INFLUENCE OF QUANTIZATION NOISE ON DFT BASED DSP ALGORITHMS

Similar documents
Myth or fact? You can increase the force of your solenoid by removing turns. (Part 2 of 3) Paul H. Schimpf Eastern Washington University

Back EMF Design of an AFPM Motor using PCB Winding by Quasi 3D Space Harmonic Analysis Method

Measurement of RMS values of non-coherently sampled signals. Martin Novotny 1, Milos Sedlacek 2

The Periodic Ambiguity Function Its Validity and Value

N2-1. The Voltage Source. V = ε ri. The Current Source

Discussion #7 Example Problem This problem illustrates how Fourier series are helpful tools for analyzing electronic circuits. Often in electronic

Low-Complexity Time-Domain SNR Estimation for OFDM Systems

Design of FIR Filter using Filter Response Masking Technique

A New Buck-Boost DC/DC Converter of High Efficiency by Soft Switching Technique

where and are polynomials with real coefficients and of degrees m and n, respectively. Assume that and have no zero on axis.

Sliding Mode Control for Half-Wave Zero Current Switching Quasi-Resonant Buck Converter

(2) The resonant inductor current i Lr can be defined as, II. PROPOSED CONVERTER

10! !. 3. Find the probability that a five-card poker hand (i.e. 5 cards out of a 52-card deck) will be:

An Efficient Control Approach for DC-DC Buck-Boost Converter

Figure Geometry for Computing the Antenna Parameters.

Performance Analysis of Z-Source Inverter Considering Inductor Resistance

THE UNIVERSITY OF NEW SOUTH WALES. School of Electrical Engineering & Telecommunications

Experimental Investigation of Influence on Non-destructive Testing by Form of Eddy Current Sensor Probe

The phenomenon of incoherent signal interference

ISSN: [Reddy & Rao* et al., 5(12): December, 2016] Impact Factor: 4.116

Real-time Self Compensating AC/DC Digitally Controlled Power Supply

Analysis and Implementation of LLC Burst Mode for Light Load Efficiency Improvement

Analysis of Occurrence of Digit 0 in Natural Numbers Less Than 10 n

Trigonometry: Angles between 0 and 360

Analysis of a Fractal Microstrip Patch Antenna

ECEN326: Electronic Circuits Fall 2017

Discrepancies Between Euclidean and Spherical Trigonometry. David Eigen

Diagnosis method of radiated emission from battery management system for electric vehicle

Available online at ScienceDirect. Procedia Engineering 100 (2015 )

TECHNICAL REPORT: CVEL Maximum Radiated Emission Calculator: Power Bus EMI Algorithm. Chentian Zhu and Dr. Todd Hubing. Clemson University

Parameters of spinning AM reticles

Demosaicking using Adaptive Bilateral Filters

Radiation resistance

1 Performance and Cost

UNCERTAINTY ESTIMATION OF SIZE-OF-SOURCE EFFECT MEASUREMENT FOR 650 NM RADIATION THERMOMETERS

HYBRID FUZZY PD CONTROL OF TEMPERATURE OF COLD STORAGE WITH PLC

A Virtual Instrument for the Electric Power Monitoring in the Distributing Network

Analysis and Design of a 1MHz LLC Resonant Converter with Coreless Transformer Driver

Cyclic Constellation Mapping Method for PAPR Reduction in OFDM system

IEEE Broadband Wireless Access Working Group < Modifications to the Feedback Methodologies in UL Sounding

Spectrum Sharing between Public Safety and Commercial Users in 4G-LTE

ABSTRACTT FFT FFT-' Proc. of SPIE Vol U-1

Optic Cable Tracking and Positioning Method Based on Distributed Optical Fiber Vibration Sensing

51. IWK Internationales Wissenschaftliches Kolloquium International Scientific Colloquium

Chapter 9 Cascode Stages and Current Mirrors

ICAN meeting, Palaiseau, 11/10/2012 Some temporal aspects of Passive cavity-dumped enhancement for ICAN

Small Loop Antenna and Duality Theorem

Exercise 1 (MC Question)

Image Enhancement. To process an image so that output is visually better than the input, for a specific application.

