Transformer Design for “Zero” Impedance Amplifiers N. R. GROSSNER® A rigorous discussion of a practical method that can reduce weight and volume of an output transformer for zero-impedance output stages without increasing distortion. 1, Eliminate the output. transformer entirely. Apparently thie procedure a: stitutes new problems for old ‘2 Uso special eleeuitry. to. over the “limitations” of the output. tran former. MacIntosh’ and Peterson? using high-quality transformers of potial de sign have achieved. notable succes, Devise cinouitey” whieh overtomes transformer ‘limitations,”” using sull and. comparatively inexpensive output transformers, ‘This article is eoneernerl with a development fn this category. The “Zero-Impedance Transformer” ‘This writer’s experienee indicates that 4 compartively small output transformer specifically designed for the “sero” im- pedance output stage* provides per- formance characteristics at least com: parable with the “large” transformer designed for the conventional high-qual: ity feedback amplifies ‘The advantages this type of trans. former affords when fed by a zero. impedance stage are us follows 1. Small size. Typical weight reduction ‘approximately "40. Yor cent (Seo ‘anLE'T). Te appears feusible to desig, a high-quality ‘output transformer ‘only ‘nes the size of a O0eyelD *David Bogen Corp, Transformer Divi sion, 29 Ninth vey N.Y. 14, NY. “Ataetntonh & Gow, Deseription and analysis of @ new S0-swatt amplifier elt- cuit,” AUDIO ENGINEERING, Dee. 104. TA. PL Pe ‘neve push-pull aim plier ‘cireuit”” General Radio" Baper Imenter, O Sr, ‘Miller, "Combining five feodback,”™ Blectronte, ire and ‘Mareh, A, Wilkins, Pat, Ponding: ‘Gon trolled Positive Becaback. AUDIO. ® MARCH, 1956 power transformer of the same power ating ‘The smallest. size is" obéainnd by uning a B supply with perfect reg tion "sud fixed baa 3. Wider bandwidth 5. Lower inter tancereduees ““avitehing™” transients ‘L Lower effective primary capaci tance, High transformer. input, eapact {ance limite the amount of low-distortion high-frequency" power froin those elaes AB, and B, amplifiers that utilive larg en bitter windings. The foregoing. advantages, contib ute tothe foantity of Mighpertormanee wide-band “Classy simpler employing’ a regalated 1 axp Dy eee ) esl) ity (3) Primary BO yoda w XL ‘The voltage drop aeross (q+ R,) is 1a) (e+ fe) ~itul a +R) (3) +1.) (Re + R,)] ~jla(Re+ By) ‘oy Since Ty js the quadrature magnetizing curment “containing. the distortion har = [le TABLE 1 POR CAMINA- stack Wan 1 E012 1% 328 35 E25 18569 70 est i 866 former appears tie Zero i dance output rept continued eed. for the typo of laborato ‘winding schedule often ‘used fn" high-quality Inego transformers, 3, The continued employment of the same highquality core maleri sm uted than What of the large’ traneformer, foot is higher than the small Tk.ty franeformer which it may euperteiay resemble beeause of ita smaller. size Distortion Low distortion has a decisive effect ‘on the size of the ontpnt transformer. ‘The equations pertaining to distortion ed on the equivalent Iow-frequeney cient (Pig: 1), tll Exciting current, Le=le~ ilu in-phase core loss current o where Io and Jy.= quadrature magnetizing eurrent total primary eurrent TYPICAL WEIGHT REDUCTION DATA (PEOANCE WoW REBUETION is 155-359 V3 439 ve 38% monies, it is clear that the distortion producing voltage drop Ty(Rg+2,) subtracted from the input voltage By produces a component of the voltage Ey, having a distortion term across the pri- mary reactance X;, If we assume Ty, is all harmonies, then the maximum frac. tional distortion Dp appearing across the effective primary reactance is Fig. 1. Low-frequency equivalent circuit. 