doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 1 Rishi Mohindra, MAXIM Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: OFDM PHY proposal for SUN Date Submitted: May 1, 2009 Source: Rishi Mohindra, MAXIM Integrated Products Contact: Rishi Mohindra, MAXIM Integrated Products Voice: , Re: TG4g Call for proposals Abstract: PHY proposal towards TG4g Purpose: PHY proposal for the TG4g PHY amendment Notice:This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release:The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P Slide 1 doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 2 OFDM PHY Proposal for g IEEE 802 Interim Session Montreal, Canada May 2009 doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 3 Contents Motivation for using OFDM TDD Framing structure to support 100,000s of nodes OFDM based PHY Proposal for IEEE g OFDM parameters & Symbol structure 2-Ray channel simulations: comparison of FSK and OFDM Transmit spectrum, ACPR Application Issues Conclusions doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 4 Motivation for OFDM OFDM offers unmatched performance in adverse multi-path conditions when multipath channel coherence bandwidth >> subcarrier spacing, forward error correction and interleaving techniques completely recover the packets error-free. E.g. when relative path delay is 10us in a 2-ray model, there is only 1 deep null in a 75kHz OFDM modulation bandwidth i.e. only 1 subcarrier is affected out of a total of say 14 for a 16-FFT case. FSK & GFSK systems completely break down under these large multipath conditions (for similar modulation bandwidths and data rates) OFDM tolerates large multipath delays by using cyclic prefix. It has no impact on modulation bandwidth and very minor impact on data rate. There is almost no ISI. GFSK systems need to ensure multipath delays dont exceed 10-20% of symbol duration (e.g. 2-ray equal power model shown later). For a given maximum multipath rms delay spread Td, there is a limit to the GFSK data rate of approx 0.2/Td. Beyond that the ISI increases significantly. OFDM maximizes spectral efficiency (e.g. 2.5bits/s/Hz) and improves transmitter ACPR (nearly 20 dB better than GFSK in the adjacent channel). OFDM offers best compromise between battery life and Data Rate by minimizing the active transmit or receive time for a given data payload. doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 5 Motivation for OFDM Occupied Channel bandwidth of OFDM can be increased to mitigate flat fading (due to small multipath delay spreads), without affecting performance under large delay spreads (since subcarrier spacing and cyclic prefix are independent of channel bandwidth). FSK & GFSK systems dont have this advantage. If their bandwidth is increased, they cant operate under large multipath delays (since bit rate < 0.2/Td). Extremely high throughput enables easy mesh networking with longer battery life. Can fully re-use IEEE802.11a/g Phy technology for 64-FFT OFDM. 16-FFT OFDM is a greatly simplified sub-set. OFDM allows data interleaving in both frequency domain (across subcarriers) and over time domain (across symbols). FSK, GFSK etc dont allow frequency domain interleaving and are sensitive to channel frequency nulls. More than one null in the modulation band can destroy the entire packet. This happens for larger multipath delays, when they start exceeding 20% of the symbol duration. doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 6 Motivation for OFDM Frequency hopping and antenna diversity can mitigate multipath effects for FSK & GFSK, but OFDM will also benefit from it Switching over to another spatially separated node (e.g. another home) for the mesh link can help, but that could fall into a flat-fading channel-null at a different hop frequency. For OFDM the simple remedy is to increase the occupied channel bandwidth. FSK & GFSK cant increase their bandwidth significantly since their symbol duration decreases and they become more sensitive to larger multipath delays that can obliterate the packet (see 2-ray models covered later for FSK and OFDM). OFDM hardware complexity (including gate count and power) can be scaled with data rate and channel bandwidth: without affecting the multi-path performance which only depends on subcarrier spacing and symbol duration that can be kept constant, independent of channel bandwidth 16-FFT can be used in 75kHz occupied channel, or 64-FFT in 265kHz occupied channel. Both work equally well for >10us multi-path delays for raw data rates from 288kb/s to 1.152Mb/s for 64-QAM doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 7 Motivation for OFDM High OFDM symbol rates and throughput allow 100,000s of nodes to communicate to a single Base Station in star network over much shorter frame durations compared to other modulation schemes. Up-link signaling overhead can be tremendously reduced Complete up-link packets of only 2 OFDM symbols are needed per node for sending time critical data. It includes training sequence and a 288-bit raw payload (for 64-FFT) No interference occurs between nodes due to non-overlapping OFDM symbol based slotted time structure, even with 2000 nodes transmitting each second. Just by changing the subcarrier spacing from 5kHz to 10kHz, the OFDM symbol rate doubles, and it will allow 4000 nodes to communicate per second using a 2-symbol packet. Transmit power control (TPC) is not required for the OFDM network operation. It is only useful for reducing emission into other channels that may be occupied by other networks. There is no loss in SNR or noise floor increase with nodes communicating per second