doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 1 Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [DS-UWB Proposal Update] Date Submitted: [July 2004] Source: [Reed Fisher(1), Ryuji Kohno(2), Hiroyo Ogawa(2), Honggang Zhang(2), Kenichi Takizawa(2)] Company [ (1) Oki Industry Co.,Inc.,(2)National Institute of Information and Communications Technology (NICT) & NICT- UWB Consortium ]Connector’s Address [(1)2415E. Maddox Rd., Buford, GA 30519,USA, (2)3-4, Hikarino-oka, Yokosuka, , Japan] Voice:[(1) , (2) ], FAX: [(2) ], ] Source: [Michael Mc Laughlin] Company [decaWave, Ltd.] Voice:[ ], FAX: [-], Source: [Matt Welborn] Company [Freescale Semiconductor, Inc] Address [8133 Leesburg Pike Vienna, VA USA] Voice:[ ], Re: [] Abstract:[Technical update on DS-UWB (Merger #2) Proposal] Purpose:[Provide technical information to the TG3a voters regarding DS-UWB (Merger #2) Proposal] 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

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 2 Outline Merger #2 Proposal Overview –DS-UWB + Common Signaling Mode (CSM) + MB-OFDM Complexity/Scalability of UWB implementations Spectral control options for DS-UWB Performance

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 3 Overview of DS-UWB Proposal Merged Proposal #2 has a fundamental goal of DS- UWB and MB-OFDM harmonization & interoperability through a Common Signaling Mode (CSM) –High rate modes using either DS-UWB or MB-OFDM Best characteristics of both approaches with most flexibility A piconet could have a pair of DS and a pair of MB devices –CSM waveform provides control & interoperation between DS-UWB and MB-OFDM All devices work through an MAC –User/device only sees common MAC interface –Hides the actual PHY waveform in use

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 4 The Common Signaling Mode: What Is The Goal? The common signaling mode (CSM) allows the MAC to arbitrate between multiple UWB PHYs –It is an “etiquette” to manage peaceful coexistence between the different UWB PHYs –Multiple UWB PHYs will exist in the world DS-UWB & MB-OFDM are first examples –CSM improves the case for international regulatory approval Common control mechanism for a multitude of applications Planned cooperation (i.e. CSM) gives far better QoS and throughput than allowing un-coordinated operation and interference –CSM provides flexibility/extensibility within the IEEE standard Allows future growth & scalability Provides options to meet diverse application needs Enables interoperability and controls interference

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 5 What Does CSM Look Like? One of the MB-OFDM bands! MB-OFDM (3-band) Theoretical Spectrum Proposed Common Signaling Mode Band (500+ MHz bandwidth) 9-cycles per BPSK “chip” Frequency (MHz) DS-UWB Low Band Pulse Shape (RRC) 3-cycles per BPSK “chip”

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 6 CSM Specifics We have designed a specific waveform for the CSM –BPSK modulation for simple and reliable performance –Length 24 spreading codes using 442 MHz chip rate –Harmonically related center frequency of 3978 MHz –Rate ½ convolutional code with k=6 –Provides 9.2 Mbps throughput Extendable up to 110 Mbps using variable code and FEC rates MAC works great with CSM –CSM can be used for control and beaconing –Negligible impact on piconet throughput (beacons are <1%) Requires negligible additional cost/complexity for either radio –MB-OFDM already has a DS mode that is used for synchronization This proposal is based on both MB-OFDM and DS-UWB operating with a 26 MHz cell-phone crystal –Very low cost yet terrific phase-noise and accuracy (see GSM spec)

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 7 Overview of DS-UWB Proposal DS-UWB proposed as a radio for handheld with – low-cost, – ultra high-rate, – ultra low-power, BPSK modulation using variable length spreading codes –Scales to 1+ Gbps with low power - essential for mobile & handheld applications Much lower complexity and power consumption

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 8 Overview of DS-UWB Proposal 11 … … Wavelets are modulated with BPSK or QPSK Symbol is made with an N-chip code sequence Code is ternary (+1, 0, -1) Two wide 50%-bandwidth contiguous bands Each captures unique propagation benefits of UWB Bandwidth and Center Frequency Programmable Low band provides long wavelet High band provides short wavelet Wavelet = 3 cycles, packed back-to-back N-chips GHz Result is Not-spiky in either Time or Frequency Domain time volts

