| The Datasheet Archive - 100 Million Datasheets from 7500 Manufacturers. |
Originally, video filters were passive circuits surrounded amplifiers.
Top Searches for this datasheetActive filters video Originally, video filters were passive circuits surrounded amplifiers. Smaller, more efficient designs currently achieved combining amplifier with filter. Sensitivity analysis predistortion methods developed 1960s have, moreover, overcome poor performance that gave early video filters reputation. High-performance amps specialized software enable design wide-bandwidth active filters, those advantages address requirements specific application. video filters, particular application signal format nuance each circuit design. major video applications follow. Antialiasing filters: These devices placed before analog-to-digital converter (ADC) attenuate signals above Nyquist frequency, which half sample rate ADC. These filters usually designed with steepest possible response reject everything above cutoff frequency ITU-601 applications others, such performance achieved using analog filters combined with digital filters oversampling ADC. applications such graphics, very little filtering required. Reconstruction filters: Also called (sinx)/x zero-orderhold correctors, these filters placed after digital-toanalog converter (DAC) remove multiple images created sampling, though remove clock. Reconstruction filters seldom selective antialiasing filters, because DAC's hold function also acts filter-an action that lowers required selectivity, introduces loss response. available video formats RGB, component video, composite video, graphics. applications formats require video filter "phase linear," condition specified parameter called group delay (delay versus frequency). degree phase linearity required depends application video format. example, antialiasing filters component formats more tightly specified than reconstruction applications composite video. Requirements various applications formats specified NTSC, PAL/DVB, ITU, SMPTE, VESA. This article compares different filters determine optimum design given application format. Rauch Sallen-Key realizations compared their GBW-to-cutoff ratios, using predistortion element sensitivity techniques achieve accuracy design. Those filters considered are: ITU-601 antialiasing filter 20MHz antialiasing reconstruction filter HDTV reconstruction filter Filters their characteristics Whether used antialiasing reconstruction, filter must have lowpass characteristic pass video frame rate. should, therefore, wary AC-coupling. Lowpass filters categorized their amplitude characteristic name polynomial that describes (Bessel, Butterworth, Chebyshev, Cauer). Figure shows these characteristics normalized 1-rad bandwidth. Typically, filter with best selectivity minimum number poles minimize cost) would chosen, additional need phase linearity limits available choices. Phase linearity group delay filter's phase linearity specified envelope delay group delay (GD) versus frequency. flat group delay indicates frequencies delayed same amount, which preserves shape waveform time domain. Thus, absolute group delay important variation group delay. separate specification called channel-to-channel variation, which specified "time coincidence," should confused with group delay. INSERTION LOSS AMPLITUDE CHEBYSHEV CAUER BUTTERWORTH BESSEL NORMALIZED FREQUENCY DELAY TIME GROUP DELAY Figure Amplitude group delay frequency various filter types normalized 1-rad bandwidth. Though desirable video, much group-delay variation acceptable, why? answer depends application video format. example, ITU-470 specifies group delay very loosely composite video. However, ITU-601 specifies tightly ensure generational stability, both MPEG-2 compression control phase jitter before serialization. what filter characteristics considered necessary ensure phase linearity? group-delay curves Figure show peak near cutoff frequency That problem caused steep phase change near cutoff frequency. idea scale, 3-pole, 6MHz Butterworth filter group-delay variation 20ns-25ns over bandwidth. Increasing number poles filter's selectivity increases that variation. Other, more exotic filters2 used minimize group-delay variation include Bessel, phase approximation, Thompson-Butterworth, LeGendre. Nevertheless, Butterworth characteristic most often used video. Choosing After choosing filter characteristic, next step implement with actual circuit. most commonly used, single-op-amp circuits Sallen-Key configuration noninverting form Rauch configuration inverting form. important consideration amps operating wide bandwidths video applications minimum gain-bandwidth (GBW). Video signals large, typically 2VP-P, large-signal referenced. This parameter confused with 2VP-P 0.1dB GBW, which much lower. filter circuits, much larger than filter's cutoff frequency does op-amp have Rauch (inverting) filter, phase argument characteristic Arg[K(j)]inv -(c/GBWrad)(1+Rf/Ri) Sallen (noninverting) filter: Arg[K(j)]noninv -(c/GBWrad)(1+Rf/Ri) where gain-set resistors ohms, GBWrad amp's gain-bandwidth product, filter's cutoff frequency radians second. gain introducing values solve (c/GBWrad). unity-gain Rauch circuit Rf/Ri Sallen-Key