[Image] Evacuated Envelopes [Image] The On-Line Valve Amp Magazine DATE: AUGUST 1995 --------------------------------------------------------------------------- Class A Tube Amp Design Overview Author: Kurt Strain Date: 28 AUGUST 95 This is a brief (sort of) description on how to design a class A output stage for triodes in either push-pull or single-ended from notes I put together. I hope this proves informative. Single Ended Class A Triode Output Stage: This is as simple as it gets in transformer coupled tube output design, in concept. The transformer must be designed to handle the DC current at the bias point, so a special output transformer for single ended use must be used. The output transformer features an air gap that optimizes the coupling at low frequencies and the DC current that serves to lower the permeability of the core. Without the air gap, the iron will saturate under too little DC bias to accommodate the needs for a single ended triode. Too much gap will reduce the primary inductance so that the lowest frequencies will not pass without severe attenuation. A schematic for a single ended triode amp output stage is as follows: B+ o | ____ 8 )||( )||(____ 4 OUTPUT )||( )||(____ 0 | anode >--------grid cathode | +---+ | | Rk Ck | | = = There are two recognized approaches for biasing this tube: for max power output and for best sound. They are not necessarily the same. To maximize power, with a bypassed self biased stage, you need to make all the voltage swing available and all the current swing available meet at the same time. The voltage at the anode is B+ VDC, neglecting the primary resistance drop. The max primary swing will be 2 times that voltage, peak to peak, neglecting losses in the plate resistance and bias voltage. With the tube self biased and bypassed, most the AC voltage swing will end up appearing across the primary output impedance plus the plate resistance. Usually triodes have plate resistances about one-fourth that of the transformer primary impedance, so most ends up across the primary. If B+ is 350V, then the peak to peak voltage is 700V. The current that you can bias this tube to is the plate dissipation rating in watts divided by the anode-cathode voltage. The voltage at the cathode that self biases the tube will be stated for the tube, and is typically something like 60V. Thus, if the plate dissipation is 30W max and the anode to cathode voltage is 350V - 60V = 290V, then the max bias current is 30/290 = 0.103A. The bias current represents the peak current going through the transformer, and twice that will be the peak to peak current, or 0.206A. The current will swing from 0 to 0.206A while the voltage swings from 700V to 0V, roughly. This means that near optimum for power Rprimary = 700V / 0.206A = 3400 ohms. For a speaker load of 8 ohms, the turns ratio would thus be the square root of the impedance ratio, or the square root of 3400/8 = 20.6:1, 20.6 turns on the primary for each secondary turn. Now, assume the plate resistance, rp, to be 900 ohms, for example. The damping factor would thus be equal to 3400 ohms of load divided by 900 ohms of source impedance, approximately since rp is a rather dynamically shifting thing. That would make the damping factor 3400/900 = 3.8. If pentodes are used, the higher rp's of pentodes reduces damping factors considerably and global feedback is usually employed to help. A more accurate power output could be said to be Vout^2/Rprimary where Vout = (700/2*SQRT(2))*(3400/3400+900), the voltage division of the RMS voltage into the load, Vout = 196Vrms. The power is thus 196^2/3400 = 11.3W. From a current standpoint, the max RMS current is 0.103/SQRT(2) = 0.073Arms. The power output from max current is Iout^2*Rprimary = 0.073^2*3400 = 18.0W. There is more current than voltage available, so if we dropped Rprimary a little we could raise the max power available a little, or we could use the higher Rprimary to our advantage in terms of greater damping and lower distortion due to a lighter load. Besides, without more trial and error, we don't really know for sure how much voltage swing we really have available. After all, we could operate with a little positive grid voltage if we needed to. So we could stop right here and empirically find out how much we'll get. Or, when we go to look for an available transformer the closest we may find is a 3000 ohm 80mA unit, which may be optimum as found by prior experimenters despite our calculations. A little consulting from a transformer designer can help. In the case of parallel single-ended, pay particular attention since it very closely models a push-pull class A amp, with polarities switched around. A schematic for a parallel single ended triode amp output stage is as follows: B+ o | ____ 8 )||( )||(____ 4 OUTPUT )||( )||(____ 0 | +------------+ | | anode anode >--------grid---------grid cathode cathode | | +---+--------+ | | Rk Ck | | = = The current available is doubled, but the voltages remain the same. The effective load Rl = 2*Rprimary, or a half as heavy a load per tube. For a given Rprimary, the effective plate resistance is rp/2 (= 450 ohms). Using the same 30W tubes before at B+ = 350V, Ibias_total = 60W/(350-60V) = 0.207A, or Ip-p= 0.414A. Rprimary = 700V/0.414A = 1690 ohms. So you can see that parallel SE amp output transformers are lower in primary impedance, and hence have a smaller turns ratio. They also must accommodate larger primary DC bias current, 0.2 A in this case. But the power output is now about (350V/SQRT(2)*(1690/450+1690))^2 / 1690 = = 28.6 W, about twice as before. The damping factor is now 1690/400 = 4.2, almost the same. Class A Push-Pull (PP): What's the difference in the circuit model of a two tube SE amp and a two tube class A PP? Not very much, really. Here's a side by side comparison: B+ o o B+ o o |.