Welcome vacuum tube amplifier constructors - this page is presented for your information and guidance.

It is intended to provide helpful hints to save you inevitable pain and suffering in your quest for audio excellence and to help you find joy in your chosen pathway to audiophilic pleasure!!

Please note it is not intended for the novice constructor. Basic circuit theory and construction techniques are not attempted herein because it is assumed you already know that and are competent in both.

In any event, you will soon see I am not a theorist, so whatever is said is based upon a combination of the reported experiences of others, helpful guidance from my electronic guru friends, reference to published texts from people who are considered to be reliable technical experts, electron tube manufacturers' manuals and data sheets, and my own experiments, research and experiences.

I hope it is of help to you in designing and constructing the very best audio amplifier for your needs.
 

© NOTICE: INTELLECTUAL PROPERTY COPYRIGHT © D.R.GRIMWOOD 2002 - ALL RIGHTS RESERVED.
 

Copyright in all quoted works remains with their original owner, author and publisher, as applicable.

Please note that no warranty is expressed or implied - see footnote notice.

Intellectual property in the applied engineering concepts expressed in this paper remains exclusively with the author Dennis R. Grimwood.

The whole or part thereof of this paper and/or the designs and design concepts expressed therein may be reproduced for personal use - but not for commercial gain or reward without the express written permission of the author and copyright owner.

All rights reserved.

WARNING!!

Do not attempt to design and/or construct a vacuum tube audio amplifier unless you suitably skilled, qualified and/or experienced.

The Author makes no claim whatsoever as to the validity or accuracy or otherwise of any statement, information or opinion contained in these pages and no liability will be accepted for any error or omission of any kind whatsoever.

Proceed only at your own risk!!

No warranty of any kind is expressed or implied as to the workability or performance of designs, concepts or equipment described herein.

Never forget Murphy's Law:   If something can go wrong it will !!!!
 

SAFETY WARNING:
 

• VACUUM TUBES AND CIRCUIT COMPONENTS OPERATE AT HIGH VOLTAGES OF BOTH ALTERNATING CURRENT AND DIRECT CURRENT THAT CAN BE FATAL.

• VACUUM TUBES OPERATE AT HIGH VOLTAGES AND TEMPERATURES THAT CAN CAUSE PERMANENT  INJURY, DYSFUNCTION OR DISABILITY - TO YOU OR SOMEONE ELSE.

• DO NOT ATTEMPT TO CONSTRUCT, TEST, WORK ON, OR COME INTO PHYSICAL CONTACT WITH, LIVE ELECTRONIC EQUIPMENT UNLESS YOU ARE TRAINED OR OTHERWISE QUALIFIED TO DO SO.

• NEVER REMOVE A PROTECTIVE COVER FROM ELECTRONIC EQUIPMENT UNLESS YOU ARE TRAINED OR OTHERWISE QUALIFIED TO DO SO, BECAUSE LETHAL MAINS AND HIGHER VOLTAGES ARE PRESENT INSIDE.

• Extra care should be taken with circuit designs using the B+ voltages listed in this guide.

• Safety is of extreme importance not only to yourself the constructor, but also to other persons who may come into contact with your creation - particularly little children.

• Live wiring should be fully insulated to prevent electrocution of yourself and/or users.

• Tubes for use at B+ voltages over 450v DC require sockets made from materials that can withstand high voltages. Inferior materials may cause flashover or arcing, leading to burning, short-circuits and/or tube damage.

• Many tube types require a top cap plate connection and such tubes must be carefully mounted to minimise wire length, proximity to other components and risk of accidental contact. Insulated caps are recommended. Do not solder directly to tube caps.

• Most vacuum tubes are glass encased devices that  have been evacuated of air - ie they contain a vacuum atmosphere. If dropped or impacted they are likely to implode, shattering glass in any direction. Handle with care and respect. Do not drop or impact. Always pack and store in padded wrapping or the manufacturer's carton.

• When inserting into a tube into its socket always take care to align all the tube pins correctly with the matching socket. Incorrectly aligned tubes may result in high voltages being applied to low voltage circuit components or systems, and may result in short-circuit fault currents being applied with resultant expensive damage to tubes and equipment.

• To avoid serious and permanent tube damage when constructing an amplifier, always ensure the grid bias voltage is present at grids 1, 2 and 3 before applying B+ high voltages to the circuit. In this regard, cathode bias and back bias configurations are safer and more reliable designs than fixed bias.

