New Part Day: Tiny, Tiny Bluetooth Chips

The future of tiny electronics is wearables, it seems, with companies coming out with tiny devices that are able to check your pulse, blood pressure, and temperature while relaying this data back to your phone over a Bluetooth connection. Intel has the Curie module, a small System on Chip (SoC) meant for wearables, and the STM32 inside the Fitbit is one of the smallest ARM microcontrollers you’ll ever find. Now there’s a new part available that’s smaller than anything else and has an integrated Bluetooth radio; just what you need when you need an Internet of Motes of Dust.

The Atmel BTLC1000 is a tiny SoC designed for wearables. The internals aren’t exceptional in and of themselves – it’s an ARM Cortex M0 running at 26 MHz. There’s a Bluetooth 4.1 radio inside this chip, and enough I/O, RAM, and ROM to connect to a few sensors and do a few interesting things. What makes this chip so exceptional is its size – a mere 2.262mm by 2.142mm. It’s a chip that can fit along the thickness of some PCBs.

To provide some perspective: the smallest ATtiny, the ‘tiny4/5/9/10 in an SOT23-6 package, is 2.90mm long. The smallest PICs are similarly sized, and both have a tiny amount of RAM and Flash space. The BTLC1000 is surprisingly capable, with 128kB each of RAM and ROM.

The future of wearable devices is smaller, faster and more capable devices, and with a tiny chip that can fit on the head of a pin, this is certainly an interesting chip for applications where performance can be traded for package size. If you’re ready to dive in with this chip the preliminary datasheets are now available.

88 thoughts on “New Part Day: Tiny, Tiny Bluetooth Chips

  1. Most people will probably have to use their premade 12x20mm module with antenna etc. A 0.4mm QFN is on the edge of what hobbyists can handle, but a 0.35mm BGA that requires escaping from internal rows is just not going to happen without expensive laser drilled and filled/capped via-in-pad PCB processes.

    1. Sad but true. From my recent experience one event cannot work around HDI boards with 0.5oz final copper thickness, 2-sided layout with a star-shaped single-layer fanout for 0.4mm 36-BGA. The cheap manufacturing processes are just not capable of handling 80µm traces to run between the balls.
      I’ll try PCBs with parametric arrays (variable trace width / pad size) some day to see if there is at least some yield achievable by cherry-picking the best etching result.
      Still, NSMD (non-solder mask defined) patterns will fail most of the time due to poor stop mask registration (+/- 50µm).

    2. If I could work with parts so small what would I do with it? A device that only talks to itself doesn’t sound very interesting so it would need to be connected to something. I think the point was reached a while ago where the parts are smaller than the connectors so sure.. they can keep shrinking them but they will just have to be soldered onto a board which isn’t going to shrink with it as the connectors are now the determining factor in size, not the chip. This is doubly true with things like Wifi because it needs an antenna.

      1. connectivity is not just about peripheral access, the case for this chip is made by low power communications and a diversity of popular interfaces: SPI flash, Master/Slave SPI, multiple I2C master / slave transceivers and UARTs. Given the form factors of accelerometers (BMA280), IMUs (BMI160), ambient light, gas, temperature and geomagnetic sensors(BMM150), MEMS microphones, even thermopile pixels (TMP007) that all share the same length scale, allowing the system as a whole to scale down significantly.

        As the Pebble Time watch has shown, if engineered correctly the antenna may as well be the bezel, you can even mill a hole into the case and call it a slot antenna (some wifi routers start using these) so no board space is wasted.

        We’re seeing a technology-driven increase in equipment cost much like the cost explosion in semiconductor lithography machines and this now starts to hit the hobbyist market as well. We’re now at the brink of requiring proper reflow ovens and camera-assisted chip alignment tools to do the prototying (on top of hot plates, hot air stations and the trusty soldering iron) – but this is counter balanced by cameras and inspection microscope gadgets getting much cheaper (say a Raspberry Pi with two cameras that have their focus re-adjusted for macro) and open source hardware projects (paste dispensers, pick&place to name a few).

        We won’t be locked out, it’ll just get different and more fancy. X-ray PCB inspection will still suck though.

