Hi-Fi Tube Audio

Introductory Disclaimer: Before beginning, I feel obliged to inform the readers of this text that my views on high quality tube audio are mostly different and quite contrary to the "common knowledge and practice" that has come to be established on this topic. In my approach, I tend to keep a tidy, scientific engineering approach and I also strive to avoid "religious" attitudes that are also common in Hi-Fi world.

This page is dedicated to the tube power amplifier that I have built as a tribute to all experiments with tubes that I have done as a kid, while I was still learning electronics. The amplifier is presently used as part of a simple Hi-Fi system that I have at home, or at various "Hi-Fi Show-offs" that my friends organize occasionally.

The Amplifier

Tube AudioTube AudioTube Audio

Artistic design is definitely not a major attribute of this amplifier, although it has been built with a convenient component layout that can be easily upgraded later. The addition of a front panel together with some decorative covers for those "less appealing" looking components would be a good start (not to mention those ugly shiny screws that should have been replaced with nice anodized allen bolts, which of course were not available to buy in Belgrade). But I doubt that cosmetic upgrades will happen any time soon because I'm quite satisfied that the amplifier works excellent in the first place, I don't care about the looks so much as long as it can be called decent.

The amplifier has only one stereo input (a pair of RCA connectors) with volume control placed on the front, I didn't have time to build an input selector. The input has line sensitivity and should be connected directly to the signal source, I am a strong opponent of preamplifiers and all unnecessary components that tend to complicate the signal path and introduce inevitable distortions of their own. I do not want my system to play music when I have musicians play it for me, I just want my system to reproduce that music (that's the "Fi" in Hi-Fi).

Short Specification

This amplifier has been designed as an experimental platform, meaning that it can accept a number of different tubes. With its built in octal sockets it can work with (provided that an appropriate output transformer is available, as well as driver gain and bias that can be adjusted): EL34, 6L6GC, KT88, 6550, KT90...etc. The mains transformer has 5 taps that can provide different anode supply voltages: 360V, 390V, 420V, 450V, and 490V. As I presently do not have decent quality KT88 or similar tubes, I use the amplifier in the following configuration (per each channel):

Input stage:EF86, triode connected
Driver stage:EL84, triode connected
Output stage:4x EL34, triode connected, "single end" topology
Bias voltage regulator:ECC88
+390 VDC voltage regulator:PCL82

Output transformer has an 8 Ohm output only, and the amplifier delivers around 10 W per channel. I have never measured this value and it's just a guess, I don't actually care too much as long as the output stage performs to some other requirements of my own.

Electrical Design

Tube AudioThis is where all of my attention was directed to. In my amplifier I tend to keep a short signal path and simple but correctly designed amplifier stages. In order for this to function, high performance support circuitry is needed (anode, bias and filament power supply). This approach altogether benefits from superior transient response, a key quality of any audio reproduction device.

As you can see, power supplies and other support circuits make up for the largest part of the amplifier. In the upper left corner are the mains filter and the switching power supply that provides 6.3 VDC filament power. PCB in the middle-left is a terminal hub for the toroidal mains transformer. In the lower left is the soft startup controller and rectifier for the 390 VDC bus and bias voltage. PCB in the middle-right provides bias adjustment for each output tube.

In the next few paragraphs all of the major blocks of this amplifier are presented with short descriptions.

Main Anode Power Supply

Tube AudioIt has been mentioned previously that the mains transformer has 5 taps that can provide different anode supply voltages: 360V, 390V, 420V, 450V, and 490V. This enables the amplifier to accept a large number of tube types, as well as the room to experiment.

Filter design for the anode power supply is much more important than is commonly considered. Choke inductance for example is one component where "bigger is better" rule does not apply. Too large inductance will yield the filter with high Q factor and high resonant frequency that will affect the sound considerably. The whole filter needs to have a concept of operation and has to be designed carefully.

Class A single ended output stage requires a large anode capacitor, and this amplifier currently uses 2x 3300μF 400 VDC (although it was designed for 2x 2200μF 500 VDC).

Soft Startup Circuit

Tube AudioThis handy device solves the well known problem of a nasty high current surge when switching on:

- Rectifiers with large capacitors
- Powerful toroidal transformers
- Tubes with cold filaments

It has been implemented with a PIC 16F73 microcontroller that synchronizes with line voltage phase and provides driving signals for two thyristors that slowly ramp up the voltage for the mains transformer. This solution has much lower losses and is more reliable than the common solution with a series resistor and relay switch.

The microcontroller also handles amplifier power-on and power-off sequences with a soft button. At power-on, thyristors that drive the main transformer also charge the main rectifier capacitors slowly until they reach full voltage. The sequence continues by turning on the tube voltage regulators, and the last step is the filament voltage power supply for the audio tubes. Turn-off sequence starts with disconnection of the main rectifier supply while keeping the tube filaments on, to enable the output tubes to slowly discharge the bulky anode capacitors to a safe voltage. After this step, the rest of the amplifier is turned off.

