that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com - + 7 4 3 1 8 6 5 2 bias current comp 1 v bias current comp 2 fig ure 1 . 2252 equiv a lent cir cuit di a gram pin name sip pin input 1 i bias 2 gnd 3 sym 4 v- 5 cap 6 output 7 v+ 8 fig ure 2 . 2252 pin as sign ments features true rms re sponse wide dy namic range: >80 db high crest fac tor: 8 (1 db er ror) wide band width: to > 20 khz log a rith mic out put scaling low cost: $2.20 in ?000s sin gle in-line pack age matches 2180 and 2181 se ries vcas applications me ters spec trum an a lyzers com pres sors ex panders os cil la tors psychoacoustic modeling noise mea sure ment description the that 2252 in te grated-circuit rms-level de tec - tor is de signed to con vert an ac in put cur rent into a dc out put volt age. the out put is pro por tional to the log of the true rms value of the in put sig nal. the parts are housed in a space-efficient plas tic 8-pin sin gle-in-line (sip) pack age, and re quire min i mal sup port cir cuitry. based on dbx tech nol ogy and fab - ri cated in a super low-noise pro cess, the 2252 com - bines wide dy namic range with fre quency re sponse to be yond 20 khz. the log a rith mic out put is es pe - cially con ve nient for au dio ap pli ca tions re quir ing deci bel-linear scal ing. the in te gra tion time is ad - just able via an ex ter nal r/c pair. with some ex ter - nal cir cuitry, re sponse to dc is also pos si ble. dbx is a reg is tered trade mark of car il lon elec tron ics cor po ra tion that corporation ic rms-level detector that 2252 doc. 600032 rev. 02
that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com page 2 that 2252 rms-level de tec tor specifications 1 ab so lute-maximum rat ings (t a = 25c) pos i tive sup ply volt age (v cc ) +18 v neg a tive sup ply volt age (v ee ) -18 v sup ply cur rent (i cc ) 10 ma power dis si pa tion (p d ) 330 mw op er ating tem per a ture range (t op ) -20 to +75c stor age tem per a ture range (t st ) -40 to +125c rec om mended op er ating con di tions pa ram e ter sym bol con di tions min typ max units pos i tive sup ply volt age v cc +4 +12 +15 v neg a tive sup ply volt age v ee -4 -12 -15 v bias set cur rent i bias 15 24 50 m a sig nal cur rent i in i bias = 24 m a ? ? 1 ma tim ing cur rent i t 1 7.5 50 m a elec tri cal characteristics 2 pa ram e ter sym bol con di tions min typ max units sup ply cur rent i cc no sig nal ? 1 3 ma equiv. in put bias cur rent i b no sig nal ? 5 8 na in put off set volt age v off(in) no sig nal 0 +8 +16 mv sym me try volt age v sym -2 8 +18 mv out put scale fac tor e o /20log( i in / i in0 ) 31.6na< i in <1ma t a =25c (t chip ? 35c) 6.0 6.1 6.2 mv/db in put cur rent for 0v out put i in 0 i i 3 . 5 b i a s t i i 2 . 9 b i a s t i i 2 . 4 b i a s t out put lin ear ity f in = 1khz 1 m a < i in < 100 m a ? 0.1 ? db 100na < i in < 316 m a ? 0.5 ? db 31.6na < i in < 1ma ? 1.0 ? db crest fac tor 1ms pulse rep e ti tion rate 0.2 db er ror ? 3.5 ? 0.5 db er ror ? 5 ? 1.0 db er ror ? 8 ? max i mum fre quency for 1 db ad di tional er ror i in 3 100 m a ? 80 ? khz i in 3 10 m a 74 ? khz i in 3 1 m a 30 ? khz i in 3 100na 4 ? khz fil tering time con stant ( . ) 0 2 6 c i t t s out put tempco d e 0 / d t chip re: t chip = 27 c ? 0.33 ? % / c 1. all spec i fi ca tions sub ject to change with out no tice. 2. un less oth er wise noted, t a =25c, v cc = +15v, v ee = -15v. test cir cuit is as shown in fig ure 4. sym is ad justed for min i mum rip ple at v out with v in =1 v rms , 100 hz.
