ANALOG DEVICES AN-250 APPLICATION NOTE ONE TECHNOLOGY WAY e P.O. BOX 9106 e NORWOOD, MASSACHUSETTS 02062-9106 617/329-4700 Bandwidth, OFF Isolation and Crosstalk Performance of the ADG5XXA Multiplexer Series by Dan Sheehan and Matt Smith The ADG5XXA* multiplexer series is a family of single 8/16 channel and dual 4/8 channel parts. They are avail- able in both latched (ADG52XA) and uniatched (ADG50XA) versions. These are high performance mul- tiplexers that offer the following notable features: spec- ifications for both single and dual supply operation, 1 nA max leakage current, 200 ns max turn-on and turn-off times and TTL compatibility with reduced single or dual supplies down to 5 V. Multiplexers are widely used on the front end of data acquisition systems where there is an ever increasing requirement for higher accuracy. This requirement, cou- pled with the availability of increased resolution A/D converters that are capable of digitizing higher fre- quency signals, has dramatically heightened the need for a good understanding of the ac performance charac- teristics of multiplexers. Also, to be able to apply the multiplexers in applications such as RF, radar and video switching, the ac performance characteristics must be understood. The ac parameters of interest in a multi- plexer are its varying impedance characteristics versus frequency and are principally bandwidth, off isolation and crosstalk. Although the multiplexers of the ADG5XXA series are low cost, general Purpose parts, their ability to switch high frequency signals is excellent. The parts are fabri- cated in Linear Compatible CMOS (LC?MOS), an advanced process that features very low parasitic capac- itances and, consequently, improved bandwidth, off iso- lation and crosstalk performance. This application note investigates the bandwidth, off iso- lation and crosstalk of the ADGSXXA series from a prac- tical engineering viewpoint by presenting typical appli- cation circuits with results, over the dc to 20 MHz frequency range. The emphasis is very much on practi- calities and so the results are stated rather than derived, though it is intended that the reader will gain some insight into the mechanisms which affect the high fre- quency performance of the multiplexers. *ADGSO6A/ADG507A, ADG508A/ADG509A, ADG526A/ADG527A, ADG528A/ADG529A. MULTIPLEXER EQUIVALENT CIRCUIT A simplified ac equivalent circuit for a pair of adjacent switches in a multiplexer is shown in Figure 1. For high frequency applications it is essential that the parasitic elements shown in the model are taken into account. $1 | t t I $20} { 1 { t I Figure 1. Equivalent Circuit for Adjacent Switches Element |Title Cos Drain-Source Capacitance Cy Source Capacitance to GND Cy Drain Capacitance to GND Css Capacitance Between Sources Ron Channel ON Resistance Table |. Parasitic Elements The equivalent circuit simplifies further depending on which switch parameter (bandwidth, crosstalk or isola- tion) we are concerned with. These parameters will now be discussed separately. BANDWIDTH The 3 dB bandwidth frequency provides a measure of the high frequency usefulness of the ADG5XXA series in the ON state. From an ac or bandwidth perspective, the ON state equivalent circuit of a multiplexer channel can be simpli- fied to a series resistance and shunt capacitance model, as shown in Figure 2.I | | | yb 1g d t | ad Figure 2. ON Channel Equivalent Circuit The transfer function (Voy7/V,q) for the