ERIC K. PRITCHARD The Tube Sound and Tube Emulators

$0 RE TUBES MAGIC? IS THERE really a difference be- tween tubes and tran- I ea sistors? Some hear the warmthA and appreciate the full body of the tube sound, and others deride the thought. Is the magic of I. I. - the tube sound more than mere / ea

nostalgia? A recording engineer, ....„1 .4,Aa.

e1 00 14 •111 Russell 0. Hamm, could hear the Relative lame Level itoletive ler.a Level difference. Determined to find and Figure 1. components explain the difference, he began Figure 3.Distortion components for two-stage . testing for multi-stage capacitor-coupled of various technologies. His fa- amplifier. mous paper, "Tubes Versus Tran- sistors—Is There an Audible Dif- ference?" [1], shows that the 10 harmonic structures in overdrive 3 re N I 3 r• conditions for different technolo- gies are quite different, almost like 20 10 fingerprints. I.

-___...... i..4_7"

More recently, an electronics en- 14 ea gineer, Eric Pritchard, started Relative brat Level - 14 down the circuitous path to bring Retellv• tepee Leve1 Figure 2..Distortion components the two worlds together and to give Figure 4.Distortion components for two-stage amplifier.. solid state the character of tubes. for multistage - The elusive tube sound has finally coupled transistor amplifier. succumbed to an intensive re- search and development program TUBES AND THE TUBE that has produced solid-state tube SOUND emulators and tube emulator cir- In retrospect, the tube has many cuits [2]. The effort began nearly 3 re f technically superior aspects, it is seven years ago with the search for fairly linear, its operational pa- / a solid state that rameters do not vary badly, at least / sounded like tubes. After snaking / for tubes of yesteryear. / through myths and theories, the vs...... 14. 7/ -- It operates reasonably without research turned to emulating the feedback. Its gain-bandwidth // . tubes, both and . product is not low, about MHz8 for 1 111 ell a 12AX7 in generica audio stage, Relalive beer Level compared with 3 MHz for typical Russell Hamm's work provided a Figure 5.Distortion components audio operational . Tube test and an independent source of for monolithic operational parameters do not vary as badly as data to correlate the operation of amplifier with hybrid output semiconductor parameters. the triode tube emulator. stage. They do not have the widely varying input voltage threshold of nearly obsolete, and they are rap- FETs. Tubes, in fact, may be the idly becoming unavailable. Eric K. Pritchard is president of most natural amplifier. Unfortu- Deja Vu Audio, Berkeley Springs, nately, they are large, they are The tube sound has been so elu- WV 25411 fragile, they are microphonic, they sive that many consider it to be