theory of operation the that 2252 rms-level detector is designed for high performance in audio-frequency applica - tions requiring logarithmic output, rms response, and wide dynamic range. the parts compute rms level by rectifying input current signals, converting the resulting current waveform to a logarithmic voltage, and applying this voltage to a log-domain filter. current rectification figure 3 presents a simplified internal circuit dia - gram of the 2252. the input signal current, i in , flows in pin 1, the input pin. oa1 drives the base of q3 and the emitter of q1 (through v1) to main - tain pin 1 at virtual ground potential. a negative input current (flowing out of pin 1) will tend to drive the inverting input of oa1 negative, driving oa1?s output positive, turning on q3. v1 is de - signed to cut off q1 while q3 is on. therefore, neg - ative input currents are forced to flow through the collector-emitter of q3. positive i in will drive oa1?s output negative, cutting off q3 and turning on diode-connected transistor q1. positive input current is thereby forced to flow through the collector-emitter of q1. pin 4 is nor- mally connected through a 20 ? resistor to ground (see figure 4, typical application circuit , page 4, and symmetry adjustment , page 6), so the base-emitter potential of q2 is the same as that of q1. therefore, the current in the collector of q2 ( i c2 ) will mirror that in the collector of q1 ( i c1 ), which equals the positive input current. since the input impedance of oa2 is high, the cur - rent in the emitter of q4 ( i c4 ), is the sum of the currents i c2 and i c3 . the mirror action of q1/q2 reverses the positive input currents so that they add to the negative input currents in q4. the cur - rent in q4, therefore, is equal to the absolute value of the input current. mathematically, iandii c in in in cc in in in i ii ii i 312 00 0 0 0 = ? ? ? == > ? > , , , , , ? ? ? 0 . but, iiiii c cccc 4 3231 =+=+ = ? ? ? > ?< ii ii in in in in , , 0 0 = i in . see figure 3 for definitions of these currents. logging action oa2, together with q4 and q5, forms a log ampli - fier. by using two diode-connected transistors in the feedback loop of oa2, the 2252 produces a voltage proportional to twice the log of ic4 at the output of oa2. this voltage, v log , is therefore pro - portional to the log of the square of the input cur - rent, plus a bias voltage (v2). mathematically, () vvn v t i i c s log =+ 21 4 2 () () =?+ 21 21 42 vni vni v tc ts () () =?+ vni vni v tc ts 121 4 2 2 () =?+ vni vni v tin ts 121 2 2 , where v t is the thermal voltage, kt q , and i s is the reverse-saturation current of q4 and q5 (assumed to be the same in each). that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com 600032 rev 02 page 3 + - + - + - + - + + - vlog vout 7 t v6 iin ic3 ic1 ic2 ic4 i v3 oa3 q6 q5 q4 oa2 v2 v1 oa1 q3 q2 q1 20 6 4 1 - figure 3. simplified internal schematic
computing the mean in the classic mathematical definition of rms value, the time integral of the square of the signal must be evaluated over infinite time. obviously, for a practical measurement, only a finite time is available, which leads to the question of how to weight events occuring at various times. tradi - tionally, the simplest and most meaningful weight - ing is exponential in time, giving highest weight to the most recent history, and exponentially less weight to increasingly older events. this weighting corresponds to convolution in time with the famil - iar exponential weighting function, e t ? . to accomplish this weighting, pin 6 is normally connected to a capacitor and a negative current source. (refer to the typical application circuit in figure 4. in this circuit, c t is the capacitor and r t together with v- form the current source.) this current source establishes a quiescent dc bias current, i t , through q6. over time, the capacitor charges to 1 v be below v log (the potential at the output of oa2). the instantaneous emitter current in q6 is propor - tional to the antilog of its v be , which is the differ - ence between q6?s base voltage and the voltage at pin 6. the potential at the base of q6 represents the log of the square of the input current, while the emitter of q6 is held at ac ground via the ca- pacitor. since q6?s emitter current is proportional to the antilog of its v be , the current in q6 is pro- portional to the square of the instantaneous input current. note that this antilogging only takes place for dy- namic signals. for a dc input, the output of oa2 represents the square of the input current. after charging, the external timing capacitor voltage again approaches one diode drop below v log . the exact value of the diode drop will be determined by the bias current i t . however, for sudden increases in the input current i in , the current available to charge the