ON channel is: Rr Vour _ Ry + Ron (1) Vin . Ry, + Ron 1 2 Cc + Bat eR (Coe The 3 dB break frequency fo, is: f (2) Oo = Ri X Ron 27 Ro Re 7 Rel [Cp + Cr] Therefore, for de or low signal frequencies (<f,), the multiplexer channel functions as a simple resistive con- duction path (Roy = 180 typ for all parts), and any ON insertion loss (attenuation of V,,) is limited to the ratio of RiARon+R,). It is with higher signal frequencies that the complex ac impedance shown in Figure 1 becomes sig- nificant, resulting in additional attenuation of the analog input signal. From Equation 2 it is clear that the bandwidth is limited by Ron, R,, C, and Cy. Normally R, >>Roy in order to minimize the insertion loss, so the bandwidth is limited mainly by Ron, Cp and C, The ADG5XXA series of mul- tiplexers exhibit extremely low parasitic drain capaci- tance Cp. To take advantage of this low C,, the external load capacitance should be kept as low as possible. - Stray capacitance at the MUX output contributes to CL, so it is important that this be minimized by proper circuit board layout with signal line lengths as short as possi- ble. Bandwidths in excess of 40 MHz may be achieved with reasonable care.The ON resistance Ron may be minimized by operating with +15 V power supplies and low input signal levels. (See ADG5SXXA data sheets for further information on Roy variation.) The frequency response roll-off rate is typically 20 dB/decade, i.e., sin- gle pole roll-off rate. Figure 3 shows the test circuit used to evaluate the bandwidth performance of the ADG508A. Similar type test circuits apply for the other parts also. Figure 4 shows a plot of bandwidth vs. load capacitance for the ADG508A and ADG528A. The same results apply for all the input channels and over all package types. From the plots it can be seen that C, is critical to the bandwidth performance. Nevertheless, 12 pF yields excellent results, with fg >40 MHz. Similar results would apply for the ADGS509A and ADG529A, but the break frequency (fo) would be slightly higher due to the even lower C, of these parts. 15V ADG508A S1 b $2 a 88 oo AO J tA Al EN -O15V AZ GND Vss y Fite 15V Figure 3. ADG508A Bandwidth Test Circuit 800 Vin = 20V PK-PK 700 4 RL = 1k \ Ta = 25C 600 400 300 200 \ Lt Pa LOAD CAPACITANCE C, - pF N P| 1 10 100 BANDWIDTH MHz Figure 4. ADG508A and ADG528A Bandwidth Results The 3dB break frequency of the ADG506A and ADG526A, with an R, and C, of 1kQ and 12 pF respec- tively, is typically 30 MHz. Again, similar results apply for the ADG507A and ADG527A, but with a slightly higher f,. OFF ISOLATION Although the bandwidth performance (ON state) is very important, most applications involving the switching of high frequency signals are frequency limited due to reduced OFF isolation rather than degraded ON perfor- mance. OFF isolation is a measure of the ability of the multiplexer to block ac signals in the OFF state and is defined by the following equation: OFF Isolation*(dB) = 20 log ( Vin/Vour). *Due to popular convention, OFF isolation and (crosstalk) values are positive number.From an OFF isolation point of view, the multiplexer channel (or switch) can be modelled by the circuit shown in Figure 5. Ceg, which allows signal coupling from the source (S) to the drain (D), determines the OFF impedance presented by the multiplexer to ac signals. Therefore, as the signal frequency increases, Ceq passes more signal to the output. Cz, is a combination of the OFF