On eference IAA Network by Jim Hagerman You d think there would be nothing left to ay. Everything you need to know about IAA network ha already been publihed. However, a few year back I came acro an intereting chapter in a vacuum tube book[] which poke of a mythical.8µ corner in the IAA repone. Huh? That 50kHz. The eminal work[][] never mentioned thi, equalization wa to appear a Figure. Figure. Deired invere IAA repone from Liphitz. Hmmm. The mythical corner frequency i hown a f but ideally hould be miing. Or hould it? To quote from Allen Wright book[]:... look back at the graph of the recording EQ, they cut the LF and boot the HF. But do you really think they continue booting to way out pat whatever? Of coure not, they d burn out their cutter head or omething even more expenive... Thi new db point, according to a Neumann cutting amp manual, i et at.8µ which equate to 50,08Hz... [] Not only doe the cutting head repone have a pole at f, but it mut alo have one at f 5! No amplifier ha gain out to infinity. Hopefully, all cutting head manufacturer choe the ame.8µ corner for limiting gain. Where i all thi leading? It mean that the IAA equalization network in our phono preamplifier hould have a zero at.8µ, putting back ome gain before finally rolling off at higher frequencie. The legacy reference network[] hown in Figure ha an f pole at 7kHz too high of frequency. Figure. Legacy reference invere IAA network from Liphitz.
eference Invere IAA What we need i a new modified IAA reference curve to help u properly deign phono equipment. I like uing SPICE to imulate filter circuit and decided thi would be a good way to generate a new tandard. The generic filter ection hown in Figure i a imple lag-lead type with a zero and pole in the right place. Figure. Lag-lead filter ection. It voltage tranfer function i given by + C H ( ) + + C with zero and pole time contant determined by τ τ z p C C + The lower IAA zero-pole pair i at 80µ and 8µ. Uing an arbitrary capacitance value of µf, the reitance are calculated a τ z C τ p C τ 80µ.8k µ F p (.8k )( 8µ ) (.8k)( µ F ) 5. 8µ The value for a econd ection (zero at 75µ and pole at.8µ) are 7.5k and 5. ohm repectively uing a 0nF capacitor. The final circuit i hown in Figure. Note, I ued a voltage controlled voltage ource (E) to decouple the repone of the two ection, otherwie the input impedance of the econd ection would load the firt ection and alter repone.
Figure. SPICE chematic for generating reference invere IAA curve. Liting i the input text file to my SPICE imulator. Figure 5 how the reulting frequency repone of the circuit, which i alo given in tabular form in Liting. I offet the data o that gain at khz would be 0dB. I find it helpful to weep a wide frequency range of Hz to MHz a it give a better view of the total repone and what occur outide of the 0 to 0,000Hz audio band. PSPICE Input Invere IAA Curve Vin 0 ac.8k 0 5. C u E 0 0.0 7.5k 0 5. C 0n.ac dec 0 000k.print vm().probe.end Liting. SPICE input file for generating modified reference curve. Figure 5. SPICE output of reference invere IAA curve with khz et to 0dB.
Modified Invere IAA frequency db frequency db frequency db 0.00..59. 5.85 7.78 9.95.9 5. 8.8.6 5.8 9.8.67 50. 56. 6.0 70.79 79. 89. 00.0. 5.9. 58.5 77.8 99.5-9.7-9.70-9.6-9.58-9.50-9.0-9.7-9. -8.9-8.7-8.6-8.6-7.80-7.0-6.9-6. -5.8-5. -.55 -.8 -.09 -. -.5-0.70-9.88-9.05-8..9 5. 8.8 6. 5.8 98. 6.7 50. 56. 6.0 707.9 79. 89. 000 59 585 778 995 9 5 88 6 58 98 67-7. -6.6-5.88-5.5 -.6 -.8 -.0 -.6 -. -.6 -.9-0.77-0.8 0.00 0.8 0.77.7.60.07.57..7.6 5.05 5.78 6.55 7.5 50 56 60 7079 79 89 0000 0 590 0 5850 7780 9950 90 50 880 60 580 980 670 500 560 600 70790 790 890 00000 8.8 9.0 9.9 0.8.7.6.55.6 5.7 6.7 7.7 8.0 8.89 9.7 0.50.5.96.6..79.8.7 5. 5. 5.7 5.96 6.6 Liting. eult from SPICE imulation. New Network Deign Figure can be modified to achieve the deired reult. In order to hift the high frequency pole the value of mut change. Unfortunately, thi alo change the gain of the network. However, by moving part of to the input ide, we can control both gain and pole frequency independently. A reference network hould interface nicely between tet equipment and phono preamplifier. Therefore, I elected the following deign parameter: 50 ohm ource impedance (many generator do not ue the 600 ohm audio tandard) 600 ohm output impedance Dual output gain of 0dB and 60dB @khz Standard capacitance value. Figure 6 how my new modified IAA network.