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 9 DS-UWB Signal Generation Transmitter blocks required to support optional modes Scrambler K=6 FEC Encoder Conv. Bit Interleaver Input Data K=4 FEC Encoder 4-BOK Mapper Bit-to-Code Mapping Pulse Shaping Static Center Frequency Gray or Natural mapping Data scrambler using 15-bit LFSR (same as ) Constraint-length k=6 convolutional code K=4 encoder can be used for lower complexity at high rates or to support iterative decoding for enhanced performance (e.g. CIDD) Convolutional bit interleaver protects against burst errors Variable length codes provide scalable data rates using BPSK Support for optional 4-BOK modes with little added complexity

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 10 Data Rates Supported by DS-UWB Data RateFEC RateCode LengthRange (AWGN) 28 Mbps½2429 m 55 Mbps½1223 m 110 Mbps½618.3 m 220 Mbps½313 m 500 Mbps¾27.3 m 660 Mbps124.1 m 1000 Mbps¾15.1 m 1320 Mbps112.9 m Similar Modes defined for high band

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 11 DS-UWB Architecture Is Highly Scaleable DS-UWB provides low & scalable receiver complexity –ADC can range from 3 bits to 1 bit for super-low power implementation –Rake pipeline & DFE can be optimized to trade off power & cost in multipath , 5 500, Mbps Time duration of DFE scales (shrinks) at shorter range – higher rates. –FEC can scale w/data rate (k=6 & k=4) or be turned-off for ultra low power –DFE effectiveness and simplicity proven in shipping chips – 3% of area Pre-Select Filter LNA LPF GA/ VGA GA/ VGA ADC at Chip Rate ADC at Chip Rate Rake DFE LPF Cos Sin Synch/ Track Logic Agile Clock 1 to 3 bits ADC Resolution 1-16 Rake Fingers (or more) Variable Rate FEC (or no FEC) De-interleave & FEC Decode Variable Equalizer Span

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 12 UWB System Complexity & Power Consumption Two primary factors drive complexity & power consumption –Processing needed to compensate for multipath channel –Modulation requirements (e.g. low-order versus high-order) DS-UWB designed to operate with simple BPSK modulation for all rates –Receiver functions operate at the symbol rate –Optional 4-BOK has same complexity and BER performance MB-OFDM operates at fixed 640 Mbps (raw) –Designed to operate at high rate, then use carrier diversity (redundancy) and/or strong FEC to combat multipath fading –Diversity not used above 200 Mbps

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 13 Fundamental Design Approach Differences Signal bandwidth leads to different operating regimes –DS-UWB uses GHz bandwidth –MB-OFDM data BW is MHz (100 tones x MHz/tone) Modulation bandwidth induces different fading statistics –DS-UWB (single carrier UWB) results in frequency-selective fading with relatively low power fluctuation (variance) –MB-OFDM (multi-carrier) creates a bank of parallel channels that experience flat fading with a Rayleigh distribution (deep fades) Motivations for different choices –Different energy capture mechanism (rake vs. FFT) –Different ISI compensation (time vs. frequency domain EQ) These fundamental differences affect both complexity & flexibility –Significant impact on implementation, especially at high rates

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 14 Analog Complexity Equivalent analog components have similar complexity MB-OFDM Analog Components DS-UWB Analog Components Similar characteristics -Antenna -Pre-select filter -LNA -Antenna -Pre-select filter -LNA Different characteristics -Switchable UNII filter -Hopping Frequency Gen -Band filter to reject adjacent channels -Static UNII filter -Static Frequency Gen -Band filter with no adjacent channels

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 15 Implications of Switchable UNII Filter (slide copied from Doc 03/141r3,p12) MB-OFDM is proposed to use the UNII band for Band Group 2 If the operating BW includes the U-NII band, then interference mitigation strategies have to be included in the receiver design to prevent analog front-end saturation. Example: Switchable filter architecture. –When no U-NII interference is present, use standard pre-select filter. –When U-NII interference is present, pass the receive signal through a different filter (notch filter) that suppresses the entire U-NII band.  Problems with this approach:  Extra switches  more insertion loss in RX/TX chain.  More external components  higher BOM cost.  More testing time.