state parasitic capacitance of the relevant multi- plexer channel and external stray capacitances due to wiring and circuit board layout. (Vin) Figure 5. OFF Channel Equivalent Circuit The OFF state parasitic capacitance of the multiplexer channel depends on a number of factors. These factors are principally: fabrication process (LC?MOS), pin con- figuration and package type (plastic DIP, Cerdip, PLCC). It has already been noted that the LC?MOS process fea- tures very low parasitic capacitances. The optimum pin configuration minimizes the parasitic capacitance by keeping the the analog input and output pins as far apart as possible. The effect of pin configuration and package type is highlighted in the results. Minimizing the stray capacitance, to achieve lower val- ues of Ceo, is critical in achieving the optimum OFF isolation specifications. Coaxial cable should be used to transmit the analog signals and a good circuit board layout is vital. The circuit board should use short signal tracks with guard traces between them, ground planes and bypassed power supplies. In addition to Ceg, the OFF isolation is also determined to a large degree by Cp, C, and R,. The lower the value of R,, the better the OFF isolation performance (Ceqs reactance becomes less effective), but note, a low value of R, gives increased on-insertion loss (i.., ratio of Ri/f[Ron + R,]). Obviously, the choice of the R, value involves tradeoffs to suit a particular application. Large values of Cp and C, improve the OFF isolation perfor- mance but again this leads to reduced bandwidth in the ON state. The results which follow apply for an optimized circuit board layout and illustrate how some of the factors out- lined above influence the OFF isolation performance. A careless circuit board layout will give degraded results. Figure 6 shows one of the test circuits used to measure the OFF Isolation of the ADGS508A (similar test circuits for the other parts). ADGS508A Series* Results Typical OFF Isolation plots for the ADGS508A and ADG528A are shown in Figures 7, 8, 9 and 10. *ADG508A/509A, ADG528A/529A. a 1 16 Ao Al oF | ADGSO8A* 2 oN 58. u 7[s4 7%. oo-52, 10 q R, J DIP PIN CONFIGURATION SHOWN. Figure 6. ADG508A OFF Isolation Test Circuit The slope of most of the OFF isolation plots changes versus frequency, and this is very apparent from Figure 10 for different values of R,. At low frequencies the fall off rate is typically 20 dB per decade. However, the OFF channel equivalent circuit has a pole frequency at fp = W277 (CegtC_+Cp) R.. Therefore, as the signal fre- quency approaches and then becomes greater than the pole frequency, the fall off rate decreases and flattens out. The OFF isolation performance varies with package type due to the different lead frames used. Overall, the plastic leaded chip carrier (PLCC) gives the best results, fol- lowed by the plastic DIP and then the cerdip. The PLCC performs best because it has the smallest lead frame, giving the least parasitic capacitance between the input and output pins. For both the plastic DIP and PLCC packages (No Connect pins are grounded), the input channel with the worst case OFF isolation is $4 (adjacent the output, D), fol- lowed by S3 with the best performance from $1 and S5 (same results). For the cerdip package, the worst case channel is again S4, but it is followed closely by S8 and then $3, with the best results from $1 and S5.The differ- ence between this package and the plastic