capacitor c t is proportional to the square of the short-term increase in input current. the ?dynamic? antilogging causes the capacitor voltage to represent the log of the mean of the square of the input current. time constants another way of looking at this situation is to con - sider the action of q6 and c t as a first-order filter in the log domain. q6 and c t establish a single pole at a frequency determined by a) the imped - ance of q6 at the bias current i t and b) the value of c t . the time constant is given below. = c t v i t t = c t i t 0 0259 . , at 300 kelvin. the result is that the voltage at pin 6 represents the average (or mean) of the square of the input signal, averaged over the time constant . the av - eraging corresponds to convolution with the time weighting of a simple rc circuit. mathematically, this is as follows: vn iedt t t in t 61 1 0 2 2 ? ? ? ? ? ? ? , where t is the time at which the average level is computed. note that e t ? ? ? ? ? ? ? represents the exponential time weighting imposed by the log-domain filter. how fast the 2252 acquires a signal (the ?attack?), and how fast it returns to rest following a signal (the ?release?), are locked in relationship to each other by the nature of the exponential time-weighting imposed by this log-domain filter. separate attack and release adjustments are not possible within the constraint of rms response. the time response for typical values of i t and c t (the circuit of figure 4) is shown in figure 5, which shows the 2252?s response to a 100 ms, 1 khz tone burst at ~ +10 dbv followed by ~500 ms of 1 khz at ~ ?30 dbv. the top trace is the input tone burst (at 10 v/div), the bottom trace is the output at 50 mv/div. the time scale is 50 ms/div. the shape of the attack and release waveforms is determined by the interaction of the exponential response of the log-domain filter with the log-representation of the signal. the straight-line decay follows from the fact that the natural release of the exponential time weighting is a decaying ex - ponential in the linear world. this maps to a straight line in the log representation. the attack in the photo appears exponential, but actually fol - lows the () 1 ? ? e t shape of the attack curve. the transformation from the linear to the log world steepens the apparent attack shape. the time constant, , also determines the amount of ripple (at frequency 2f in ) in the output for any that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com page 4 that 2252 rms-level detector v+ v+ out in 7 2 6 3 8 1 5 4 2252 v+ out cap v- sym gnd ibias in v- v- t r t c sym 50k 24k 47k 20 22m rf 1k 2m2 560k 1u rb 10u 20u cin 10u 10k rin figure 4. typical application circuit ( 15v)
given in put fre quency, f in . larger val ues of t re - duce rip ple at the ex pense of lon ger at tack and re - lease times. for f i n > > 1 4 p t , the rip ple volt age at the out put is given by: v r v f t i n ? 4 2 p t , where v r is the rms rip ple volt age. taking the square root the square root por tion of the root-mean square is im plied by the con stant of pro por tion al ity for the out put volt age: it is not com puted ex plic itly. this is be cause, in the log rep re sen ta tion, tak ing the square root is equiv a lent to di vi sion by two. the volt age at pin 6 is pro por tional to the mean of the square at ap prox i mately 3 mv/db, and pro por - tional to the square root of the mean of the square at ap prox i mately 6 mv/db. out put buffering and level shifting the volt age at pin 6 is buf fered by oa3, and level shifted down by the bias volt age v3. level shift ing is re quired so that the out put volt age will be zero when the rms in put cur rent reaches a pre de ter - mined value, i in0 . this cur rent is of ten called level match , and rep re sents the 0 db ref er ence of the cir cuit. the var i ous level shifts through out the 2252 are as fol lows: v2 rep re sents one di ode drop, so the volt - age at the emit ter of q4 is +1v be . the out put of oa2 is two di ode drops higher than this, or +3v be . q6 will sub tract one di ode drop from the out put of oa2, so the volt age at pin 6 will be +2v be . finally, v3 rep re sents two di ode drops, set ting the volt age at pin 7 to 0 v. of course, the ac tual value of all these level shifts is de pend ent on the cur rents through the tran sis - tors re spon si ble for each v be . these cur rents, in turn, are de pend ent on the bias pro gram ming cur - rent in pin 2 (i bias ) and the tim ing cur rent pulled from pin 6 (i t ). this de pend ence may be given as fol lows: i i n i i b i a s t 0 2 9 = . , where i in0 is the in put cur rent caus ing 0 v out put, i t is the cur rent in pin 6, and i bias is the cur rent in pin 2. the fac tor 2.9 de rives from the ge om e try of the tran sis tors in volved. fig ure 6 plots out put volt age ver sus in put level for a 2252 in its rec om mended cir cuit con fig u ra tion (fig ure 4). in this plot, 0 dbr ? 