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 16 Band-Select Filter Complexity MB-OFDM filter complexity depends on requirements to reject adjacent-band signal energy Depends on whether design is using the guard tones for real data or just PN modulated noise DS-UWB Filter Uses single fixed bandwidth – filter provides rejection for OOB noise & RFI Bandwidth of DS-UWB > 1500 MHz Data tones Guard tones

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 17 MB-OFDM Band-Select Filter Complexity If guard tones are used for useful data, band filter must have very steep cut-off –Transition region is very narrow –Only 5 un-modulated tones between bands (~21 MHz) SOP performance also affected by filter design – rejection of adjacent band MAI for SOP If guard tones not used for data, then filter constraint is relaxed –Transition region is a wider (15 tones ~62 MHz) –Energy in guard band is distorted (not useful) –May not meet FCC UWB requirement for 500 MHz Tight filter constraint Relaxed filter constraint Filter must reject MAI for SOP Data tones Guard tones Filter response

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 18 Comparison of DS-UWB to MB-OFDM Digital Baseband Complexity for PHY Gate count estimates are based on MB-OFDM proposal team methodology detailed in IEEE Document 03/449r2 –Gate counts converted to common clock (85.5 MHz) for comparison Explicit MB-OFDM gates counts have only been reported by proposers for a 110/200 Mbps implementation –Other estimates of MB-OFDM Viterbi decoder and FFT engine are provided in IEEE Document 03/343r0 Estimates for MB-OFDM 480 Mbps mode complexity are based on scaling of FFT engine, equalizer and Viterbi decoder –MB-OFDM estimates of 480 Mbps power available in 03/268r3 –Details available in IEEE Document 04/164r0 Estimates for MB-OFDM 960 Mbps mode details are based on linear scaling of decoder and FFT engine to 960 Mbps –Assumes 6-bit ADC for 16-QAM operation DS-UWB gate estimates are detailed in IEEE Document 03/099r4 –Methodology for estimating complexity of 16-finger rake, equalizer and synchronization blocks are per MB-OFDM methodology

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 19 DS-UWB & MB-OFDM Digital Baseband Complexity Gate counts are normalized to 85.5 MHz Clock speeds to allow comparison –Based on methodology presented by MB-OFDM proposal team (03/449r3) –Other details of gate count computations in Documents 04/099 and 04/256r0 Component MB-OFDM (Doc 03/268r3 or 03/343r1) 110 Mbps DS-UWB 16-Finger Rake 220 Mbps Raw 3-Bit ADC DS-UWB 32-Finger Rake 220 Mbps Raw 3-Bit ADC Matched filter Rake [DS] or FFT [OFDM] 100K26K45K Viterbi decoder108K54K Synchronization 247K (Freq Domain) 30K Channel estimation24K Other Miscellaneous including RAM 30K Equalizer20K Total 85.5 MHz455K184K203K

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 20 Digital Baseband Complexity Comparison at ~1 Gbps Assumptions: MB-OFDM using 6-bit ADC, FFT is 2.25x & Viterbi is 4x of low rate. *DS-UWB operating with no FEC at Gbps Component MB-OFDM 960 Mbps using 16-QAM DS-UWB 2-Finger Rake Gbps 3-bit ADC width DS-UWB 5-Finger Rake Gbps 3-bit ADC width Matched filter [rake] or FFT 225K26K45K Viterbi decoder432K 0K* Synchronization 297K (Freq Domain) 30K Channel estimation24K Other Miscellaneous including RAM 30K Equalizer50K Total 85.5 MHz954K160K179K

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 21 Optional Improvement for Interference Mitigation (Approach 1):Analog type of SSA- Notch generation by using a simple analog delay line: Example ： Just Two taps delay line ＋ ＋ The output signal x(t) is given by By assuming that coefficients w 0 and w 1 is time- invariant, then its signal in frequency domain is given by Now, we set w 0 =1 and w 1 =a (a is in real value), we obtain A notch is generated at a frequency f n where |X(f n )| 2 =0, then The solutions are given by, （ m=1,2,3,… ） As you can see, the coefficient a takes +1 or -1. It leads simple implementation. where p(t) is a pulse signal, and  is delayed time by a delay line D. however, the coefficient a can take only real value. Therefore, D D w0w0 w1w1 x(t)x(t) p(t)p(t) The right figure is an example; a is set to 1 and  is set at 0.116nsec.