Filament Power Supply

Tube Audio Traditional powering of tube filaments with AC voltage straight off the mains transformer is inadequate for use in Hi-Fi equipment for the following reasons:

- AC voltage (both common and differential mode) can interfere with the signal through interelectrode capacitance
- Line frequency AC voltage causes mechanical vibrations of the filament that affect electrodes and introduce noise in the signal

It is therefore necessary to power the filaments with filtered DC voltage. High current line frequency rectifiers for this purpose tend to be clumsy, with bulky capacitors and thermal loses on the diodes (if the tube filaments are not connected in series to minimize the current). To obtain stable and filtered filament power, a "forward" type switch mode power supply has been built that delivers 6.3 VDC with up to 20 A. Due to the switching operation of this power supply, additional filtering has been applied to the output voltage to minimize high speed switching interference and EMI.

+390 VDC and -50 VDC (Bias) Voltage Regulators

Tube AudioAs electronic components, tubes are not so good as voltage regulators. But at least if designed correctly, such circuits can behave very close to a first order system, even though they possess a feedback loop as any other reference based regulator. Of course, this property is not unique to tubes only, but is very beneficial to Hi-Fi audio. Together with acceptable stabilizing performance for this purpose and the fact that I get two more nice looking tubes per channel in my amplifier my decision was clear - let's build a PCL 82 (ECL 82) and ECC 88 tube based regulators.

Rectifier for the first regulator provides +500 VDC and +550 VDC. The second voltage is needed for the screen grid of the PCL 82 pentode, to improve the range of the voltage drop. Each channel has its own regulator that powers the input amplifier and the driver stage. ECC 88 based bias voltage regulator is fed with -110 VDC.

Output Transformers

Tube AudioAs one of the most important components, these have received special attention. After extensive testing and some simulation of wide band audio transformer design, I rounded up a solution and painstakingly hand wound these two for the amplifier. They were optimized to have "controlled capacitive prevalence" when driven by tubes, meaning that not only the first zero-pole pair is as high as possible but also that capacitive effects will already limit the modulus of the transfer function to a safe low level at those frequencies where the pole is located. In another words, the first dominant pole will have a low Q factor and much less influence on the output signal. A transformer with such properties will have a smooth transient response when driven by tubes, and the system will behave closely to a first order system which is highly desirable.

Multiple terminals for different loudspeaker impedance tend to complicate the output transformer, as it is difficult to reach optimum performance on all terminals. For that reason, I have opted for only one 8 Ohm output.

Building an output transformer for the class A single ended tube amplifier bears an additional challenge - high magnetizing anode current requires a decrease in equivalent core permeability to avoid saturation. Introduction of the air gap for that purpose also lowers the primary inductance that results in higher cutoff frequency in the bass region. Since there is no point in testing the transformer outside its intended purpose, simple bandwidth measurements were performed by driving it with 4x EL 34 tubes. With the air gap adjusted for the required amplitude of anode current, these results were obtained:

-1dB Cutoff: 22Hz - 120kHz
-3dB Cutoff: 12Hz - 200kHz

The anode impedance of this transformer was set to around 5000 Ohm, a rather high value for 4x EL34 but with an important reason. The greatest shortcoming of any tube amplifier is its high output impedance (low damping factor), a widely neglected fact in tube Hi-Fi world. Low damping factor signifies the inability of the amplifier to precisely control the motion of the loudspeaker membrane, most notably in bass region. Just average semiconductor based amplifiers have bass that sounds quick, tight and controlled, while the average tube amplifier gives smeared out, "broad" sounding bass. Although this gives an impression of a warm and more gentle (widely known as "tube-like") sound, it is actually highly improper. Any transient musical instrument with high dynamics and rich bass spectrum can show you this, like rock-and-roll bass drum, timpan, concert bass drum, bass guitar or a contrabass. In order to get a good, tight, defined bass out of a tube amplifier, the designer must sacrifice maximum output power to a certain extent. One should keep in mind that a 3dB power difference (twice the power) is quite audible, but not really "that" audible, while on the other hand it can mean a lot to the quality of sound.

How Does It Sound?

Pioneer A-656 that I've been using before is a decent amplifier, but it is certainly far from high performance Hi-Fi. When I plugged in the tube amplifier for the first time, I thought that there was something wrong with the system, the guitar was so open, warm and spatial that I couldn't belive the difference. To cut the story short, although this first trial run was supposed to be a start of further experiments with different tubes and other changes on the amplifier, this first configuration worked so good that it remained basically unchanged up to this day.