26 db January/February 1994 drift, they burn out, they are mysterious and beyond the capa- bility of modern instrumentation. age. Other devices, pentodes and This is partly true. The tube sound all semiconductors, have a sharply is often more subtle than the oscil- rising current in the saturation re- loscope display. Low level harmon- gion that then bends over into a ics are not visible, but are audible. constant current region. The triode Distortion meters do not consider plate characteristic insures that individual harmonics, but our ears for moderate-to-high impedance do. Standard audio tests do not tell circuits the grid conduction always the whole story. limits the negative excursion of the Russell Hamm moved beyond plate for any reasonable load line. the Total Harmonic Distortion This contrasts with pentodes and measurement and developed a test semi-conductors which saturate that separated different types of for small loads. This is evident in the pentode which microphone preamplifiers: triode, Figure 6.Complete Tube Emula- has the second harmonic rising pentode, transistor, and opera- tor for a 12AX7. tional amplifier. This test meas- substantially later than it rises in ured and plotted the percentage of the triode preamplifier. each harmonic as a function of This is quite evident in operational overdrive. amplifier based preamplifier, Fig- THE TRIODE TUBE The harmonic character of these ure 5. A transistor amplifier pat- EMULATOR preamplifiers, Figures 1 through 5 ent, in an attempt to sound like The full triode emulation circuit, [1], are quite different, virtual fin- tubes, has reduced the extent of shown in Figure 6, has all of the gerprints of the various technolo- feedback to avoid these problems needed features: grid, plate, and gies and their respective circuits. [6]. Tubes, being more natural am- cathode terminals, grid-to-cathode The triode curves, Figure 1, show plifiers, need less feedback and significant second harmonic gener- consequently do not generate as conduction, grid-to-plate capaci- tance, gain, and the non-linear net- ated by the bias shifting in the cou- much of these high, harsh, discor- works. The gain is created by two pling capacitor created by grid con- dant harmonics. operational amplifiers such as the duction. The pentode curves, Tubes, particularly triodes pro- dual op amp, 0PA2604. This op- Figure 2, show the grid conduction duce significant second harmonics, erational amplifier has a sufficient delayed by the plate load curve go- Figure 1. The second harmonic has gain-bandwidth product, about 10 ing through the saturation region two sources: the non-linearity of MHz, to simulate audio tube cir- well below the knee into the high tube characteristics and the inter- cuits. The grid terminal drives the plate resistance region. The semi- action of the coupling capacitor and negative input of the first op amp conductor preamp curves, Figures the grid-to-cathode diode [7]. The to produce the needed inversion, 3 through 5, show no equivalence non-linear characteristics are the while the second op amp creates to grid conduction. The operational plate resistance and the gain. The the feedback for the first. R1 deter- amplifier, Figure 5, shows the plate resistance and the gain pro- mines the proportion of distortion rapid rise in distortion created by duce harmonics at all signal levels. in the voltage gain. R2 then deter- high gains and substantial feed- These harmonics are superseded mines the voltage gain. back. by those created by the interaction The second harmonic provides of the coupling capacitor, grid-to- The choice of voltage gains is lim- punch in contrast to the blanket of cathode diode, and . The ited by the selection of the grid con- the third harmonic [1]. Conse- grid-to-cathode diode charges the duction components. The grid quently, these Figures show that coupling capacitor when the grid drive circuit cannot produce so the triode, Figure 1 initially pro- conducts. The resulting change in much grid-to-cathode voltage that vides a blanketed punch that fades charge creates an offset that shifts the first operational amplifier goes into a lot of punch. The pentode is the bias from quiescent conditions. into negative saturation, yet a primarily blanketed with a little This bias shift alters the duty cycle large output voltage swing is desir- punch. Semiconductors vary from of the resulting waveform. A duty able. Thus, there is an engineering quite blanketed to completely cycle which is not 50-50 produces choice between the diode voltage blanketed. even harmonics. The two harmonic drop, the power supply voltage, The next remarkable feature is sources plus the low feedback com- and the gain. the generation or rather the lack of bine to produce harmonics that oc- The reason why the tube emula- high order harmonics. The high or- cur over a wide range of input. This tor does not behave as its opera- der harmonics, especially odd ones, range is far wider than found in tional amplifier is that operational put a discordant edge on the signal. typical transistor or semiconductor amplifier is kept out of negative A prime source of high order har- designs. By comparison then, they saturation by the grid-to-cathode monics is feedback. Although feed- distort too fast. diode and the positive saturation back corrects for amplifier errors, The grid conduction plus the effects are not transmitted by an it also attempts to correct for power unique plate characteristics give output resistance that becomes ef- supply limits. The feedback slams triodes the soft clip characteristic. fectively infinite. Of course, this re- the output against the power sup- Triode plate resistance is unique quires that the tube emulator plate ply rails and creates sharp corners because the plate current sweeps supply be at or below the op komqed/A.Ionuorqp amp pool that translate into high harmonics. upward with increasing plate volt- saturation level. L of input levels and frequencies show a good match at various over- drives, frequencies and loads, see Figures 11-20. A close examination of these figures show slight differ- ences in the curvature in some por- 1:1 ( tions of some of the waveforms. This is reflected in slight differ- ences in harmonic levels. Harmonic analysis showed that the tube emulator erred in the di- rection of a more ideal tube, Figure 9.0utput (plate) charac- teristics of a 12AX7 triode tube slightly more rounded waveforms Figure 7.Low-Noise Tube Emu- emulator. and consequently less intense high lator for a 12AX7. Horz: 5 volts I div. order harmonics. Vert: 0.5 milliamps I div, Step: .05 volts. The low-noise implementation, shown in Figure 7, is viable for flat response stages. A difference be- 4 LV 30K tween these circuits is the feed- back connection between the two operational amplifiers is attenu- ated by R6 and R7 so that the noise of the second op amp is also attenu- ated. This circuit drops the noise from 3 dB over a single operational amplifier to about 1 dB higher. The restriction is that the cathode ter- minal is grounded. This corre- sponds to a cathode bypass capaci- tor large enough to bypass the lowest frequency of interest. This does not upset the bias since this arrangement corresponds to a bias of about 1.3 volts on a 12AX7. The non-inverting implementa- tion, shown in Figure 8, is also vi- able for flat response stage. Since the amplifier is non-inverting, the diode network must have the oppo- _MAXI site polarity and the circuit cannot use the input-to-output capacitor. The output characteristics of the tube emulator, shown in Figure 9, Figure 10.Comparison Tester for 12AX7 versus a Tube Emulator are similar to the plate charac- teristics of a 12AX7 triode. The output characteristics are quite ac- curate around normal load lines. THE TUBE EMULATOR The inaccuracy in the region of si- MICROPHONE Figure 8.Non-Inverting Thbe multaneous low currents and low Emulator. voltages is not important since the PREAMPLIFIER emulator is not operated there. The complete tube emulator of These characteristics do not re- Figure 6 was used to build a micro- semble the current limited charac- phone preamplifier, Figure 21, to teristics of pentodes nor any semi- conductor. compare harmonic structures against Russell Hamm's findings. A tube emulator was operated The preamplifier is a paraphrase alongside a 12AX7 to compare out- of a generic two-stage triode ampli- put waveforms, see Figure 10. Af- fier. The circuit topology is the ter adjusting the tube bias, the in- same except for the cathode fol- put and output gains and the lower is replaced by a unity gain emulator grid-to-plate capaci- buffer. The impedances are low-