DIP is due to a different lead frame configuration. The plots also show the OFF isolation performance for all the input channels ($1 to S8) driven simultaneously. From the plots for the ADG508AKN (see Figure 8), it can be seen that S4 couples more unwanted signal into the output than the remaining seven channels put together. Therefore, by sacrificing the use of Channel 4, the user can obtain excellent OFF isolation results from the remaining seven channels. Alternatively, to achieve even higher isolation from the seven channels, $4 may be utilized as a NORMALLY ON grounded channel. With this scheme, S4 is selected when all the other channels are off. This effectively grounds the output (via Ron) giving improved OFF isolation results which are inde- pendent of R,. Hence large values of R, may be selected which minimize the ON insertion loss without degrading the isolation. The improvement offered with this scheme is included in Figure 10. Similar results apply for the ADG509A and ADG529A, except at high frequencies where the OFF Isolation is marginally worse due to the lower Cy of these parts.120 N Vin = 20V PK-PK R= 1kA NY C, = 12pF 100 _ Ty = 25C Mw NI NI MLO hs MN Ms 2 80 N KI PS , SS MI z NS Ns E 60 MA [i si on ss | 4 SS ss & N it RN 40 cs s4 a 81 TO 8 (SIMULTANEOUSLY) 20 a 10k 100k 1IM 10M 100M FREQUENCY - Hz Figure 7. ADG508AKP and ADG528AKP (PLCCs) OFF Isolation vs. Frequency 120 Vin = 20V PK-PK NI R. = 1k NA N C, = 12pF 100 N NS Ty = 25C ~ Pe MN *. Sh MN ~ oR ots NL OA t e . DSA Ts W 2 NAA TB. Ny 9 SS Ne MN MA E 60 S10R 55 7 g ~ 4 A 3 SW Ys. Mm 2 NA x. pe SS i SS ~ o , a 40 = =: Peed SW S4 $1TO S8 20 {SIMULTANEOUSLY} | 0 10k 100k 1M 10M 100M FREQUENCY-Hz *ALL CHANNELS EXCEPT $4 {SIMULTANEOUSLY) - Figure 8. ADG508AKN and ADG528AKN (Plastic DIPs) OFF Isolation vs. Frequency 120 . Vin = 20V PK-PK se ie Ry = 1k SJ C, = 12pF 100 F > ~ Ta = 25C NJ MN N s my 80 ho Po P<] m g NS IN MN i INS NJ NY S 60 MA Pm s10R $5 = 3 S | || 5 ISA ~~ a SS [pa S3 OR S7 ix 40 MISS oO ~* Sa MSS $8 el s4 20 5 siTo se -| (SIMULTANEOUSLY} 0 10k 100k 1M 10M 190M FREQUENCY - Hz Figure 9. ADG508ABQO and ADG528ABOQ (Cerdips) OFF Isolation vs. Frequency 120 T T T vet Vin = 20V PK-PK q C, = 12pF < T, = 25C 100 }- NI CHANNEL 8 (S8) NJ NN g 80 R, = 10k0 os L Y NY ~ GROUNDED 84 z SELECTED Bo mM NS a NY R, = ON a ne SA MA rs R, =10k0 N 5 40 PS NS hd MN 20 0 10k 100k 1M 10M 100M FREQUENCY - Hz Figure 10. ADG508ABQ and ADG528AB0 (Cerdips) OFF Isolation vs. Frequency for Different R, ADG506A* Series Results Typical OFF isolation plots for the ADG506A and ADG526A are shown in Figures 11 and 12. The ADG507A and ADG527A wouid yield similar results except at the high frequency end where the results would be margin- ally worse due to the lower C, of these parts. Since the ADGS5O6A series has no analog input channel on a pin which is adjacent to the output, the results for this series surpass those for the ADG508A series. For both the DIP and PLCC packages (N.C. pins connected to GND), the input channel with the worst case OFF isola- tion is $8, followed by S7, with the best performance from $1 and $9. In general, comments concerning the ADG508A series results apply for the ADG506A series also. 120 . T T TT Ja Vey = 20V PK-PK . RL = 10 4 ~ 4 a = 12pF 100 T, = 25C NON oN ne NEN NY g SOS aS . Ms mM PISS z MN NY NN Ni & 60 aS s10Rs3 4 pI N N] 6 ~N MN Speed 57 B NONE Ps pomene] SE 5 a MN PPh s1 To si16 (SIMULTANEOUSLY) 20 0 10k 100k 41M 10M 100M