• Always ensure large HV electrolytic capacitors are fitted with suitable bleeder resistors to automatically discharge them after the amplifier is switched off - otherwise you may expose yourself to lethal current if you accidentally touch a live conductor.

• Persons familiar with assembling and servicing personal computers and suchlike - all of which feature fully insulated construction - must appreciate that tube amplifiers and ancillary equipment operate at high-voltages and have unprotected open construction that is exposed to the touch - particulaly components connected to the AC mains supply.

Now that you have been suitably warned, let us proceed together to explore the world of vacuum tube audio.

INDEX:

INTRODUCTION:

STEP 1:    DETERMINE HOW MUCH POWER (WATTS RMS) IS NEEDED
STEP 2:    SELECT THE LOUDSPEAKER
STEP 3:    DETERMINE WHAT KIND OF "SOUND" YOU WANT.
STEP 4.    DETERMINE IF NEGATIVE FEEDBACK IS REQUIRED OR NOT REQUIRED.
STEP 5:    DETERMINE THE OUTPUT LOAD IMPEDANCE
STEP 6:    SELECT THE TUBE COMPLEMENT
STEP 7:    SELECT CLASS OF OPERATION
STEP 8.    DRIVER STAGE CONFIGURATION
STEP 9.    CHASSIS AND COMPONENT LAYOUT AND WIRING
STEP 10:  B+ OPERATING VOLTAGE
STEP 11:  POWER SUPPLY
STEP 12.  APPEARANCE

INTRODUCTION:

"HIGH-FIDELITY" is a term that literally means "TRUE FAITHFULNESS" - to the original.

For just on one hundred years, and particularly the eighty years since the advent of electric recording, countless researchers and engineers have endeavoured to attain that elusive standard of performance described as "High-Fidelity".

In modern parlance, a term such as "audio excellence" would be used to describe that sought after performance.

Others might simply say "peerless" or "the best".

If we review hi-fi audio amplifier circuit designs we can see much commonality between them, but most might be described as
"variations on a theme".

In fact, many hi-fidelity amplifiers have attained their superior performance simply by taking a proven design, tweaking it (optimising
circuit values) and using the very best components available at the time of manufacture. That approach is still pursued today by both commercial designers and enthusiasts alike.

This paper takes a fresh look at some of the important elements essential to designing and constructing a high-fidelity amplifier.

Some of the concepts can be easily adapted to existing designs, often with very minor component or voltage changes to those chosen for an existing circuit.

IMPORTANT REMINDER:

The professional Audio Engineer designing for a commercial application must consider design elements and factors such as performance, construction and safety specifications and standards; tube types, characteristics and availability; component availability; continuity of supply; component cost; component quality; corporate vendor/supplier policies and preferences; corporate design policies; fashion ideologies that control appearance, shapes and finishes; component and complete device colour, machine tooling constraints; sheetmetal suppliers and materials; labour costs and assembly times; packaging and delivery requirements and costs; market/buyer preferences or trends; warranties and guarantees; after-sales service; and brand-reputation etc.

All of these and more impose constraints upon the professional designer when approaching the design of high-fidelity audio amplification equipment for commercial sale.

However the home constructor has no such constraints!!
 

We are blessed with more or less total freedom from all or any of the above.

We can use new, used, second-hand, salvaged, hand-me-down, or recycled components.

We are not usually constrained by original component price or cost.

We can use non-ideal, oversized or approximated components, or components that would not normally be used in such a device.

We can compromise.

We are not locked into printed-circuits and can use point-to-point wiring with confidence.

We are free to use any design we want.

We are free to have any layout we choose.

We are free to instal extra shielding wherever we want.

We are free to modify the design without having to be concerned about guarantees, warranties or product.
 

WE CAN EXPERIMENT ACCORDING TO OUR OWN IDEAS AND PREFERENCES!!!!

WE DO NOT HAVE TO COMPLY WITH THE CONSTRAINTS OF THE  ESTABLISHED CONSERVATIVE PARADIGMS IN DESIGN OR CONSTRUCTION

WE CAN REACH BEYOND THE COPY-CAT SCHOOL OF DESIGN.

WE CAN INNOVATE!!!!
 

But please, before you abandon all of the hard-won knowledge developed over a century of tube audio design and application,  do follow the essential core design rules set out in STEP 9 below: "CHASSIS AND COMPONENT LAYOUT AND WIRING" - ignore them at your own peril.

A NOTE FOR THE UN-INITIATED

Electricity normally behaves like a fluid. Hence when we observe the behavioural characteristics of water, we can see much relevance with electric current flow.