2or Dp=(RatR)/X, (9) and the distortion factor due to the transformer alone is Dy=R,/Xy, (10) ‘The EMF, Ep, across the primary shunt reactance (with the distortion eompo- nent), in turn causes the load eurrent , to flow through the secondary de, re- sistance and load resistance. Since we expeet to handle the distor- tion parameters in some detail, it is now Aesirable to establish the exact transfer ratio between input and output. Tf we solve each loop of the network (Fig. 1) using Kirehoft's Law we ean write the exact equation: (12) Multiplying these two equations and Separating the real and quadrature Ry Re )( tence Bagi) 4 1+R,/By)(1+ R,/Re+R,/B) so that the transformer distortion factor (again assuming that the magnetizing current Ty is all harmonies) is the j term in Bq. (14), D, R, 1 la TyRrR yA) or D=R,/X, (16) What is especially signiticant is that if Rg R, or better sil if RO, then the total cireit distortion is determined by the transformer alone Size of the Conventional Transformer Before attempting to reduee the size ‘of the conventional output transformer Jet us see why it is larger than a power transformer of the same rating and at the same time derive an expression for its size in terms of copper losses and distortion. Assuming 2,=,, which is the usual case transformer designed along traditional lines, its loss factor 1/Q, will be 1/Q,= R,/Xj+ Be/Xy, X;, is established after a siderations: (a) The lowest froqueney the amplifier is to pass in accordance with relative low-frequeney response in Xz, (18) number of cone "The writer is indebted to D. Widfeuer of Arma Corporation for this simple for. fiuintion of maximum distortion, 2B a= 2008 7 (a9) oe TIX, where 4/By= [1/ (Rg + H)] + 1/R; (20) (b) ‘The maximum permissible distor- tion, at high power of the flux density Byy in the pri Although the maxinnun Dy= (Rag R,)/Xy, wo the inductance, being a non-linear fune tion depends on what the effective core permeability y, (see Fig. 2), ig for a °n By, which is, in turn, fonetion of the maximum voltage across the p mary, dance with 444 NAB yf 10 ry turns, (22) lowest frequeney of operation. ‘Then to be certain that at low-frequency high power levels we do not exeeed Dy, ‘we must select a reasonable value for By to prevent Xy, from falling below the minimum established by Eq. (0). (e) In addition to the foregoing, the Toad line at high power also must be considered, At power levels near maxi mun, the effeetive load on the output uhes becomes reaetive, the load line he- comes “elliptical,” and Ry inereases due to phase shift in the feedback loop. ‘The impedance Z of X;, in parallel with Ry, reduces the effective plate load impe- dance to a value lower than the optim (established at mi@-frequeney) required for maximum power transfer. The lower impedance inereases* the voltage drop across Rg (power wasted), thereby de- ereasing the available Tow-distortion power at the lowest operating frequency. Tt is the last two considerations which “«Bleetronic Transformers and Wiley 2a ed. 1955 pp. 158, 107, Fig. 2. Flux density ve Mv. (Courtesy Magnetic Metals Con, Cam den, No J) s 0 a then traditionally have the most bearing oon the size of X,, Therefore X, is estab. lished by Eq. (9) and the ‘Tollowing relationship X,= aR, (23) where d, is usually an integer in the range of 3 to 6 (empirically determined) in order to keep the “ellipse a3 narrow as practicable” in the high-quality ventional output transformer ‘The weight WW, of the transformer (whieh we will also refer to as. the standard”) after combining Eqs. (18) and (23) may be expressed by Wy Qu & (dyRy/2R,)"”* en Size Reduction Methods Equation (24) suggests several ways of reducing the transformer si (a) Tnevease the copper losses by K. that is, to 2R,K/R,. This however would produce moze distortion than the conventional transformer in a 2ero-im- pedance cireuit. (b) Reduce the primary reactanee by K; that is, make new reaetanee X;/K. ‘This