About a simple class D amp

Tubes are worse in so many ways. They are slow, unreliable, fragile, expensive, low-gain, inaccurate, low conductance, and only come in n-channel.

How would that reduce hum?

(The

Cute exercize, but insane.

John

With HC14, you have six gates anyway. I believe there used to be a standard design of a reasonable class D amp using one 4069 and two FETs.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

It's a shame they only come in N-vacuum, I've been thinking for many years of attempting a P type but if it works at all, it's not likely to work nearly as well as P-silicon does (i.e., closer to an order of magnitude worse performance, not just unreliable

This circuit will last >10khr on a change of tubes. At reduced ratings, 50k. Notice, BTW, I'm running the output tube at or beyond ratings, no heatsink -- semiconductor lifetime drops steeply from

100khr to fragile

A valid complaint, I am not using specifically robust tubes. Can be partially mitigated with mechanical construction.

Within an order of magnitude. The tubes are cheap, actually. It was specifically part of the design to use the cheapest tubes available. If Philips still made 600 ohm speakers, I wouldn't even need an OPT, and if you don't mind line operation, I can skip the power transformer (filament supply can be supplied from a resistor). It wouldn't be as cheap as silicon, but it would be within an order of magnitude.

I'd like to see your 2N3819 do 12mmhos. The cheapest triode-pentodes available (similar to 6U8) have gain around 12mmhos at circa 10mA. (MOSFETs have higher gain, but still within an order of magnitude, amp per amp. See below.)

Blatantly false. In fact, tubes have the most consistent matching of all active devices, within 10% on all characteristics (Gm, mu, Rp, offset, current, etc.). You can't even get gain-matched single BJTs closer than -25/+50% in the same selection.

Also within an order of magnitude. Plate "on resistance" is circa 100 ohms for this rather small sweep tube; the color TV tubes (EL519, et al.) are closer to 50 ohms or less. What's the Rds(on) of a comparable MOSFET? A MOSFET rated for 6kVds(max) and 2A Idss would be more than a couple of ohms. It may even be 50 ohms. On the other half of the cycle, the damper diode is around 50 ohms, half that for larger (color TV sized) dampers. More than silicon, but not by much since we're talking ca. 6kV PIV, 1A, which still has a fairly high Vf.

No, tubes are still not competitive. If you thought I was trying to show them as such, you were sadly mistaken! They are, however, capable of getting within not only an order of magnitude, but even approaching a factor of 2 with comparably sized silicon devices.

In fact, the two biggest drawbacks, that you didn't even name, are heater consumption and physical size.

Tubes always need a load of heat. I don't know that anyone has tried newer cathodes (e.g. LaB6, carbon nanotube field emitters) with classic designs, so we're pretty well stuck with the time-tested rare earth cathodes. It takes a lot of heat to make a lot of electrons; the output stage consumes a good 17W doing nothing. It could use less, but "on resistance" would suffer proportionally.

Tubes also do not miniturize well. I don't know that anyone has tried at least fabricating them from etched or printed sheets, a valuable lesson from the IC / micromachine industry. Even if put into a tight- fitting envelope, the cathode's emissivity limits the possible physical size. (The same can be said for silicon, its electron density is the limiting factor for size. But silicon contains orders of magnitude more free electrons than rare-earth cathodes do, and a chip like IRFZ46N can sink 100A peak through a die far smaller than a

6GK6's cathode, which might saturate at 5A.) At best, I suppose you could put something like the 6LQ6 into one of those bigger-TO-247 packages. But a comparable MOSFET would be packaged in a TO-220 (well, if the TO-220 could handle 6kV... eh, for that matter, neither can the TO-247..), or even D2PAK. Or even Zetex might try one in SOT-223, those crazy bastards. So even with modern construction, I can't see thermionic tubes matching physical size.

There are still a few rarefied uses of tubes. Besides the specialty fields (magnetrons, for instance; TWTs are still going into space; and you can pry my Trinitron CRT from my cold, dead hands), they can't be matched for sheer voltage capacity. Didn't Win Hill pick a cathode follower for some high speed ramp generator last year?

Insane? I take that as a high compliment. Thanks. :-)

Tim

Many years back I saw one that used CD4049s. The whole circuit used the 4049 as all of the gain and switching elements. Someone did it as a joke in the first place.

Many years ago I made my own class D power amplifier. It sounded very good. The whole thing was discrete. The sawtooth was very linear and the comparators were fairly fast. This made the open loop linearity fairly good. The feedback cleaned it up the rest of the way.

Class D amps is one of the things that I design as the business.

There is big problem with the open loop linearity of the class D: the effective time of switching depends on the operating point. There is a knee in the transfer curve between the continuos current and discontinuous current conduction modes. Due to that effect, the attainable open loop THD is at the order of 0.5...1%. The feedback of the 1-st order has to be shallow, and the feedback of the higher order distorts the saw. Higher clock rate is a cure for many sins, for the cost of EMI and power. It takes a lot of optimization to get the quality audio from the class D amp while maintaining the good power efficiency.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

you'll need stronger gate-drive on the mosfets also need a transformer that can handle the harmonics without overheating.

FYI, GE tried it about 40 years ago. "Thermally Integrated Micro Modules," they were called. Of course, they didn't have anything that we would call micro-machining, so they were built out of metallic disks and ceramic rings for insulation, and rather than trying to include tiny filaments, the whole assembly was built into a vacuum oven and run at dull-red heat, with the differential emissivities of the cathode and plate materials making the electron flow. The point of the exercise, I gather, was to build very rad-hard electronics, a great concern at the time. I never found out what became of the project. That, and cold-cathode magnesium oxide emitters, which seemed promising.

Semiconductor-vacuum devices are still very interesting (microtip or equivalent electron source, vacuum gap, anode) but so far none are practical.

John

I have a Nuvistor in my desk drawer that's pretty small. I also have one of those ceramic-stack devices, but can't find it at the moment... need to dump out the desk drawer onto the kitchen counter and do a "sort" ;-)

...Jim Thompson

--
| James E.Thompson, P.E.                           |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |

  "When anyone resorts to personal attacks, it is almost always
   because they are losing an argument."     -The Happy Heretic