    1. As an addit; mace get beat me to it.

      To add substance, the external part BOM is not insignificant; (QFN version)
      20 capacitors
      2 crystal networks
      5 or 6 inductors (can omit 1 at the expense of power)
      5 ferrites

      Big bonuses worth thinking about;
      Battery management; input voltage 1.8 – 4.2v (lithium chemistry)
      Integrated Bluetooth profiles for common sensing applications (temperature, heart rate)

      You might be able to use the BGA version with a manufactured board if you “write off” the 7 trapped pins, accepting the loss of SPI including SPI FLASH and a mixed signal input (VSS tying to ground and ignoring the test point)

      1. Yes your tiny package is really tiny, but it’ll take similar amount of board space to place the part and/or break it out unless you are using bleeding edge manufacturing.

        Ditto for QFN if you don’t have blind & buried vias. The middle heat sink pad pretty much kills any possibilities for breakout within the package footprint. Having said that I do like standard QFN packages (not the dual rows one). The are much easier to handle, inspect, align, and reflow/solder manually. I have done multiple 0.4mm pitched QFN on $10 Chinese PCB proto PCB with no issues..

        1. The thing is, with a chip that tiny it means they can bake thousands on a single wafer, which means it’s dirt cheap.
          So all we need is some manufacturer to deliver it on a slightly larger PCB so we don’t need to go crazy trying to connect it and there we go, a nice device for 3 dollars or what have you.

          1. Packages can be a lot larger than the die it hosts. For example, 2″ long 40 pin DIPs often hold a single die that’s 0.2″ or so. That’s a rather extreme example. Also, other posters pointed out there’s other package variations for same die, so the situation for a hobbyist might not be so dire.

        2. Well, you never break out BGAs *within* the package footprint either. The outer row or two are almost always directly routed out. The difference between BGA and QFN with exposed pad is that interior balls prevent routes on *all layers* if they’re tight pitch. With the exposed pad you don’t need to make the ground connection that obstructing.

          You could easily break out that QFN on a 2-layer board to standard 0.1″ pitch headers (and with a solid ground plane on the back, too) with a single power input and all extras (crystals, inductors, antenna, etc.). The board would be slightly larger footprint than the headers since you’d need to the 1.2V to get around, but it’s really no big deal. This is where the integrated switching regulator helps – you’ve only really got one power trace to deal with.

          1. Unlike a leaded part where you can breakout and route with a part’s foot print, a QFN doesn’t allow for that because of the slug in the middle. Unless you are using bleeding edge PCB, the clearance requirement between vias are going to be more than the pitch. So you have to fan out much farther and stagger the vias. In the end you’ll need a much larger area than the package size.

      2. One of the crystals is optional (the RTC). A lot of the others are coming from the integrated switching regulator. Whether or not this is a benefit depends on application.

        For the “make it sleazy and kinda sorta work” crowd, a fair number of those parts could probably be trimmed and it’d be mostly okay.

      1. I don’t agree. The ball count is so low that it probably wouldn’t be that hard to line up.

        Left-right, up-down orientation isn’t that hard to get right. The problem with larger chips is getting it right on all four sides, at once – because slight adjustments rotate the chip, which translate into big errors at the end. In this case it’s only 5 balls deep in any dimension, so a slight misalignment really won’t have a chance to grow into something significant by the edge.

        The obvious extreme example are those 4-ball BGA packages, which are actually really simple to solder on.

          1. As mentioned, single gate inverters, single transistors, voltage references… just about anything with 4 or less pins.

            Pretty much anything under 9 balls (or really, anything with less than 3 rows or columns – only one dimension really matters) is a total joke to solder on (using reflow/hot air, obviously). Really their only downside is that you can’t recover/reuse them. To me they’re easier than QFNs because with practice you can actually feel the balls settle into the solder mask gap, and you don’t need to worry about paste volume or anything like that. Practice a bit on cheap parts to get the reflow timing down, and it’ll work every time.

    1. Perhaps in this age of teeny-weeny packages, we need smaller human beings with tiny fingers to hold wee soldering irons. I believe the Chinese are working on cloning a 1:8 scale workforce for the high-tech industry.

        1. No, that’s the later project. The one for making the tiny operators to put inside cell phones. They’re fed on a slice of highly processed seaweed, fed in through a SIM-like slot.

    1. Not till someone builds a breakout and that would be a pretty empty board. The pin pitch on your standard bread board means you will have a lot of dead space. But man could you do some cool stuff with this.

    1. >TI and other reliable chip manufacturers

      Stellaris(TI) left a bad taste in my mouth, they randomly dropped support for a few -new- chips we designed boards around. Honestly if you’re going to call one less reliable than the others you best have a personal experience or anecdote that helps us see why, else you’re just spouting random slander and libel.

