The Class – H Amplifier | diyAudio

I will use your nomenclature for consistency. The saturated rail switch will be referred to as class H, and the linear follower (modulated rail voltage) as class G.

Prior art on class G (3772606):

http://patimg1.uspto.gov/.piw?Docid…ageNum=&Rtype=&SectionNum=&idkey=637608C562C2

An earlier example for CRT deflection (3622899):

http://patimg2.uspto.gov/.piw?Docid…ageNum=&Rtype=&SectionNum=&idkey=2E6E84FB42ED

Early class H (3319175):

http://patimg2.uspto.gov/.piw?docid…=3319175.WKU.%26OS=PN/3319175%26RS=PN/3319175

‘Commutation’ was Bob Carver’s nomeclature for a rail switching design. I believe he coined this term for the diode on the lower rails, also refered to as an ORing diode (as it allows the output stage to use power from either the lower, or the higher supply, as needed). These diodes switch, or commutate as needed. The fast switching, soft recovery types, or newer high voltage Schottky type are best here.

The theoretical class B efficiency is 78.8%, a three rail design like the M1.0 is 87% (after Pol HSU). When you consider bias in a class AB design, 60% is more in line. Pol HSU found that the rail switcher showed even higher efficiency driving reactive loads vs resistive.

At full power, class G and H are slightly less efficient than class AB. However, the majority of the time we are not at full power.
Worst case for the class AB would be at about 50% voltage out, where the rail switchers are king.

Most Carver are three rail designs, most Crest are two rail designs. AB International has an interesting two-and-a-half rail design. Instead of a third set of transformer windings, associated rectifiers, and filter caps, they use 0V as a rail in class H. No improvement on a resistive load line, but much better into a reactive load line, and a reduction in the Vce required for the outputs.

IMO class G sounds better than class H. One example of this is the QSC MX2000, a 1KW per channel design. The original was class G, the ‘A’ version class ‘H’. The original sounds better, and the used market value is higher for that very reason.

Why do any class H?

Cost, it is much lower for class H.

The later Carver M1.0 is an interesting hybrid, being class G on the lower rail, and class H on the higher rail. Good sonics at a reasonable cost.

The Early Effect doesn’t seem to be a real issue.

An early example of the ‘Tracking Down Converter’ or ‘BASH’ type (3426290):

http://patimg2.uspto.gov/.piw?docid…=3426290.WKU.%26OS=PN/3426290%26RS=PN/3426290

Hi Jens,I will use your nomenclature for consistency. The saturated rail switch will be referred to as class H, and the linear follower (modulated rail voltage) as class G.Prior art on class G (3772606):An earlier example for CRT deflection (3622899):Early class H (3319175):’Commutation’ was Bob Carver’s nomeclature for a rail switching design. I believe he coined this term for the diode on the lower rails, also refered to as an ORing diode (as it allows the output stage to use power from either the lower, or the higher supply, as needed). These diodes switch, or commutate as needed. The fast switching, soft recovery types, or newer high voltage Schottky type are best here.The theoretical class B efficiency is 78.8%, a three rail design like the M1.0 is 87% (after Pol HSU). When you consider bias in a class AB design, 60% is more in line. Pol HSU found that the rail switcher showed even higher efficiency driving reactive loads vs resistive.At full power, class G and H are slightly less efficient than class AB. However, the majority of the time we are not at full power.Worst case for the class AB would be at about 50% voltage out, where the rail switchers are king.Most Carver are three rail designs, most Crest are two rail designs. AB International has an interesting two-and-a-half rail design. Instead of a third set of transformer windings, associated rectifiers, and filter caps, they use 0V as a rail in class H. No improvement on a resistive load line, but much better into a reactive load line, and a reduction in the Vce required for the outputs.IMO class G sounds better than class H. One example of this is the QSC MX2000, a 1KW per channel design. The original was class G, the ‘A’ version class ‘H’. The original sounds better, and the used market value is higher for that very reason.Why do any class H?Cost, it is much lower for class H.The later Carver M1.0 is an interesting hybrid, being class G on the lower rail, and class H on the higher rail. Good sonics at a reasonable cost.The Early Effect doesn’t seem to be a real issue.An early example of the ‘Tracking Down Converter’ or ‘BASH’ type (3426290):