In Part 1 of this article (GA 2/97), I used computer simulations in studying the characteristics of various stages and tubes types to find an elegant solution for a full-featured tube DAC output amplifier. In this second part, I will introduce a practical design with regulated power supplies and a printed circuit board that you can use to replace semiconductor output stages in existing CD players or outboard DACs.
As you can see from the simulation results in table 2, it is difficult to drive low-impedance loads with low distortion using tubes without transformers. Another common remedy for this is negative feedback, which is precisely what happens if we replace IC1 in figure 1 with the high gain cascode amp.
Circuits K and L in table 2 provide well over 100x amplification, very wide bandwidth, and low output impedance. They are able to drive a 1kohm passive filter, which makes it possible to omit the output buffer entirely. The output filter is still able to drive subsequent tube amps with ease, and it is not affected by moderate amounts of interconnecting cable capacitance.
Figure 6 shows a complete simulator schematics for the cascode connected DAC output amp. This circuit closely matches the configuration of figure 1 , except that the output buffer IC2 is discarded as unnecessary. Rf is the feedback resistor, which determines the current-to-voltage conversion ratio. The output voltage and input impedance are approximately:
Since the amplification is over 300 for circuit L, the error using these simplified formulas is less than 0.3%.
Now that the cathode follower is inside the feedback loop, and the distortion is very low. The low-pass filter must stay outside the feedback loop, hence the output impedance of the whole amp is 1kohm. To get realistic bandwidth results, stray capacitances CIN, CS and cable capacitance CL are included in the model.
This amp achieves superb simulated performance: The closed-loop bandwidth prior to the filter is 23 MHz, rise time is 16ns, and the slew rate 150V/µs. With these figures, the high-frequency performance rivals that of modern op amps. Not bad for an amp using an allegedly extinct technology! Power supply rejection is 55dB, thanks to split resistors R8, R9 and C11. The input impedance is about 10ohms, and distortion (for what the simulated figure is worth) is now negligible, less than 0.001.
Impressive as the specs are, I was not very happy with this design. It needs several high-voltage, low-noise power supplies. Correcting the unacceptable distortion performance by negative feedback did not meet my audio design principles. In fact, it is a vacuum tube clone of figure 1 , where brute force overcomes the component imperfections. It lacks the simple elegance of well-designed tube equipment.
The real complication in this amp is the need to drive the relatively low-impedance filter network. Using a 10kohm filter would be an option, but that demands yet another CF output buffer. No thank you! Then I had a bright idea: why must the filter follow the amp? Wouldn't it be easier to filter the signal first, and then amplify it?
Eureka! I was so accustomed to the solid-state topology in Figure 1 that I couldn't see the easier way. The passive filter can just as well handle low level signals; there is no need for preceding amplification.
In this scheme, the demands for the amplifier stage are much revised. Very low output impedance it no longer needed. High gain is still desirable, but low distortion is now mandatory, since you can not use negative feedback. In circuit B, the 12AX7 mu follower seems to be the amp of choice - it's simple and has very low distortion and excellent PSRR. Low bandwidth or high input capacitance do no harm here, and 1.3kohm output impedance is fine for driving a downstream tube preamp.
One way to arrange the DAC load resistor and the low-pass filter is to combine a 25-35ohm resistor and the familiar 1kohm filter, creating a resultant load suitable for circuit B. Alternatively, you can scale the filter values for lower filter impedances, see table 3. Based on Mouser's catalog, 25ohms seems to be the lowest practical impedance. Below that, the series resistances of the commonly available inductors become too high.
Using the passive filter network with a current-output DAC presents a new problem. The input impedance of the filter remains nominal up to the corner frequency, but beyond that it is inductive: the input impedance doubles for every octave, an unacceptable behavior for the DAC. Adding an RC network to the input keeps the impedance controlled also at high frequencies. With the constant-current drive signal from the DAC, it is advantageous to make the input impedance as small as possible at high frequencies, which enabled me to tweak 10dB more attenuation at 353kHz. Table 3 lists practical values for various filter impedances.
Figure 7 is the schematic for the second version. The actual amplifier part is remarkably simple, with only a handful of components. The input to the DAC is DC-connected, but C102 and C104 determine the low-frequency response - about 1Hz (-3dB).
