An Audio Amplifier Design Philosophy
By David Berning
( Courtesy Meta-Gizmo.com )
I became interested in audio at a very young age because I liked music, but was too uncoordinated to be able to play a musical instrument. My first teen-age attempt at building a piece of audio equipment was a rather difficult project: an open-reel tape player. It was a rather comical device, being made mostly of wood and including an old washing machine motor, but it worked. While the tape head (a salvaged used device that I did not make, and the only part on this machine that was actually made for the purpose) could drive the mic input of my sister’s guitar amplifier, it did not sound right and this launched me into the amplifier-building career.
I began my amplifier life with tubes because I could get them free, along with tube-related parts, from old discarded TV sets. I can clearly remember sitting in high-school English class drawing 6SN7 circuits when I was supposed to be reading Shakespeare.
Over the next couple of years I built several amplifiers. The first was a single-ended 6V6 stereo unit that was similar to the audio-output designs used in the TV sets that I had cannibalized. I enjoyed listening to this for a while, but then sold it to a friend and used the money to buy some push-pull output transformers for the next amplifier: a push-pull 6V6 design. As I learned how to make amplifiers technically better, and could afford to spend money on parts, I began to make transistor amplifiers. I built a very elaborate transistor preamp that included stepped tone controls and a cathode-ray tube display, and thought that I had the best preamp that could be made.
A Time to Listen
Then one day my ego was shaken. I had just fixed an old Fisher preamp that had a faulty filter capacitor, and I hooked it up to my system in place of my transistor preamp to see if I had fixed the hum problem. At the time, I was listening to a recording of the seashore, one of a series of sound environment recordings on the Atlantic label. The roar of the pounding surf dominates this recording, but the sound of foghorns from ships can be occasionally and faintly heard. I had listened to this recording many times while studying for my university classes, and I was aware of one relatively loud and two very soft foghorns during the 30-minute recording. As I listened to the Fisher, I started hearing more foghorns than I had ever heard before. I started the record over and counted 26 foghorns! With the transistor preamp placed back into the system again, only the original three foghorns were clearly resolved. What could possibly be wrong with the transistor preamp, it measures perfectly?
This incident was instrumental in turning my amplifier designs back to tubes, and the first of two incidents that shook my reliance on measured parameters as being all that matters in the design of amplifiers. The second incident began when a friend built an amplifier from a schematic that I had given him. This friend really liked the sound of his completed amplifier, and wanted to show it off at a Macintosh Amplifier Clinic. My friend came back from the clinic very disappointed because the measured harmonic distortion was high. I told him that I could fix that by adding negative feedback, and he brought the amp over for the modification. He happily took his amp home, but when he listened to the amplifier it no longer sounded good. Both of us listened and compared the amp with and without feedback, and I agreed that the sound was better without the feedback, even though the amplifier measured better with it.
While I still believe that measurements are useful and should always be made during the design of an amplifier, these incidents taught me that that there is a good deal of art that needs to go along with the science.
A Time to Invent
My first radical departure from conventional amplifier design came with my Screen-Drive amplifiers that I developed as a result of the energy shortage around 1973. By that time I knew that I liked my tubes, but they just were too wasteful in terms of energy. I found that I could get the same tube linearity at one tenth the idle current normally used if I drove the tube from the screen as opposed to the grid. I further found that this only works well on TV sweep-type tubes and not very well on audio tubes, and it requires a much more powerful driver stage. I patented the screen-drive circuit (US patent 3995226, Nov ’76), which was initially based on a transistor drive for the output tubes. With this circuit I could get high power outputs with relatively little heat. Aside from the heat, the reliability of the output tubes increased dramatically because most audio output tube failures are caused by grid to cathode shorts brought on by the relatively close spacing of these elements, and this new design has these externally shorted. An unexpected additional advantage of the low idle current is that it is much easier to keep the output transformer out of saturation.
Sonically, the sound is much cleaner and more transparent at low listening levels when the transformer is not in saturation. The higher the idle current (more Class A) used, the harder it is to keep the absolute dc flux balance within the one to two milliamps range where saturation occurs in most push-pull output transformers. At loud listening levels, the advantages of the low idle currents go away, as the transformer is going in and out of saturation all the time due to both the non-symmetric nature of music signals and gain differences between the push and pull sections of the amplifier. At loud levels, more Class A is an advantage because the low-idle current Class B operation produces more high-order harmonics due to crossover distortion.
From a design philosophy standpoint, I prefer having the best performance out of an amplifier at lower listening levels because the amplifier spends most of its time at low levels, and I don’t like to listen to music at loud levels on an extended basis. Furthermore, the low heat and low idle current provides for extremely long tube life, whereas the conventional bias currents cause rapid deterioration of the tubes and sound with time. Hence, my design philosophy also includes making amplifiers that are very stable with time, sonically. For these reasons my current ZH270 model amplifier uses the screen drive, even though it does not use output transformers.
A Time to Dream
I spent a long time building amplifiers with output transformers, over 25 years. I hated their performance limits. I hated the development and prototype phase of transformers for new products. I hated their weight. I had already freed myself from the hum-inducing problems of line-frequency power transformers with the inclusion of my first switching power supply used in my TF-10 preamp introduced in 1979. I had freed myself of the weight of the line-frequency power transformer in the EA-2100 amplifier, introduced in 1983. My new switcher in that amplifier improved low-frequency stiffness of the power supply by a factor of better than ten over the best large transformer design. Furthermore, the regulation meant that the bias adjustments would remain stable.