28 db January/February 1994 1994 January/February db 28 tance, the waveforms for a variety ered by a factor of 10 so that the Figure 11. Output Waveform Figure 13. Output Waveform Figure 15. Output Waveform Comparison with tube at +5 Comparison with tube at +14 Comparison with tube at +11

Figure 12. Output Waveform Figure 14. Output Waveform Figure 16. Output Waveform Comparison with tube at +8. Comparison with tube at +17. Comparison with tube at +11.

parts. The biasing, however, is monic rises first. The second har- lower than proportional simply be- monic rises from a lower level at cause that is the character of the the same time, then overtakes and tube emulators. After finding the dominates. The fourth rises about appropriate biasing, the tube emu- 4 decibels later than the second. lator tested like Russell Hamm's The remaining harmonics remain generic two-stage triode preampli- below 5 percent, at least for 12 fier. It did not correlate with the decibels of overdrive. This matches pentode, transistor, or operational the description by Russell Hamm amplifier based preamps. for a triode preamplifier: 'Me out- The test designed by Russell standing characteristic is the Hamm is simple. The harmonic dominance of the second harmonic percentage for harmonics 2 followed closely by the third. The Figure 17. Output Waveform through 7 are plotted against over- fourth harmonic rises 3-4 dB later, Comparison with tube at +11. drive. The input level that creates running parallel to the third. The 1 percent third harmonic distor- fifth, sixth, and seventh remain be- tion becomes the 24 dB reference low 5 percent out to the 12 dB over- emulator and for the triode ampli- level. The result for the tube emu- load point." [1]. The harmonic per- fiers are quite different from the lator microphone preamplifier is centage curves for the tube curves for amplifiers made with shown in Figure 22. The third bar- pentodes, Figure 2, for ,

