Software Defined Radio`s next wave: Fully
Transcription
Software Defined Radio`s next wave: Fully
Software Defined Radio’s next wave: Fully Digital Radios FWIC 2015 Bryan Donoghue 22 June 2015 P2120-P-016 v1.2 What was Software Defined Radio (SDR) meant to be? Software / Maths can directly manipulate signals at radio (GHz) frequencies Transmitter ..0101101… DAC Software / firmware / digital logic 22 June 2015 Gigahertz Radio Frequency Receiver ..1101001… ADC 2 P2120-P-016 v1.2 What is Software Defined Radio today? Software / Maths can directly manipulate signals at MHz frequencies – Real SDR at GHz is hard – Radios remain at least half analogue – Today’s radios provide many of the benefits of true SDR 22 June 2015 3 P2120-P-016 v1.2 Today’s Software Defined Radios have driven wireless handset performance Data rates have increased 1000-fold from GSM to LTE Enabled by 20 years of Moore’s Law gains: – Massive increase in SDR baseband complexity – Limited benefits to RF e.g. pre-distortion and linearization Data Rate Evolution 1000 100 Mbps 10 1 0.1 0.01 0.001 GSM 22 June 2015 GPRS EDGE WCDMA 4 HSPA HSPA+ LTE LTE-A P2120-P-016 v1.2 What’s the downside of analogue-heavy radio architectures? Analogue radios are costly to design – Analogue design process is lower productivity than digital or software – Analogue designs do not port easily between processes or geometries Analogue circuits do not scale well with Moore’s Law – Digitals circuits get smaller, cheaper and lower power with each new geometry 22 June 2015 5 P2120-P-016 v1.2 Where next? Old solution is running out of steam – Difficult to squeeze more bits/Hz – Already getting close to Shannon limits New Trends – Greater flexibility in use of spectrum – Greater spatial re-use of spectrum 22 June 2015 6 P2120-P-016 v1.2 Trend 1: Greater flexibility in use of spectrum All new wireless initiatives and standards are considering flexible use of spectrum – Often joining together (‘aggregating’) disparate fragments – It’s the only way that 100 billion Internet of Things ‘things’ will be viable! 4G ‘Soft Cell’ may simultaneously use licensed and unlicensed spectrum: Small cell in unlicensed spectrum Macro network using licensed spectrum More flexible SDR is needed to support multiple bands and protocols 22 June 2015 7 P2120-P-016 v1.2 Trend 2: Greater spatial re-use of spectrum Achieving higher spectral density than 4G requires more spatial processing – A common theme in “5G” technologies is therefore Massive MIMO 100s of radios form 1000s of signal paths Antenna Base station Phone Reflective object Phone Building multiple analogue radios on a single chip is disproportionally more difficult – A fully-digital radio would allow on-chip integration more readily 22 June 2015 8 P2120-P-016 v1.2 What is Fully-Digital SDR? 22 June 2015 9 P2120-P-016 v1.2 Delta-Sigma converters are key Advantages – Low Cost – Low Power – Digital-silicon friendly Typically limited to 10’s of MHz 22 June 2015 10 P2120-P-016 v1.2 GHz Delta-Sigma DAC in a Digital ASIC Sample rate must be >> bandwidth – Low-Pass Converter: – Needs 30GHz sample rate to create 1GHz signal – Band-Pass Converter: – Needs 2GHz sample rate to create 1GHz signal How to output samples in digital ASIC? – Use SerDes at 3, 6, 12 or 28GHz How to compute a high-SNR and configurable Delta-Sigma stream at GHz? – Unsolved problem until ‘Pizzicato’ 22 June 2015 11 P2120-P-016 v1.2 Breaking the computational bottleneck for Delta-Sigma DACs Complex DSP calculations at GHz Feedback loop is bottleneck Conventional approaches to parallelism fail ‘Pizzicato’ creates new approach to Delta-Sigma parallelism – Can achieve 100-fold or more speed-up – Generate RF signals directly at GHz 22 June 2015 12 P2120-P-016 v1.2 Pizzicato Experimental Test System Runs on mature-technology FPGA and generates high-quality