FREQUENCY - Hz Figure 11. ADGS506AKN and ADG526AKN (Plastic DIPs) OFF Isolation vs. Frequency *ADGSO6A/507A, ADG526A/527A.120 Fy ~ wd x] <A 100 * oN WN, MN a N , he MN NN 80 en, ~ Pe a NJ ~s v sw SIAN 5 IN N. MN 5 60 ~ S10R S94 ~ se a MIN) a 87 i Spee SB o 49 a pwnd $1 TO S16 (SIMULTANEOUSLY) 20 o 10k 100k iM 10M 100M FREQUENCY - Hz Figure 12. ADG506ABQO and ADG526ABQ (Cerdips) OFF Isolation vs. Frequency CROSSTALK Crosstalk, which is also called Channel-to-Channel Crosstalk, is another frequency-dependent factor. It is basically channel-to-channel isolation and occurs due to the OFF state parasitic capacitances of the multiplexer. Figure 13 shows a simple crosstalk model for a two- channel multiplexer. Assuming negligible ON insertion Figure 13. Crosstalk Equivalent Circuit loss due to Ron1 and R,, Vous should equal Vint. How- ever, Ceg (see OFF isolation section for details) couples some portion of Vi2 into Voy7. Also, but to a lesser degree, additional signal coupling (not shown in Figure 13) occurs between the switches due to parasitic capac- itances within the die and between the package leads. With Vin1 connected to GND, crosstalk for the circuit shown in Figure 13 is defined by the following equation: Crosstalk (ABs) = 20 log (Vin2Vout!. The error signal couples into a parallel combination of Ris Ron, Cp and C, versus R,, Cy and C,_ only for OFF isolation. Because Roy is typically 1809, crosstalk results are generally 10 dB better than OFF isolation results. While maintaining good bandwidth and OFF isolation performance, crosstalk can be minimized by choosing a multiplexer with a low Roy. A good circuit board layout that shields the analog signals and keeps stray capaci- tances to a minimum is also vital. Figures 14 and 15 show two of the test circuits used to measure the crosstalk of the ADG508A (similar test cir- cuits used for the other parts). TS 1 Ao arp o +15v V ADGSO8A* +15V oHen 3 Az 15V 0-9 "1 Vs, GND ore ne, 4st Yoo ps _____ o +18V 2x 500 O.1pF ==10pF ech noeha, OS 4x 500 $3 oo 84] 11 $4 + ~ $7]10 o sea JL *DIP PIN CONFIGURATION SHOWN. Figure 14. ADG508A Crosstalk Test Circuit (One OFF Channel Driven Only) wa 1 AO At 16 v7 ADG508A* 2 +15V OJEN AZ 15 3 14 ~18V Ot es Gn wre mr 4}si Voo 3 +15V 502 O.1mF 10nF a ons a No SEE Ea 1 HH Wee Ee soft itt po =s SS SSS SSS aay t i: ty! 9 fan ijt-------------J Th54 s7Poil! ea aPp ras aaa SSS a r= avr 50n ht 9 ! v, ae Wy Wl Vour =4 oo SBE = tl THe tt Ht tft Cq. R, it cn aa if L il W ih tt te ift hle---~------ 2 ee LLL ajl ! DIP PIN CONFIGURATION SHOWN. Figure 15. ADG508A Crosstalk Test Circuit (7 OFF Channels Driven) ADGS5OB8A Series Results Typical crosstalk plots for the ADG508A and ADG528A are shown in Figures 16,17 and 18. In Figure 14, the crosstalk is measured by driving S4 only, with S3 selected and the remaining OFF channels connected to GND via 50 9. Due to the proximity of $4 to the output pin, driving this channel gives the worst case crosstalk results (for one OFF channel driven only) see Figure 16 Plot b. For example, driving $3 only with S4 selected would give far superior results typically 15dB better in the case of the ADG508AKN and ADG528AKN. Crosstalk can also be measured by driving all of the remaining OFF channels simultaneously, with any one channel selected as in Figure 15. Under these conditions the worst case performance (see Figure 16 Plot a) occurs when $1 or $5 is selected and the best performance (see Plot c) results when S4 is