Electricity is a force. Hence it can have all manner of forms - not just regular AC (Alternating Current) or DC (Direct Current).

It will flow when there is a pressure difference between the two ends of a conductor or a circuit.

It will not flow unless there is a "circuit". A circuit can be created by hard-wiring, or by electro-static or inductive coupling through the air.

A "circuit" can be created by adjoining components, wires or even through the air. Never assume that because a hard-wired circuit is not evident, high-frequency AC current or electromagnetic forces cannot be present to influence circuit behaviour.

It will always try to find the shortest path. This attribute creates problems with devices and wiring in high-voltage and/or high-current circuits. In the context of vacuum tube amplifiers, "high-voltage" means anything above about 450 VDC. Above that, life becomes more and more difficult as all manner of unexpected phenomena occur in the amplifier.

Electricity flow can be controlled electrically, electronically, electro-magnetically, electro-statically or mechanically.

Power Out = Power In minus Losses. This works like a garden hose. What comes out = what goes in less friction in the pipe.
In an amplifier, AC power out = AC power in minus conversions losses from AC to DC then DC back to AC.

THE VACUUM TUBE

The vacuum tube is to an amplifier as the carburettor is to a motor vehicle engine.

If there is no input regulating device, an engine will simply run at full RPM and power output - usually to self-destruction, so the carburettor serves to regulate fuel/air gas mixture and flow into the engine as a means of regulating engine power output.

By means of controlling carburettor operation, the driver determines the speed at which the engine operates - to produce power for transmission through the drive train to the driving wheels.

Similarly, the vacuum tube is an electro-mechanical electronic regulating device that uses a small input effort to produce a comparatively large output power.

Just as we would not claim the butterfly valve in a carburettor produces the engine power output we would not also claim the input signal voltage to a vacuum tube produces the amplifier power output.

Importantly, the vacuum tube, like the carburettor neither produces power nor consumes it (excepting for heating requirements and efficiency losses).

In considering high-fidelity audio amplifier design principles, it is first vital to understand the vacuum tube serves two fundamental purposes or functions:

The first primary function of the vacuum tube is to electronically regulate the Direct Current (DC) flowing through it in such manner that it varies in direct proportion to the applied ALTERNATING CURRENT (AC) signal voltage. This is effected by injecting the AC signal voltage into Grid #1 of the vacuum tube. (Other Grids can be used but Grid#1 is the most common configuration).

It is the variable Direct Current flowing in this circuit that, when applied into a LOAD, produces the Alternating Current OUTPUT.

Thus the Alternating Current OUTPUT applied into a LOAD produces the useable output from the amplifier.

To simplify our understanding, we could regard the Direct Current circuit as the SOURCE of power - or INPUT POWER, and the Alternating Current circuit as the LOAD - or OUTPUT POWER.

Hence the vacuum tube could may be classed as a TRANSDUCER because it converts energy from one form - ie Direct Current (DC) to Alternating Current (AC). On the other hand, the complete Amplifier is not, because it uses an AC input to produce an AC output.

The resultant Alternating Current output means that a single vacuum tube provides a common current path for BOTH Direct Current and Alternating Current circuits.

In the case of the DC circuit, that portion of the current path between Plate and Cathode terminals of the vacuum tube, is used to REGULATE the DIRECT CURRENT flow in the whole circuit under constant supply voltage conditions by means of the Control Grid (Grid #1). A negative DC voltage (Grid Bias) is applied to it so that it will present a negatively charged element that controls the current flow to the desired level by interfering - in a deliberately controlled manner - with the ability of the Plate to attract electrons from the Cathode.

Because the vacuum tube has negligible internal DC resistance or AC impedance (see tube rectifier characteristics for typical values) and it is a current path in the circuit supplied by the Power Supply, it is essential to insert a LOAD into the circuit to limit total current to aceptable limits - otherwise the tube would behave as a low-resistance metallic conductor.
 
 

The second primary function of the vacuum tube is to provide a return current path for the power supply source input power - which is in DIRECT CURRENT form.

Without a return path no current will flow in the power source circuit. That is to say, current can only flow in the circuit when the vacuum tube conducts.

This is demonstrated by the fact that if the power supply voltage is applied to the Cathode and Plate terminals of a cold vacuum tube, current will not flow. However if a fixed resistor - having the same resistance value as that specified for the design value of AC load impedance for the particular power output vacuum tube(s)  - is inserted into the circuit in series with the vacuum tube and the tube terminals bridged with a wire conductor, then DC current will flow to a value determined by the resistor.