Journal of Applied Science and Agriculture

Regeneration of frequency response functions from poles and zeros: a discussion with implications for cepstrum-based operational modal analysis

The mathematical structure known as a graph has the valuable feature of

Determination of The Winding Inductances Of A Two-Phase Machine.

DESIGN AND PARAMETRIC EVALUATION OF RECTANGULAR MICROSTRIP PATCH ANTENNA FOR GSM APPLICATION

Multiagent Reinforcement Learning Dynamic Spectrum Access in Cognitive Radios

Maximum Power Point Tracking Techniques for Efficient Photovoltaic Microsatellite Power Supply System

Design of A Circularly Polarized E-shaped Patch Antenna with Enhanced Bandwidth for 2.4 GHz WLAN Applications

BATTERY AND SUPERCAPACITOR CHARGER BASED ON LLC CONVERTER

WIRELESS SENSORS EMBEDDED IN CONCRETE

AUTO-TUNED MINIMUM-DEVIATION DIGITAL CONTROLLER FOR LLC RESONANT CONVERTERS

FFH and MCFH Spread-Spectrum Wireless Sensor Network Systems Based on the Generalized Approach to Signal Processing

Analysis of the Radiation Resistance and Gain of Full-Wave Dipole Antenna for Different Feeding Design

Analyze Power Supply Rejection Ratio of LDO Regulator Based on Accurate Small Signal Model

Analysis, Design, and Performance Evaluation of Asymmetrical Half-Bridge Flyback Converter for Universal-Line-Voltage-Range Applications

Throughput Performance of Quantized Proportional Fair Scheduling with Adaptive Modulation and Coding

A Novel Chaos Based ASK OOK Communication System

Controller Design of Discrete Systems by Order Reduction Technique Employing Differential Evolution Optimization Algorithm

Adaptive Line Enhancement and Spectrum Analysis D. W. TUFTS PROCEEDINGS LETTERS 169

School of Electrical and Computer Engineering, Cornell University. ECE 303: Electromagnetic Fields and Waves. Fall 2007

Short-Circuit Fault Protection Strategy of Parallel Three-phase Inverters

Proceedings of Meetings on Acoustics

stanchion paper roll mandrel base FIGURE P3-12 ramp

An Ultra Low Power Segmented Digital-to-Analog Converter

Synchronous Machine Parameter Measurement

ScienceDirect. Simplified Generation of Electromagnetic Field within EMC Immunity Test Area

Fault severity diagnosis of rolling element bearings based on kurtogram and envelope analysis

Development of Corona Ozonizer Using High Voltage Controlling of Produce Ozone Gas for Cleaning in Cage

Impact of bilateral filter parameters on medical image noise reduction and edge preservation

Assignment 0/0 2 /0 8 /0 16 Version: 3.2a Last Updated: 9/20/ :29 PM Binary Ones Comp Twos Comp

Statement of Works Data Template Version: 4.0 Date:

Volume 1, Number 1, 2015 Pages 1-12 Jordan Journal of Electrical Engineering ISSN (Print): , ISSN (Online):

IMAGE QUALITY ASSESSMENT USING THE MOTIF SCAN

Control Limits of Three-Phase AC Voltage Controller under Induction Motor Load A. I. Alolah Ali M. Eltamaly R. M. Hamouda

ANALYSIS OF CIRCULAR MICROSTRIP ANTENNA ON THICK SUBSTRATE

A Study Of a Wide-Angle Scanning Phased Array Based On a High-Impedance Surface Ground Plane

An SMC based RZVDPWM Algorithm of Vector Controlled Induction Motor Drive for Better Speed Response with Reduced Acoustical Noise

Synchronous Machine Parameter Measurement

OTA MEASUREMENTS OF SMALL RADIO TERMINALS BY A SPHEROIDAL COUPLER

Variance? which variance? R squared effect size measures in simple mediation models

A Transmission Scheme for Continuous ARQ Protocols over Underwater Acoustic Channels