would also produce move distortion from the small’ transformer than the ge one used in a zero-impedanee stage, (e), Make the distortion of the small same as the large one A special this will transformer th in the zero-impedance stage. procedure for accomplishing be deseribed. A Special Design Procedure Equation (15) tells us we tain zero distortion if 2, 0 eed to zero, Although this is not phys MILT, Stat, ‘Magnetic Circuits and Transformers!” Wiley: 1943, p. 228, dnerein it fe: demonstrated. that Q se Hat, 1K, lava factor by which the Tixear jon of the transformer is changed, therefore ‘proportional to. volume'?* Cand weight?) AUDIO. © MARCH, 1956ieally possible, we ean make 2, very low by deerensing it from its enstomary value B, =, and increasing R,. This is tanta mount to putting most of the copper Tosses into the secondary. In transformer vernucular—run the secondary “hot” fand the primary “cold.” The transformer designer can now proceed as follows: Keep the new transformer copper eff- cieney the same as that of the standard, Int make new transformer primary de. resistance By=R,/K, (a) where Kis the factor representing the degree to whieh we expect to reduce the size of the standard, and new trans- former secondary resistance Ry Ry=R,(2-1/K). Hy + Ry=Ry+R, New transfor same as Targe (26) wy sr primary Qy is th so that O1=Q.=Ki/Be (23) where XyoXi/ie ‘Transformer secondary Qn=Xa/Ruy (30) n= (RaQ) /TR(2-4/K)). (31) Now total new-transformer dissipation 1/Qx is 1/Qx=(1/9,) +K(2-1/K)/Q, (82) =2K/Q, (33) 1/Qx=2KR,/(alt,) (34) And snultiplying (34) by (24, 0/0.=K, (35) ‘The weight ratio of large and small transformer is then W,/We=K (36) Examine the consequences of this pro- cedure. The total distortion with our standard transformer is Dy=(Ro+ R/S, (9) an of our modified transformer is R/K Deri M/S, (8) And comparing the distortion factors Dy and D, by dividing (9) by (37) we have Dy/Dg=1+(Ra/B,) (38) ‘The intrinsic distortion of the standard transformer is the same as that of the ler transformer. But when the standard is used with a substantial source resistance, it yields more distor- tion than our new smaller transformer in a zero-impedanee output stage. Here we have a teebnique that looks very promising. However, having ig- nored a number of important. param- ters, we should try to see what prac- AUDIO. © MARCH, 1956 tieal limits smust be assigned to K by studying each negleeted parameter 1, lox density and non-linear nature ot es 2, Power supply regulation at low Arequeneless Pore Tosss 44. Temperature formers 3 Bandwith; and 6, The Class’ By. high-troqueney re sponse ial regulation rise of the trans 1. Flux Density and Distortion (a) When X;, is reduced by a factor of KK, the flux density is generally in fereased (sce Hq. 22), 80 it is essential to insure that yy in the equation L=Bd9NtAlO*/t (89) where 1=length of magnetic path oes not fall below the value needed to maintain 1,=1,/K. (b) Te is also necessary now to reeon- sider the previous formulation for dis. tortion which was based on the supposi tion that all of the magnetizing eurrent was harmonic, Because, as Partridge hhas shown, the pereentage of third and fifth harmonies in the magnetizing enr- rent is funetion of the operating flux density. Table IT summarizes such data TABLE II Typical Silicon-stcel_Magnetizing Current Harmonic Components with Zero-Impedance Source? 51000 10/000 (On the bacis of this data we shall define Kg as the factor by which the distortion inerenses when operating at B, (due to size reduction) rather than Byy (the fx Gensity of the standard transformer). Then Kuk/%, (40) and Dy (Rat RX, _ (5, Ra Dee (Xa) Ke” (OR, If, for example By=5000 gauss and By=10,900 gauss then Ky 45 (noglecting fifth harmon tion (11) is therefore a more accurate version of Eq. (98). For help in designing the new trans. former, ‘Table TIT lists likely values of Ky and the corresponding values of Rj/Be, and DzDy TN, Partridge, «Harmonie Distortion in af. transformers,’ Wireless 21g Bopt-Nov., 1042. PIR Lee, op ey p. 163, TABLE 11h RELATIVE DISTORTION ® cE Os 2 10 05 1s 19. 