300b lovers

I have mixed feelings about cathode follower drive: the output impedance of the CF is low (probably 100 ohms or so), but the peak current available is no different than anode drive. Considering the load is dominated by the Miller capacitance of the 300B (about 80 pF), the CF will definitely extend the small-signal bandwidth, but will have no effect on the large-signal bandwidth (also known as slew rate) which is determined by the (linear) current available to charge a capacitive load.

@lynn_olson I think you have the highlighted bits wrong.

A CF driver can deal with a lot of capacitance in the grid of the output tube. We use a single 6SN7 section to drive 14 such grids (in our MA-1 amplifier) and it does it with no worries even in class A2 (or AB2, if the amp is subjected to a low impedance load) where grid current is present, with good linearity.

The peak current available is higher because the coupling is more efficient and the output impedance of the CF so much lower. When AC coupling (anode drive) it is very difficult to get the driver to be able to handle grid current in the output section (transformers are good at this though)!

The large signal advantage is several: no blocking distortion at overload since there’s no coupling cap (so overload recovery is instantaneous) and the Voltage amplifier sees a very high impedance load so it has a much easier time doing its job (so it can be lower distortion). This allows for the coupling cap used between the Voltage amplifier and driver to be a small value, which is advantageous because there’s less inductance associated with the coupling cap, so it can sound better and also offers better layout options. This aspect helps with HF bandwidth but also helps if you want LF bandwidth since the coupling cap value is so small.

In our MA-1 we have full power to 2Hz using a 0.1uF capacitor.

Another advantage is as a fixed bias scheme, its extremely stable since the impedance controlling the power tube’s grid is so low. Put another way its more reliable.

My advice, since its obvious you’ve not tried it, is to do so. If for an SET I would limit the LF timing constant of the Voltage amplifier’s coupling cap since SETs have such terrible problems with elliptical load lines at low frequencies.

Something both transistor and tube amps share are performance limitations set by current available to drive a capacitance. In a transistor amp, that will be the dominant pole capacitor associated with the second voltage-gain stage. The current available to charge that capacitance sets the slew rate of the entire amplifier. It should be mentioned that slew rate is similar to hard clipping; in slewing, linear operation has ceased, and there is no relation between input and output. As in hard clipping, there is a large region that is pre-slewing, with increased distortion but still a relation between input and output. With sinewave stimulus, the region of maximum dV/dT (rate of change) is actually around the zero crossing, which of course generates lots of high-order distortion harmonics.

Hard clipping and slewing have quite different origins; hard clipping is the result of one or more stages getting too close to the power rail, abruptly shutting off the gain stage. This can be either hard or soft, depending on the amount of feedback as well as the shutoff characteristics of the active stage. Transistors typically have a quite narrow shutoff range, with 0.7V being typical. Tubes are usually considerably broader, around 10 to 30V, depending on the device. This is also why tube rectifiers have a softer switching characteristic than solid-state.

Slewing, by contrast, is part of the amplifier running out of current, not voltage. Specifically, current available to charge a capacitance. Now, 80 pF isn’t much capacitance, but tube circuits are inherently high impedance (compared to solid-state) and operate at fairly low currents (again, compared to solid-state). The appearance on a scope are triangle waves, instead of flat-topped sine waves.

The somewhat arcane descriptor often seen in op-amp specifications is "large-signal bandwidth". This is another way of seeing slew rate: you measure output just below clipping and increase the frequency until the output begins to decline (which is the result of massive slewing). This measurement is simple enough for an op-amp and done all the time, but it can destroy a solid-state amp, so it usually calculated from quick measurements of slew rate.

Small-signal bandwidth is quite different and is usually measured well below clipping ... 1 watt is a common reference point. This is the result of various lowpass functions in the amplifier ... in a transistor amp, there is often a simple RC filter at the input that scrapes off unwanted RFI, which causes distortion in the audio band in solid-state circuits. Tubes are less prone to this but it is still not desirable to amplify AM radio signals at 500 kHz. In the tube world, the dominant lowpass is often set by the output transformer, which behaves like a 2nd-order (or higher) lowpass filter around 50~100 kHz. If a feedback circuit is wrapped around an output transformer, there needs to be compensation in the feedback network that phase compensates for this ... that’s the shunt capacitor you see around the feedback resistor ... but it must be tuned for that specific transformer, not just anything.

The Raven and Blackbird are non-feedback amplifiers, with the cathodes bypassed so local feedback does not apply, either. So there are no stability criteria or load stability issues. The distortion is simply the distortion of the matched pairs used in the preamp or amplifier. It has similarities to a SET amplifier in terms of a simple harmonic structure, but the pair-matching and balanced operation reduce distortion by about 30~35 dB ... without feedback or any associated stability or settling-time issues.