43 mv. fig ure 7 plots out put volt age for sev eral dif fer ent con - stant-amplitude fre quency sweeps for the same cir - cuit. the ver ti cal di vi sions are 60 mv apart, rep re sent ing ap prox i mately 10 db in cre ments. full au dio band width is main tained over a 60 db dy - namic range. cur rent pro gramming all the in ter nal cur rent sources in the 2252 are slaved to the cur rent in pin 2, i bias . as men tioned above, the choice of this cur rent af fects i in0 . i bias is nor mally set by a re sis tor from pin 2 to ground (v cc would do also). note that the volt age at pin 2 is nor mally v ee + 2.1 v, where v ee is the neg a tive sup ply volt age. sym me try ad just ment the rec ti fier (q1 through q3 and oa1) de pends on the match ing be tween q1 and q2 for ac cu rate re - that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com 600032 rev 02 page 5 fig ure 6. 2252 dc out put vs. ac in put level fig ure 7. 2252 dc out put vs. fre quency at var i ous levels fig ure 5. tone burst re sponse
pro duc tion of pos i tive-half in put sig nals. the base of q2 is brought out to pin 4 in or der to al low ad - just ment for mis matches be tween these tran sis - tors and for the in put off set volt age of oa1. pin 4 should be con nected to a vari able low-impedance volt age source ca pa ble of sup ply ing a few mil li volts (v sym ). the base of q1 is con nected to its col lec tor through a 20 w re sis tor to al low match ing be tween the base im ped ances of q1 and q2. a 20 w source should be used at pin 4 for op ti mum match ing. in put bias cur rents and volt ages oa1 will rec tify any cur rents flow ing in its feed - back net work, in clud ing the bias cur rent of oa1 it - self. in put sig nals be low oa1?s bias cur rent will be ob scured, with a steady dc out put ?floor? the re - sult. the in put bias com pen sa tion for oa1 largely can cels oa1?s bias cur rent, im prov ing low-level per for mance. even if over com pen sated, any un cor - rected bias cur rent (pos i tive or neg a tive) in oa1 re sults in a limit to the low-level res o lu tion of the 2252. any dc off set volt age at pin 1 will cause dc in put cur rents to flow if the in put is dc cou pled. the dc off set at pin 1 is typ i cally 8 mv, so the in put should be ac cou pled (as shown in fig ure 4) for most ap pli ca tions. if low-level per for mance is not re quired, dc cou pling may be used with out fur ther mod i fi ca tions to the ap pli ca tion cir cuits. how ever, dc cou pling and good low-level per for mance are pos si ble with ad di tional ex ter nal cir cuitry, as de - scribed in the sec tion on dc mea sure ments on page 9. pos i tive equiv a lent in put bias cur rent in oa2 (af ter cor rec tion) will also add to the out put level read - ing. this is be cause the equiv a lent in put bias cur - rent is sup plied through q4 and q5, which adds to the dc level at the out put of oa2. oa2, like oa1, has in put bias com pen sa tion. while gen er ally ex - tend ing low-level per for mance, the com pen sa tion can cause oa2?s net in put bias re quire ment to be - come neg a tive , tend ing to re verse bias q4 and q5. the cir cuit of fig ure 4 uses a 22 m w re sis tor, rf , to sup ply in creas ing dc in put cur rent as the 2252 out put drops. this forces cur rent to flow in q2, which over whelms any neg a tive in put bias at oa2 and pre vents the cur rent in q4 and q5 from re - vers ing. rf is re spon si ble for the flat ten ing of the low-level re sponse shown in fig ure 6 (page 5). the feed - back cur rent pro vided en forces a lower limit to the re solv able in put cur rent. with out this feed back, those 2252s where the oa2 in put bias is overcorrected will ex hibit a sharp down ward swing in out put level at low in put sig nals (see fig ure 8). for many ap pli ca tions, such low-level re sponse is not nec es sary. in such cases, rf may be de leted, as shown in fig ure 11. see fur ther dis cus sion of this sub ject in out put con sid er ations , (page 8). ap pli ca tions the in tended pur pose of the 2252 is to com pute the log of the time-weighted root-mean-square of an in put cur rent sig nal. sev eral prac ti cal con sid er - ations ap ply when at tempt ing to make full use of the 100 db dy namic range of the de vice. in put con sid er ations the 2252 in put is in tended to ac cept a cur rent: as is clear from fig ure 3 (page 3), pin 1 is a sum ming junc tion sup plied with in ter nal neg a tive feed back. max i mum in put cur rent is on the or der of 1 ma, lim ited by in ter nal de vice char ac ter is tics. the min - i mum prac ti cal in put cur rent is de ter mined by the ef fects of in ter nal bias cur rents (whether from oa1, oa2, or the ef fect of v1). at cur rents as low as 31 na, ap prox i mately 1 db er ror in read ing will re sult from interal bias cur rents. that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com page 6 that 2252 rms-level de tec tor fig ure 8. low level out put from sev eral 2252s (no rf) v+ out in 7 2 6 3 8 1 5 4 2252 v+ out cap v- sym gnd ibias in v- v- t c r t 20 22m rf 1k 2m2 750k 1u rb 10u 20u cin 10u 10k rin fig ure 9. ap pli ca tion cir cuit with out sym me try con trol