1 2 133 1 1s 15 2 2 20 2 ts of 15 and 2 for Ky are delib- chosen to represent the probable ‘maximum distortion increase if the flux density doubles due to-2 choiee of K == 2, ‘Table IIL indicates that the small trans former produces the same or less dis. tortion than the large, provided 1 (and thereby Iq) is not made too large. (e) Since there is nsnally some degree ‘of output tube unbalanee, polarizing de, in the primary has two important celfeets: (1) the effective wy is decreased, and is now obtained from the family of curves in Fig. 2 after estimat- ing the ampere-turns per inch, #2: W=NI/i us primary turns Zing de. current length of magnetie path, inches (2) the magnetizing: earrent now eon- tains even harmonies as well as ‘odd. It is therefore desirable when reducing the size of the standard transformer that H be kept the same (or sn easily nec 2. Power Supply Regulation Power supply regulation is quite im- portant to the high-level low-frequency performanee (a) In Fig. 3 neglect Ro (to be studied Inter), and 2, (since Ry € Ry First assume perfect power supply 2 ulation so that the output tubes must ish the following V (8) Byt zn, a VA=P,-JP/d (48) VA=P(1~j/d) (46) where P= real watts audio 3. Low-frequency impedance, 2»Equation (45) tells us that at high power levels P,/d reactive VARS will be supplied to the magnetio eore. Since the VA furnished by the output eomes ultimately from the plate supply, it is apparent that distortion is higher at high power levels when using an unrega- lated power supply. We can now attempt to establish su able criteria for d,=X,/(KRy)- ‘The VA ratio between small and large transformers is VA, 1-iK/d, Va,” 1-i/a, VETTE (yy) Vie ia If we assume certain values of d, we can tabulate (Table IV) the VA increase TABLE 1V VA INCREASE & CWA, 1010 Md aise 1.055) 022) ate 045 2 18 127 104 i os 159 202 demanded of the power supply at high: level low frequencies. If, for example d,=3 at lowest use- ful frequency and is redueed by i= 2 (a weight saving of 65 per cent) only 1 per eent more VA is demanded from the power supply at that frequency. Under these conditions the measured dis- tortion in the zero impedance eireuit at the lowest frequency is found to be un- usually low, and easily comparable to the distortion figure for the lange trans- former in a conventional feedback am- plifer () When Rg =0, but the power sup- ply is not perfectly regulated, experience indieates that a quite sizeable but less Aramatie size-reduction is feasible. Ex- amination of Eq. (41) has suggested to the writer a more modest value of in ‘this ease. Experience has shown that with a choice of K'= 1.4, a weight saving of about 40 per eent is obtainable with a distortion figure equal to or lower than that of the large transformer in anu regulated conventional feedback ampli fier. 3. Core Loss Analysis of the effect of increasing core loss (decreasing Re) when K=2 indicates a nogligibly small change in high-level low-frequeney performance, 0 core loss may safely be ignored ex- cept perhaps when K>2. This would be the case where we sueceeded in re- ueing the output transformer to the size of a 60-cps power transformer of the same power rating. 4, Temperature Rise Since our (low-distortion) conven- tional output transformer is much 30 larger than its equivalent power trans. former of the same rating, it runs “eool,” that is its temperature rise ig in the ap- proximate range of 10°-20° C. If we use an approximate equation for temperature rise"? 0,= eP,/w*! 