All-CMOS Bluetooth/ZigBee Transceiver for Personal Networks

According to researchers at the Interuniversity Microelectronics Centre (IMEC), the number of “things” connected to the Internet in 2008 was more than the entire population of the Earth! And by 2020, it projects billions of devices connected to the Internet. Eventually, IMEC researchers say, the number of short-range radios will outnumber smartphones and tablets.

Such wireless innovations applied in smart buildings, smart metering, personal healthcare with portable diagnostics, and monitoring that can be worn, will challenge designers to conserve energy and extend battery life without compromising performance. All of this, plus complying with standards for interoperability where applicable.

To help get us to this world of ubiquitous wireless, IMEC used this year's ISSCC gathering in San Francisco to describe a CMOS-based Bluetooth Low Power/ZigBee 802.15.6 combo radio architecture for low-power, personal/body-area networks (PAN/BAN) that is worth examining for its design choices. This is one of the first development ideas that the IMEC research team has conceived toward their goal of a more efficient set of ULP (ultra-low power) radio ICs.

In general, widespread, use-it-everywhere wireless will be enabled by circuit integration. IMEC cites the following statistics that highlight how these wireless devices (and the technology behind them) are sneaking into our lives. Mixed signal: 10 papers published at ISSCC conference 2013, seven patents (filed but not granted in the last three years) Radio and DSP patents: 12 papers published at ISSCC conference 2013, 20 patents (filed but not granted in the last three years). These are sure signs of coming innovations to foster RF, processor, and mixed-signal integration, coupled with the drive for lower power technologies and architectures.

To help put all this in perspective and explore possible routes to meeting these on-going design challenges, I had a discussion recently with Harmke deGroot, Program Director, Ultra-low Power Circuits at the Holst Centre/IMEC, (Eindhoven, The Netherlands). Ms. deGroot, one of the ISSCC paper's co-authors, and her team, are working with various IC suppliers, like Renesas, to enhance ULP for short range communications while extending battery life in portable systems like those in PAN/BAN.

IMEC researchers predict that power and data will be simultaneously sharing a single antenna at long distances (They have already achieved 15-20m in 2012 with their Rectenna design). Others are also working on these power/data transmission modes. MIT is working on a mat that sits in your garage with an integrated coil over which you drive your electric vehicle to begin charging.

The PAN/BAN solutions were a subject of a 2013 ISSCC paper[1] that attracted my interest because they were proposing a 2.4GHz radio that would be compliant with three wireless standards for PAN/BAN, with the caveat that it also improve energy efficiency of most existing radio designs. Figure 1 shows the block diagram for such a radio.

Figure 1

Simplified block diagram of the multi-standard ULP transceiver[1]

Simplified block diagram of the multi-standard ULP transceiver[1]

IMEC will use a CMOS process for radio integration with the transceiver discussed in this ISSCC paper being based on TSMC's 90nm technology. Figure 2 gives a good view of how all this will be squeezed onto the IC.

Figure 2

The layout of the highly integrated ULP transceiver IC[1]

The layout of the highly integrated ULP transceiver IC[1]

Normal analog integration is challenging enough, but to integrate RF VCOs (voltage controlled oscillators), PLLs (phase locked loops), ADCs, LNAs (low noise amplifiers), etc. — that's a real challenge. Crosstalk at these high frequencies can be problematic. EMI/RFI can plague a design if not done properly. The die itself should have good practices to minimize radiated energy as well as to de-sensitize the circuitry from radiated RF signals interfering with the IC function.

If possible, it's always best to make an IC whose die does not have to be encased in an EMI shield. Besides eliminating the cost of the shield, the die can more easily dissipate heat energy into the air.

The designers wanted to achieve an energy-efficient radio architecture, so a suitable local oscillator (LO) frequency plan was selected. Some other efficiency enhancement techniques were employed in the critical RF circuits (such as a push-pull mixer and a digitally-assisted power amplifier).

More clever techniques were employed by the designers, like the receiver (RX), which uses a sliding-IF architecture to reduce the power consumption by avoiding the need for a high frequency quadrature LO.

In addition, to save power and eliminate the lossy balun, the researchers used a single-ended LNA that also functions as an image rejection filter, and chose a push-pull mixer structure which improves both the transconductance efficiency and noise performance.

The delicate balance of a creative design that preserves performance needed while conserving power is an endeavor that has always involved trade-offs. Size can be one of those trade-off specs, yet the designers were able to keep the die at a core area of under 4mm2 .

IMEC's design achieved the best RX sensitivity versus energy efficiency performance of current state-of-the-art ULP 2.4GHz RXs while achieving a 3 to 10X improvement in energy efficiency.


