InterBand

This is actually a prototype a did months ago, when first thinking about the project. Back then I had started with the idea of only using two lugs per side on the hinge, where the screws were set in permanently, and the modules slid together sideways. A cylinder held in place by the jumper (here shown as a piece of black cardboard) would then hold the two inner lugs apart. It worked reasonably well, but essentially relied on the jumper staying on place (being a piece of cardboard, it did not).From there, I've moved on to the idea of using 3 lugs on each side, holding themselves in place when attached. I was initially skeptical that these would be strong enough, as when using a screw with 4mm of thread, each lug is only just wider than 1.5 mm. I'm still getting used to SCAD, so these were drawn up in SolidWorks. Printing at 0.15mm seemed to work okay, they are suprisingly solid around the hinge points.

Something that probably needs pointing out here is that these basic module shapes are flat ( as opposed to a concave base). This is going to be more important with longer modules, but I'm yet to figure out a good way of printing them with such a shallow overhang. Of course, the whole thing could be tipped on its side, but then the lugs become small features overhanging horizontally, which in my experience never work out well.

To progress with the design of the electrical connection between modules, I figured I really just needed to start mocking things up and trying them. To that end, I drew up a rough version of an FPC jumper using the Molex SlimStack connectors I've mentioned before.

Though I haven't done any FPC design before, I'm aware that the rough design above needs a fair few changes, most notably curvier, smoother trace corners (honestly, a real pain in KiCad), and retaining pads leading out from the connector pads under the coverlay (also a pain).

The only cheap(ish) vendor I've found so far that offers a prototype service for FPC boards is Seeed Fusion. The list of design constraints for FPC on their site is pretty sparse though, so I sent them a support email explaining what I wanted to do, along with a copy of the gerbers for the strip shown above. If anyone is interested, apparently the minimum copper to edge distance is 0.2mm, and the required gerber stack is the same as their standard PCBs (no mention of a stiffener layer or anything). The reply also indicated they don't allow FPCs to be panelized, something that would be absolutely necessary for these small jumpers (it wasn't clear on what constitutes panelizing, I suspect having one large square and cutting them up with a craft knife would be fine). If anyone reading this has any other ideas as to where small jumpers like this could be made, please, let me know.

It turns out that getting the connections across the jumper while maintaining the rotational symmetry in the end connectors of the modules is a more complex problem than I'd first thought. The simplest method is to use a basic 5 pin connector on the ends of the modules, and use a crossover jumper. As long as FPC is being used though, the design is ideally single sided. This means that there are at least 2 traces going around the edge of the connector, increasing the width of the jumper.

Almost all stacking connectors use two rows, and using a 10 pin double row connector instead allows for the design I used in the KiCad screenshots above. This layout adds another axis of symmetry to the connector, simplifying the design of the module PCB by allowing any of the connections to be pulled off either side of the connector. It does require a trace passing around each edge of the connector though (the screenshots above show a version that connects the center pins separately).

One other method I've found since then involves shifting the swapping of pins to the module PCB, allowing for a very compact jumper. Here the module PCB provides both pin orientations, one to each row of the connector, and one row on each end of the jumper is masked off, insulating it from the traces underneath. I'm not sure how successful this method would be, as it relies purely on the insulation of the soldermask (in the case of FPC, the coverlay) to separate the lines, and has fewer attachment points to hold the connector to the circuit board.

The point of the physical part of the InterBand standard is to only define the shape of the end of modules, allowing them to happily interconnect with a range of modules without unnecessarily restricting the shape or form of the rest of the module. For instance, battery modules would probably lend themselves to a flatter, thinner form intended to sit underneath the wrist, possibly with a lug on the side to house a USB micro port for charging. Smaller sensor modules would happily sit on the side of the wrist, being shorter and allowing a greater curvature.

There are several forms this hinge could take, and like the electrical connection, I'm yet to firmly settle on one. In general, it seems like a good idea to use two hinge sections, one on each side, with a gap in the middle for the electrical connection. While I'd like the body of the modules to be 3D printable, the hinge axis itself is something that should probably be a separate component. The goal is to have the hinge be similar in size to conventional watch bracelet links, 22mm wide and no more than 4mm high. At these sizes, a pin in a 3D printed hole would not be strong enough.