Inspired by the experience gained from the several POOGE articles in TAA, I implemented low-noise regulated supplies for the plate voltages and filaments, even though the support circuitry then contained more components than the actual amplifier.
Only one supply voltage with is needed. I devised a high-voltage regulated supply, a modified version of the preamp power supply Joe Curcio used in his "Daniel" preamp10). I replaced the LF356 op amp with a low-power type designed for single supply operation. Its input voltage range includes the lower supply-voltage line.
IC101 acts as a voltage follower. The output voltage tracks the heavily filtered input on pin 3 provided by resistor R109 and C105. Diodes D101 and D102 ensure that the input voltage rating is not exceeded under any condition. R110, R111, and C107 ensure that the op amp stays stable with the heavy capacitive loading of C108. R112 limits the current output to about 10mA, and diode D103 clamps the output voltage during start-up and overload conditions.
Figure 8 shows the preregulator used as the raw plate power supply. Zener diode D9 and MOSFET Q1 ensure that regulator IC1 always operates within its voltage ratings. R6 causes a constant 5mA current to flow through the zener diode chain D10-15, which sets the output voltage.
During testing, a tiny slip with a probe caused a momentary short to ground in the zener chain (D11-D15). As a result, R5 flamed, and the resulting carbon residues caused an arc from the 300V input to ground. It vaporized the remains of R5, burned a hole in the board, and nearly set my workbench to fire. If you can't find a special nonflammable resistor for R5, use a fast fuse in the high voltage supply line.
R4 and C12 provide a smooth rise for the regulator input voltage. During power-on, the plate capacitors C108/208 are initially charged by R4 through zeners D9, D10, and clamp diodes D103/203. Hence all the caps charge slowly, and the low voltage semiconductor circuits are happily unaware of the destructive high voltage they are riding on.
During power-on, the amplifier output generates a momentary high voltage while capacitor C104 is charging. This is not harmful for tube preamps, but might cause an annoying thump from the speakers if the preamp does not contain a mute circuit. Timer IC3 and relay RL1 keep the outputs shorted for about 30 seconds after power-on, which keeps the speakers quiet during the turn-on period.
The filaments are heated with a regulated voltage. A soft-start circuit (C19, R10 and Q2) ramps the filament voltage up to the nominal 6.3V in 10 seconds, which prevents high inrush currents when power is initially turned on. Note that the filament supply is floating, bypassed with capacitor C8.
Timer IC4 (Figure 9 ) controls the +300V input voltage for the preregulator using relay RL2. Immediately after power-on, RL2 is unenergized and capacitor C5 charges through resistor R1. Contact RL2B applies the raw plate voltage to the preregulator only when the filaments are fully heated - after about 15 seconds. At the same time, R1 is shorted by RL2A.
The rectifiers for the plate supply are soft-recovery types to keep radiated noise to minimum11). Schottky barrier rectifiers used in the filament supply generate even less RFI than the soft recovery diodes, because they completely lack the transient reverse current phenomenon.
The printed circuit board uses a copper plane to keep ground-related noise to a minimum. The track widths and clearances are ample to permit home-brew etching. Be sure to check the board carefully for proper insulation widths between the ground plane, tracks, and pads.
Check the operation on the preregulator (Q1, IC1) and filament regulators before assembling IC101 or 201. The actual output voltage is not very important. If it is far from the nominal 260V, check (with caution!) the voltages in the zener diode chain (D10-15). Replace any zeners that are too far off their specification.
The mechanical assembly into the existing equipment is the trickiest part of this project. The heat generated by the tubes prohibits the assembly into closed or inadequately ventilated enclosures. The printed circuit board has a special feature for upgrade installations: the tubes are installed on the solder side of the board, so you can install it on a wall or cover plate, with the tubes protruding through holes in the plate. Most of the dissipated heated is now generated outside of the enclosure; and the installation is also visually appealing.
Place the output-amp board in the vicinity of the DAC chips. Disconnect the existing output amps by pulling off the original output op amp chips or cutting the board traces. Connect the DAC outputs to J101/102 and J201/202 with short twisted wires.
If you can't fit the amp board into an existing equipment, you can always get a separate cabinet for the output amp. Mount it on top of the player or DAC and connect the input to the amp with high quality shielded cable. It is not pretty, but the warm glow of the tubes will certainly divert attention from this exotic piggy-pack solution. An aluminum heatsink bracket reduces filament-regulator heat.