I was certainly aware of the output-transformerless amps, particularly the Futterman design produced by NYAL, and some copies of the original Futterman circuit that I had investigated. I did not want to make anything like that because I knew that such a large number of hot tubes would have reliability concerns. I also did not like the lack of symmetry between the push and pull sections in these designs. I had previously made some custom OTL direct-drive electrostatic speaker amplifiers and was fully aware of the magic that could be had by OTL when it could be properly matched to the speaker in impedance. But conventional tube OTL simply cannot properly impedance match to 8-ohm speaker loads unless it is designed for ten kilowatts (reasonably efficient power transfer) or higher. So I continued with my output transformers until 1995, and then ceased all amplifier production so that I could have time to dream.
My first dream was a resonant switching amplifier with all-transistor switching in the power sections, but tube in the analog voltage to frequency conversion part. The great thing about this concept was that it could emulate a Class A amplifier in that a characteristic of the resonant switching power circuit is that it contains the full load circulating current at all times, but without much power loss. I do not know of anyone who has attempted to build such an amplifier before or since. I did not pursue the more conventional pulse-width modulation because I had concerns about crossover notch distortion that these circuits are subject to. A prototype of this resonant amplifier worked, but had certain audible distortions that I could not eliminate.
First and foremost an output transformer, or any transformer for that matter, is an impedance-matching device. Impedance matching is used to transfer a source of power to a load so that both the power source and the load are operating under optimum conditions for maximum efficiency. What other kinds of impedance matching devices are there? Some such mechanical devices include an automobile transmission, bicycle gears, and a simple lever. Besides the transformer, a dc-to-dc power converter is an electrical type of impedance matching device. The dc-to-dc converter can convert a high voltage at low current into a low voltage at high current, or vice versa. Dc-to-dc converters form the basis of the switching power supply, but usually contain added complexity for regulated power supply applications. In its simplest form, the output voltage or current of the dc-to-dc converter is always related to its input by a proportionality constant which is just like the turns ratio in a transformer. Like the transformer, impedance matching goes as the square of this constant.
All current Berning amplifiers use the dc-to-dc converter to form the basis for impedance matching between the tubes and the speaker. The patent for this technology is US patent 5612646 (March ’97), and is reprinted on the Berning web site at
This patent describes the many advantages of this amplifier over the transformer-coupled amplifier, so this will not be repeated here. Unless the reader is well versed in switching power supply design, it is unlikely that he or she will understand exactly how it works, and it is too complicated to try to explain it here. Perhaps the best explanation to read is the one given by Chuck Hansen in his technical review of the ZH270 that appeared in Glass Audio, vol. 12, issues 1&2, 2000. This is also reprinted on the Berning web site. In his review, Mr. Hansen reaffirms the success that this new technology has with preserving the tube’s transfer characteristics at the speaker.
But What Does It Sound Like?
Various people have different philosophies on how an amplifier should sound. This is part of what makes audio fun. Transistor guys hold up their spec sheets and say that tubes are too colored and the true reference is their zero-measured distortion amplifier. Tube guys say that the transistors are unmusical and unnatural sounding. Let’s face it: real live sound is lost as soon as it is converted to electrical signals at the microphone. The electrical signals are further distorted in the mixing, recording, storage, and playback processes.
I can say that audio transformers do influence the sound of amplifiers by acting as frequency-dependent filters, and iron has rather severe hysteresis-induced nonlinearities that show up mainly at low frequencies where flux swings are large. As bad as this seems to look from a technical standpoint, when these distortions are added to those already present on the recording, the net result may not be bad because subjectively some of these effects can partially cancel. This does not mean that you are recovering any of the original musical information; you are at best making one mess more palatable. It may, however, still be easier to hear certain musical information if high-frequency hash is filtered.
I feel that certain triodes are the most linear voltage amplification devices available. This view is supported by measurements of amplifier circuits using various amplifying devices taken by Eric Barbour in "The Cool Sound of Tubes" IEEE Spectrum, pp. 24-35, August 1998. While transistor amplifiers in whole generally have lower distortion figures than tube amps, this is because there is a much heavier reliance on negative feedback in those units than is used in tube amps. I have already described my own findings in regards to how high feedback can degrade the subjective sound of an amplifier.
So how does the Berning Zero-Hysteresis Output-Transformerless (ZOTL) circuit sound? To me, the dominant characteristic is that it is fast and clean. If you want to hear the true sound of power tubes properly impedance matched to low-impedance speakers, this technology is the best technology available at this time to do it. Note that this is an enabling technology, and the particular implementation will affect the sound. For this reason, I am producing several different amplifier models, including both single-ended and push-pull units. If the 70 watt per channel ZH270 at $4500 was perfect for all applications at 70 watts and below, I would not be making a Siegfried 300B, rated at only 6 watts for $6950, with higher measured distortion.
Finally, I offer some comparisons between the measured performance of audio output transformers and my dc-to-dc impedance converter. The following figures are taken from white papers that are on my web site, and additional figures and discussion are presented there.
This pair of photos shows the looping due to magnetizing current and hysteresis distortion in an output transformer (top) vs. the lack of these artifacts in the Berning impedance converter (bottom) for a push-pull pair of 6L6s connected in the triode mode. A curve tracer is used in the ac mode for the measurement as the transformer can not handle a dc measurement that would be needed it only one tube was used, even though the Berning impedance converter can handle the dc of one tube.
This pair of photos compares the Berning impedance converter applied to an actual special wide-bandwidth amplifier circuit to an output transformer in the same circuit using a 10kHz square wave. Notice that the rise time is significantly slower with the transformer. Tube to speaker impedance matching is 5000 ohms plate-to-plate to 8 ohms in both cases.FONT>
"I wish I could score everything for horns." - Richard Wagner. "Our writing equipment takes part in the forming of our thoughts." - Friedrich Nietzsche