The approach Don and I take are borrowing elements of a modern SET and classical Western Electric designs from the 1930’s, getting the distortion as low as possible at the device level. This is where balanced operation and transformer coupling come in. The transformers of the 1930’s didn’t have much bandwidth, but they didn’t need it, since signal sources usually topped out at 8 kHz. Nowadays, of course, we need at least 30 to 50 kHz, which is where custom-designed transformers come in, using design tools not available in previous decades.

To recap, there are different challenges associated with low and mid-frequency distortion vs high frequency distortion. HF distortion is very often caused by nonlinear current delivery into a capacitance, and stray capacitance is everywhere in audio design. Sometimes you can reduce the capacitance using various methods such as cascode circuits, or pentodes (which are electrically similar), or take the alternate approach of increasing the drive current severalfold.

The 300B is a bottleneck in many SET amplifiers. The 80 pF load isn’t so bad, but the 300B requires 70 to 80 volts to clip it, and if the driver circuit is A2 capable, 100 volts. And ... the 300B has lower distortion than many, if not most, driver circuits, which defeats the entire purpose of using an expensive DHT like the 300B.

What looked like a simple problem is not simple at all, if you want to hear what the 300B actually sounds like, instead of a distorting driver stage. You have to deliver extremely low distortion into a capacitive load, over a range of hundreds of volts (if using PP output devices). This is no longer trivial. The common RC coupling seen in many amplifiers may not be up to the task.

We found transformer coupling with dedicated power tubes, themselves operating balanced Class A mode, gave the lowest distortion. Transformer coupling also allows A2 drive, with the 300B smoothly transitioning into the positive-grid region with no glitching. Although the 300B is not rated for A2 operation, we’ve found no indications of harm, although steady-state operation into A2 might overheat the grid, so not suitable as a guitar amp.

Now if feedback enters the picture, the design criteria all change. Forward gain goes up by as much as 10~20 dB, different parts of the circuit get optimized, and stability at high frequencies, particularly transient overload, become important. Nested loop feedback (2nd-order or higher) gives even lower distortion, but long settling times (after transient overload) can be problematic (because the different loops have different recovery times). You can have even more fun with modern feedforward techniques, but now we need serious computer modeling to pull that off and still have a stable amplifier.

The Raven and Blackbird are non-feedback amplifiers, with the cathodes bypassed so local feedback does not apply, either. So there are no stability criteria or load stability issues. The distortion is simply the distortion of the matched pairs used in the preamp or amplifier. It has similarities to a SET amplifier in terms of a simple harmonic structure, but the pair-matching and balanced operation reduce distortion by about 30~35 dB ... without feedback or any associated stability or settling-time issues.

@lynn_olson Our OTLs are zero feedback too. We get a similar reduction in distortion over SETs.

Slewing, by contrast, is part of the amplifier running out of current, not voltage. Specifically, current available to charge a capacitance. Now, 80 pF isn’t much capacitance, but tube circuits are inherently high impedance (compared to solid-state) and operate at fairly low currents (again, compared to solid-state).

The load the driver sees in our OTLs is considerably higher than just 80pf; more like 200pf. We tie the cathode resistors of the driver tube to B- which in our smaller amps is about -300V. We put some current though there too, so the 6SN7 in question has to have a -GTA or -GTB suffix to handle both the current and Voltage to which is subjected. But they hold together in that circuit for years and even decades.

Driving a 300b in an SET has a lot in common with sitting on a park bench by comparison. We built an OTL using four 300bs about 25 years ago using the same kind of driver circuit. It worked fine. There's no problem at all using this technique as seen in the schematic that was posted above.

Ralph, you’re the acknowledged OTL expert. How does an OTL amp work without feedback? I’d like to know. The last I checked, tubes designed for series regulator use like the 6080, 6AS7, or the Russian 6C33C have a Zout on the cathode side somewhere around 100 ohms. I’m not an expert on this class of tubes, but I wouldn’t expect any tube to have a Zout in the 8 ohm or less range.

(For the reader following along: a Zout of 8 ohms gives a damping factor of 1, a Zout of 2 ohms a damping factor of 4, and so on. Zout in the 0.1 ohm range, typical of transistor amps, gives a damping factor of 80. It should be mentioned that damping factors much greater than this are kind of pointless, since the speaker cable, which is in series with the DCR of the lowpass inductor in the crossover, will typically have a DC resistance of 0.1 ohm or so. I don’t expect speaker cables to use superconductors any time soon, so we’re stuck with copper or silver at room temperatures.)

I’ll admit this is kind of an idle curiosity since I will never design an OTL amplifier. The amps I’m interested in use transformers to solve various circuit-design problems. But I’m always curious how things work, whether solid-state, vacuum tube, or Class D, or hybrids of all three (like ZOTL’s).