circuit Rf/Ri Thus, same phase error, Sallen-Key requires half Rauch circuit. required gain increases, they converge leave little advantage Sallen-Key terms GBW, other issues must considered well. Group-delay problems with component video formats applications sensitive group-delay variation. degree sensitivity depends number signals their bandwidths. Composite NTSC/PAL only signal, with group delay specified ITU470. Those requirements easily met. component video each have multiple signals. signals have equal bandwidths while component-video signals not, making group-delay matching easy with RGB, difficult with component video. Because signals have half bandwidth luma signal, their group delay double that signal3. solution slow down signal adding delay stages. Another solution equalize bandwidths doubling sample rates which raises 4:2:2 sampling rate 4:4:44, allowing signal treated RGB. additional samples discarded during antialiasing averaged reconstruction applications. other component-video format, S-VHS, somewhat confusing. channel same YPbPr, chroma signal looks like should bandpass filtered rather than lowpass filtered. YPbPr signals, bandpass filtering causes group-delay timing problems and, therefore, should implemented. Unless analog encoding done, lowpass filtered with same filter. S-VHS more forgiving bandwidth than problems caused trying equalize delay. S-VHS typically seen reconstruction applications, which main concern correct timing between Predistortion, bandwidth, element sensitivity Anything less than infinite GBWrad/c ratio causes closed-loop poles filter move. That actual filter often exhibits lower bandwidth than does paper design6. This compensated increasing design bandwidth, which known predistortion. Formulas Sallen-Key Rauch circuits (listed Tables allow calculate design bandwidth that provides actual bandwidth needed. Component tolerance must then taken into account. determine component tolerance, sensitivity function7 needed: gives ratio between change value part consequent change parameter example, Table shows that SallenKey circuit (vs. Rauch circuit) large sensitivity variations That means Sallen-Key less tolerant parasitics than Rauch. point that lets effect predicted, then design created accordingly. Next, some typical designs considered. Table Component sensitivities including predistortion formulas (Sallen-Key realization, 1rad/sec) Sensitivity Function SR1Q SR2Q SC1Q SC2Q SRaK SRbK Gain -1/2 -4.5 -9.5 -9/14 9/14 Gain -1/2 -1/2 Gain -1/2 -4.5 -5.5 -1/2 (actual) (actual) (design)[1 1/2(3 1/Q)2 GBW] (design)[1 1/2(3 1/Q) GBW] (design)[1 GBW] (design)[1 GBW] Table Component sensitivities including predistortion formulas (Rauch realization, 1rad/sec) Sensitivity Function SR1Q SR2Q SC1Q SC2Q SR3K SR1K SR3Q Gain -1/2 -1/6 -1/2 Gain -1/2 -1/2 Gain 100) -1/2 -1/2 (actual) (actual) (design)[1 2GBW] (design)[1 2GBW] Design antialiasing filters antialiasing filters, selectivity determined template ITU-601 like Figure specified bandwidth 5.75MHz ±0.1dB, with insertion loss 12dB 6.75MHz 40dB 8MHz, with group-delay variation ±3ns over 0.1dB bandwidth. Such performance difficult analog filter alone, oversampling modifies requirements 12dB 27MHz 40dB 32MHz. Using software normalized curves8, find that 5-pole Butterworth filter with -3dB bandwidth 8.45MHz satisfies requirement selectivity, though group delay. latter, delay stage needed, which important op-amp parameter 0.1dB, 2VP-P bandwidth9. That number should used equations accurate design. schematic this application, with curves showing gain group-delay characteristics, based oversampling (Figures 3b). video considered next. VESA does specify templates antialiasing reconstruction filters. resolution (1024 85Hz) sampling rate 94.5MHz Nyquist frequency 47.25MHz. >35dB attenuation Nyquist frequency, Rauch realization 20MHz, 4-pole Butterworth filter (Figures used. Again, MAX4450/4451 chosen their excellent transient response large-signal bandwidth (175MHz 2VP-P). Reconstruction filters Reconstruction filtering after among more poorly understood applications. Some designers think reconstruction filters introduced remove sample INSERTION LOSS +0.1dB -0.1dB -3.0dB -12dB GROUP DELAY -40dB 5.75MHz 6.75MHz 8.0MHz FREQUENCY 5.75MHz FREQUENCY Figure This filter template illustrates antialiasing requirements accordance with ITU-R BT.601-5 standard. 5-POLE FILTER BUFFER GAIN DELAY EQUALIZER 78.7 270pF 120pF 43pF 680pF 18pF 91pF MAX4451 Figure This schematic output response represent 5-pole, 5.75MHz Butterworth filter ITU-601 antialiasing, using Rauch circuit with delay equalizer. 