Chapter 5: Trigonometric Functions of Angles

LLR Reliability Improvement for Multilayer Signals

4 Trigonometric and Inverse Trigonometric Functions

Spatial Coding Techniques for Molecular MIMO

ONE-WAY RADAR EQUATION / RF PROPAGATION

Digital Simulation of FM-ZCS-Quasi Resonant Converter Fed DD Servo Drive Using Matlab Simulink

Noise Attenuation Due to Vegetation

Experiments with the HoloEye LCD spatial light modulator

77 GHz ACC Radar Simulation Platform

Spread Spectrum Codes Identification by Neural Networks

Transcription:

XVIII IMEKO WORD COGRESS Metoogy fo Sustinbe Deveopment Septembe, 7, 6, Rio de Jneio, Bzi IFUECE OF QUATIZATIO OISE O DFT BASED DSP AGORITHMS Mtin ovotný, Mioš Sedáček Czech Technic Univesity in Pgue, Pgue, Czech Repubic, {SedceM, ovotnm5}@fe.cvut.cz Abstct: This ppe des with nysis of infuence of quntiztion noise on estimtion of pmetes of sign gined by vious goithms bsed on Discete Fouie Tnsfom (DFT) fo the cse of coheent nd non coheent smping. Theoetic nysis, numeic simution esuts nd expeiment vidtion e pesented. The nyzed methods e comped fom the point of view of thei sensitivity to quntiztion noise. Keywods: DFT, uncetinty nysis, quntiztion noise.. ITRODUCTIO A digit mesuing instuments nd systems wok with digitized signs, i.e. signs discete both in time nd in mgnitude. Smping nog sign in time esuts in peiodicity of sign fequency spectum, sign mpitude quntiztion eds to coesponding component in uncetinty of mesued sign. Quntiztion noise is dditive noise of input sign smpes nd is usuy supposed to be unifomy distibuted white noise, with stndd devition diecty popotion to the quntiztion step of the used quntize (nog-to-digit convete). Its men vue is zeo in most cses (if ounding is used in A/D convesion). Since A/D convesion is pesent in digit mesuing instuments, quntiztion noise is wys pesent in sign smped to be pocessed by the DSP goithm impemented in given instument. Tht is why nysis of its infuence to mesued vue uncetinty is vey impotnt tsk in mesuement. Mny ppes wee theefoe devoted to nysis of infuence of vious components of the digit mesuing instument stuctue on mesuement uncetinty. The smping convete ws nyzed in [], infuence of ADC quntiztion, smping jitte nd micopocesso finite wodength in bsic bsic FFT goithms fo coheent smping nd without sign windowing ws nyzed in [], [3], numeic method bsed on compute simution ws seected to nyze mesuement of men sque vue of input sign in [4], [5]. Spectum nysis of digitized sign beongs to the most common tsks in digit sign pocessing. The DFT goithm epesents the bsic DSP goithm fo the spect nysis. Since the mesued sign is most fequenty smped non-coheenty, windows e used to suppess the ekge. Vious goithms cn be used to obtin estimte of sign pmetes fom the DFT spectum of windowed sign. This ppe is focused on uncetinty nysis of some of these goithms. Quntiztion noise is supposed to be the min souce of uncetinty. Sign pmete estimtes e ffected besides quntiztion so by ekge. The infuence of ekge on estimted pmetes ws discussed in [6].. DFT AD QUATIZATIO OISE Since the estimte of sign pmetes is bsed on pocessing of DFT spectum, uncetinty of DFT spectum infuences the esutnt uncetinty of the estimte. Fo the cse of non-coheent smping, cosine windows e most fequenty ppied to the smped sign. Cosine time window w cos (n) is defined s w cos π ( n) cos n, n,,..., () whee is the window ode, is window ength nd e coefficients specific fo ech window. Fequency spectum of windowed sign cn be computed using the DFT goithm nd is given by n πn k X ( k) x( n) w( n) e () whee is DFT ength nd x(n) e input sign smpes. The smpes e ffected by quntiztion noise with stndd devition u n V EOB whee V is the fu sce nge of the used ADC nd EOB is its effective numbe of bits. This quntiztion noise cuses uncetinty of modue u M of DFT spect ine M(k) X(k) [7], um un nnpg (4) whee nnpg is nomised noise powe gin defined s [8] nnpg n ( w( n) ) Phse uncetinty cn be expessed s [7], (3) (5) un uϕ ( k ) nnpg (6)