the gain of the in put op amp (oa1) is suf fi cient to pro vide ef fec tive log ging for r in 3 10 k w . how ever, with r in < 10 k w ., low level and high fre quency per for mance will suf fer due to the fi nite open-loop gain of oa1. this is be cause oa1 is re quired to swing across a ?dead zone? be tween turn ing on q1 and q3. the dead zone is re duced by v1. but, v1 is small ( ? 0.5 v) to main tain low-level track ing, so oa1 must still swing through sev eral tenths of a volt at each re ver sal of the in put po lar ity. the smaller the value of r in , the higher the loop gain de manded from oa1 for ac cu rate rec ti fi ca tion. there fore, for good high-frequency per for mance, r in should be 10 k w or larger. oth er wise, choose r in based on the de sired in put volt age at level match, e in0 , and i in0 as fol lows: r i n e i i n i n = 0 0 the neg a tive in put of oa1 typ i cally rests at +8 mv. if dc cou pled, this will cause an in put cur rent to flow which will ef fec tively set a low-level ?floor? be - low which read ings will be ob scured. there fore, ac cou pling is re quired if low level sig nals are to be ac cu rately ob served. choose the value of the ac cou pling ca pac i tor (c in in fig ure 4) based on the value of r in and the de sired low-frequency limit. c i n r f i n c = 1 2 p , where f c is the de sired 3 db-down point. (for dc cou pling, see the sec tion on page 9, dc mea sure ments .) sym me try ad just ment as noted ear lier, the rec ti fier re lies on the match - ing of q1 and q2 for pre cise re pro duc tion of pos i - tive-half in put cur rents. q2?s base is brought out to pin 4 to al low ad just ment of this match. pin 4 should be con nected to a 20 w volt age source ca pa - ble of sup ply ing from ?4 mv to +20 mv. the ap pli - ca tion cir cuits in fig ure 4 (page 4) and fig ure 11 (page 6) are typ i cal. to set the sym me try, ap ply a low-frequency sine wave to the in put. nei ther the fre quency nor the level are crit i cal: 100 hz at near level match is usu ally a good choice. ob serve the out put wave - form with a ?scope while ad just ing the sym me try trim. with proper ad just ment, the rip ple in the out put will be al most pure sec ond har monic of the in put fre quency. no fun da men tal fre quency should be pres ent in the out put. an other method would be to sense rip ple in the out put via a nar row bandpass fil ter cen tered at the fun da men tal feed - ing an ac volt me ter: ad just the trim for min i mum volt age read ing. the ac tual volt age re quired for proper sym me try de pends on the in put off set volt age of oa1 (typ i - cally +8 mv), and the v be mis match be tween q1 and q2 (< 6 mv). for less crit i cal ap pli ca tions where pre cise rec ti fi ca tion is not re quired, pin 4 may be con nected to a volt age match ing the in put off set volt age of oa1, pref er a bly through a 20 w re - sis tor (to match the 20 w in q1?s base). the sim - plest cir cuit con nects pin 4 to pin 1 through a 20 w re sis tor, as shown in fig ure 9. this con nec - tion en sures that the v be of q1 will equal that of q2, but does not al low for ad just ment for any mis - match in the two de vices. when us ing this con fig u - ra tion, one should try to keep the bias pro gram ming cur rent be low 20 m a, to en sure sta - ble op er a tion. time con stants both the ca pac i tor (c t ) and cur rent source (i t ) con nected to pin 6 con trol the time con stant over which the rms value of the in put cur rent is eval u - ated. ei ther may be var ied, but a few prac ti cal con - sid er ations in flu ence the choice of val ues. first, the in put bias cur rent of oa3 in fig ure 3 will add to the charg ing cur rent i t . for small val ues of i t , this will af fect the ac cu racy of the re sult ing time con stant. the in put bias cur rent for oa3 is typ i - cally 100na, so i t should be kept above 1 m a. at the other ex treme, i t flows from oa2 and through q6 un der steady-state con di tions. dy - namically, as was men tioned on page 4 in com - puting the mean , the cur rent which charges c t is pro por tional to the square of the short-term in - crease in i in . a sud den 30 db in crease in in put causes a 60 db in crease in charg ing cur rent. for ex am ple, if i t is 10 m a, the peak charg ing cur rent will be 10 ma. how ever, if i t were 100 m a, the peak charg ing cur rent called for would be 100 ma. the de vices within the 2252 will not sup port