Do not install the line transformers in the same enclosure, since common unshielded chokes may pick up hum due to stray magnetic fields from mains transformers. If this happens, replace L101/201 and L102/202 with shielded cans. The coils should be rated to at least 50mA current to prevent any harmonic distortion caused by magnetic saturation.
A 220V transformer secondary can generate +300V input voltage with a full-wave bridge (figure 9 ). Another option is to use a 110V secondary with a voltage doubler (figure 13 ). Thus an 1:1 isolation transformer can serve as the mains transformer.
I have not designed a circuit board for the mains power supply, since the acceptable dimensions of the board depend on the mechanical assembly. You can easily mount the components on a perforated board and wire them point to point.
Connect the transformer primary to a suitable point after the existing line filters, fuses and power switch. Keep the high voltage supplies away from the DAC board, and do not connect the power supply grounds together, except through wires to J102 and J202. It is best to use a separate pair of isolated output jacks for the tube output.
I have made provision for easy experiments with various tube types and filter impedances. The filter in Figure 7 is optimized for Burr-Brown PCM1702 DACs with 1.2mA peak output current. Depending on your DAC's characteristics, you can select a more suitable filter impedance from table 3. 36.5ohm filter is suitable for DACs with 1mA output, and the 51.1ohm version is ideal with 12AT7 tubes. Using 6DJ8 or 6922 tubes is possible with the 100ohm filter. Table 4 shows the necessary changes in component values with different tube types. Table 5 is the parts list for components in the circuits of Figures 6-9 and Figure 13.
The tubes in the prototype proved to be slightly suscebtible to external hum fields. As expected, the noise level is also appreciably higher than with the solid state version. Both are quite inaudible with normal listening levels, becoming perceptible only with the full gain of my preamp. You might want to purchase special low-noise tubes for this application. Grounded tube shields will completely eliminate the hum, but then you'll hide the glow of the beauties.
Computer simulations may seem quite absurd and academic with regard to tubes. I hope that this case shows how useful such a simulator may be even for a hobbyist. It also proves that pure intuition can not be replaced by a computer; new ideas and innovations are still made only in "between one's ears".
I hope that this study for extending high performance tube technology up to the boundary between digital and analog domains will encourage others to experiment further. I would be most happy to hear from your experiences and novel solutions. Happy listening and experimenting!
10. Joe Curcio, "Daniel, a Vacuum Tube Preamp", TAA 2/85, p. 7.
11. Rick Miller, "Measured RFI Differences Between Rectifier Diodes in Simple Capacitor-Input Power Supplies", TAA 1/94, p. 26.
Jukka Tolonen has
over 20 years experience in designing medical instruments, industrial
measurement and control systems, and integrated circuits for several
Finnish companies. His audio hobby started at the age of 15, when
he built his first amplifier. He has been a tube addict since
1992. He can best be contacted by email at 
.
He also maintains a personal World Wide Web site at http://www.megabaud.fi/~jtolonen/.
With easily available components, 25ohms is the lowest practical impedance. Component designations refer to Figure 7.
| R101 | L101 | C101 | L102 | R106 | C103 |
| 100ohm | 390uH | 33nF | 120uH | 22ohm | 22nF |
| 51.1ohm | 200uH | 68nF | 68uH | 10ohm | 47nF |
| 36.5ohm | 150uH | 100nF | 47uH | 8.2ohm | 56nF |
| 24.9ohm | 100uH | 150nF | 33uH | 7.5ohm | 68nF |
Component designations are those of Figure 7.