0,= temperature rise, °C. (4s) where =a constant P= losses w= weight ‘The relative temperature rise of the new transformer will he So that our new transformer rise for K=2 will be in the range of 20° to 40° C. A temperature rise of 40° to 55° C is permitted for military and com- mercial transformers, respectively, with Class A. (105° C final temperature) in- sulation, On a thermal basis, therefore, the conventional transformer may be re- duced in weight by at least 65 per cent. 5. Bandwidth Tf we neglect the various winding capacitances and shunt core loss Tc, the following bandwidth equation js in- formative (sce Fig. 4) Te () and (a3) 20 (33) where M=coil mean length turn b=winding length @ ulation between windings a=tolal eopper depth of wind- ings Substituting the equations for Zy and Try into (30) we get fy R, Berfeom Rt (50 = = —=—srermme ing the geomiy of the eons if Rg, — By substituting Ry and Ry in (55), we get for the bandwidth of the small transformer nyik+r, (2g “Re (06) 4, (GR, + By) ite Dividing (37) by (54) ite (x7) Fig. 4. High-frequency equivelent circuit R, L, (50) where low-frequeney Ry=RgtR+R+R, (52) 41/Ry=1/(Ro+ Ry) + A/{Re* Ry) (20) ‘otal Teakage inductance, leakage reactance L,=primary shunt induetanee ‘The relative high-frequency response is Bee, op cit. p, 60. This equation, while" intended “for” ange transformers serves our purpose here. SM stat "Magnetic Cireuits and ronsormera Mey 14 yt “oo Terman, “Radio Bnginters” ian ok? Moral: 1048p. 988, Fi. 26 Ey (8) impedance output stage has at Teast K times the bandwidth as the large. 6, Class B, High-Frequency Performance ‘The “K-modified” transformer has im- portant beneficial results in a class B, amplifier with zero source resistance and a well regulated plate supply. (a) The interprimary (half-primary to half-primary) leakage reactance is lower. If K= 2, the dreaded “notch” due to switching transients'? is moved up an ‘oetave, approximately. (Continued on page 69) AP, Sah, ‘Quasictransiente in Class B auclio-froquéney amplifiers." Prov. TEE Nowe 1036.7 : AUDIO © MARCH, 1956AUDIO. © MARCH, 1956 | TRANSFORMER DESIGN (rom page 30) (b) Keeping interprimary inductance very low by using a billlar winding tech. nique usually increases the effective pri- mary capacitance, Higher ¢ tum reduees the available power at low distortion, ‘The performance is then quite analo- ‘gous to that at low frequencies where VA=P,~5PQ/a (43) Analysis indieates, in a similar way, that at high frequencies volt-amperes Vly niust be furnished (at high power levels) in uecordance with VAy=P, + iPo/dy (60) where dy=Xe/Ry. (on) and Xp=elfective primary eapaei tive reactance ‘We ean now recall that eapacitanee, be- ing basically a measure of length, has 18 proportionality C2 vies wis, (62) where V is volume and 7 the weight, 2, then TY is reduced by 2"? = \d the capacitance of the smaller transformer is reduced by a factor 2/8 4.41, almost a half oetay (¢) The combination of greater band- width, lower interprimary leakage and lower eapaeitance ean now contribute substantially to the desizn of an econom ieal high performance B, amplifier Limits of K For over a decade it has heen known that a low generator impedance obtained with negative feedback wonkd improve the performance of medigere output transformer. But, unfortunately, feed Daek amplifiers using such transformers perform poorly at high power levels, The writer has used small values of KC (in the range of 14-1.6) and maintained adequate high-level primary induetanee for low distortion in eonjunetion with grain-oriented laminations, and ex ployed winding techniques common to the traditional output transformer ‘Tests on stable zero-source-resistance {attained by controlled positive fecd- Iiack) amplifiers with output transform crs designed in accordance with the pro- cedures outlined in this article eonfirm the advantages stated herein,