Class-g / class-h

[dhaen]

I've been looking for examples of class-g / class-h implemented using valves as the output devices, but cannot find any. This might be because of the search terms, or maybe nobody has...
Any links or references are welcome.

This is a purely academic exercise .... at the moment...

[Miles Prower]

^^^^

Tubelab (George) is working on something like that. Class H or G was never implemented back in "the day" since this requires lots of solid state electronics to pull it off.

[smoking-amp]

This scheme I think would qualify (as class G?) if the SS amp is running class AB, since the SS amp output backs up thru the tube output xfmr unless the tube conducts. So one could put a minimum B+ supply in the tube amp part, and the SS amp will supply the B+ variation needed for signals.

Depends on the gain settings of one amp versus the other as to how much "B+ overdrive" the SS amp will push back thru the OT. Ie, set the gain of the SS amp too high versus the tube feedback network ratio, and the tube amp will see its B+ vary with signal.

Of course, if you set the gain of both amps equal, then the tube amp will see constant B+, and will conduct with constant voltage across the tubes.

The tube amp, via its feedvack network, still defines the final output signal.

http://www.diyaudio.com/forums/showt...030#post871030

Don

[tubelab.com]

Tubelab (George) is working on something like that. Class H or G was never implemented back in "the day" since this requires lots of solid state electronics to pull it off.

I built a SET amplifier that used a single 6336A tube per channel to make 38 watts at 5% distortion, 24 watts at 1%, and 10 watts at 0.33%. The key was a DSP controlled "modulated" switchmode power supply. The same amp will do over 20 watts with a 6AS7 tube. This amplifier was built for and submitted to a Microchip dsPIC design contest hosted by Circuit Cellar magazine. In a sea of microprocessor controlled gizmos, this vacuum tube amplifier won one of the prizes. The amp was finished just in time for the contest deadline (October) and hasn't been touched since due to seriously ill family members.

The magazine has stated that they are running a special issue highlighting all of the contest winners, and they have asked me to write a future article about the design, so it will come off of the shelf soon for some more experiments. I have also been experimenting with an "all tube" amplifier design incorporating some of the concepts developed during the design of the DSP amp. It will not have the efficiency improvements afforded by the DSP controlled SMPS, but should capture some of the distortion improvements. That design can be found here:

augmented cathode follower?

The entire project submission including schematics and source code can be found on the Circuit Cellar web site here:

http://www.circuitcellar.com/microch...rs/MT2209.html

Ever see a tube amp that you tweak with a laptop?

[dhaen]

George's idea is more akin to what I was imagining, with modulated rails and bias but I had no idea that a solution would be quite as complex. The word "impressed" would be a complete understatement!

I'm afraid my micro-controller program experience is almost zilch. Whenever I've sat down to learn, I've allowed myself to be distracted by something seemingly more interesting (often of the glowing variety)....

George, did you choose a cathode follower output stage for a particular reason?

[smoking-amp]

John,

The two amplifier scheme I pointed out IS a modulated rail design. It's just not obvious unless you analyze it. The SS amplifier is providing a low voltage modulated rail output and that is fed backwards thru the output tranny to the tube amp to become HV modulated rails for each tube. The settings of the two feedback network gains is what determines how much B+ modulation will show up in the tube amp. Also the OT xfmr impedance ratio is important.

You probably are looking at it as a mostly SS amp with a little tube power added on. But that is the way ALL of the modulated rail designs work. The modulator is always effectively in series with the final amp stage and the modulator DOES put out most of the power itself. But the final amp stage (the tube amp here) is still what controls the final voltage output.


If you look at the Technics Class A+ designs, which use a Class AB amp to provide modulated rails for a low voltage Class A stage, you will see the series connection clearly.

The scheme I pointed out is actually a very elegant design since it avoids having to deal with HV modulators.

This is similar to the way the Berning Impedance converter generates HV for its tubes by feeding low voltage backwards thru the OT xfmr. Just not using a switching xfmr in this case. (So we have to supply the minimum DC B+ in the tube amp. since we can only push AC rail modulation back thru the conventional xfmr.)

The Berning imp. converter even could be made into a tracking rail design with PWM added to the switching imp. conv. The ripple free Cuk converter would be ideal there. (I have pointed this out in the past as a way to make a dynamically VARIABLE RATIO switching output xfmr. I better not see any patents filed on these! These are declared open to DIYers.)

Don

[Wavebourn]

Originally posted by smoking-amp


The tube amp, via its feedvack network, still defines the final output signal.

http://www.diyaudio.com/forums/showt...030#post871030

Don

How much of SS power will be dissipated on output tubes, and how this power will be reflected on working points of output tubes?

I mean, they don't have zero output impedance, right?