this high a cur rent de mand, so tim ing will be less ac cu - rate un der this con di tion. (the steady-state out - put volt age will be ac cu rate, since over time the er ror will be ?for got ten? due to the time-weighting e t - t . how ever, the dy namic re sponse will be in cor - rect.) the best per for mance nor mally re sults when 5 ma < i t < 50 ma. c t is usu ally cho sen to be a con ve nient value which re sults in the de sired time con stant t , and an i t within these lim its. see time con stants (page 4) for the re la tion ship be tween t , i t and c t . since c t is com monly 1 m f or larger, this ca pac i tor is of ten a po lar ized elec tro lytic or tan ta lum. v6 is nor mally pos i tive ap prox i mately 2v be ( ? 1.4 v), which de ter mines the po lar ity of c t . leak age in c t will add to i t , so low-leakage electrolytics or tantalums are pre ferred. tan ta lum, once pre ferred for long-term sta bil ity and low leak age, has been largely superceded by pre mium elec tro lytic types which are avail able in low-leakage, high-stability ver sions. when us ing an elec tro lytic, how ever, it is good prac tice to se lect a volt age rat ing higher than ab so lutely nec es sary (at least 15 v is rec om - mended). when a re sis tor to the neg a tive sup ply is used as the i t cur rent source (as in the cir cuits shown here), de ter mine its value by the equa tion: that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com 600032 rev 02 page 7
r t v 1 . 4 i e e t = + , where v ee is the neg a tive sup ply volt age. bias pro gramming as shown in fig ure 1 (page 1), pin 2 is con nected to the neg a tive sup ply through three di odes in se - ries. its typ i cal volt age is v ee + 2.1 v. the cur rent in these di odes, i bias , con trols the cur rent sources through out the 2252, and there fore af fects the band width of the de vice, as well as i in0 . be cause it is re spon si ble for pro gram ming so many cur rents, i bias should be re stricted to a rel a tively nar row range about the nom i nal 24 m a. normally, i bias is cho sen to fine-tune i in0 , the 0 db ref er ence point. i 8 . 4 b i a s i t = i i n 0 2 . the fac tor 8.4 de - rives from the ge om e try of the de vices in the 2252. as with the value of r t , r b is de ter mined by the de - sired i bias and the neg a tive sup ply volt age, as fol - lows: r b v 2 . 1 i e e b i a s = - . power sup ply by passing the 2252 is not par tic u larly sen si tive to power sup ply im ped ances, but the high charg ing cur rents which are pos si ble in c t must be kept out of the pc board ground and volt age sup ply sys tem for good re sults. the ap pli ca tion cir cuits show a 1 k w re sis tor in se ries with the pos i tive sup ply pin (pin 8), and a 10 m f ca pac i tor by pass ing this pin to ground. the ground end of this ca pac i tor is in - tended to be con nected di rectly to the ground end of c t . one (and only one) point on this trace should be con nected to the cir cuit ground sys tem, since high cur rents flow in this trace dur ing in put sig nal tran sients. fig ure 8 il lus trates the flow of cur rent, i charge , dur ing an in put tran sient. note that i charge is con fined to a lo cal loop, flow ing from c b through oa2 and q6 (within the 2252) into c t , and re turn ing di rectly to c b . by keep ing the re turn trace short, and con - nect ing it to sys - tem ground at only one point, high cur rents are kept out of the over all ground sys tem, while c t main tains its ref er ence to ground po ten tial through the sin gle con nec tion to ground. an other by pass ca pac i tor is re quired at pin 2 (the i bias pin). typically, with i bias ? 24 m a, the im ped - ance at pin 2 is ap prox i mately 3 k w to v?. if the in - put sig nal ca pac i tively cou ples into pin 2, i bias will be mod u lated at the in put sig nal fre quency. this mod u la tion adds sig nal volt age (as rip ple) to the out put at pin 7. (re call that, ide ally, the 2252 pro - duces rip ple only at the sec ond har monic of the in - put sig nal.) when the 2252 is used to con trol the gain of a volt age-controlled am pli fier (vca) such as the that 2180 or 2181, this fun da men - tal-frequency rip ple will cause sec ond har monic dis tor tion in the vca. by passing pin 2 to the neg a - tive sup ply shunts any sig nal-related cur rents around the 2252 bias sys tem, pre vent ing this rip - ple from occuring. out put con sid er ations the out put of the 2252 (pin 7) pres ents a low source im ped ance, but it is some what lim ited in avail able cur rent drive. re ferring to fig ure 1 (page 1), the neg a tive cur rent source at the out put pin is ap prox i mately 10 i bias . there fore, for the typ i cal i bias of 24 m a, the max i mum cur rent sink at pin 7 is 240 m a. (the 2252 will source con sid er - ably more than this cur rent if re quired.) since