I can say that the amplifier surpassed my expectations, first of all it had such a quick, powerful and controlled bass that it easily blew away a couple of transistor based amplifiers that I compared it with. The spatiality and "presence" of the middle range spectrum were also quite audible. Precision and definition of sound details in the highs were just amazing, I have witnessed many audio show-offs but up to this day I haven't found a more precise amplifier than this one when fed with a high quality source (although I have heard some good amplifiers that have come very close, but not close enough). And last but not least, this amplifier sounds excellent all the way from jazz and classical music to the rock-and-roll and heavy metal (that I mostly listen to). This fact I take as the testimony of a correct design, and I strongly object the practice of saying "this amplifier/loudspeaker/whatever-device is good for jazz/classical/whatever-music and not for this/that". In short - if it's designed good than it should play everything properly.

One of my friends has a pair of Cabasse Egea II loudspeakers and a Sony Esprit 779ES CD player, a good platform to test this tube amplifier against the Accuphase E-307 and Perreaux 350 Classic Series amplifiers. Accuphase partly knocked itself out of the race for the flat "reference" amplifier by showing noticeably tone-processed low-mids (with all tone controls turned off). But we can overlook this incident as this amplifier truly has nice and crisp highs for a transistor amplifier and it already blew away a couple of less fortunate tube amplifiers in that aspect.

In front of a couple experienced listeners and me, all three amplifiers exhibited virtually identical bass reproduction which I call an excellent result for the tube amplifier. Quickly after this it was noticed that it also pulls out in front of the two other competitors with slightly nicer details and better definition in higher end of the spectrum. As an overall conclusion of this test, I may say that all three amplifiers work excellent and differences shrink considerably when it comes to real Hi-End audio. But you can still hear the difference and enjoy the fact that your effort has paid off.

Why Tube Audio?

First, it just looks good and alive, not another dull old black box amongst a pile of your components. And if built and maintained correctly - it also works astonishingly good. If you're interested in my global opinion on the question "What is better - tubes or transistors?", my short answer it - they're the same, but I personally like tubes more because they're fun. But it's not just that simple, and if you are interested please take a look at the graph below:

This graph represents my conclusion after all experiments up to this date, based on the generalized listening experience. The loosely defined parameter "quality" should also be taken as general as possible, otherwise this discussion would get too complicated. Before all else, we must take into consideration that the human ear is truly an extraordinarily sensitive instrument when regarding some properties of sound, and this is what gives us every right to analyze and tune the finest details of audio electrical circuit performance. Next, it should be pointed out that if you analyze the electrical topology of both types of amplifiers, their expected performance is quite consistent with the graph. In tube amplifiers, schematics related to signal path are rather simple, and one may expect that this simplicity will result in lower probability of signal distortion. This is partially true, and steeper quality curve of the tube amplifier confirms it. But simplicity also comes with numerous shortcomings, so additional effort is needed to step up the performance after a certain level. This usually relates to the design of support circuitry that is very important when working with tubes, a fact that is widely neglected leading to a large number of amplifiers performing just average. Semiconductor based amplifiers are naturally complex when it comes to signal path, giving rise to much wider class of problems related to signal amplification. But systematic, conceptual and well done design can't just fall short of the expected results, so both types of amplifiers are fully able to attain similar high levels of performance when enough effort is invested in them.

Common belief says that tube amplifiers are globally better than semiconductor ones, and graph data for the same effort seem to confirm it. But my findings do not concur with this, in front of my ears I have seen expensive tube amplifiers bite the dust of the similarly priced good amplifiers based on transistors. The reason for this is that design effort is not proportional to price, neither it scales the same with two types that we're discussing. Tube amplifiers mostly rely on their fame and glory when it comes to fixing a price tag on them, as well as on the exclusive artistic finish on the outside, but in fact they mostly suffer from lack of electrical design effort. On the other side, inferior marketing position of semiconductor amplifiers is usually an incentive for designers leading to the fact that in terms of sound quality, the outcome of the race between tubes and transistors is tight.

But what about the tubes, weren't they supposed to be outdated and vastly inferior to semiconductors in their electrical properties? How come tubes are even able to achieve such bold and competitive levels of sound quality? Well it seems that there is some magic about tubes after all. First, tubes are actually very good at what they do - in terms of electronic components they are very fast and robust. Second and most important, they are able to form simple circuit topologies that have excellent transient responses, the one and only property in audio amplifiers that gives them a large head start over transistors. Nonsense about tube amplifiers having lower THD figures and thus sounding better are totally misleading. Tubes have many deficiencies that have put them out of use, mainly their high price per unit and horrible scalability (inability to form compact integrated circuits). But when it comes to audio - they work good and look good, I wouldn't need more.