[tubelab.com]

George, did you choose a cathode follower output stage for a particular reason?

The cathode follower was chosen for its high PSRR (power supply rejection ratio). The digital power supply solution that I have used here has 10 bit resolution. This means that the power supply will not be exactly what you expect. A cathode follower gives you about 30 db of isolation from the power supply imperfections. For this reasons most SS class H amps use source follower output configurations.

I had no idea that a solution would be quite as complex.

This particular implementation was meant to be a learning experience for me, and it was an entry into a dsPIC design contest. The dsPIC was required for this project, and it will allow for some serious experiments in the future.

The modulated supply concept can be implemented with mosfets or even vacuum tubes, but you will loose the efficiency enhancements afforded by the SMPS. I have working (sort of) examples of both designs that will be on my web site after I perfect them.

It is also possible to build the modulated supply without a microcontroller. You would need to modify the control circuitry of a conventional SMPS controller IC circuit. Look at some of the digital modulators being used by the high power ham radio operators for ideas. One example can be seen here:

http://www.classeradio.com/modulate.htm

[smoking-amp]

Hi Anatoliy,
"How much of SS power will be dissipated on output tubes, and how this power will be reflected on working points of output tubes?"

As George just pointed out, the usual approach is to use followers, preferably pentode or hi-Z bipolar/Mosfets to get maximum PSRR. It is really, the Hi Z plate/collector/drain side that is giving the isolation from power supply. But follower mode also lowers the controlled side Z to help as well. Using a triode will work, but PSRR will suffer. Follower mode in the tube amp is not required with pentodes. With triodes, it would be necessary to get good PSRR.

So, as an example, first suppose the feedback network for the tube amp specifies Vout/Vin as 20. Using the same Vin for the SS amp, we set its feedback network to Vout/Vin as 20 also. Now the SS amp will be putting just the right voltage for the Tube amp to be happy. So the tube amp will be maintaining constant voltage across its pentodes (whatever the internal B+ supply is providing). The current thru the pentodes will be varying as the signal however, and the OT Z ratio determines how much current shows up there (and to be shown later, how much % power is delivered by the tube amp relative to the SS amp) (tube biasing will determine any additional DC component of current from the internal B+). The tube amp is not just loafing in this case (with a constant voltage across the tubes) however, it has to actively control the required current draw, so the grids are operating dynamically. Its a real amplifier, just not putting out extra power.

Now suppose the feedback network for the SS amp is set Vout/Vin = 22. The tube amp will now be unhappy with the signal output by the SS amp, so will allow the excess back primary voltage to additionally drop across the pentodes, ie, we now see effective B+ variation in the tube amp. (its magnitude depends on the OT Z ratio and how much excess voltage the SS amp generates.) Remember, the two amps are in series, so the tube amp can "eat" the output voltage of the other by dropping more V across the tubes if it sees fit to.

If we were to use triodes in the tube amp, their output Z will set a max voltage that can be "eaten" since the excess voltage appearing across the tube will eventually cause sufficient current flow to match the required load current draw. PSRR will suffer due to the plate resistance as well, so SS sound will bleed thru as well.

The OT Z ratio is set in a way to determine how much power will be delivered by the tube amp versus the SS amp. For a nominal 8 Ohm speaker load, the secondary winding on the tube amp xfmr will normally be considerably less than the usual 8 Ohm impedance. This is because of the rule that power out is V*V/Z. To provide10 Watts from the tube amp and 70 Watts from the SS amp, the peak voltage needs to increase from 47.3 Volts to 50.6 Volts. Notice that the increase of 3.3 Volts would not lead to 10 Watts if put into an 8 Ohm load. It has to be put into a lower Z in order to generate 10 Watts. The reason being that additional current is flowing here due to the SS amp.

I think I gave a formula for determining the tube amp secondary Z in the referenced thread, but its not too hard to work out the formula from the usual Ohms laws.

Don

[smoking-amp]

Whups, forgot the third case.

If the SS amplifier feedback is set to Vout/Vin = less than 20, then the tube amp has to provide boost voltage to get the output up to snuff (Vout/Vin=20). So it now consumes some of its own internal DC B+ to generate the extra output like a normal amplifier would. So plates start swinging in phase with signal.

(in the earlier case, where the SS amp was putting out excess amplitude, the tube plates swing in opposite phase to the normal amplifier operation, effectively "eating" the extra power. Ie, plate voltages increasing when current increases.)

This 3rd case (SS putting out less than required), or the zero plate voltage variation case (SS putting out exactly that required), would correspond to the normal tracking rail amplifier cases.

But the excess output from the SS amp case demonstrates that the tube amp can correct the output to its liking no matter what distortion or variance the SS amp kicks out versus what the tube amp wants to see for output. (well, up till it runs out of internal B+)

Don