3. METHODS OF PARAMETERS ESTIMATIO Vious methods e used to obtin sign pmetes of muti fequency signs (fequency, mpitude nd phse of individu hmonic o intehmonic components) fom DFT spectum in coheent nd in non-coheent smping. Sinusoid input sign is supposed in the nysis pesented hee. Resuts of this nysis e vid fo ech sinusoid component of muti sinusoid sign (sign composed of sum of sinusoid components). 3.. Coheent smping Windowing is not used in the cse of coheent smping of muti hmonic sign. Ampitude of spect component V i of the mesued sign in this cse is diecty popotion to the modue of the coesponding DFT spect ine V i k V M () k whee M(k) is the modue of DFT spect ine epesenting the i-th hmonic component nd M() epesents the spect ine coesponding to the men vue of sign V. The uncetinty due to quntiztion noise of these pmetes cn be found fom (7) u n u ( Vi ) i (8) un u ( V ) (9) The uncetinty of estimte of phse spectum ines cn be computed by (6) fo nnpg (ectngu window). 3.. on coheent smping In cse of pocessing of non coheenty smped sign (non-intege numbe of sign peiods smped), windowing is used to educe the ekge effect. Vious goithms cn be used fo detemining sign pmetes fom the DFT spectum of windowed sign. Two of them e mentioned in the next sections. 3.. RMS vue estimtion bsed on sign enegy in the window min obe Estimte of RMS vue of hmonic spectum component cn be found [8] s X RMS f f M ( f k ) + M ( f nnpg fk f fk f ) k () Hee f nd f e fequencies defining the min obe of the window spectum shifted to the sign hmonic component nd contining + DFT mpitude spectum ines. M(f k ) e components of mpitude spectum nd nnpg is nomized noise powe gin (5). Uncetinty nysis of this goithm is compicted becuse of nonzeo covince between spect ines in the min window obe. The nysis incuding this covince ws pesented in [7] nd the esutnt fomu fo uncetinty of RMS vue is un ux RMS EBW () whee EBW is equivent-noise bndwidth of the squed window (equ to fo ectngu window) nd given by [9] n EBW () n 4 w ( n) w ( n) 3.. Windowing nd intepotion in fequency domin Anothe method is bsed on sign windowing nd intepotion in fequency domin (W/IFD). Intepotion bsed on Rife Vincent windows of the fist css tht we nyze beow is descibed in []. This method is bsed on finding the exct fequency of ech spect component of the mesued sign nd on known window spectum shpe. Knowedge of fequency (in cse of non-coheent smping not pced on DFT gid) ows subsequent estimte of mpitude nd phse of the investigted spect component. The uncetinty nysis focused on the impotnt speci cse of using Hnn window is pesented in []. Estimte of exct fequency In the intepotion in fequency domin bsed on Rife Vincent windows of the fist css (RV), finding the exct fequency is bsed on tios of the oc mximum vue M(k) to its neighboing vues of the mpitude spectum M ( k + ) M ( k ) β, β (3) whee M(k), M(k+) nd M(k ) e modues of the DFT spectum of sign. Estimtes of fction fequency δ etive to fequency bin (see Fig.) coesponding to tios β nd β e given by ( + ) β ( + ) β δ, δ (4) + β + β whee is ode of cosine window (see ()). RV windows e subset of cosine windows fo the concete coefficients vues, see Tb. 5). Fig. Pincipe of intepotion in fequency domin