DSSS and CDMA suffer from self-induce noise floor increase and require a good TPC in order to mitigate this to some extent doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 8 16-FFT and 64-FFT OFDM proposals Use 5 kHz subcarrier spacing as example. Both 16-FFT and 64-FFT will allow >10-20 us of multi-path maximum excess delays with no impact on throughput using Convolutional encoding, interleaving and FER. 16-FFT allows 288 kb/s raw data rate in only 75 kHz modulation bandwidth (100 kHz channels) using 64 QAM. 64-FFT provides Mb/s in just 265 kHz modulation bandwidth with 64 QAM. Modulation code set (BPSK to 64QAM and r=1/2 to r=3/4) can be adapted for individual nodes based on range and interference or channel conditions Extremely degraded RF Transceiver phase noise can be implemented for BPSK and QPSK. 16-FFT has -121 dBm Ideal-Receiver sensitivity in AWGN for BPSK r=1/2, for 75kHz occupied channel (add noise figure and implementation loss to it). 64-FFT 64-QAM r=3/4 achieves -98 dBm sensitivity in a 265kHz occupied channel (based on IEEE802.11a receiver sensitivity definition). doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 9 Basic TDD frame structure For Mesh networking, existing mesh optimized frame structures and protocols can be used. IEEE802.11a type PPDU frame format (packets) can be used. For a Star network, a Frame can comprise of a short Down-link segment followed by long Up-link segment, followed by a short Slotted Contention segment. Local node to coordinator or Mesh communication segment can be allocated within this frame. Each node can transmit 2 or more OFDM symbols per frame. At 4000 OFDM symbols per second (for a 250 us OFDM symbol duration), up to nearly 2000 nodes can communicate per second using a 1-second TDD frame structure. A super-frame of say 60 frames can support up to 120,000 nodes over 60 seconds. With 10kHz subcarrier spacing (125us OFDM symbols), up to 240,000 nodes can transmit in 60s. Each frame, a node can send 288 uncoded data bits in a 64-QAM 64-FFT OFDM symbol in a 500 us transmit burst interval that includes one OFDM training symbol. More OFDM symbols can be allocated to a node for a larger payload. doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide FFT Down-link OFDM Subcarriers (also up-link OFDM subcarriers for 2 nd or greater symbol number) Frequency F = 5 kHz 75 kHz Over 15 sub-carriers (incl null SC) Pilot SC Data SC Null SC N SD : # data subcarriers12 N SP : # pilot subcarriers2 N ST : # total subcarriers14 F : subcarrier frequency spacing 5 kHz Occupied bandwidth75 kHz Channel spacing100 kHz doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide FFT Down-link OFDM parameters (also up-link OFDM subcarriers for 2 nd or greater symbol number) N SD : # data subcarriers48 N SP : # pilot subcarriers4 N ST : # total subcarriers52 F : subcarrier frequency spacing 5 kHz Occupied bandwidth265 kHz Channel spacing400 kHz Re-use IEEE802.11a parameters, scaled for 5kHz subcarrier spacing: Keep structure of Short and Long training sequences of Down-link as in IEEE802.11a/g. For 16-FFT OFDM Down-link Short and Long training sequences, use corresponding subcarriers 7 to +7 of IEEE802.11a/g doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide FFT Up-link OFDM Subcarriers for 1 st Symbol Frequency F = 5 kHz 75 kHz Over 15 sub-carriers (incl null SC) All reference subcarriers Null SC N SD : # data subcarriers0 N SP : # pilot subcarriers14 (reference subcarriers with known randomized phases) N ST : # total subcarriers14 F : subcarrier frequency spacing 5 kHz Occupied bandwidth75 kHz Channel spacing100 kHz For DQPSK or DPSK modulation, these reference SCs can be the starting symbol doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide FFT Up-link OFDM Subcarriers for 2 nd symbol and beyond Frequency F = 5 kHz 75 kHz over 15 sub-carriers (incl null SC) Data SC Null SC N SD : # data subcarriers12 N SP : # pilot subcarriers2 N ST : # total subcarriers14 F : subcarrier frequency spacing 5 kHz Pilot SC In order to mitigate fast fading due to moving traffic in an urban canyon or alley, the positions of the 2 pilots can be cyclically shifted over the 7 subcarriers on each side of the null subcarrier each OFDM symbol. This will allow a slow but continuous channel estimation. For DQPSK or DPSK modulation, pilots are not required doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 14 Down-link OFDM Symbol Up-link OFDM Symbol Time T FFT = 200us T CP1 = 50usT GUARD = 25us Time T FFT = 200usT CP = 50us T SIGNAL = 250us Training symbol of 250us Node # n-1Node # n... Data symbol Node # n-1 T CP2 = 25us (short cyclic prefix) doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 15 For 3 or more Up-link OFDM Symbols for a node Time T FFT = 200us T CP1 = 50usT GUARD = 25us T SIGNAL = 250us Node # n-1Node # n... T CP2 = 25us Node # n T CP = 50us Trainning symbol for channel estimation Data symbols 25us guard (blank) interval kept for nodes timing error margin T CP1 is used for Base Station receiver AGC doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 16 Timing Parameters T FFT : IFFT & FFT interval200 us T SIGNAL : OFDM symbol duration (up-link & down-link) 250 us T GUARD : up-link guard interval (no signal)25 us (used for timing error margin) T CP1 : up-link cyclic prefix interval50 us (used for base station AGC) T CP2 : up-link cyclic prefix interval25 us (between 1 st and 2nd symbols per node) T CP : down-link cyclic prefix interval50 us (also between up-link 3 rd and 4 th symbols and beyond for the same node) T SHORT : down-link short training sequence500 us T LONG : down-link long