Tube amplifier with active DC balancing

Introduction

Motivation

The aim to design a tube amplifier originated from different motivations. First of all I started my activities in audio with the design and assembly of tube amplifiers for a musical band. The second reason was the increasing interest of the audio community for tube amplifiers that have always remained to deliver a "better perceived sound quality" in spite of elevated levels of measurable distortion due to very low levels of feedback. Thirdly, like many people I really like the looks of a tube amplifier shining in the dark (see above) with the glowing cathodes and the electrons that flow in a space that can be observed from the outside through the holes in the anode electrode. With a badly tuned power tube you can even see these electrons as a blue shining gas around the cathode as I remembered well from the past with tube amplifiers driven in positive control grid current by a bass guitar.

This all made me decide to design a tube amplifier with a mix of modern and ancient technology to see if the application of knowledge on control of transistor amplifiers and active mechatronic systems can improve the musical performance of tube amplifiers. Not being a purist I allowed the use of operational amplifiers and transistors for active control of the DC balancing and current biasing of the tubes.

Features

Key features of the design are a fully balanced signal path to reduce distortion without feedback. For the output transformers I applied ring-core types from Amplimo that are designed by Menno vd Veen. I also like balanced signal paths, especially between equipment to avoid groundloops and other problems.

Two active control circuits are applied, one for the bias setting of the power tubes and one for setting the anode DC voltage of the driver tubes in order to guarantee a sufficient voltage headroom.

Finally the filaments of the signal amplifying stages are supplied by a rectified DC voltage to prevent hum.

Feedback

The design is directed to achieve the maximum of perceivable "tube-sound" to clearly distinguish from state-of-the-art transistor amplifiers. Nevertheless the amplifier is not completely without feedback. First of all one should realise that "no feedback" is practically impossible as even a cathode resistor provides some kind of linearising feedback. Furthermore the application of mild levels of mainly local feedback proves useful for reducing spread in functionality due to tolerances and aging.

The following feedback loops are applied in this design:

Local feedback around the first high-gain differential signal gain stages to linearise the signal and create an undistorted high-voltage driving signal for the power tubes with a very low output impedance of ≈200 Ω.
Cathode and screen-grid feedback (ultra-linear configuration) for the power tubes to increase linearity and reduce the output impedance to 5 Ω, giving a damping factor of 1.6 with an 8 Ω loudspeaker.
Open-loop frequency response flat from 20 Hz to 20 kHz so no frequency dependent (loop-)gain over the audio band.
Overall feedback of only 8 dB (factor 2.5) to further reduce the output impedance to 2 Ω (damping factor of 4). With this level, deviations of the loudspeaker impedance from minimum (5.5 Ω) to high levels at resonance (40 Ω) will result in less than 3dB change of the output voltage.

Also another execution has been examined without the local feedback as mentioned under (1:) but with larger overall feedback and an open-loop first pole at ≈ 5 kHz. The perceived sound quality of this version was however less preferred by several listeners, even though the measured distortion figures of this execution were better than the figures of the final execution. In the measurements section these results will be shown for comparison. The perceived difference underlines the often stated finding that distortion levels in the order of 1% are not necessarily bad and the distortion in tube amplifiers is often appreciated.

Applied target loudspeaker

It should be noted that the target loudspeaker which is used with the amplifier is a special execution of the Scanspeak Reference monitor (Design: Rolf Smulders, published in Elektuur Hifi-luidspreker 11) modified by Ruud Janssen of Speakerland to a floor standing version in oakwood with the help of one of his people. It was a one-off special execution that I bought in a lazy moment and never felt sorry about.

This loudspeaker is a bass-reflex unit with the following dynamic properties for the woofer as modelled with the Excel spreadsheets (English Excel) that can be downloaded at Scanspeak. The upper graph gives the response with a series impedance of 2 Ω of the amplifier. These curves are slightly different from the curves in the Elektuur article, due to the modifications.

Scanspeak reference monitor

Results

Fully according to the expectations this tube amplifier has its own sound. As such it is not a simple signal "magnifier" but just another musical instrument in the total chain that transfers signals from the original source to the listener. Overall it's character is best appreciated when listening to music with ony a few instruments and voices like Jazz and classical music with small ensembles. With large orchestra's the sound becomes a bit less defined and fuzzy, most probably because of the intermodulation distortion. Also the voices sometimes become a bit sharp due to harmonic distortion but aside from these small flaws it is really a wonderful sounding amplifier. The low noise and hum figures result in a very quiet environment. It radiates both a pleasant atmosphere without causing listening fatigue and even the famed "warmth" of tubes is present because of the limited damping at low frequencies.