1/7 AN1751 application note ? emi filters: recommendations and measurements september 2003 - ed: 1 p. merceron and p. rabier with the development of wireless telecommunications, consumer products and cellular phones are sub- jected to radio frequency interference and may generate electromagnetic interference. this is in addition to electrostatic discharge the user can apply when touching a connector like a bottom connector on a cel- lular phone. in the past, filtering was achieved by discrete devices (capacitors, resistors) and esd protection was done by discretes diodes. cellular phone size drop and enhanced features require faster signals and more and more integrated devices which are becoming very sensitive to esd or emi/rfi. discretes devices impose a well designed layout to minimize parasitic effect of pcb inductances while in- tegrated passive and active devices suppress most of these inductances due to very short tracks between passives on the die itself. 1 ghz 100 mhz 10 mhz db ipad ipad discretes discretes fig. 1: frequency response comparison between discrete and ipad? filter. discrete filter will behave like a rejection filter but the rejection frequency will be depending on parasitic in- ductances while the ipad filter will act like a low pass filter. emif filters have three main functions, the first is of course to filter emi/rfi, the second is to protect inputs and outputs against esd and the third one is to transmit data from inputs to outputs. emif datasheets pro- vide data and curve measured in specific conditions. the goal of this note is to explain test conditions for emif devices. 1-frequency response emif target is to pass low frequency signals and to reject frequency higher than 800mhz especially 900mhz, 1.8 and 2.4ghz. attenuation curve provided in specification shows : n simulation thanks to the aplac (or p-spice) tool. this is done before the die design to be sure the device will fit customer requirements n measurement done on demonstration board.
AN1751 - application note 2/7 aplac models take into account of die, bumps and via for the ground connections. it does not consider pcb track in the application. figure 2 represents an example of aplac model for one filtering cell. rsubump cin rsubump cout model = d02 model = d01 rs i1 o1 die model gnd lbump rbump cgnd lgnd rgnd via model bump model fig. 2: aplac model for die, bumps and pcb via. p-spice is probably the most wellknown simulation software in electronic industry. limits of p-spice is reached when trying to simulate rf signals because of the time it takes for each simulation. it is also impossible to simulate crosstalk phenomena. aplac has been developed to avoid all these p-spice limitations. concerning measurements, the first step is to calibrate the equipment. this is why demoboards are delivered with a calibration kit shown in figure 3 while the ipad? device is on another board ( figure 4 ). fig. 3: calibration board. 50 w 50 w vg fig. 4: measurement condition on demoboard. if test equipment is not calibrated, non negligible error can occur as the attenuation measurement will correspond to the one of the ipad? + board + connections. furthermore measurement is done with a 50 w load while some applications may have other impedances.
AN1751 - application note 3/7 2- esd and latch-up measurements i/o lines of a cellular phone must be protected against esd. most popular esd standard is the iec61000-4-2 having a surge generator defined in figure 5 . r1 c1 gnd gnd + + -- to pin c2 l1 r2 l2 fig. 5: iec61000-4-2 generator and the result on a non protected integrated circuit die. iec61000-4-2 specifies c1 charged up to 8kv (contact) and 15kv (air discharge). the mil-std 883e method 3015.7 is also a reference. all external pins (bottom connectors, microphone jack) may be subjected to these kinds of surges. if no protection is used, result will be the destruction of the internal silicon chip. destroyed i/o is generally a short circuit as silicon melt on a very small area as shown in figure 5 . if esd protection device is the minimun to prevent failure, layout is also very important as very high di/dt of surge will generate a high ldi/dt. this means even with a protection device, an integrated circuit can be destroyed because of layout problem. figure 6 explains differences between 2 layouts. l1 l2 l3 l6 v ic v in v cl ic to be protected correct layout l1 l3 l4 l5 l6 ic to be protected v ic v in v cl l2 wrong layout ic to be protected ic to be protected ic to be protected ic to be protected fig. 6: two layouts for two very different results. knowing all pcb tracks are equivalent to an inductance, in the first case the ic will see a voltage: v ic = (l1+l2+l3+l4) x di/dt + v cl with a track lengh, for protection device connection, of 2 cm (1cm from side to side), 35 thickness, 0.5mm wide (microstrip track) then l2+l3 = 8nh considering a 15kv esd surge surge having di/dt = 50a/0.7ns the result is v ic = 570v
AN1751 - application note 4/7 overvoltage being directly linked to track lengh, voltage across ic may be much higher with few more centimeters. most of ic will not like it. with the correct layout, no high current is flowing through l6 and l3, then v ic =v cl in the first case, ic can be destroyed while with the second layout ic is perfectly protected. figure 7 gives input and output voltages of an emi filter with the suitable layout. the demoboard shows the esd is applied on input (v in ), while emif output (v out ) is connected to the ic to be protected. in out e emi 325 gnd vout measurement vin measurement in out e emi 325 gnd in out e emi 325 e emi 325 a a a a esd test board for emi fig. 7: esd demoboard and measurements on emif10-1k010f1. figure 8 shows the equivalent schematic with the esd surge generator, the emi filter and the load which corresponds to the device to be protected. rg vg rd rd v br v br r i/o vg rd vg r load vout vin vin vin fig. 8: equivalent schematic of emi filter connected to esd surge generators.