selected.The plastic DIP package gives better results than the cerdip package (similar to OFF isolation). The results for PLCC are not shown but would be approximately 5 dB better than for plastic DIP. Figure 18 shows that the crosstalk performance improves as R, reduces. Similar results would apply for the ADG509A and ADG529A. However, at the high frequency end, the results would be marginally worse due to their lower Cp. 120 T Viu = 20V PK-PK R, = 1k OG = 12pF 100 4 Tj =2cL} tf ~ Is 80 SSS E o ml Sk Sat = NS = 60 > B ~ 3 5 SOY NS 40 c b a 20 "10k 100k 1M 10M 100M FREQUENCY ~ Hz a. ALL CHANNELS DRIVEN EXCEPT CHANNEL 1. WHICH IS ON. 6. S4 DRIVEN; 53 "ON"; REMAINING CHANNELS CONNECTED TO GND (SEE FIGURE 14). . ALL CHANNELS DRIVEN EXCEPT CHANNEL 4, WHICH IS ON. Figure 16. ADG508AKN and ADG528AKN (Plastic DIPs} Crosstalk Results vs. Frequency 120 v TT Vau = 20V PK-PK R, = 1k C, = 12pF 1 100 T, = 25e +14 BS wo NAY NOS : ASN V de SS z g TA z= S o NEN NN: SS MN e 20 }- Oo 10k 100k 1M 10M 700M FREQUENCY - Hz a. ALL CHANNELS DRIVEN EXCEPT CHANNEL 1, WHICH IS ON. b. S4 DRIVEN; $3 ON; REMAINING CHANNELS CONNECTED GND (SEE FIGURE 14). . ALL CHANNELS DRIVEN EXCEPT CHANNEL 4, WHICH IS ON. Figure 17. ADG508ABQ and ADG528ABQ (Cerdips) Crosstak Results vs. Frequency 120 Vin = 20V PK-PK C, = 12pF Ta = 25C 100 PS a0 Sy mn a SONY 2 iN MAR, = 1002 3 60 NE < a & R, = 10k0 MN 40 HY 8 NSe RL = ka 20 | o 10k 100k 1M 10M 100M FREQUENCY - Hz ALL CHANNELS DRIVEN EXCEPT CHANNEL 1, WHICH IS ON. Figure 18. ADG508AKN and ADG528AKN (Plastic DIPs) Crosstalk Results vs. Frequency for Different R, ADG5O6A Series Results Typical crosstalk plots for the ADG506A and ADG526A are shown in Figures 19 and 20. The ADG507A and ADG527A yield similar results except at high frequencies where the results would be marginally worse due to the lower Cp of these parts. Since the ADG506A series has no analog input channel and output on adjacent pins, the results for this series surpass those for the ADG508A series. $8 is the channel nearest to the output, and therefore driving S8 with any other channel selected contributes most to poor crosstalk results. In general, comments on the ADGS5O8A series results apply for the ADG506A series also. 120 F T UeTT Vig = 20V PK-PK R, = tk RS c= tapF ! 100 SJ Ta = 25C [J NS a0 SSN * SA NY $ SS 3 50 S ~ b Ni a SW 2 NSN 40 Ee 4+ a 20 oe 100k 1M 10M 100M FREQUENCY - Hz a, ALL CHANNELS ORIVEN EXCEPT CHANNEL 1, WHICH IS ON. b. S8 DRIVEN: $7 ON; REMAINING CHANNELS CONNECTED GND (SIMILAR TO FIGURE 14). . ALL CHANNELS DRIVEN EXCEPT CHANNEL 8, WHICH IS ON. Figure 19. ADG506AKN and ADG526AKN (Plastic DIPs) Crosstalk Results vs. Frequency 120 T T TT Vig = 20V PK-PK R, = kd C, = 12pF 100 . Ta = 25C SY 20 NON . J a INN s NWN z 60 MS NN g NWN 40 A bec a 20 1 J Q 10k 100k 1M 10M 100M FREQUENCY Hz a. ALL CHANNELS DAIVEN EXCEPT CHANNEL 1, WHICH [S ON. b. S8 DRIVEN; $7 ON; REMAINING CHANNELS CONNECTED TO GND (SIMILAR TO FIGURE 14). , ALL CHANNELS ORIVEN EXCEPT CHANNEL 8, WHICH IS ON. Figure 20. ADG50GABQ and ADG526ABQ (Cerdips) Crosstalk Results vs. Frequency COMMENTS The results furnished in this application note are based on measurements from three fabrication lots. With the same conditions and device type, the results varied very little from lot to lot and from device to device. For exam- ple, the results for both OFF isolation and crosstalk var- ied by typically 1 dB only. For bandwidth, the variation in results was less than 0.2 MHz.