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 22 DS-UWB systems X Spreading codeCarrier frequency x(t) fcfc c(t) b(t) Tx model 1 XX p(t) Pulse signal 4.3GHz (EES) c(t)=[ ] 4.3GHz (EES) c l (t), c(t)=[ ] X x(t) fcfc c(t) Tx model 2 XX p(t) X long code c l (t) b(t) Assumption: Chip rate of a long code is the same as bit rate. Narrow and Repetitive (Scrambler) Spreading codeCarrier frequencyPulse signal Narrow and Repetitive Example: Note: These notches are diminished by a bi-phase modulation. Optional Improvement for Interference Mitigation (Approach 2): Analog type of SSA- Notch generation by using a spreading code

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 23 Optimization of coding rate and spreading factor Data rateFEC RateCode LengthRange (AWGN) 110Mbps1/2618.3m 220Mbps1/2312.9m Original VS-DS-UWB Data rateFEC RateCode LengthRange (AWGN) 110Mbps 1/4313.9m 1/3416.1m 3/4916.9m 220Mbps 1/3211.4m 2/3412.9m The other combinations FEC Rate=1/2: [53,75] FEC Rate=1/3: [47,53,75] FEC Rate=1/4: [53,67,71,75] > >= Constraint length is fixed to 6 (Have you already optimized the combinations ?)

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide Mbps90% Outage Range (meters) Mean of Top 90% Range (meters) CM CM CM CM Simulation Includes: 16 finger rake with coefficients quantized to 3-bits 3-bit A/D (I and Q channels) RRC pulse shaping DFE trained in < 5us in noisy channel Front-end filter for Tx/Rx dB Noise Figure Packet loss due to acquisition failure Multipath Performance for 110 Mbps

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 25 Multipath Performance for 220 Mbps Simulation Includes: 8 finger (16 finger) rake with coefficients quantized to 3-bits 3-bit A/D (I and Q channels) RRC pulse shaping DFE trained in < 5us in noisy channel Front-end filter for Tx/Rx dB Noise Figure Packet loss due to acquisition failure 220 Mbps 90% Outage Range (m) 8-finger rake 90% Outage Range (m) 16-finger rake Mean of Top 90% Range (m) 8-finger rake Mean of Top 90% Range (m) 16-finger rake CM CM CM

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 26 Multipath Performance for 500 Mbps Simulation Includes: 16 finger rake with coefficients quantized to 3-bits 3-bit A/D (I and Q channels) RRC pulse shaping DFE trained in < 5us in noisy channel Front-end filter for Tx/Rx dB Noise Figure Packet loss due to acquisition failure 500 Mbps90% Outage Range (m) Mean of Top 90% Range (m) CM CM

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 27 AWGN SOP Distance Ratios Test Distance 1 Interferer Distance Ratio 2 Interferer Distance Ratio 3 Interferer Distance Ratio 110 Mbps15.7 m Mbps11.4 m Mbps5.3 m AWGN distances for low band High band ratios expected to be lower –Operates with 2x bandwidth, so 3 dB more processing gain

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide Mbps 1 Interferer Distance Ratio 2 Interferer Distance Ratio 3 Interferer Distance Ratio CM CM CM Multipath SOP Distance Ratios Test Transmitter: Channels 1-5 Single Interferer: Channels 6-10 Second Interferer: Channel 99 Third Interferer: Channel 100 High band ratios expected to be lower (3 dB more processing gain)

