An Amp with no R's and no C's

This report describes a small (2 watt) SE power amplifier that uses no resistors, no capacitors, no batteries and no solid state devices anywhere in the amplifier. There has been some discussion as to the limited sonic qualities of these devices, so I decided to see if it were possible to totally eliminate these "bad" devices. The resulting amplifier utiilizes -only- tubes, transformers and inductors in its circuitry, including the power supply. (There is no filter capacitor in the supply).

The Power Supply

The first challenge was to see if a capacitorless supply was realistic. Energy can be stored in an inductor as well as in a capacitor, so this could work. The problem is that whereas a series R, with C to ground filter has a decreasing output impedance with frequency, a series L with R to ground has a relatively high output impedance. So we will need to use a shunt regulator (VR tubes) to improve the filtering as well as keep the output impedance low.

To eliminate the "R" component, the built in effective supply DC resistance must be high enough to prevent the VR tubes from conducting too heavily when the line voltage is too high (or load current too low) and from extinguishing when the line is too low. If we use a tube rectifier, this will provide some relatively high output impedance due to the tube voltage drop, and using a relatively small power transformer will provide a relatively high impedance (poor regulation) as well. If we combine this with a relatively high DC resistance in the filter choke, this might provide sufficient current limiting. The initial test with a 6X4 rectifier and small 750VCT transformer and a high values choke (50Hy) showed an output voltage of 260 volts at 50mA, and 250 volts at 60mA. (And, I should point out, about 20 volts of ripple).

If we design the power amp for an output stage current of 22 mA, an input stage current of 12 mA (from a "tap" on the voltage regulator), then it should be possible to place 30mA thru the "top" VR tubes, extract 12mA and have 18mA flow through the "bottom" VR tubes. This will maintain regulation as long as the instantaneous load current (from the audio signal) does not exceed an additional 15 or so mA. Using a 10k output transformer for the output and transformer coupling the interstage should do this.

This allows us to use an 0A3 (75V), an 0C3 (105V) producing 180 volts and another 0A3 to produce 255 volts. Notice that with 22 mA in the output stage, this means we will have about 33 mA flowing in the VR tubes. At low line, this decreases to about 26 mA and at high line increases to 40 mA, which is the maximum rating of the VR tubes. So we are "safe", as long as we select VR tubes that sum to approximately "nominal" 255 volts. [All VR tubes will regulate at slightly different voltages, so we will have to pick some, if we happen to get a bad "mix".

As it turns out, the VR tubes reduce the ripple to a measured 1.1V at the 255V point (which with my set of tubes was 256.3 volts). The ripple at the 75 volt point (74.1 volts) was 350mV. This was considered acceptable for initial tests.

The supply CAN be made with no R and no C! The power supply parts count is remarkably low, consisting of 1 transformer, 1 6X4, 2 0A3, 1 0C3, 1 50Hy choke, and fuse, linecord and switch.

The Amplifier

Sovtek provides a high gm low rp dual triode called a 6N30Pi. This tube appears to be beefy enough to provide a couple of watts output in SE class A. With about 265V plate voltage and 15 volts bias, slightly over 2 watts can be obtained into a 10k load. Distortion is relatively low (about 4%) at this point. The second section can be used as an input stage, so that somewhat less than a volt provides full output.

So how can we bias this thing without batteries and without resistors?

If the input stage is run at about 70 volts, and biased at about 3.6 volts, we get about 12mA flowing.

I mentally toyed with using various diodes for bias, but in the back of my mind was a thought: Can  I kill 2 birds with 1 stone, so to speak?

If we use another triode as a "feed forward" amplifier, driving its grid positive, then the plate voltage and impedance ought to be low enough to provide the cathode bias for the output stage. So this quickly lead to an investigation. Is there a triode that properly biases the input stage driven into the grid (so the grid draws current) and at that level of positive grid current properly biases the output stage?

As it turns out, there are several choices. Too low mu (for instance 12B4) resulted in too low a bias for the output stage, too high mu (for example 6DJ8 with both sections paralleled) resulted in too much current in the input stage to properly bias the amplifier. Three choices turned up:

1. 6S4. This provided almost perfect bias, but the effective plate resistance was high enough so that the amplifier output impedance was about 20 ohms. Otherwise, a very good choice.
2. 12GN7 triode connected. If the plate voltage on the input tube could be run at 41.5 volts, this would be perfect. Good bias, low effective plate resistance for good damping in the total amplifier, and very low distortion. The amp distortion 1 dB below clip was less than 1% and almost all second order.
3. 12BH7, both sections paralleled. I ended up with a Tung-Sol part as it sounded better as well as measured better than either GE or Sylvania. This provided adequate bias, low enough plate resistance and low overall distortion (which measures 2% 1 dB below clip). This is what I ended up using. Notice the 2% distortion is lower than the 6N30Pi by itself provides. This is due to the forward bias characteristic of triodes... they curve opposite the normal negatively biased operation, providing partial compensation of the transfer characteristic.

Overall tubes investigated for biasing: 12AV7, 12B4, 6S4, 12BH7, 6S4A (incidentally, not the same as 6S4), 6CG7, 6DJ8, 6N1P, 6N30Pi, 5814A, 12AU7, ECC99, 12AY7, 12BY7, 12GN7, 6AH6, 6AQ5. (I tried the pentodes in pentode and triode mode; for power supply considerations, triode mode connection worked best. If I had "unlimited" voltages available, 6AG7 also worked well in pentode mode).

The overall circuit looks like this:

Notice the amplifier section adds the following components to the power supply list above: 1 interstage transformer (I used a Lundahl 1660), 1 output transformer (I used a Magnequest FS-100), 1 6N30Pi, 1 12BH7, 1 100Hy inductor, input and output connectors.

Performance:

Power output: 1.5W. Can be raised to 2W with 280V on the output tube.

Sensitivity: 0.8VRMS at clip.

Frequency Response: -2dB ref 1kHz at 20Hz and 20kHz.

Distortion: 2% at 1.2 watts. Mostly 2nd. Slight 3rd, 4th present.

Output impedance: about 11 ohms as the circuit is shown.

How's it sound?

I had the option of using either the power supply shown, or hooking it up to stiff "bench" supplies. The sound was very similar in both cases, being ever so slightly more "open" with the supply shown. Didn't like the residual hum, though (which is equivalent to that produced by a DHT with AC filament). Many folk would consider this hum level acceptable.

Even though the output impedance was relatively high, there was no flabbiness to the bass response; it was nice and crisp on the transmission line speakers I used for the tests.

The overall sound is pleasant. It is odd in that the amplifier sounded quiet (low volume) for a given power level; almost reminiscent of a solid state amplifier. It retained the nice open sound of a tube amp, though. Vocals clear, open. Instruments detailed. Certainly SE sounding.

Clipping was very benign on the bench supplies, almost invisible until you get to very hard clip conditions. With the power supply shown, hum level increased near clipping, but as the signal also increased (by definition), it wasn't noticeable. If I was particularly listening for that effect; I could tell.

Summary

It's an interesting amplifier, well worth the effort to string it together.

-Steve