Figure 18 .Output Waveform Figure 19.0utput Waveform Figure 20.0utput Waveform Comparison with reactive load. Comparison with reactive load. Comparison with reactive load. 6Z V66L Aioniqed/AJonuorqp Aioniqed/AJonuorqp V66L 6Z 30 dbJanuary/February 1994 Figures 3 and 4, 5, harmonic levels. Although the pen- hew tested. h i ch mseventh i c rsignificant, Consequently, ophamplifier on e p harmonicsthird r eamp l harmonic. if version,certainly the as rapidly i egeneration r only well. stode The operational rising notmonic is fiers is dominate. a andampl tube, delayed fifth smaller show if ier The the a and and or dominate s solid , second zero smaller, even third harmonic har- tube emulatorHamm testsimilarity amplifier microphone between and the pream- the Russellstate examples, withcording tubethate mu emulators manycrophone classics la t of o r ptube r the eamp preamplifiers because l may professional character iftriode be i eplifier they r i re-created nd i is stagessource ca by t re- e a evident s was tube and for that two-stage the in they paraphrase both. both triode had of two the plifierexample,use two-stage is a generic used preamplifiers. in two-stage Pultecsimilarly. equalizersamplifiers am- For However, tended the to inherent be designed VELOP MENT the need fora tube nd T h based le t r on compo- i c s l im i t e r / c omp r e s - ers,ers, anda limiters, pp powernents l i c a amplifiers. [1], t i onplifiers. w h compressors, esors r the ev e o r tube as v e r well emulator l oad equaliz- s as has microphone pream- c haracto ter easily is t icsare create . likely: synergistic microphone combi- preamplifi- They translateTheamplifier, second thethe tube output first against emulator of because published a with tube itnations its could data. of solid be tested state and tube havepower approximatelyemulator MOSFETsemulate so phasea pp that a l splitter as power i ca theyindependently, either t output ion s be- p e to c stage. 6L6s if drive i c c omp outside on en t s of a guitar or EL34s. The third tube emulator A Side Bar TUBE EMULATOR DE- Russell Hamm did not specify The good emulation of tube mi- Just as Russell Hamm pointed to Tube emulators allow designers The triode tube emulator became also show a faster rise in the Figure 5. These Figures 3 through and operational Figure 5, has ampli- has only the absolutely necessary essenceis of a tube characteristics. latedworkscombination hybrid built SIP of into and non-linear application an epoxy net- encapsu- entirespecific length and components. is 0.71 inches This SIP has 15 pins on 0.1 inch centers along its Figure 21.A microphone preamplifier using triode tube emulators. Additionally, the tube Figure 22.At left THD and at right the tube emulator preamplifier topologies: Neumann The triode tube emulator is a U67, Neumann M49, built into Neumann U67 housings in a emulator is being low-cost triode emulator that variety of circuit and AKG C12. harmonic distortion components for 2, 30 c'0 (-) 0 0 The applicationhigh and about specific 0.125 compo- inches thick. andnents the grid provide conduction the gain compo- elements nents. MENT PRODUCT DEVELOP- various tube emulator circuits. ceivedwatt very andp rooemulators an well. f 11 -o f watt, - The con have cep general been have t p ro used been t o con- in typ a re- e g u i ta r thetors emulatorsgood because testwellemulator in the platform inaccepted allcept guitar its has regions. and as also for prototype exercises tubes.amplifiers. the the worked overall emula- The guitar quite low-cost These tone ampli- amplifiers, has been a 111 fier. The guitar amplifier makes a TUBE EMULATOR The triode and output stage tube Figures 6 through 8 show The triode emulators have been tube. Census data shows an in- used in microphone preamp proto- Although tubes are crease from $911 million in 1986 to types. This preamp has the syner- available from former $1.415 billion in 1991. But this gistic quality of dual chains, one market is also under attack from tube emulator and the other stand- Communist Block