J'ai trouvé un lien direct vers l'article de Keroes pour un Williamson à la sauce Hafler. Il est beaucoup plus lisible que les scans que j'ai mis au dessus.
C'est un pdf que je vous conseille de sauvegarder.
http://www.technicalaudio.com/pdf/Gordo ... n_6550.pdfEncore une fois ce circuit est un Killer musicalement parlant.
C'est la base du montage utilisé par Kevin Hayes sur les VAC.
Et ceux qui ont écouté des VAC en saisiront toute l'importance.
A noter que si on réduit la CR globale comme sur le circuit Eckland, il est possible de virer la compensation R17/ C6 et R18/C7.
Sur un push pull de 300B, il faut augmenter le swing en jouant sur les charges d'anodes de la double 6SN7 driver. On peut également adopter une 6BX7 à la place, ce qui donnera un driver plus robuste.
Le principal avantage des DHT est leur faible résistance interne en comparaison des Tetrodes à faisceau dirigé en UL. (avec en plus une meilleure linéarité, point fort des DHT)
Les conséquences sont importantes musicalement parlant: cela permet de réduire encore de façon nette la contre réaction globale pour une impédance de sortie inchangée.
De mémoire, on descend en dessous de 1 ohm pour l'impédance de sortie (8 de DF sur charge de 8 phm) avec seulement 6 db de contre réaction (provisoire, je vais vérifier...) sur un PP de 300B. Et pour disposer d'environ 50W, il suffit de doubler le push de 300B (en adaptant bien sur le transfo de sortie, il faut remonter dans ce cas environ 2000 à 2500 ohm sur les anodes).
Le seul problème est de trouver 8 tubes 300B capables de tenir le choc.
http://www.vac-amps.com/R30_UltimateAud ... ponent.htmextrait:
The Renaissance Thirty/Thirty produces a grand total of 32-watts/channel from a pair of
300Bs (VAC imported Chinese 300Bs) per channel. Their exceptional linearity, even when
compared to other triodes and to triode-strapped pentodes or beam power tubes
(EL34, KT88), was one of the main reasons VAC decided to use 300Bs in the Renaissance
amplifier series. This linearity is clearly demonstrated when plate current is plotted as a
function of grid voltage for different tube types; the 300B displays a noticeably straighter
and more uniform characteristic, in essence a truer transfer function for the signal.
According to Hayes, "the greatest possible linearity and purest fidelity results from
avoiding cut-off and saturation." By choosing to operate the Renaissance Thirty/Thirty in
Class-A1 mode, the 300B tube never reaches cut-off (the region at the lower end of the
dynamic characteristic curve where the tube ceases to conduct currect) at any time at any
levels up the maximum it operates at. Each 300B is self-biased and protected by a special
"sentry" circuit that senses a runaway tube without any electrical connection to the signal
path or B+ rail (which adds grain).
Another reason is the 300B's low impedance. A 300B delivers exceptional performance
at less than 450 V while comparable triodes such a 845 or 211 typically require 1,000 or
more volts. The 300B's low internal impedance and voltage requirements allow the
construction of a superior output transformer design with less step down and insulation,
allowing better coupling to the loudspeaker. In the case of the Thirty/Thirty, the 14 pound
output transformer uses 21 sections and bifilar winding with match points for speakers from 1
to 8 ohms.
The final reason for selecting the 300B is related to the question of static vs. dynamic
damping. Normally, the best way to achieve a high amplifier damping factor is through
the use of negative feedback which Hayes feels, "entails major comprises. Negative
feedback may reduce distortion but it multiples the order. Higher damping (though the use
of negative feedback) factors don't necessarily translate into better speaker control." While
conventional wisdom is that greater control is achieved when the amplifier "corrects" for
back EMF generated by the loudspeakers extraneous motion or overhang, Hayes points out
that the problem with this theory is that, "the error signal is in itself erroneous and is not
representative of the sound being produced." For example, the acoustic output and the
cone motion may differ somewhat. Or the voice coil motion and overall cone motion differ
due to flexing and breakup. Another serious consideration is the voice coil's large
inductance phase shifts the signal. Finally, there's the problem created by the when the
crossover network further phase shifts the signal and mixes in the different back EMF
from other drivers, resulting in a summed signal that does not represent what any one
driver is doing. So Hayes feels that, "the best overall speaker control and sound fidelity
results from the highest damping obtained without negative feedback and the use of highly
linear triodes with low impedance such as the 300B."
The Renaissance Thirty/Thirty's input circuitry is a direct descendent of the original
Williamson circuit. This gives pure, direct coupled, balanced input amplification and phase
splitting. The circuit is completely push-pull save for first 6SN7/5692 octal twin triode
which VAC operates under conservative conditions to minimize distortion. According to
Kevin Hayes, "the modified Williamson phase splitter is unique in that the same electrons
flow through a single triode creates the antiphase push and pull signals, resulting in a
precision and balance possible up to several megahertz."