the out put volt age nor mally swings be tween 300 mv, the 2252 will di rectly drive loads of greater than 1.5 k w with out dif fi culty. in the sec tion on in put bias cur rents and volt - ages (page 6), it was noted that the 2252 out put is gen er ally pro por tional to the log of the ac rms in - put cur rent. but, at low lev els (<20 na in put cur - rent), the in put bias com pen sa tion in in ter nal opamp oa2 in fig ure 3 (page 3) can cause a sharp in crease in the out put change with in put, as shown in fig ure 8, (page 6). since this ef fect is de pend ent on the ac cu racy of the bias com pen sa tion, some 2252s will ex hibit this be hav ior, but most will not. for most ap pli ca tions, where con sis tent low-level re sponse is desireable, the 2252 should be con fig - ured to sup ply its own ?cor rec tion? sig nal to pre - vent this dis con ti nu ity from occuring, as shown in the cir cuit of fig ure 4, (page 4). in this cir cuit, a 22 m w re sis tor (rf) is con nected from the out put of the 2252 back to the in put. when the out put volt age is near 0 v, rf con trib utes neg li gi ble cur rent to the in put of the de tec tor. for pos i tive out put volt ages, the cur rent in rf is very small com pared to the ac in put sig nal which causes a pos i tive out put, so its ef fect is again neg li - gi ble. how ever, as the ac in put sig nal drops in level, the dc out put goes neg a tive, sup ply ing more and more in put cur rent through rf, which tends to drive the dc out put pos i tive. the point at which the two ef fects will bal ance is de pend ent on i bias , i t , and the size of the feed back re sis tor. in the cir - cuit of fig ure 4, the out put will ?bot tom out? at ap - prox i mately -300 mv, where rf sup plies ~14 na of dc cur rent to the in put. rf es tab lishes a sta ble ?floor? in the 2252?s re - sponse. how ever, this self-supplied lower limit will ob scure very low-level in put sig nals which could be mea sured by those 2252s with good in put-bias cor rec tion at oa2 of fig ure 3. if ac cu rate low-level that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com page 8 that 2252 rms-level de tec tor v+ within 2252 oa2 part of 8 6 charge i 1k b c t c q6 fig ure 10. de cou pling sup ply transients
re sponse is quite im por tant, the 2252s may be se - lected, and rf may be omit ted, as shown in fig - ure 11. the 2252s should be se lected on the ba sis of ac cu rate low-level, low-frequency per for mance (see fig ure 8, page 6). the cir cuit of fig ure 11 may also be used with un - se lected 2252s for ap pli ca tions where ac cu rate low-level per for mance is un nec es sary. how ever, the rel a tively low cost of a 5% 22 m w re sis tor usu - ally ar gues in fa vor of its use. dc mea sure ments as noted ear lier, the 2252 is pri mar ily in tended for mea sur ing the rms value of ac sig nals. this fol lows from the 2252?s rel a tively high in put off set volt age (typ i cally +8 mv). if the in put is dc cou pled, the in put off set volt age will cause a dc cur rent to flow in the in put, which will in ter fere with pre cise low-level rec ti fi ca tion, ul ti mately pro duc ing a ?floor? be low which in put sig nals will be ob scured. for typ i cal val ues of r in ? 10 k w , this dc in put cur - rent would be 800 na, up set ting rec ti fi ca tion for ac sig nals be low ap prox i mately 8 m a, and al most com pletely ob scur ing sig nals be low 800 na peak. c in in fig ures 4 and 11 blocks this dc cur rent, ex - tend ing low-level per for mance to well be low 100 na. how ever, as shown in fig ure 12, with a few added parts, it is pos si ble to ex tend re sponse to dc with lit tle loss in low-level ac cu racy. the es sen tial pur pose of the added cir cuitry (oa1, oa2 and as so ci ated parts) is to buffer the in put off set volt age (at pin 1) and add it to the in put sig - nal. the cir cuit of fig ure 12 op er ates as fol lows. as has been noted, pin 1 is a vir tual ground. the o - retically, the volt age at pin 1 is only the dc off set of the 2252?s in ter nal in put am pli fier, v off(in) . prac - ti cally, the volt age at pin 1 will con sist of this off set plus a small ac sig nal de ter mined by the value of r in and the loop gain of the in ter nal am pli fier. oa2 is in tended to buffer v off(in) with out draw ing any ad di tional cur rent from the pin 1 node. for the out put of oa2 to be an ac cu rate rep re sen ta tion of v off(in) , oa2 must be a low drift, low in put bias cur rent opamp, and should have ei ther low in put off set volt age (< 0.1 mv) or be trimmed to have low in put off set. typ i cal choices would in clude an op07 or an lm108a with a trim for off set volt age. the volt age at oa2?s out put is di vided by 2 through the two 10 k w re sis tors r 1 and r 2 . the 1 m f ca pac i tor by passes any ac sig nal at the out put of oa2 so that only