1. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 1 Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [What is really fundamental?] Date Submitted: [12 September, 2004] Source: [A. Batra, J. Balakrishnan, A. Dabak, S. Lingam] Company [Texas Instruments] Address [12500 TI Blvd, MS 8649, Dallas, TX 75243] Voice:[214-480-4220], FAX: [972-761-6966], E-Mail:[batra@ti.com] Re: [FYI] Abstract: [This document examines what is fundamental.] Purpose:[For discussion by IEEE 802.15 TG3a.] Notice:This document has been prepared to assist the IEEE P802.15. 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 P802.15.

2. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 2 What is Really Fundamental? Anuj Batra, Jaiganesh Balakrishnan, Anand Dabak, Srinivas Lingam Texas Instruments 12500 TI Blvd, M/S 8649 Dallas, TX 75204 September 12, 2004

3. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 3 Motivation  This presentation is going to look at the some of the “fundamental” issues concerning both proposals: 1. Is the 6-dB gap for 480 Mbps MB-OFDM a fundamental gap? 2. Implementation losses associated with the DS-UWB.

4. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 4 Is the 6 dB Gap Fundamental?

5. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 5 Fundamental Concepts in 802.15.3a?  According to Document 15-04/022r0: “Rayleigh fading for MB-OFDM cannot be mitigated by any amount of added signal processing”  High rate modes degraded by 6 dB or more relative to AWGN  “Rayleigh fading performance does not improve with process technology or added digital processing”  The DS-UWB authors have REPEATEDLY stated that this is a fundamental problem with Multi-band OFDM.  Question:  Is this a fundamental concept that we cannot violate?

6. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 6 Latest MB-OFDM Proposal (1)  Guard tone mapping was added to the MB-OFDM proposal in order to address concerns raised by the Task Group.  Exact mapping of tones is shown below:  Equivalent to:  Frequency-domain spreading for the lower rates of MB-OFDM system.  Excess BW used by single-carrier systems.

7. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 7 Latest MB-OFDM Proposal (2)  Mathematically, the mapping of the data on to the Guard Tones can be written as follows: where P n are the Guard Tones and C n are the Data Carrier Tones.  Let us now consider the effect that the Guard Tones has on the system performance of the MB-OFDM solution for the 110, 200, and 480 Mbps modes.

8. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 8 Simulation Parameters  Assumptions:  System as defined in 03/268.  Clipping at the DAC (PAR = 9 dB).  Finite precision ADC (4 bits for 110, 200 Mbps and 5 bits for 480 Mbps).  No attenuation on the Guard Tones.  Degradations incorporated:  Front-end filtering.  Multi-path degradation.  Shadowing.  Clipping at the DAC.  Finite precision ADC.  Crystal frequency mismatch (  20 ppm @ TX,  20 ppm @ RX).  Channel estimation.  Carrier/timing offset recovery.  Carrier tracking.  Packet acquisition.

9. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 9 System Performance with Guard Tones  The distance at which the Multi-band OFDM system can achieve a PER of 8% for a 90% link success probability is tabulated below: *Includes losses due to front-end filtering, clipping at the DAC, ADC degradation, multi-path degradation, channel estimation, carrier tracking, packet acquisition, etc. Range * AWGNCM1CM2CM3CM4 110 Mbps21.5 m New: 12.0 m Original: 11.4 m New: 11.4 m Original: 10.7 m New: 12.3 m Original: 11.5 m New: 11.3 m Original: 10.9 m 200 Mbps14.8 m New: 7.4 m Original: 6.9 m New: 7.1 m Original: 6.3 m New: 7.5 m Original: 6.8 m New: 6.6 m Original: 4.7 m 480 Mbps9.1 m New: 3.2 m Original: 2.9 m New: 3.0 m Original: 2.6 m N/A

10. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 10 Improvement with Guard Tones  System performance improves for both channel models:  CM1: 2.9 m  3.2 m (+0.9 dB improvement).  CM2: 2.6 m  3.0 m (+1.2 dB improvement).  Using the fact that shadowing contribution is ~3.9 dB to the overall degradation, the gap from AWGN to the 480 Mbps mode using Guard Tones has already been reduced by ~0.8 dB!  This analysis shows that the Rayleigh fading for MB-OFDM can be mitigated by additional signal processing.  Gap of 6 dB in fading is NOT a fundamental issue.

11. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 11 Examine Implementation Losses Associated with the DS-UWB Proposal

12. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 12 Implementation Losses (1)  Let’s review the performance results given by the DS-UWB authors:  Simulation results used to show an implementation loss of 0.8 dB, but now show an implementation loss of 0.4 dB.  Some examples of implementation loss:  Degradations due to finite-precision ADC.  Degradations due to timing synchronization errors.  Degradations due to carrier frequency synchronization errors.  Degradations due to channel estimation errors.  Degradations due to finite-precision effects in the digital domain.  In contrast, the MB-OFDM proposal has shown simulation results with a 2.5 dB of implementation loss for 110 Mbps. Assumptions / Value DS-UWB AWGN Link Budget Table DS-UWB AWGN Simulation Results (1)* DS-UWB AWGN Simulation Results (2)** Range18.3 meters22.2 meters23.4 meters Noise Figure6.6 dB Implementation Loss2.5 dB * Not Given, Inferred Value = 0.8 dB * Not Given, Inferred Value = 0.4 dB * Results extracted from IEEE 802.15-04/099r2. ** Results extracted from IEEE 802.15-04/483r2.

13. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 13 Implementation Losses (2)  Question: Is it possible to design a system with an implementation loss of 0.4 dB?  In this presentation, we start by looking the degradations that are finite-precision ADC degradations and caused by timing synchronization errors.  In follow-up presentations, we hope to look at the some of the remaining categories:  Carrier-frequency synchronization errors.  Channel estimations errors.  Etc.

14. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 14 Degradations Due to a 3-bit ADC (1)  System model:  Simulation assumptions:  Considered: 3-bit and 20-bit ADC.  Channel: AWGN.  K = 6, R = 1/2 convolutional code (as specified by DS-UWB authors).  Rates: 110 and 200 Mbps.  Perfect channel estimation.  No frequency offset / 0 ppm crystal error.  No fixed-point effects other than ADC.

15. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 15 Degradations Due to a 3-bit ADC (2)  Simulation results:  At a BER = 10 –4, loss due to a 3-bit ADC is ~0.4 dB.  Still need to examine the performance degradations due to finite- precision ADCs in multi-path channel environments.

16. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 16 Nyquist Sampling  Nyquist theorem states that sufficient statistics can be obtained if and only if a signal is sampled at twice its largest bandwidth.  Is there a fundamental limitation in sampling a system with sub- Nyquist sampling?  Yes. The answer is Aliasing.  Aliasing may result in destructive interference:  Results in loss in base-band sampled signal energy.  Destroys the flatness of the signal spectrum and requires signal processing to invert the channel.

17. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 17 Sub-Nyquist Sampling  Assumptions for analyzing the impact of 1X sampling:  DS-UWB system with a chip rate = 1326 MHz, excess BW of 50%  a max BW of ~2 GHz.  AWGN Channel  0 ppm crystal mismatch between transmitter and receiver  UNKNOWN PROPAGATION DELAY  Ideal Rake to collect energy in all the paths.  Perfect equalization (which does not exist in practice) – DOES NOT INCLUDE IMPACT DUE TO ISI.  Maximum loss in signal energy can be as high as 1.25 dB.  This is a true FUNDAMENTAL limitation of chip-rate sampling for the DS-UWB system.

18. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 18 Additional Items for Analysis  We hope to have additional simulations / analysis in the future that examine the effects of the following items on implementation loss:  Carrier-frequency synchronization errors.  Channel estimations errors.  Etc.

19. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 19 Other Implementation Issues for DS-UWB  Are there any problems that may arise when multiple piconet overlaps?  Recall: all DS-UWB piconets use the same spreading code.  Only difference between piconets is a carrier-frequency offset of just 13 MHz.  Q: Is the 13 MHz carrier frequency spacing enough to separate multiple piconets?  A: No. As well will show, the resulting SINR (signal-to-interference and noise-ratio) appears to FADE every 6 chips or so.  Thus, the desired piconet sees the equivalent of a highly-time selective fading channel.  Thus, the advantage of using a large BW channel is LOST for a DS-UWB system.  More details are shown on the next slide.

20. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 20 Multiple Piconets  DS-UWB System Assumptions:  Chip Rate of desired Piconet = 1326 MHz.  Chip Rate of Interfering Piconet = 1339 MHz.  Spreading factor of Interfering Piconet = 6 (i.e., 114 Mbps)  SNR of desired signal = 15 dB.  dint/dref = 1  AWGN channel for both desired and interfering piconet.  Zero sampling offset (Best case assumption, but not practical)  0 ppm crystal mismatch (Best case assumption, but not practical)  Instantaneous SINR shows deep fades.  This is a true FUNDAMENTAL limitation of using the same spreading code in the impulse radio like DS-UWB system.

21. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 21 Conclusions (1)  Rayleigh fading for the Multi-band OFDM system can be mitigated by additional signal processing.  So-called “Fundamental-Gap” of 6 dB in fading is NOT a fundamental issue after all.  DS-UWB authors show an implementation loss of 0.4 dB in their simulation/performance results:  Shown that a 3-bit ADC results in a 0.4 dB loss when compared to 20-bit ADC. Shown that a timing synchronization error can be high as 1.25 dB.  Finite-precision ADC degradations and timing synchronization errors are ONLY TWO OF THE COMPONENTS that make up implementation loss.

22. doc.: IEEE 802.15-04/0533r0 Submission September 2004 Anuj Batra et al., Texas InstrumentsSlide 22 Conclusions (2)  In addition, we have shown that the DS-UWB system experiences an equivalent highly-selective fading channel when multiple piconets overlap.  The so-called “ultra-wideband BW advantage” (using a very large BW) disappears for the DS-UWB system.