Resutnt estimte of fequency dispcement (fction fequency) δ is detemined s vege of δ nd δ δ + δ δ (5) Estimte of fequency of coesponding hmonic component is theefoe f ( k + ) f δ (6) whee f is fequency bin ( f f s /), f s is smping fequency nd is ength of DFT. Estimte of mpitude nd phse Estimte of mpitude nd phse of spect component is bsed on intepotion of sign DFT spectum to find the ppoximte vue of oc mximum of Discete-Time Fouie Tnsfom (DTFT) spectum. The DTFT spectum is continuous function of fequency, see Fig.. Spectum of the used window hs to be known. Spectum of cosine window () is η jπ η W ( η) e sin( πη) (7) π η whee η is e numbe fom the intev <, >, is ength of the DFT nd e window coefficients. Sign windowing in the time domin coesponds to the convoution of sign nd window spect in the fequency domin. Since sinusoid sign FT spectum is Dic puse pced on sign fequency, the bove-mentioned convoution mens ony shifting window spectum to the sign fequency. An estimte of mpitude of the investigted spect component cn be found fo the known δ s π A (8) D whee nd D γ ( δ ) δ sin( πδ ) (9) γ ( δ ) () δ The estimte of phse is bsed on the ineity of phse of cosine windows. Fo the known δ it cn be obtined s π ϕ nge( X ( k)) π δ + sign( D) () whee nge(x(k)) is phse of the DFT spectum (). This expession epesents the phse eted to the beginning of smping. If we e inteested in phse diffeence ony, this diffeence cn be found s the diffeence of phse vues of the windowed DFT spect (spect ekge hs the sme effect on phse vues of both signs hving the sme fequency). Uncetinty nysis Since the intepoted DFT spectum is ffected by quntiztion noise, sign pmetes bsed on this spectum e s we coupted by quntiztion noise. Estimte of uncetinty due to quntiztion cn be deived by using the w of popgtion of uncetinties []. Uncetinty of fequency A estimtes of pmetes depend on fction fequency bin δ. Theefoe so uncetinty of estimtes of these pmetes depend on uncetinty of δ. Tht is why it is necessy to find fisty the uncetinty of fction fequency bin δ. As cn be seen fom (3) (5), δ depends on some tios of modues of DFT. We sh theefoe fisty find uncetinty of tios β nd β. Fom the w of popgtion of uncetinties thee is um u ( β i ) ( β i + β i ) () whee i,, u M is uncetinty of the DFT modue (4) nd is coetion coefficient between neighboing two modues of DFT spectum (M(j) nd M(j+)). The cn be expessed s [7] + n nnpg n n+ (3) Knowing uncetinty of β nd β, it is possibe to deive uncetinty of δ nd δ u u + δ) u ( β ) ( β) (4) + ( + δ ) u ( β ) ( β ) (5) + ( Hee is the cosine window ode (). To obtin esutnt fomu fo uncetinty of δ it is necessy tke into ccount the covince between δ nd δ. These vues e dependent on β nd β tht e detemined by DFT modues (3). Since we know the covince of these modues, it is convenient fisty to deive the coetion coefficient between these quntities ( β, β ) β β ( β + β ) + ( β + β )( β + β ) (6) Hee is coetion coefficient between two modues of DFT spectum M(j-) nd M(j+). It cn be expessed s [7] + + nn+ 4 n (7) nnpg It cn be shown tht the demnded coetion coefficient between δ nd δ is δ, δ ) ( β, ) (8) ( β