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 29 Conclusions Our vision: A single PHY with multiple modes to provide a complete solution for TG3a –Base mode that is required in all devices, used for control signaling: “CSM” for beacons and control signaling –Higher rate modes also required to support 110 & 200+ Mbps: –Compliant device can implement either DS-UWB or MB- OFDM (or both) Increases options for innovation and regulatory flexibility to better address all applications and markets DS-UWB is shown to have equal or better performance to MB-OFDM in all modes and multipath conditions – for a fraction of the complexity & power

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 30 Back-up slides

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 31 DS-UWB systems Notch generation by using a spreading code X Spreading code Carrier x(t) fcfc c(t) b(t) Frequency domain Output spectrum is given by convolution Example: Tx model XX p(t) Pulse signal Spectrum of a pulse signal Spectrum of a spreading code Convolution

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 32 Experimental result by UWB Test bed Notch generation by using a spreading code MATLAB results UWB testbed outputs

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 33 All-Digital Architecture DS-UWB Receiver DS-UWB Digital architecture provides scalable receiver complexity –ADC can range from 3 bits to 1 bit for super-low power implementation –Rake & DFE can be optimized to trade off power & cost in multipath –FEC can scale data rate or be turned-off for low power operation –DFE effectiveness and simplicity proven in shipping chips Pre-Select Filter LNA LPF GA/ VGA GA/ VGA ADC at Chip Rate ADC at Chip Rate Rake DFE LPF Cos Sin Synch/ Track Logic Agile Clock 1 to 3 bits ADC Resolution 1-16 Rake Fingers (or more) Variable Rate FEC (or no FEC) De-interleave & FEC Decode Variable Equalizer Span

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 34 Scalability to Varying Multipath Conditions Critical for handheld (battery operated) devices –Support operation in severe channel conditions, but also… –Ability to use less processing (& battery power) in less severe environments Multipath conditions determine the processing required for acceptable performance –Collection of time-dispersed signal energy (using either FFT or rake processing) –Forward error correction decoding & Signal equalization Poor: receiver always operates using worst-case assumptions for multipath –Performs far more processing than necessary when conditions are less severe –Likely unable to provide low-power operation at high data rates ( Mbps) DS-UWB device –Energy capture (rake) and equalization are performed at symbol rate –Processing in receiver can be scaled to match existing multipath conditions MB-OFDM device –Always requires full FFT computation – regardless of multipath conditions –Channel fading has Rayleigh distribution – even in very short channels –CP length is chosen at design time, fixed at 60 ns, regardless of actual multipath

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 35 Interference Issues (1) Hopped versus non-hopped signal characteristics –ITS and FCC studies are underway Goal is to see if interference characteristics of MB-OFDM are acceptable for certification (using DS-UWB/noise/IR for comparison) Use of PN-modulation to meet 500 MHz BW –Recent statements by NTIA emphasize importance of minimum –Desire is to ensure protection for restricted bands –DS-UWB bandwidth is determined by pulse shape and pulse modulation Spectrum exceeds 1500 MHz –MB-OFDM bandwidth for data and pilot tones is 466 MHz, guard tones are used to increase bandwidth to 507 MHz Guard tones “carry no useful information”, only to meet BW req’t. See authors statements in /267r1 (July 2003, page 12)

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 36 NTIA Comments on Using Noise to meet FCC 500 MHz BW Requirement NTIA comments specifically on the possibility that manufacturer would intentionally add noise to a signal in order to meet the minimum FCC UBW 500 MHz bandwidth requirements: “Furthermore, the intentional addition of unnecessary noise to a signal would violate the Commission’s long-standing rules that devices be constructed in accordance with good engineering design and manufacturing practice.” And: –“It is NTIA’s opinion that a device where noise is intentionally injected into the signal should never be certified by the Commission.” Source: NTIA Comments (UWB FNPRM) filed January 16, 2004 available at

doc.: IEEE /140r5 Submission July 2004 Kohno NICT, Welborn Freescale, Mc Laughlin decaWave Slide 37 FCC Rules Regarding Unnecessary Emissions FCC Rules in 47 CFR Part 15 to which NTIA refers: “§ General technical requirements. (a) An intentional or unintentional radiator shall be constructed in accordance with good engineering design and manufacturing practice. Emanations from the device shall be suppressed as much as practicable, but in no case shall the emanations exceed the levels specified in these rules.”