic connections and ancillary trouble required to make the whole business work. Simple, right?

I've got the schematic done (including correcting issues found during placement), and even got all the parts placed in (more or less) their final positions. I was also able to shrink the board to 3 x 3.4 inches. I know I can shrink it farther for the final, but this is good enough for a prototype.

I've now finally got the configuration nailed down (see the block diagram on Github), and I'm mostly done laying out the circuit. It's a monster -- I suspect I'm now north of 150 components, but I haven't counted yet. More than eighty of them are capacitors of various descriptions. I've got the four power rails (+10V, -10V, +5V, +3V3) powered from a single LiPo cell. I've got the major chips into the schematic and wired together.

I've still got the following elements to finish in the schematic before I can start the massive layout task ahead:

-- MAX11300 temp sensors

-- CC3100 flash

-- RF components & antenna

-- CC3100 crystals

Grab the latest off Github if you'd like to feel my pain.

I've also concluded the the production version of this board is going to have to have components on both sides to keep it from being enormous. If I'm going to bite that bullet, I might as well push most of the components down to 0402 or even 0201 in order to keep the overall size of things under some sort of control. However, I will only do that once I know I have a working circuit overall.

Now that the triple whammy of Makerfaire/Toorcamp/school project is over, I now have some time to work on this. I've created a Github repository for it, and I'm in the process of uploading core documentation for the chips I'm using so I have a consistent version of it to work from.

I'm still working out the details of what the analog front end should look like. In an ideal world, I'd have some sort of magical tuneable anti-aliasing and reconstruction filters. In my circuit of the imagination, the input would also have nice low-drift high-impedance programmable gain and the output would have high current drivers. I think I have to give up on most of that simply due to the complexity, size, and expense of implementing a circuit that does all of those things and is also low noise, low distortion, etc. I think I will break this tie in favor of building a circuit of minimal size and providing shield kits that allow users to add things like filters, drivers, and screw terminals. 

Next step is to put together a high-level block diagram that shows the functional blocks of the circuit and how they relate to each other. Then I can actually have an image for the project!