Fig. Uncetinty of fction fequency bin δ s function of nomized fequency, Hnn window used Bsed on pevious fomue the esutnt uncetinty of δ cn be expesses s u ( δ) u ( δ) + u ( δ ) + u( δ) u( δ ) ( δ, δ ) (9) Fig. shows simution esuts nd theoetic vues of stndd uncetinty of δ due to quntiztion (sinusoid sign, V pp 9.9 V, zeo men vue, DFT ength 5, ADC fu-sce nge V V nd its effective numbe of bits EOB., epetitions, not incuding bis of δ cused by sideobes of spectum coesponding to mioed spect component). Quntiztion noise ws modeed by unifomy distibuted dditive noise. The uncetinty of fequency estimtion of sign is ccoding to (6) u( f ) u( δ ) f (3) Uncetinty of sign component mpitude Uncetinty of mpitude of sinusoid component of sign cn be found by ppying the w of uncetinty popgtion on eqution (8). We get π u ( A) A u M ( k) A A + u M ( k) δ M A + δ (, δ ) u ( δ ) + (3) whee u(m(k),δ) is the covince between M(k) nd δ nd it cn be expessed s u ( ) M ( + ) β β u, δ (3) ( + β ) ( + β) Afte expessing deivtives in (), the esutnt uncetinty of mpitude cn be witten s u ( A) + C D π A u u ( δ ) C D M + u(, δ ) (3) Fig.3 Uncetinty of sign mpitude estimtion s function of nomized fequency, Hnn window used whee D D C + δ sin( πδ ) δ cos( πδ ) nd D is defined by the eqution (9). ( δ ) (33) Dependence of uncetinty of mpitude estimtion on nomized fequency is depicted in Fig. 3. Uncetinty of phse Uncetinty of phse is ccoding to () detemined by uncetinty of DFT phse spectum (6) nd uncetinty of δ (9) u ( ϕ) uϕ( k ) + π u ( δ) (34) whee u φ(k) is uncetinty of the DFT phse spectum (6) nd u(δ) is uncetinty of fction bin δ (9). The fction fequency bin δ depends on the modue of DFT spectum. The mutu covince between fction bin δ nd the phse spectum ws expeimenty found to be negigibe. Dependence of uncetinty of phse estimtion on nomized fequency is depicted in Fig. 4. Fig.4 Uncetinty of sign phse estimtion s function of nomized fequency, Hnn window used

4. COMPARISO OF THE DESCRIBED METHODS Using windows eds to the ekge eduction, howeve it infuences so the uncetinty of esutnt pmete due to quntiztion. Theoetic vues of uncetinty wee veified by mesuements. Tb. to Tb. 4 ow compison of pesented theoy nd mesuements. The HP345A geneto ws used fo genetion of sinusoid with defined mgnitude nd fequency. Sign ws digitised by ow-cost mutipexed -bit ks/s DAQ pug-in bod (63E of tion Instuments). Instuments wee contoed by GPIB, MATAB tooboxes Dt Acquisition nd Instument Conto wee used. Mesuement conditions wee sinusoid sign, V pp 9.9 V, zeo men vue, DFT ength 5, smping fequency khz, ADC fu-sce nge V V nd its effective numbe of bits EOB.. Uncetinty ws evuted s stndd devition of mesuements, bis of δ cused by sideobes of spectum coesponding to mioed spect component ws not tken into ccount, becuse it is not cused by quntiztion noise. Sign fequency ws seected so tht k+δ 5.4 (6 fo the cse of coheent smping). It ws found tht fo the cse of noncoheent smping, uncetinty of the RMS vue by using ectngu window cn incese moe thn -times. When using othe windows, the chnges of uncetinty with δ is ess thn %. Tbe. Stndd uncetinty of RMS vue in cse of coheent smping (p 3.). Stndd uncetinty (V) theoetic mesued Rectngu (no window) 5.9-5 5.9-5 Tbe. Stndd uncetinty of RMS vue computed fom window spectum min obe (p 3..). Window ode Stndd uncetinty (V) theo. mes. Hnn 7.6-5 7.6-5 Hmming 7.3-5 7.4-5 Bckmn 8.3-5 8.4-5 7 Tems Bckmn His 3 9. -5 9. -5 Tbe 3. Stndd uncetinty of RMS vue computed by mens of W/IFD method (p 3..). Window ode Stndd uncetinty (V) theoetic mesued RV, css (Hnn). -5.5-5 RV, css.6-5.8-5 RV, css 3 3. -5 3. -5 RV, css 4 3.4-5 3.6-5 Tbe 4. Stndd uncetinty of fction fequency bin δ computed by mens of W/IFD method (p 3..). Window ode Stndd uncetinty (-) theoetic mesued RV, css (Hnn) 6.6-5 7.4-5 RV, css 7.4-5 7.6-5 RV, css 3 8. -5 8. -5 RV, css 4 8.6-5 8.7-5 Tbe 5. Coefficients of windows Rife-Vincent, css, fo window ode Rife Vincent window coefficients - - 3 4 / / -- -- -- 3/8 4/8 /8 -- -- 3 /3 5/3 6/3 /3 -- 4 35/8 56/8 8/8 8/8 /8 Tb. 5 pesents vues of coefficients of windows Rife- Vincent, css, fo window odes up to fou. Thei vues e tken ove fom [] but chnged so tht the mximum vue of the window is equ to. 5. COCUSIO Windowing suppesses ekge on the one hnd, but inceses uncetinty of estimted pmete due to quntiztion (comped with coheent smping) on the othe hnd. The highe ode of window, the highe ekge suppession, but so the highe uncetinty due to quntiztion. Howeve, s cn be seen fom the tbes, the uncetinty inceses etivey sighty with window ode incesing. Uncetinties nyzed in this ppe epesent the uncetinty component due to quntiztion noise ony. The uncetinty component cused by ong-nge ekge fom othe sign components (e.g. the mioed component in cse of sinusoid input sign) wee mentioned in [6]. ACKOWEDGMET This esech ws suppoted by the esech pogm o. MSM684775 "Resech of Methods nd Systems fo Mesuement of Physic Quntities nd Mesued Dt Pocessing " of the CTU in Pgue sponsoed by the Ministy of Eduction, Youth nd Spots of the Czech Repubic, nd by the esech pogm o. /5/H3 "Resech, deveopment nd optimiztion of mesuing systems nd mesuement uncetinty estimtion by thei ppiction" of the CTU in Pgue, sponsoed by Gnt Agency of the Czech Repubic.