dc is pres ent at the noninverting in put of oa1. as suming that the in - put to the en tire cir cuit is con nected to a 0 w volt - age source, the volt age gain from the noninverting in put of oa1 to its out put will be 2.0, so the com - bined gain from the out put of oa2 to the out put of oa1 is unity. in this fash ion, the dc volt age at the out put of oa1 is equal to v off(in) plus the in put volt age. un der no-signal con di tions, no cur rent flows in r in . with sig nal, this cur rent is v r i n i n . of course, oa1 must also be a low-offset opamp, though its in put cur - rent is not as crit i cal as that for oa2. typ i cal choices here would in clude an op27 or an lm308a type with off set-voltage trim ming. an ad di tional con sid er ation is that r 1 , r 2 , r 3 and r 4 should be pre ci sion tol er ance types, or the volt - age ap plied to r in will not ac cu rately rep re sent v off(in) . the source im ped ance of the volt age to be mea sured will also af fect the gain to v off(in) . for well-defined, but non-zero source im ped ances, re - duce the value of r 3 by the value of the source. for un con trolled, non-zero source im ped ances, the in - put sig nal should be buf fered. that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com 600032 rev 02 page 9 v+ v+ out in 7 2 6 3 8 1 5 4 2252 v+ out cap v- sym gnd ibias in v- v- t c t r sym 50k 24k 47k 20 1k 2m2 1u rb 560k 10u cin 20u 10u rin 10k fig ure 11. ex tended low-level range with se lected 2252s v+ v+ out in 7 2 6 3 8 1 5 4 2252 v+ out cap v- sym gnd ibias in v- v- + - + - t c t r rin r 1 10k r 10k 2 r 4 10k r 10k 3 oa2 oa1 oa2 oa1 1u 10k sym 50k 24k 47k 20 rf 22m 1k 2m2 1u rb 560k 10u 10u fig ure 12. mea suring dc and ac signals
with this tech nique, the fre quency re sponse is ex - tended to dc, while a slight loss in low-level ac cu - racy re sults from the ad di tional bias cur rent drawn by oa2. tem per a ture-compensated mea sure ments in cer tain ap pli ca tions, the 2252 may be used for high-precision mea sure ments where the 0.33% / c vari a tion in out put scal ing may be un ac cept able. for these sit u a tions, the out put should be tem per - a ture com pen sated through the use of a 3300ppm / c re sis tor. such re sis tors are avail able from rcd com po nents, inc., 3301 bed ford st., man ches ter, nh, usa [(603)-669-0054] and koa/speer elec tron ics, po box 547, brad ford, pa, 16701 usa [(814)-362-5536]. fig ure 13 shows a typ i cal cir cuit. for good re sults, r1 should be lo - cated phys i cally close to the 2252 to op ti mize tem - per a ture track ing. un usual ap pli ca tions since the 2252 con tains a pre ci sion rec ti fier fol - lowed by a log ging stage, it is pos si ble to think of other ways to ap ply the part be yond its con ven - tional use as an rms-level de tec tor. log am pli fier in ap pli ca tions where pre ci sion log ging is re - quired, the rec ti fier can be en tirely by passed by driv ing the in put with a sin gle-sided dc sig nal, and the logged out put used di rectly. in such cases, the tim ing ca pac i tor (c t ) con nected to the pin 6 should be re duced to a small value (1~10 nf) to quiet noise in the out put while avoid ing sig nif i cant de lay in the re sponse. log-responsive peak de tec tor where peak de tec tion is de sired, the 2252 may be con fig ured with very fast time con stants (as in the log am pli fier above), and fol lowed by a peak hold am pli fier. the ad van tage of us ing the 2252 for this ap pli ca tion is that the 2252 pro vides rec ti fi ca tion and log ging be fore the peak de tec tor, which al lows ex tremely wide dy namic range in the re sult ing sig - nal. closing thoughts that cor po ra tion wel comes com ments, ques tions and sug ges tions re gard ing this de vice, its de sign and ap pli ca tion. our en gi neer ing staff has ex ten - sive ex pe ri ence in ap ply ing this part to com mer cial cir cuits. we are pleased to of fer as sis tance in op ti - miz ing your cir cuitry to adapt the 2252 to your par tic u lar ap pli ca tion. please feel free to con tact us with your thoughts and ques tions. that corporation; 45 sumner street; milford, massachusetts 01757-1656; usa tel: +1 508 478 9200; fax: +1 508 478 0990; web: www.thatcorp.com page 10 that 2252 rms-level de tec tor v+ v+ out in 7 2 6 3 8 1 5 4 2252 v+ out cap v- sym gnd ibias in v- v- c t 10u r t 2m2 oa1 r1 2k +3300ppm r2 2k sym 50k 24k 47k 20 rf 22m 1k 1u rb 560k 10u cin 20u rin 10k fig ure 13. tem per a ture com pen sating the 2252 i k l g e typ. f b d c 1 a n m h n 17.78 0.3 0.700 0.012 j millimeters 19.5 +0.2/-0 1.25 0.65 0.85 2.54 0.2 0.9 1.2 5.8 +0.2/-0 2.8 +0.1/-0 10.5 0.5 1.3 0.3 3.5 0.5 inches 0.77 +0.008/-0 0.049 0.026 0.033 0.100 0.008 0.04 0.05 0.23 +0.008/-0 0.11 +0.004/-0 0.413 0.02 0.05 0.012 0.14 0.02 item a b c d e f g h i j k l m fig ure 14. sip pack age out line
|
