WP_Term Object
(
    [term_id] => 18
    [name] => Intel
    [slug] => intel
    [term_group] => 0
    [term_taxonomy_id] => 18
    [taxonomy] => category
    [description] => 
    [parent] => 158
    [count] => 319
    [filter] => raw
    [cat_ID] => 18
    [category_count] => 319
    [category_description] => 
    [cat_name] => Intel
    [category_nicename] => intel
    [category_parent] => 158
    [is_post] => 1
)

What, SD doesn’t have enough pins?

What, SD doesn’t have enough pins?
by Don Dingee on 04-16-2014 at 6:00 am

I was in a Twitter conversation over the weekend with some very smart people, and one of the discussion points was how slow and painful the formal standardization process can be. One suggestion was that IoT companies should “just do it”, creating specification-by-implementation. While I get the zeitgeist of the open sourcers, and I’m not a big fan of the plodding pace of most standards bodies, experience says it’s just not that easy – especially for hardware.

Intel, after watching the first decade of the IBM PC unfold, realized that selling a lot more microprocessors meant taking control of the form factor. By establishing the ATX specification in 1995, they not only solved a bunch of nagging problems with the AT form factor, but launched a new de facto standard that enabled a sweeping transformation of the supply chain for PC components.

There is no buzzkill for designers quite like ordering a new development board, then spending huge amounts of time chasing down a compatible power supply, funky cables, and a suitable enclosure – just to get a prompt. In the ATX era, standardizing mounting holes, expansion slots, I/O panels, and a power connector format meant any compliant board could be brought up quickly and easily using multi-sourced enclosures, cables, and power supplies.

We still see this thinking reflected in many maker modules today. Most use a barrel jack for fairly common AC/DC power supplies, and micro USB connectors for ease of connecting a development host and downloading code quickly. As these boards have gotten smaller and smaller, the need for a standard enclosure has diminished; most designers use a development board to get prototypes running, then design their own board fit to the exact device dimensions for production.

Point being, one would think Intel would get this right before Brian Krzanich stood up at CES 2014 holding a really tiny thing between his thumb and forefinger, an SD Card-ish Edison module almost ready for IoT prime time. Did Intel forget the lesson they taught the market? As it turns out, there is more to small form factors than soldering small parts to a small board – a lot more. That image everyone ran with from CES? It’s been replaced.


You might be thinking: “WTF? That’s not an SD Card. Bait and switch?” I look at a few of my scars – for instance, I’m the co-creator of the EBX specification – and feel their pain.

The advantage that Intel had in creating ATX was the design win pipeline was already running; all Intel had to do was keep feeding in new Socket 7 processors at one end, and the sheer velocity of demand created a draft that pulled massive volumes through without any extra pressurization. Nobody balked at ATX as the delivery vehicle for Intel processors.

When the design win pipeline isn’t up and running, designers take a good, hard look at how to connect your module to their world. Standards organizations stress out over connectors, spending a lot of energy on their selection and pinout definitions, with good reason.

Not that an SD Card form factor is a bad idea; in fact, Electric Imp is already finding some success in IoT designs with it. The imp001 packs a Wi-Fi interface with an ARM Cortex-M3 core, backed by a cloud connection that allows a device to get talking quickly. The pinout is deceptively simple, with 9 pins: two for power, one for ID, and six that can be multiplexed for a variety of functions.


image courtesy LX Group

That should be a big clue to a small problem. In a recent blog post on the “slightly larger than an SD card” version of a new Edison module, Mike Bell of Intel disclosed very few details other than they’ve switched processors – not Quark, but a 22nm Silvermont dual core Atom – and defined a small 70-pin connector.

While most of the tech blogosphere got all lathered up about a possible 22nm dual core Quark delay and Silvermont not fitting on SD, the real problem after running the idea up the flagpole might have been the pin count and connector. A more functional module requires a lot more functions at the connector, and Intel may have realized their miss very early.

Small form factors get eaten alive if the connectors are too expensive, unreliable, a beast to assemble, or present a crappy pinout with signal integrity or muxing issues. I’ve seen way too many good ideas go down in flames because of problems at the connector. The new and enhanced Intel Edison promised this summer will be worthy of following, even without Quark in the first round – one thing to watch closely is if this form factor catches on, and the design win pipeline starts to gain speed.

lang: en_US