33pF 22pF EOUT 5.6pF 115pF Figure This schematic output response depict 4-pole, 20MHz Butterworth filter graphics antialiasing, using Rauch circuit. clock, nothing further from truth. When signal sampled, samples composed multiple recurring signal images centered harmonics sample clock. reconstruction filter removes baseband sample. antialiasing filter served purpose, output looks like image Figure then samples right should removed. Thus, reconstruction similar antialiasing except that, because each sample exists only instant, holds each clock period, thereby creating familiar staircase approximation sloping line. hold function corresponds digital filter whose characteristic10 similar that Butterworth Bessel filter (Figure Notice that response decreased half sample frequency. second objective reconstruction filter restore that loss, which requires amplitude equalizer like circuit shown Figure equalizer based delay stage response like Bessel filter. designed from sample rate (Fs). Figure shows DAC's frequency response with without amplitude equalizer. Like delay stage, included reconstruction filter. hold response also pole centered sample clock, which completely removes clock. Nevertheless, most reconstruction applications refer clock attenuation figure merit. that function reconstruction filter understood, designed. AMPLITUDE EOUT FREQUENCY SAMPLE FREQUENCY NYQUIST FREQUENCY component's values function sampling frequency DAC. Fsample R1/10 R1/50 Figure typical output spectrum shown terms sampling (FS) Nyquist (FN) frequencies. CORRECTION ZERO ORDER HOLD INSERTION LOSS (IL) 0.071 -0.5 -1.0 WITH COMPENSATION AMPLITUDE |H(F)| WITHOUT COMPENSATION -1.5 -2.0 -2.42 -2.5 NORMALIZED FREQUENCY 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 NORMALIZED FREQUENCY Figure "hold" function produces (sinx)/x response with nulls multiples sampling frequency. Figure output shown with without (sinx)/x correction provided amplitude equalizer circuit (a). most common requirement NTSC/PAL reconstruction attenuation >20dB 13.5MHz >40dB 27MHz, where depends applicable video standard. 3-pole Butterworth with Sallen-Key configuration chosen reasons. First, gain (+2) drives back-terminated cable. Second, group-delay variation adjusted optimize performance without delay equalizer. (Figures 8a-8d show NTSC designs, including gain group-delay characteristics.) These applications usually include digital amplitude correction DAC, which easily added, necessary. Illustrating circuit XGA, 20MHz, 3-pole Butterworth filter Sallen-Key configuration includes Figure circuit amplitude correction (Figures 9b). Complementing antialiasing filter Figure this filter gain drive back-terminated coaxial cable. last application reconstruction filter HDTV. Based templates SMPTE 296M, center frequency 29.7MHz. Amplitude correction usually included, group-delay compensation must added. resulting 30MHz, 5-pole Sallen-Key filter (Figure >40dB attenuation 74.25MHz, well groupdelay stage with gain drive back-terminated coaxial cable. Practical aspects active video-filter design Whether filters designed hand, with software, with combination these approaches, actual response exactly what wanted. cause discrepancy between calculated response actual response obtained using standard component values. That error minimized choosing standard (5%) capacitor values deriving resistor values from them. reason practical-capacitors with tolerance acquired, only values, although 47pF VERSION MAX4450 NTSC VERSION 100pF MAX4450 78.7 150pF NOTE NOTE 330pF NOTE NOTE 47pF 82pF NOTE LOAD NOTE ADJUSTS GROUP DELAY Figure reconstruction filters with group-delay adjustment, version amplitude group-delay responses shown (c), NTSC version amplitude group-delay responses shown (d). 330pF 4.12 220pF 10pF 20.5 10pF ROUT EOUT 27pF Figure This 3-pole, 20MHz Butterworth filter reconstruction includes (sinx)/x compensation. output response shown (b). XU1a 7.5pF XU1b 8.2pF XU1c XU1d 33pF 69pF 150pF 27pF Figure 5-pole, 30MHz reconstruction filter HDTV includes amplitude correction DAC. resistors that combine values with tolerance available. Such components give best approximation most precise amplitude response. Once built, filter unstable oscillate. that case, short input ground continues oscillate. stops, impedance high. Lowering design impedance should eliminate oscillation. continues, note whether oscillation near filter cutoff frequency just below. that case, oscillation probably components parasitics. oscillation above cutoff frequency, probably circuit layout. Good layout seems art, based simple principles. important have clean supply voltage solid ground, meaning filtration with low-ESR capacitors sometimes with regulator. loop formed bypass-capacitor connections must small, resulting parasitic inductance will resonate with capacitance. good ground plane essential good analog design, bandwidth increases, parasitic capacitance that detune filter. avoid that problem, remove ground plane beneath offending part(s) traces. 4The 4:2:2 sampling originally indicated number times color subcarrier oversampled. ITU-601 replaced subcarrier frequency with 3.375MHz. 4:2:2 sampled 13.5MHz 6.75MHz, respectively. 5For noninverting case, inverting case, Rf/Ri 6E.J. Kennedy, Operational Amplifier Circuits: Theory Applications. Defined H.W. Bode Network Analysis Feedback Amplifier Design Nostrand, Princeton 1945). 8Taylor Williams, cit. 9MAX4450/51 data sheet available www.maxim- ic.com. 10The Sinc function mathematics (sinx)/x. similar article appeared June 2003 issue EDN. References filter response drops -3dB cutoff frequency. Taylor Williams, Electronic Filter Design Handbook, McGraw Hill, ISBN 0-07-070441-4. 3Ibid. Other recent searchesW5641A - W5641A W5641A Datasheet NJW1103 - NJW1103 NJW1103 Datasheet MPBG119B - MPBG119B MPBG119B Datasheet
Privacy Policy | Disclaimer |