advancing technology, the flat ard solid state. This provides the countries, their solid video display. These replacements user with the rich, warm tube for picture tubes will appear in sound or the clean, transparent state revolution television sets in two or three solid state sound and the flick of a should occur faster years. The tube industry will loose switch. a substantial portion of its market. since the technology The loss of the video display mar- Additionally, the tube emulator ket plus the expected semiconduc- is being built into Neumann U67 is now available from tor revolution in the former Com- housings in a variety of circuit to- the West. munist Block countries will pologies: Neumann U67, Neu- accelerate the decline of audio mann M49, and AKG C12. tubes through a chain reaction. A Side Bar The latest threat is the flat video Fewer tubes will require less met- display that will replace picture al. The foundries will impose mini- THE TUBE MARKET tubes shortly. The picture tube is mums, charge higher prices, and 50 to 80 percent of tube markets The reason for creating the tube ship inferior material. Tube prices [4,5]. Although tubes are available emulator is the declining supply will go up and the quality will go from former Communist Block and quality of tubes. Although down. Then more semiconductors countries, their solid state revolu- some believe that the tube supply will be used, etc. tion should occur faster since the will prevail and continue to make technology is now available from high quality tubes, others differ. REFERENCES the West. Hartley Peavey of Peavey Elec- 1."Tubes Versus Transistors—Is The audio or receiving tube mar- tronics: "It's just a matter of time There an Audible Difference?" ket has been in decline for years. before these countries catch up to by Russell Hamm, Journal of the The U.S. Department of Com- the west, and stop making cheap merce Bureau of Census found Audio Engineering Society, May but reliable tubes. This is why so that the receiving tube market 1973, Volume 21, Number 4, pp. many companies, including Pea- dropped from $39.8 million in 1986 267-273. vey, are working feverishly to cre- to $21.6 million in 1991.[3] This 2.Patents by Eric K. Pritchard, ate the solid state equivalent of a market segment is so small the U.S.: 4,809,336,4,995,084, tube amp." [3] Paul Meisenzahl of Census is likely to stop collecting 5,133,014, plus one pending, plus Yamaha: "One of the big concerns this data. The Electronic Industry foreign patents for most customers is the availabil- Association reports a drop in im- 3.Musical Merchandise Review, ity of replacement tubes. No one ported receiving tubes from $8.3 June 1992, pages 16-27, Larkin- wants to invest good money in an million in 1985 to $2.67 million in Pluznick-Larkin, Inc., Newton,MA amp and then have to go through 1991.[4] The receiving tubes of all 4."Semiconductors, Printed Cir- all sorts of trouble when the tube types imported in 1991 numbered cuit Boards, and Other Electronic burns out in six months or a year." less than 0.77 million units [4]. Components", MA36Q(91)-1, Oc- [3] Cathy Duncan of Seymour Dun- This is very low. By comparison, tober 1992, U.S. Government can: "We do extensive computer large semiconductor manufactures Printing Office, Washington, DC testing on all the tubes we receive, do not consider manufacturing less 5.Electronic Industry Associa- and we send most back to the ven- than a million units of a single type tion Market Services Department, dor." [3] per year. 2001 Pennsylvania Ave., The tube market is constantly The ever increasing capability of Washington, DC 20006-1813 qp qp being eroded by the advance of semiconductors is a continuing 6.U.S. Patent 4,987,361 by Brent solid state technology New semi- threat to the tube market. The only K. Butler conductors are replacing tubes con- growth area in the tube industry 7.U.S. Patent 4,995,084 and pat- stantly. has been the cathode ray or picture ents pending by Eric K. Pritchard El LC V66[ kloniqed/AJonuor kloniqed/AJonuor V66[ LC