training sequence500 us doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 17 Timing and Synchronization For Mesh networking, existing mesh optimized frame structures and protocols can be used. For a Star network, each node is pre-allocated a 2 or larger OFDM symbol slot in every 1-sec frame, or in every 60-sec super frame, or once over a larger time interval. Initial entry or out-of-turn access is done through a contention process. After entry, an ISI-free slot is allocated (at least one OFDM symbol slot before and after the nodes packet are empty). Base Station then informs the node the amount of time-shift required. After the time- shift is done to align the OFDM packet correctly, the node can be allocated a another starting slot that can immediately follow the packet of another node (without a blank symbol in between). This is the basic ranging mechanism. At each node, a 32kHz crystal oscillator keeps running continuously, and its frequency error is calibrated regularly with respect to the base station frame timing. A node that transmits once each frame, has to maintain a timing accuracy that is better than the 25us guard interval between symbols of different nodes. This has to be done after considering the propagation delay between the node and the base station. Combined with the above mentioned ranging mechanism and 32kHz oscillator calibration, the slot accuracy improves to better than a microsecond. doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide kb/s GFSK BT=0.5, for 2-ray channel with 1us relative delay Channel Response Signal present at channel peak. BER: no impact from Eye distortions for this specific situation. No distortion to Eye. doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide kb/s GFSK BT=0.5, for 2-ray channel with 1us relative delay Channel Response Signal present at channel null. Large drop in received signal power! Flatter fading! BER: increased impact from Eye distortions. Significant impact on zero-crossing jitter dB increased S/N required due to distortion, jitter and fading. doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide kb/s GFSK BT=0.5, for 2-ray channel with 5us relative delay Signal present at channel peak. BER: small impact from Eye distortions for this specific case. Reduced Eye opening. Channel Response doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide kb/s GFSK BT=0.5, for 2-ray channel with 5us relative delay Signal present at channel null. Significant drop in received signal power! Flatter fading! BER: large impact from Eye distortions and jitter. Extremely reduced Eye opening that will require huge increase in S/N. Channel is not usable for reliable communication. Channel Response doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide kb/s GFSK BT=0.5, for 2-ray channel with 10us relative delay Signal present at channel peak. Frequency selective fading occurs. BER: very large due to fully closed Eye. High S/N will not help. Channel is not usable. Channel Response doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide kb/s GFSK BT=0.5, for 2-ray channel with 10us relative delay Signal present at channel null. Frequency selective fading occurs. BER: very large due to fully closed Eye. High S/N will not help. Channel is not usable. Channel Response doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 24 Worst case 16-FFT OFDM Subcarrier EVMs for 2-ray multi-path, 10us relative path delay, equal powers Only one subcarrier is destroyed in this 2-ray channel null. Packet is recovered error-free using FEC. doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 25 PA output spectrum for 16-FFT OFDM Red graph: without PA non-linearity ACPR = -45 dBc in adjacent 100kHz channel using 80kHz wide receiver channel filter Blue graph: PA with 6.5 dB backoff (saturation power to rms transmitted power ratio), Rapps model, rho=2 EVM = dB with 64-QAM ACPR = -33 dBc in adjacent 100kHz channel using 80kHz wide receiver channel filter doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 26 Adjacent Channel Power for 100kb/s GFSK and 288kb/s OFDM Relative power in 150kHz wide 5 th order Butterworth bandpass GFSK receiver channel filter. Signal: 100kb/s GFSK BT=0.5 Relative power in 80kHz wide 5 th order Eliptic bandpass OFDM receiver channel filter. Signal: 16-FFT OFDM, 5kHz subcarriers, 14 subcarriers used, Rapps PA doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 27 Application Issues Will SUN cover gas and water meters or will they be separate networks with mostly battery operated nodes. If separate, are they covered under 15.4g ? Should 3 co-located networks be supported (water, gas, electricity ?) If yes, should they operate in 868MHz band in Europe with only 600kHz available spectrum. Will this dictate global channel spacing or is that band specific (e.g. 915MHz and 2.4GHz bands can allow wider channels). Should Star, Mesh or both be supported. What is the transmitted power and expected battery life of battery operated nodes How many transmissions per day, how many payload bits per transmission Protocol and synchronization overheads Are transmit-only devices supported? If yes, whats the impact on frame structure & protocol doc.: IEEE g Submission May 1, 2009 Rishi Mohindra, MAXIM Slide 28 Conclusions Raw data rate Modulation bandwidth Ideal Sensitivity Max multipath delay spread for 2-ray equal power channel Required Adjacent Channel frequency separation for 33 dBc ACPR # nodes talking in 1 second per base station Frequency domain interleaving for multipath frequency selective fading Relative battery life (data rate adjusted for identical channel spacing for