RF Integration – The Final Frontier

I saw a teardown of the new BlackBerry Z10 the other day. It’s pretty typical of smartphones, with a lot of computing capability. The teardown highlighted the major components. The digital stuff, including the cellular baseband processor and the apps processor and associated memory, fits into three packages:

  • Samsung K3PE0E000A – Multichip Memory – 2GB Mobile DDR2 SDRAM
  • Samsung KLMAG2GE4A – Multichip Memory – 16GB MLC NAND Flash, Controller
  • Qualcomm MSM8960 – Snapdragon S4 Baseband/Applications/Graphics Processor

That latter chip, done in 28nM CMOS, includes a pair of >1GHz processors, LTE/3G/2G modem, graphics processor, and a ton of I/O and memory. Lots of different functions, but hey — it’s all digital, right?

The analog componentry, including interfaces to allow humans to use the device, keep the device powered up, and get voice and audio into and out of it takes a few more chips.

  • Qualcomm WCD9310 – Audio Codec
  • Qualcomm PM8921 – Power Management IC
  • ST Microelectronics LIS3DH – MEMS Accelerometer
  • STMicroelectronics LSM330DLC – 3D Accelerometer & 3D Gyroscope
  • Synaptics Clearpad 3203 – Capacitive Touchscreen Controller

And that partitioning makes sense if you think about it. Funny thing about analog partitioning — the chips tend to use lots of unique processes and make integration hard. And power-management as a stand-alone function makes sense rather than putting all the high-energy switching and heat-generating linear regulators on the same chip as something that might be sensitive to noise, thermal gradients, etc.

But what struck me was the lack of integration in the RF section… all for good reasons, I guess, but just look at what it takes to get the radio part done!

  • Qualcomm RTR8600 – GSM/CDMA/W-CDMA/LTE RxD Transceiver + GPS
  • Texas Instruments WL1273L – Single-Chip 802.11a/b/g/n WLAN, Bluetooth, and FM
  • Inside Secure SECUREAD IC5C633I4- NFC Solution Module
  • TriQuint TQP6M9017 – Dual-Band WLAN Module
  • RF Micro Devices RF7252 – CDMA/WCDMA BAND 2 Linear Power Amplifier Module
  • RF Micro Devices RF7303 – LTE/UMTS/CDMA BAND 3 Linear Power Amplifier Module
  • Avago ACPM-5017 – LTE Band XVII Power Amplifier
  • Avago ACPM-7051 – Quad-Band GSM/W-CDMA Bands I & V Power Amplifier
  • Sony CXM3582UR – SP10T Antenna Switch

The move to direct-conversion promised that RF would finally jump on the Moore’s Law curve by eliminating non-integrated discrete passive SAW and ceramic RF and IF filters in favor of integratable low-pass analog and digital filters. And for the small-signal RF stages, that’s worked out OK. But how will we get all those power amplifiers (and associated passive filter and matching components) integrated or eliminated?

7 comments on “RF Integration – The Final Frontier

  1. eafpres
    March 7, 2013

    I was teasing Martin Rowe over on The Connecting Edge about the evolution of optical communications and he retorted that we'd still have all that electrical stuff on the chipsets, to which I said “optical quantum computing is just around the corner, then we can be rid of those pesky electrons.

    Martin then set me up of an onslaught by saying this “eafpres,

    “So if I understand you correctly, all electrical engineers who will develop will really have to be trained as optical digital engineers. Thus, the only real electrical engineers will the the analog people. There are those who already argue at only analog engineers are real electrical engineers even today.”

    Seriously, though, do you think that the proliferation of so many wireless standards and rapid evolution of use models argues against more RF integraiton in the near term?

  2. Netcrawl
    March 7, 2013

    Interesting article! thanks for the share, the latest teardown ofthe new  BlackBerry smartphones revealed the mobile phone's massive computing capabilities- its packed with sophisticated chips. I believe, a new type or breed of technology is about to emerge-something new! its optical quantum computing. Althought ist still in infancy and stage one- its will probbaly revolunize the way we build our stuff in the next couple of years. It replace those nasty electrical stufdfs on our chipsets.


  3. patrick_m
    March 8, 2013

    I don't think it 'argues' against the integration of RF, eafpres, but what it does is add urgency to the need for programmable wideband filters to handle multiple standards, from 900-MHz 802.15.4 up to 5-GHz Wi-Fi and all else in between. MEMS filters have been eyed as a solution here for some time, but the different processes required for that are an obstacle, though I do hear of progress being made there.

    Or, a direct-to-digital approach, which puts a lot of onus upon the A/D designers, and we're not quite there yet. Whatever happens, Doug's pointing to a huge opportunity for integration!

  4. eafpres
    March 8, 2013

    @Patrick–yes, and you also need tuneable matching networks because the s-parameters of antennas across all those bands aren't the same, or you might be switching among antennas, so you can either switch (and duplicate) the whole matching network component set or find a way to tune for each band.  If you had a front end integrated that could sense s parameters and match as well as tune the filter in a dynamic way, that would be pretty useful.

  5. Brad Albing
    March 8, 2013

    eafpres – let me know when you get that s-parameter sense ckt designed.

  6. Netcrawl
    March 11, 2013

    There's one thing we need to get here: we need to cover as many bands as possible. The industry are moving to a level where it requires more bands and faster data rates, this is a serious matters, because this could puts more stress on the component.

Leave a Reply

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