AN1751 - application note 5/7 using emi filter, protection is done in two steps. the first esd diode will limit the v in voltage to: v rg v rd vg rg rd in br = + + v rv rd v rrd out br in io = + + / taking into account o f:r>>rd,rg>>rdand rload >> rd (open circuit on the test board). with vg = 15kv, rg = 330 w applied to the emif10-1k010f1 board (rd = 1 w ,v br = 8v and r i/o =1k w ) v in = 53v and v out = 8.4v with a correct layout, maximum output voltage is closed to v br . early ageing and destruction of ic is often due to latch-up phenomena which is mainly induced by dv/dt. thanks to its rc structure, emi filters provide high immunity to latch-up by integration of fast edges. measurements done on figure 7 show very clearly the high efficiency of the structure. 3- crosstalk measurements 3a - digital crosstalk crosstalk phenomena are due to coupling between 2 lines (2 filters in our case). coupling factors ( b 12 and b 21) shown in figure 9 increase when distance between lines decreases, particularly in silicon dice where distances between components are very short. in the example above the expected signal on load rl2 is a 2 x vg2, in fact the actual voltage at this point has got an extra value b 21 x vg1. this part of the vg1 signal represents the effect of the crosstalk phenomenon of the line 1 on the line 2. this phenomenon has to be taken into account when drivers impose fast digital data or high frequency analog signals. the disturbed line will be more affected if it works with low voltage signal or high load impedance (few k w ). rg1 vg1 r l1 ab 1 x vg1 + 12 x vg2 rg2 vg2 r l2 ab 2 x vg2 + 21 x vg1 drivers receivers line 1 line 2 fig. 9: crosstalk phemonema.
AN1751 - application note 6/7 for digital crosstalk, test is done thanks to a square pulse generator connected in vg1 through a 74hc04 gate. an oscilloscope is connected to vg2. figure 10 shows the measurement circuit used to quantify crosstalk effect in a classical digital application. rg1 vg1 b 21 x vg1 + 5v line 2 line 1 ? 74hc04 ? 74hc04 fig. 10: digital crosstalk measurement circuit. figure 11 shows with a signal from 0 to 5v and rise time of few ns, the impact on the disturbed line is less than 40mv peak to peak: no data disturbance was noted on the concerned line. the measurements performed with falling edges gives an impact within the same range. fig. 11: digital crosstalk results (emif10-1k010f1). 3b - analog crosstalk figure 12 gives the measurement circuit for analog crosstalk. in figure 13 , the curve shows the effect of cell on filter i1/o1 on cell i2/o2 for the emif10-1k010f1. in usual frequency range of analog signals (up to 100mhz) the effect on disturbed line is less than -47 db
AN1751 - application note 7/7 4- conclusion emi filter are designed to suppress emi and rfi noises and are also used to protect sensitive systems against electro-static discharges. the use of an z-r-z structure provides a very stable voltage on the output during esd discharge when layout has been correctly routed. rf measurement must be done carefully after using calibration board provided with emif demoboards. emi filters on flipchip are the most efficient topology to filter and protect integrated circuit by saving pcb area. 50 50 vg w w fig. 12: analog crosstalk measurement circuit. information furnished is believed to be accurate and reliable. however, stmicroelectronics assumes no responsibility for the consequences of use of such information nor for any infringement of patents or other rights of third parties which may result from its use. no license is granted by implication or otherwise under any patent or patent rights of stmicroelectronics. specifications mentioned in this publication are subject to change without notice. this publication supersedes and replaces all information previously supplied. stmicroelectronics products are not au- thorized for use as critical components in life support devices or systems without express written approval of stmicroelectronics. the st logo is a registered trademark of stmicroelectronics. all other names are the property of their respective owners. ? 2003 stmicroelectronics - all rights reserved. stmicroelectronics group of companies australia - belgium - brazil - canada - china - czech republic - finland - france - germany - hong kong - india - israel - italy - japan - malaysia - malta - morocco - singapore - spain - sweden - switzerland - united kingdom - united states www.st.com fig. 13: typical analog crosstalk result (emif10-1k010f1).
|