| Tube | Ia | R102,R202,R104,R204 | R103,R203 | P101,P201 | P102,P202 |
| 12AX7 | 1mA | 430ohm | 47kohm | 1-2 | 1-2 |
| 12AT7 | 5mA | 62ohm | 10kohm | 1-2 | 1-2 |
| 6DJ8/6922 | 5mA | 549ohm | 10kohm | 2-3 | 2-3 |
| Qty | Item | Description |
| Tubes: | ||
| 2 | V101,V201 | 12AX7 or ECC83 |
| Semiconductors: | ||
| 2 | IC1-2 | LM317T |
| 2 | IC3-4 | NE555 |
| 2 | IC101,IC201 | TLC2201 or TLC271 |
| 1 | Q1 | IRF821, IRF822 or equivalent |
| 1 | Q2 | 2N2907A |
| 2 | Q101,Q201 | 2N2222A |
| 4 | D1-4 | GI854 |
| 4 | D5-8 | 1N5822 |
| 1 | D9-10 | Zener 10V 0.4W |
| 5 | D11-15 | Zener 47V or 51V 1W |
| 4 | D16-17,D103,D203 | 1N4448 |
| 4 | D101,D102,D201,D202 | 1N4148 |
| Resistors: | ||
| 1 | R1 | 33k 5% 2W wirewound |
| 1 | R2 | 511k 1% 0.25W metal film |
| 2 | R3,R6 | 249 1% 0.25W metal film |
| 1 | R4 | 10k 5% 2W wirewound |
| 1 | R5 | 100 1% 0.25W metal film (nonflammable) |
| 7 | R7,R110-112,R210-212 | 1k 1% 0.25W metal film |
| 1 | R8 | 4.7 1% 0.25W metal film |
| 1 | R9 | 249k 1% 0.25W metal film |
| 1 | R10 | 562k 1% 0.25W metal film |
| 1 | R11 | 330k 5% 0.5W carbon film |
| 2 | R101,R201 | 24.9 1% 0.25W metal film |
| 4 | R102,R104,R202,R204 | 430 1% 0.25W metal film |
| 2 | R103,R203 | 47k 1% 0.25W metal film |
| 4 | R105,R205,R109,R209 | 100k 1% 0.25W metal film |
| 2 | R206,R106 | 7.5 1% 0.25W metal film |
| 2 | R207,R107 | 1M 1% 0.25W metal film |
| 2 | R108,R208 | 33 1% 0.25W metal film |
| Capacitors: | ||
| 4 | C1-4 | 10nF 1250V polypropylene |
| 1 | C5 | 150uF 350V Al Electrolytic |
| 1 | C6 | 10000uF 16V Al Electrolytic |
| 2 | C7-8 | 10nF 100V CERAMIC |
| 1 | C9 | 10nF 600V ceramic |
| 2 | C10,C17 | 100nF 400V polycarbonate |
| 4 | C11-12,C108,C208 | 10uF 350V Al Electrolytic, axial |
| 2 | C13-14 | 100nF 63V polyester |
| 5 | C15-16,C18-20 | 47uF 16V tantalum |
| 2 | C101,C201 | 150nF 100V polypropylene |
| 2 | C102,C202 | 470nF 100V polypropylene |
| 2 | C103,C203 | 68nF 250V polypropylene |
| 2 | C104,C204 | 2.2uF 250V polypropylene |
| 2 | C105,C205 | 100nF 400V polypropylene |
| 2 | C206,C106 | 100nF 63V ceramic |
| 2 | C107,C207 | 3.3pF 63V ceramic |
| Miscellaneous: | ||
| 2 | L101,L201 | Inductor 100uH (Mouser #434-23-101) |
| 2 | L102,L202 | Inductor 33uH (Mouser #434-23-330) |
| 1 | RL1 | Signal relay, 2PDT, 12V coil (Mouser #431-OVR-SH-212L) |
| 1 | RL2 | Power relay, 2PDT, 12V coil (Mouser #528-7660-63) |
| 1 | TR1 | Transformer 220V/220V, 10VA |
| 1 | TR2 | Transformer 220/10V, 10VA |
| Optional: | ||
| TR1 | Transformer 110V/110V, 10VA | |
| 1 | TR2 | Transformer 110/10V, 10VA |
| 2 | C21-C22 | 330uF 250 Al Electrolytic (optional for voltage doubler) |
The part numbers in parenthesis are given as examples that fit mechanically on the PCB. They are not specially recommend over similar types.
Figure 6 : Simulation model for the cascode DAC output stage.
Figure 7 : Schematic for one channel of the mu follower DAC output amp. Component numbers for the other channel start from 200.
Figure 8 : Plate power supply preregulator, filament regulator and turn-on mute circuit.
Figure 9 : Mains power supply, 220V secondary in TR1.
Figure 13 : Optional main power supply, 110V secondary in TR1, and voltage doubler.
Figure 12 : Component locations. Note: Tubes V101 and V201 are installed on the solder side. The jumpers for the filament settings are hardwired for 12AX7/12AT7.
Comments, questions and feedback are highly appreciated:
This page was last modified on 24.08.97.