      Also to be fair I have nothing but praise for their MSP430 series.

    2. Reliable as in the product reliable or the manufacturer not delivering consistently? Because I’ve done very long part build runs in high ambient temperatures using an ATMega chip on a 3D printer.

      I call BS on Atmel being unreliable without corroboration.

    1. eeew no onboard flash? totally agree, yeah no thanks. you’re going to need a few more mm^2 to route and place a SPI flash…been misled enough times after realizing that a smaller chip needs tons of external components that I don’t buy that nonsense anymore. they’re basically the only major BLE/Bluetooth SMART chipset I know of at this point that doesn’t have internal FLASH ROM…they’re pretty behind the times here

      1. Cool. I’m glad he wrote it. It’s just the “non-commercial purposes” well, plainly sucks.

        It leads to situations we’re currently seeing in the FAA, where doing a drone flight video, and uploading it to youtube and turning on advertising is “commercial”. Exactly how far, legally speaking, does “non-commercial” go? How much is “commercial”? $0.01, $0.1, $1, $10, … If I am given a handful of nRF24L01+ boards, does that count as commcercial?

        I work for a public University, and I was interested in using it in that setting. Yet, we charge students a hell of a lot. Is that commercial? Our uni is in a large part, funded by the state.

        A GPL or similar license would have made much more sense.The GPL is legally unambiguous, combined with tried and tested in court.

        1. Also, if you’re going to make money off a product with his work in it, why shouldn’t you pay him? I’m sure you could discuss terms if you email him. With a GPL, you have obligations to release source, a nice simple license might be better.

    1. Reimplmentation is not a problem, you have all the research done for you.

      // see BT Core Spec 4.0, Section 6.B.3.2
      // see BT Core Spec 4.0, Section 6.B.3.1.1

      All the configuration values can be copied as is because magic numbers can not be copyrighted.

  2. I really hope we start getting some very tiny arduinos with these tiny wireless chips on them. itty bitty processor with a BT4.0 radio would really open up for some very cool small devices like tiny data loggers that you can read the stored data with your phone ,etc…

    1. You could use this and an ESP8266 on the same small pcb. Write custom firmware for the ESP and you could easily have a serial bluetooth/wifi radio for under $10 total. Break out the right pins and it would still be programmable!

      Someone should get on that, and while they’re at it break out some more of the ESP8266’s pins.

  3. This isn’t a flash based part, so it may have 128K of SRAM to work with, but it will chew it up with your code. The firmware is in ROM, but I’d imagine it requires a chunk of SRAM for patch tables and fixups. The part may be small, but you need to stick an SPI flash part on your board to store your application code. The part may be small, but not exactly a minimal BOM design.

    If you are doing a BLE design, do yourself a favor and go with a flash based part and avoid parts that use a ROM based BLE stack. ROM is great for bootloader functionality, but a real pain for radio protocols.

  4. Just for those Microscale-lovers, Freescale has some VERY TINY package options on the Kinetis family too. They claim to have the smallest Cortex M0 with a 1.6×1.99mm 20WLCSP package. Theyr smallest wireless options are a bit bigger tho, i think the 32QFN with 5x5mm ist as small as they go with the MKW30Z160VHM4 (Kinetis+BLE).

    1. I’ve seen those; Kinetics KL0x or similar, in 0.4mm pitch BGA. They are insane! SiLabs also recently released a (formerly EnergyMicro) EFM32HG part in 36-pin 0.4mm BGA. This is slightly larger, but packs in a 3.3v regulator, USB transceiver and good-enough-for-usb oscillator on-die.

    1. That part would also require a lot less external board space, since it presumably has all the caps and crystals packed under that heatsink as well. In the end, that’s probably the build style most people will use in any but the largest build scales.

      The chip shown in the article is basically for people who want to make modules like these.

  5. I was really looking forward to seeing what Atmel would bring to the BTLE scene but I’m a bit disappointed.

    It’s looking like the TI chips with the ARM cores doesn’t need the overpriced IDE like the previous part so when the cheap modules hit aliexpress it will probably be the best option for hacking.

    Interestingly the EFM Blue Gecko BGM111 module is fairly cheap from mouser, though the QFN style module might be tricky to solder without a reflow oven or hot plate.

    1. I bought a toaster oven for $10 at a resale shop. Maybe I paid too much but I’ve done a few dozen boards with it, no failures yet. I also use a thermocouple meter to watch the actual temps, sometimes the temperature set slider is wrong.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s