  1. 1. “A 1.9nJ/bit 2.4GHz Multistandard (Bluetooth Low Energy/Zigbee/IEEE802.15.6) Transceiver for Personal/Body Area Networks”, Yao-Hong Liu, Xiongchuan Huang, Maja Vidojkovic, Ao Ba, Pieter Harpe, Guido Dolmans, Harmke de Groot, 2013

Related posts:

5 comments on “All-CMOS Bluetooth/ZigBee Transceiver for Personal Networks

  1. eafpres
    June 22, 2013

    The design you highlighted seems a nice achievement.  For me, given the massive industrial efforts in innovative chip designs, including heavyweights like TI along with many others, I'm confused by IMECs choice to spend resources here.  Why spend public funding on wireless technologies when the industry is doing a great job of putting huge numbers of choices in front of design engineers already?

    I wanted to comment on one point regarding die-design.  While I would agreee that it is desireable to avoid needing shielding, it may not be wholly accurate to say that adding a shield degrades heat dissipation.  There are several compnaies that provide shielding with thermal interface materials between the inside top of the can and the chip, and have nearly continuous contact around the perimeter (vs. other designs with lots of “feet”), thereby making the shield can a highly effective heat spreader.  Proper board design to capitalize on this can result in lower temperatures than a bare chip.

  2. Steve Taranovich
    June 22, 2013

    Excellent point regarding the dual use of a shield and heat-spreader, easpres. My thought was to save money, little cost as it might be, for a shield plus assembly step to install with proper IC design thermally and EMI-wise—if that can be effectively done. Otherwise, your point is well taken and certainly valid.


    I will let imec comment on your other point—stand by

  3. Imec_nl
    June 24, 2013

    Thanks for liking our design.

    Internet of things, making all devices (also small sensoric ones) talk to each other and deliver context information and smartness to your personal environment, dictates that all devices, also very small sensoric devices, are able to wirelessly communicate to each other in the near future. By 2020, the number of short range radios sold per year will be higher than the number of mobile radios. Though there are many radio solutions already in the market today, most of them cannot be used for long lifetime sensor applications running on a small battery or harvester. Radios on the market today are either low power enough to work on a coin cell or harvester,  OR standardized, but then not realizing a long lifetime when running on a coin cell or harvester.

    In sensoric nodes, the overall percentage of the system power consumption consumed by the radio today is often as high as 50-85%. This leads to an enormous market demand for short range radios which are (multi-)standardized for interoperability, and which have a power consumption 3-50 times lower than radios on the market today.

    As imec, we do not deliver radio ICs to the market, but help semiconductor companies to realize their goals and roadmaps. Only approximately 15 percent of our funding is public and that part is normally spent on longer term research and infrastructure. The other 85% is covered by industrial contributions, e.g. for this radio design by semiconductor companies who like to productize our prototype design.

    We contribute to the internet of things revolution by delivering new innovative short range radio designs where we combine our IC technology, application and IC design competence. While a company like TI is without a doubt able to have a full research and design team on this topic, nearly all semiconductor companies will need short range radio solutions as it is going to be an add-on IP block in nearly every design within a few years from now. Not for many of these companies it will a differentiator in the market, it will just become a basic requirement. This makes short range radios an excellent topic for an imec program: we develop the generic solutions, and many companies can benefit from the results (at a fraction of the cost and risk compared to in-house R&D), adapting them to their individual needs.

    Hope this explanation helps, when we are successful you will indeed get even more choice in radios as a design engineer, but they will have significantly lower power consumption and better specifications!

  4. eafpres
    June 24, 2013

    @IMEC–thank you for the additional details.  All of your points are valid assuming the predictions of IoT are accurate!

    Perhaps we are touching on a relatively new distinction being drawn between M2M and IoT.  In the M2M space, which in my definition includes short-range wireless like  Zigbee, there has been significant emphasis on adding capabilities and integrating the wireless functions with micro processors or micro controllers to make a single chip solution for a diversity of applications.  On the other hand, the design you are describing is a building block that would be integrated into, say, a revision of an existing, non-wireless design.  I note that your radio design has no processor or memory associated with it, so those functions would be external, if needed.

    Can you comment on how this design would come to market?  Would it eventually be open source, or will it reside at particular foundarys which would be available to customers?  I noted elsewhere references to a partnership with Renesas on these designs, but am unclear if this design would belong to Renesas or be made available through some other model.

    Thank you again for your comments.  IMEC is involved in a lot of cutting edge technologies, perhaps this forum can get early views of your other work in the future.

  5. Imec_nl
    June 24, 2013

    Our industrial partners bring our designs to the market as part of their Product offering. Of course they add memory, microntroller etc. to the production design so the whole radio (and in some cases even the complete sensor solution) will be in one package. Unfortunately I cannot comment on their timelines. Our designs will not become available as open source. However we do sometimes contribute to the community by making proposals for new standards, writing white papers etc.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.