REFERECES [] J. Jkubiec, K. Konopk, Uncetinty popgtion mode of A/D mesuing chin, Poc. of th intention IMEKO TC-4 Symposium pes, Ity, Septembe 998, pp. 83-836. [] G. Bett, C. iguoi, A. Pietosnto, Uncetinty nysis in fst Fouie Agoithms, Poc. of th intention IMEKO TC-4 Symposium pes, Ity, Septembe 998, pp.747-75. [3] G. Bett, C. iguoi, A. Pietosnto, Popgtion of uncetinty in discete Fouie tnsfom goithm, Mesuement, vo. 7,, pp. 3-39 [4]. occi, C. Muscs, E. Ghini, Evution of uncetinty in digit pocessing of quntized dt, Poc. of th intention IMEKO TC- 4 Symposium isbon, Potug, Septembe, pp.-6. [5]. occi, C. Muscs, E. Ghini, Evution of uncetinty in mesuements bsed on digitized dt, Mesuement, vo. 3,, pp. 65-7 [6] M. ovotny, M. Sedcek, Mesuement of RMS vues of noncoheenty smped signs, Poc. of 3th intention IMEKO TC-4 Symposium Athens, Geece, Octobe 4, pp.3-35. [7] M. ovotný, D. Sepičk, M. Sedáček, Uncetinty nysis of the RMS vue nd phse, not pubished [8] O. M. Somon, The Use of DFT Windows on SR nd Hmonic Distotion Computtions, IEEE Tnsctions on Instumenttion nd Mesuement, vo. 43, no., pp. 94-99, Api 994 [9] P. Cbone, E. unzi, D. Peti, Windows fo ADC Dynmic Testing vi Fequency Domin Anysis, IEEE Tnsction on Instumenttion nd Mesuement, vo. 5, no.6, Decembe, pp.57-575, August 989. [] G. Andi, M. Svino, A. Tott, Windows nd intepotion goithms to impove eectic mesuement ccucy, IEEE Tnsctions on Instumenttion nd Mesuement, vo. 38, no. 4, pp. 856-863, August 989. [] D. Agež, Popgtion of Uncetinty in the Intepoted DFT, Poc. of IMTC 3, Vi, CO, USA, My 3, pp. 48-53. [] Guide to the Expession of Uncetinty in Mesuement, Intention Ogniztion fo Stnddiztion, Geneve, Switzend, 993