Figure 6. Modified invere IAA circuit. Optimizing component value i eaily done by iterative SPICE imulation, but a good tarting point i needed. Approximate value can be calculated by utilizing known boundary condition. Since output impedance hould be 600 ohm, and the two output are 0dB apart, we can write + + 600 0dB 0. Solving we get ( 0.)( 600) 60 600 60 50 At high frequency the capacitor appear a hort circuit and the network implifie to a reitor divider compried of 5,, and. From Figure 5 we ee that high frequency gain i about 7dB higher than at khz. Since deired khz gain i 0dB, our high frequency gain will be about db. The gain equation db 0log A HF i rewritten and olved a A HF 0 db 0 0 0 0. High frequency divider gain i then given by A HF + + + 0. and we can olve for 5 5 + 600 5 600 A 0. HF.k At low frequency we have the oppoite effect and the capacitor appear a open circuit. The divider again i reitive and ha a gain of 60dB. Thi i written a
A LF + + + + and the equivalent erie reitance of and i olved a Our high frequency pole at.8µ i equal to the erie reitance of,, and 5 time the equivalent erie capacitance of C and C. Thi capacitance i given by Finally, we have four equation and four unknown I ll pare you the math, i olved a and a and the capacitor a + 5 0.00 ( + ) k eq + 999 5 597 (.8µ ) C C.8µ C eq. nf C C + + 5 600 +.k + C 75µ C C eq eq 80µ.nF 597k ( C )( C ) ( C ) eq Ceq ( C ) ( C ) 50k eq 597k 50k 57k 75µ C.5nF 80µ C 5.8nF Thee are only tarting value. For a bet fit real world deign I optimized for the nearet % reitor and tandard capacitor value (a hown in Figure 6). Comparion: Old v. New A a reference network the performance mut be pretty good. Figure 7 how the frequency repone error relative to Liting. My new network i accurate to within +/-0.dB over the full frequency range. For comparion I ran an error imulation of the Liphitz network in Figure. It i extremely good up to about 0kHz where it drop off a expected due to the miing.8µ corner. The Liphitz circuit can, however, be modified by adding a.8k reitor a 5 at the input to repoition the high frequency pole.
Figure 7. SPICE generated error repone for Liphitz and Hagerman network. To check the enitivity of component value I ran a few more imulation varying capacitance by 5%. Error remained within +/-0.dB. I hope thi information i helpful to deigner of high end audio equipment. eference [] A. Wright, The Tube Preamp Cookbook, Vacuum State Electronic, 995. [] S. Liphitz, and W. Jung, A High Accuracy Invere IAA Network, Audio Amateur, 980. [] S. Liphitz, On IAA Equalization Network, JAES 979. [] AN-: Three High Accuracy IAA/IEC MC and MM Phono Preamplifier, Analog Device, 99 [5] M. Gile, Audio/adio Handbook, National Semiconductor, 980 [6] A. Wright, Secret of the Phono Stage, Sound Practice #5, 998 Source The network of Figure 6 i available a a kit (order #KF-) from Old Colony Sound Lab for $5. The two channel kit come complete with PCB, % metal film reitor, % polypropylene capacitor, connector, and intruction. Old Colony Sound Lab PO Box 876 Peterborough, NH 058-0876 60-9-67 http://www.audioxpre.com