More on Class-XD™

My last entry sputtered to an end without my ever mentioning the star of this website—the triode. Don't worry; I will put the “tube” back into the Tube CAD Journal soon enough. Even when no tubes are in sight, tube circuits are always at least implicit in the Tube CAD Journal. How's that?

In general, a sharp divide separates solid-state from vacuum-state electronics, but only on a physical level. Topologically, the division is much more rounded and shallower, if not totally nonexistent. Indeed, the only sharp, unbridgeable divide is the P-channel versions of the MOSFET and FET, and the PNP version of the transistor that finds no equivalent in the vacuum tube world.

In other words, just about any technique/topology can be translated from one device technology to another. For example, I have performed the electronic-alchemist’s trick of converting glass into silicon and solid-state into vacuum-state many times in this journal: the tube-based White cathode follower, SRPP, Broskie cathode follower, and Aikido amplifier into solid-state circuits; the solid-state-based Taylor and the Macaulay amplifiers into tube-based amplifiers. Here is an incomplete list:

http://www.tubecad.com/2004/blog0013.htm
http://www.tubecad.com/2004/blog0016.htm
http://www.tubecad.com/2004/blog0023.htm
http://www.tubecad.com/2005/February/blog0037.htm

Last time, I had a good time teasing Audio Advisor and Cambridge Audio; the former by deflating its sales copy’s egregious exaggeration and the latter by refuting its claim to originality.

But don't get me wrong, I don't claim to have invented crossover-notch shifting. In fact, the technique of forcing a push-pull amplifier into single ended operation over a small portion of its output, so that its crossover notch is shifted away from the zero crossing, is at least 20 to 30 years old. Just about anyone who has seriously worked with OpAmp ICs knows about the technique of adding a pull-down resistor or constant-current source to an OpAmp’s output to force the internal output stage to undergo a crossover-notch shift. My only claim to originality in my 2002 essay on Mixed-mode Amplifiers was pointing out that the crossover-notch-shifting constant-current source could be housed externally to the amplifier to which it is attached, with its own external power supply and enclosure, something of a possible high-end audio accessory that would bestow a single-ended glory on existing push-pull, solid-state amplifiers.

Well, I am grateful to Tube CAD Journal reader James in Australia, who sent the link to the Cambridge Audio’s British patent. In the patent, we read:

I guess this means that all those OpAmp-based circuits built for the last thirty years that used pull-down resistor and constant-current source on the output will have to be desoldered and anyone who built an amplifier based on my 2002 essay or the headphone amplifier based on my 2001 schematic will have to revert their amplifiers back to conventional push-pull mode and live with the crossover-notch distortion at the zero-crossing, as this patent owns it all. Or does it?

Let's be frank, everyone except patent holders and the patent office will admit that the patent system is broken. Crazy, silly, obvious, and prior-art existing ideas receive patents every day. Remember the madness of granting Amazon.com the patent for one-click purchasing? Well, I should have patented two-click purchasing, and you should have patented three-click purchasing, and someone else four-click purchasing… then every time an online purchase is made it will take eight clicks. Eight clicks? Don’t I mean 4,132,482 clicks? No, as the patent office usually takes a year of careful deliberation before granting each patent—which behooves such subtle, complex, and clever inventions—it will be quite sometime before we get to 4,132,482 clicks.

Here is a suggestion: all patent applications should cost $2,000; and all applications are immediately approved; then if someone can show that the invention has already been patented or is obvious or doesn’t work as promised or finds prior art, then that worthy individual (or company or institution) is awarded $1,000 as a reward from the patent office and the patent is thrown in the trash. If two years pass with no dethroning, a long-term patent number is assigned.

Returning to crossover-notch shifting
We have seen how in push-pull solid-state amplifiers the crossover notch can be shifted away from the zero crossing, which involves creating a mixed-mode amplifier, wherein the power amplifier shifts from single-ended, class-A mode to push-pull, class-B mode. So, how do we do it with tubes? Well, in the same 2002 essay on Mixed-Mode Amplifiers, such a design is laid out. The schematic below came from that essay.

This topology loads one output tube with a constant-current source, while the other output tube is cut off. If a resistor were used in place of the third pentode, the current draw through the resistor would be in the same current phase (but the opposite voltage phase) as the active output tube. This is not the way to go, in terms of efficiency, although I am not convinced that it doesn’t have some merit in terms of distortion. But how do we achieve the inversion of the resistor’s current draw, as in the Cambridge Audio amplifier? Below is one possibility.

Or, instead of driving the screen, we drive the grid, which would allow triodes to be used instead of pentodes.

Transformer crossover-notch distortion
In a push-pull, transformer-coupled amplifier, the output transformer’s core exhibits hysteresis, a lagging of the magnetic core material when faced with a polarity inversion of the magnetic field. In other words, a type of crossover-notch distortion. One of the great advantages to transformer-coupled single-ended amplifiers, on the other hand, is that the output transformer does not undergo a hysteresis glitch at the zero crossing point, as the core is always magnetized in one direction only. Will the above two schematics address this problem? No. Is it a big problem? I believe it is, but some testing would be helpful.

For example, we could build an OTL amplifier of sorts to which we then attach an output transformer (air-gapped) via a large-valued coupling capacitor. Then we listen and measure, but not at full output, but rather at 1W of output. Next we add a resistor that bridges the secondary to the B+ voltage. Now the secondary will experience a DC bias current, which will shift the hysteresis-induced crossover-notch away from the zero–crossing point. My bet is that the resistor-laden version wins.

Next time
I actually planned on writting much more. So expect more very soon. And I must point out that I am running low on mono PCBs.

//JRB