So far I've found two options for the hinge axis: either use standard watch band pins, protruding all the way through the hinge, or small screws (likely M1.6). It's reasonably easy to get hold of either of these. Watchband pins have the advantage that they're fairly low profile, but I suspect that they rely on fitting tightly into a hard wearing hole to retain their position effectively (not how I'd describe the ABS and PLA used in 3D printing). Another disadvantage of these pins is that they obstruct the middle of the hinge, meaning any electrical connection would have to go around the center axis of the hinge, rather than being able to pass through it as it moves.

Small screws would have to be used in pairs, one on each end of the hinge, and can self tap relatively easy into 3D printed plastic. A potential problem with them is that the thread might slowly wear on the sliding half of the hinge (possibly even loosening the screw)

Of course, having a triplet of interleaving lugs on each side of the hinge isn't the only option. I've briefly thought having a short bar snap into a sort of cup (think a closed lug with a slot cut into the side of it), but am pretty sure that trying to print this at diameters under 4mm would fail.

The electrical connection and bus between the modules is a big part of what would make the InterBand standard successfull. The current plan involves 5 connections: GND, 3.3V, SDA, SCL, and INT. GND and 3.3V are pretty self explanitory, and SDA and SCL essentially implement I2C or TWI. Even though these protocols support multiple hosts, the intention behind InterBand is to have a single host module with many slaves.

This host module would generally be the communication interface with the outside world, whether that is a simple Bluetooth connection or a more complex display interface. The INT line is my attempt to try and support low power features in supported modules. The idea here is that unless a module is actively doing something, it can enable any low-power modes it supports, and wake up when the INT line is asserted. Any module can assert the line, and when it is pulled low all modules must wake up. If it was the host that asserted INT, it then does what it needs to (communicating with a slave module), before letting the line go. If the host wakes up due to a slave asserting the line, it polls them, finding who is asserting it and sorting out what it needs. This scheme should allow very basic modules to be made of ICs that natively support I2C or TWI, but also add a bit more flexibility for those with a microcontroller on board.

Actually getting these 5 circuits connected between the modules is another problem entirely. There are several requirements of such a connection: First, it must be flexible, as in able to flex over the hinge between modules. Sharp, acute angles aren't necessary, but the physical hinge should limit movement before the flexible connection is damaged. Second, it must be rotationally symmetric (as must the physical hinge).

By this, I mean that both ends can happily connect to the same end of another module (no distinction between "leading" or "trailing" ends). This is to avoid putting limitations on the configuration of modules in the band - individual modules may be designed to work with a particular orientation with respect to the wrist or arm, and as long as they have an inside edge (which would normally be slightly concave to curve around the wrist or arm), they will be able to connect to other modules in both orientations. The third requirement is that the connection is a removable, re attachable part. Modules should be able to be mixed around in the band with relative ease, and a soldered connection between them kind of prohibits this.

Flexible, easily removable electrical connections are hard. My first thought regarding this was to use some sort of pogo pins, with a conductive target pad on the other side of the hinge, but even a cursory amount of thought down that path revealed that the sliding connection as the hinge flexed could momentarily interrupt the connection - not good for a high speed bus. Additionally, dust and other foreign particles would be free to get in the way and obstruct the connection.

The next idea was standard flex cable jumpers. These seem to be readily available and the connectors that the ribbon is inserted into are quite compact, but so far I have been unable to find any shorter than 25mm (to not intrude too much on the space in the modules this really needs to be under 15mm or so). The right-angle insertion nature of these connectors also isn't ideal, as they would need to be inserted in sideways while the hinge is slid into place, complicating attachment process.

The leading solution at the moment is to use custom flexible jumper strips, with a low profile stacking connector at each end. This has the advantage of being attached from the inside of the band, rather than tangentially. The connectors themselves are readily available, and I'm currently looking at these, though they're not exactly cheap. There are two ways these jumpers could be constructed, either FPC (more expensive), or two thin PCBs with a soldered cable (cheaper, less compact).