|. ____ 8 |.| ____ 8 ) ) ||( ) ) ||( ) ) ||(____ 4 OUTPUT ) ) ||(____ 4 ) ) ||( ) ) ||( ) ) ||(____ 0 ) ).||(____ 0 | | | | +-+----------+ | +----------+ | | | | anode anode anode anode Vin >--------grid---------grid Vin >------grid -Vin >-grid cathode cathode cathode cathode | | | | +---+ +---+ +---+ +---+ | | | | | | | | Rk Ck Rk Ck Rk Ck Rk Ck | | | | | | | | = = = = = = = = These are nearly identical circuit models, just shifts in polarities. Note the polarity differences in the inputs to the tubes and in the polarity dots of the primary windings. That's about all the differences there are. If Rprimary = 1500 ohms in the case of one primary winding shown and two tubes as in the left case shown above, then the effective load in every case for every tube would be Rl = 3000 ohms. The difference is in how we total the primary impedances. For push-pull, the windings are reversed in polarity. Now, the number of turns from plate-to-plate just doubled, and there's an effect due to mutual coupling, like an autotransformer. N more turns produces N^2 more impedance, plate-to-plate. So doubling the number of turns makes a four times more plate-to-plate impedance. The plate-to-plate impedance will thus be 4*1500 = 6000 ohms. But since in a push-pull each tube is assisted in driving the load by the other tube, the effective load is 2*1500 = 3000 ohms, or Rpp/2. The power delivered to the speakers is the same. To calculate Rpp, or sometimes Raa, find Rprimary for the single ended case and multiply by 4. Rprimary is the same as using half the full center-tapped primary winding of a push-pull. Rpp uses both Rprimary halves, which quadruples the impedance. The Class B Case: Class B is more difficult to compute power for than class A. In class A, bias is going through the output tubes all the time. Power is a virtual constant, so the power supply load moves very little. This is a great advantage, and probably one of the sonic advantages to class A. You can design a power supply filter with less concern for massive changes in current demands. Not true for class B. Class B allows only one side of the transformer to power the load at any time. The voltage across the full primary winding will look the same because the autotransformer effect induces the full voltage swing despite only one device conducting. But the lower power dissipation of idling class B or class AB allows more current swing to take place. And these large swings will amplitude modulate the power supply. It is adviseable to be have a more overdesigned power supply for class AB as a result, with more capacitance and more current capacity. Unfortunately, this limits your use of the type of power supply capacitors as a result, probably requiring large electrolytics. Also, in class B, since you do not have the assistance of the other tube during conduction, the effective load then will be Rprimary, not 2*Rprimary as in class A, and virtually no load when not conducting. Class A push pull still has the disadvantage of requiring a split phase driver, as well as more turns in the transformer, and SE has the disadvantage of requiring DC bias through the transformer. OK, you cover the case of biasing to maximize power, but I couldn't find where you explain how to bias, for best sound. Could you elaborate? First, I did mention that raising the primary impedance above max power level can help improve damping and lower distortion. But beyond that, experimentation and experience will tell better. Mike LaFevre of Magnequest reported that "Stereo Sound" magazine, a Japanese publication, listed the "optimum" operating points for various direct heated triodes in a single ended amp, in terms of sound performance. They are listed below: * Vaic VV30B: Plate dissipation rating = 65W max Vp = 430V Ip = 70mA Vg = -70V Rprimary = 3K Pout = 10W Pd = 30W, 46% of max Note: this is best sounding when the tube is run at medium to high power. * WE 300B: Plate dissipation rating = 27W max Vp = 400V Ip = 60mA Vg = -87V Rprimary = 3.5K Pout = 10.5W Pd = 24W, 89% of max Note: this is best sounding when the tube is run at high power. * 2A3: Plate dissipation rating = 15W max Vp = 250V Ip = 60mA Vg = -45V Rprimary = 2.5K Pout = 3.5W Pd = 15W, 100% of max Note: this is best sounding when the tube is run at high power. * 845: Plate dissipation rating = 100W max Vp = 430V Ip = 62mA Vg = -51V Rprimary = 5K Pout = 4W Pd = 27W, 27% of max Note: this is best sounding when the tube is run at low power. * 211: Plate dissipation rating = 75W max Vp = 480V Ip = 40mA Vg = -20V Rprimary = 10K Pout = 2.6W Pd = 19.2W, 26% of max Note: this is best sounding when the tube is run at low power. So, the best sounding operating point can be very much different from the best power output operating point. But it is a matter of opinion. Maybe you can see reasons why the 300B is popular. I happen to think the 2A3 is an exceptional sounding tube, from my experiments comparing it to the 300B, and I prefer it. Here's how I run each of three parallel 2A3's: Vp = 305V Ip = 47mA Vg = -65V Rprimary (effective) = 4.8K Pout = 4.3W (at clipping) Pd = 14.3W, 96% of max It just turns out I need to do this to fit a 1600 ohm 150mA transformer with 3 2A3's, and the results are better than if I dropped Vp, sound-wise. Run 2A3's and 300B's hot, seems to be the best rule. --------------------------------------------------------------------------- tubes@hillier.demon.co.uk © M.J.Hillier 1995