Class-g / class-h

[isaacc7]

Hey George,

Any chance (or interest) of incorporating Broskie's "Sliding fixed bias" scheme for even more efficiency? You can see it here.

Isaac

 

[Tubelab_com]

I experimented with this scheme back when Broskie wrote it up. I found that the output impedance of a tube amplifier changes with the bias current leading to distortion if the bias is varied dynamically. Of course this can be cured by adding negative feedback. There are also some time constant issues. I lost interest in the scheme as he presented it shortly after these initial experiments. The hardware I developed during the past year is certainly capable of many dynamic biasing schemes, and I will no doubt be experimenting along these lines in the future.

The autobiasing scheme presented in the original Broskie article is "backwards" from what I believe is needed. A class A amplifier dissipates the most power when it is sitting idle. The peak to average ratio in most music is 10 to 20db. This means that most of the time an audio amplifier is operating between 1% and 10% of its rated power, and that is assuming that the volume control is up all the way. The scheme that Broskie presented would see little use in normal listening situations. What we need is a way to reduce the power dissipated at idle, but be able to crank it up as needed to produce the high power transients. The easiest solution is of course class AB, and even class B operation. It is well known that these methods extract a sonic penalty and are not useful for SE. I have built a nice sounding class B screen driven P-P amp that was about 75% efficient at full power (80 watts). No unusual tricks were employed. Each output tube dissipated about 4 watts at idle.

I am exploring the classic "class H" method of varrying the voltage applied to a cathode follower in step with the audio signal currently. It should be possible to vary the current through a common cathode stage in a similar manner, but I havent gone there yet.

 

[Wavebourn]

Class A is less needed on high powers than on low powers. I've found that A+C transistor amps sounds better than traditional AB amps dissipating the same power, and the secret is a distortion pattern on lower powers that is closer to what we expect from sounds psycho-acoustically. Dynamics matter. When something sounds loud it usually is richer of harmonics than when it sound softly. When strings decay the higher is order of harmonics the faster they go down. The same with human voices. The same with all reflecting surfaces that cause reverberation, so when we hear "an acoustics of a room" we hear sounds that are going softer and softer, with less and less harmonics. But when we listen to sounds through an amp that increases distortions of softer sounds we understand that the sound is artificial even if such distortions are very small when measured. If you hear a static sound of such tone and volume you can't say how distorted it is, but if you hear how it goes softer by loudness but louder by specter you hear that something is wrong!

So, what is the purpose of an adaptive bias?
To get energy saving while maintaining high sound quality like in class A amplifiers, right?

In case of AB or A+C we have already an energy saving, and switching from A to B (or adding C) happens instantly; errors are instantly compensated by a feedback loop that the faster, the deeper, and the cleaner, compensates the better. Making an amp that reacts after the sound level inevitably screws down dynamics because some time is needed for such a reaction.

...modern music is compressed. Even old records remastered are compressed. People got used to smashed dynamics. It is easier to listen to such a music in cars, in offices, in a public transport.
But when we speak of tube amplifiers we speak of a different levels of quality: if we say spending $1,000.00 on parts to build an amp that sounds like a stock $50.00 SMD rug from BestBuy what is the reason?

So, do we need to save an energy loosing a sound quality when we are spending lot of money and time to build amplifiers?

It is more reasonable to go and buy a SMD rug designed on a computer screen and mass-printed in China...