in-band signals Exploits FPGA Serializer/Deserializer (SerDes) for high rate output Hints at what is possible 22 June 2015 13 P2120-P-016 v1.2 Project Pizzicato Pizzicato DAC performance We achieved ~65 dB between sine wave peak and noise floor ..with a passband of around 100 MHz wide 22 June 2015 14 P2120-P-016 v1.2 Project Pizzicato Within the passband, we can transmit any combination of signals Example: 14 carriers spread over 80 MHz Pizzicato excels in transmitter tests 22 June 2015 15 P2120-P-016 v1.2 What are the advantages of Pizzicato transmitter? Very high Nyquist Frequency - 14GHz with today’s SerDes Very low cost (e.g. 7 cents of silicon in 28nm) Very flexible SDR transmitter – Programmable Centre Frequency – Programmable Bandwidth Multiple transmitters in a single chip – Massive MIMO applications Easy to port between Silicon Geometries (it’s just Verilog plus a SerDes) Digital-only solution scales with Moore’s Law 22 June 2015 16 P2120-P-016 v1.2 Project Pizzicato There are limitations today Analogue filtering of the output signal is required – Reduces unwanted emissions and wasted power We only want this bit Pizzicato is a transmitter only – A receiver is even more complex 22 June 2015 17 P2120-P-016 v1.2 Project Pizzicato Amplifying the output Pizzicato outputs ~0dBm peak – Equivalent to a quiet Wi-Fi access point Pizzicato can drive a classic RF power amplifier via an analogue filter Pizzicato For ultimate flexibility and performance a purely digital amplifier is desirable – Class-S amplifiers are potentially more efficient than conventional power amplifiers Amplified bit stream This is an area of on-going research… 22 June 2015 Pizzicato 18 P2120-P-016 v1.2 What are the technology implications of a Fully-Digital Radio? The design process changes – Higher productivity digital IC and software tool-flow – More design re-use of Digital-RF IP modules – Availability of flexible, frequency-agnostic modems – More fluid boundary between RF and software-defined domains Analogue doesn’t go away – Fewer, more elemental components (Filters, LNAs, Power Transistors) – RF design knowledge migrates into the DSP domain 22 June 2015 19 P2120-P-016 v1.2 What are the market implications of Fully-Digital Radio? What if the radio becomes an IP core? – Will silicon IP companies displace chipset manufacturers? What are the effects of a ‘soft radio’? – Design cost of new radio architectures drops – More innovation and customisation, particularly in standards-exempt bands What happens if “RF is Easy” – New entrants in the chipset market? – Commoditisation of chipsets? – Value transfers to the hard-bits e.g. protocol stacks? 22 June 2015 20 P2120-P-016 v1.2 Further thoughts for the future Baseband modems today bear no resemblance to the modems of thirty years ago – Digital implementation has allowed baseband processing to change dramatically – Simple modems needed analogue implementation for transmitter and receiver (tones and filters). These are now replaced by abstract mathematical concepts such as MIMO, SCFDMA, and OFDMA But wireless standards today make assumptions about the characteristics of the radios to be used that are wholly influenced by our “analogue thinking” What if future radios were as mathematically abstract as basebands are today? – Will techniques like carrier aggregation (effectively multiple radios) appear crude when we look back in ten years time? 22 June 2015 21 P2120-P-016 v1.2 Cambridge UK Boston USA Singapore www.CambridgeConsultants.com Cambridge Consultants is part of the Altran group, a global leader in Innovation. www.Altran.com 22 June 2015 The contents of this presentation are commercially confidential and the proprietary information of Cambridge Consultants © 2015 Cambridge Consultants Ltd. All rights reserved. Registered No. 1036296 England P2120-P-016 v1.2