In the case of a Plate loaded output transformer configuration, the output AC circuit load also is shared with the DC circuit - ie is common to both circuits. This is achieved by installing the LOAD in series with the Tube Plate and Cathode terminals.

The value of the DC power so consumed by the circuit will be equivalent to the maximum AC power output plus circuit losses.

Fundamental to hi-fi amplifier design is the principle that maximum power output of an amplifier is limited to the "prospective power" of the power supply - ie the maximum power (Volts x Amperes) the power supply can deliver in any instant of time into its load.

Actual maximum signal power output (including any distorted component) cannot exceed the "prospective power" capability less circuit losses. Circuit losses are always very significant, even at at audio frequencies, ranging from about 23% minimum to about 70% of power supply DC input power.

Thus what is of most significance to us is that assuming a constant value of load in the circuit, Actual amplifier power output is regulated by the vacuum tubes in the circuit - not by its load.

Another important concept to understand is that electron flow in vacuum power tubes does not operate in quite the same way as in the electroplating process. In that process using Direct Current, particles of a metal material are transferred from the Cathode to the solution, thence to the Anode - which is the article being electroplated. The Cathode metal is consumed by the process and eventually there is none left.

However in a vacuum tube this process does not occur - ie during normal tube operation significant amounts of Cathode material do not deposit onto the Plate. This tells us that the electron flow in a vacuum tube may be nothing more than a static stream of conductive particles that assemble in the tube to bridge the Cathode to the Anode (Plate).

Hence, the performance and "sound" of an audio amplifier should be far more dependent upon circuit design and componentry than what goes on in the vacuum tube itself.

Practical experience though tells us differently and there is provable difference in sound between different tube types.

Finally, the audio amplifier is not quite the static solid object that it appears to be.

Electricity, as a force, is fluid.

That means it neither does not, nor cannot, anchor itself to a solid object. It can only move from one solid object to another (eg a set of contacts) or move within or through a solid object to another (eg a wire or cable or X-Rays), or through a fluid such as air (electric arc) or water (electric current), or through a vacuum (eg an electron beam such as in CRT) or space (eg microwaves or light).

So long as there is a potential difference between two points in a circuit, current will flow. However for current to flow, there must ALWAYS be a circuit.

Thus if electric current passes through a conductor, there MUST be a return path back to the source so that the circuit is complete and equilibrium is restored.

This is why no current flows when a positive and negative polarity voltage is available from a pair of terminals but there is no load connected. Or why an electric lamp does not light up until the power switch is turned "on".

In audio amplifier applications, it is the LOAD that normally forms the return current path.

However, the return path MUST be electrically separate to the forward path - otherwise a short-circuit will result.
 
 

Kirchhoff's Current Law

The algebraic sum of currents entering and leaving any point in a circuit must equal zero - ie no matter how many paths
into and out of a single point in a circuit, all the current leaving that point must equal the current arriving at that point.

Thus the extremely important principle to note is that in an audio amplifier, the current returning to the source does not have to be the same current that left it.

In other words, current can be drawn from an alternative source, or sources, to complete the circuit such that current back equals current forward.

Having said this, in the case of an audio amplifier, we find a whole collection of individual circuits within it - many of which SHARE current paths.

Hence the designer must pay close attention to what is transpiring separately in each individual circuit - as well as the inter-action and inter-dependence between individual circuits and sets of individual ciruits.

Note that Kirschoff's Law applies separately to BOTH AC and DC circuits - therefore each AC or DC circuit should be analysed independently to all other circuits, as well as interactively with all other circuits in the amplifier.

It is this phenomenum of inter-action and inter-dependence between circuits that causes much angst in audio amplifier design, because not a great deal of attention has traditionally been given it.

Example 1:    the concept of negative loop feedback from loudspeaker to input is a good example of how to ruin an otherwise fine set of circuits, because the loop feedback circuit bridges several individual circuits and by so doing, creates a time-delayed interaction and inter-dependence that is otherwise not there.

Example 2:    the alternating current and direct current circuits that exist about each vacuum tubes are separate but also together. Each is dependent, inter-dependent and inter-active with the other.

Example 3:    the B+ bus usually provides a common source of power to all stages of the amplifier, resulting in a situation where the current in each stage affects the voltage available to each other stage.

Example 4:    losses in the output transformer and power transformer must be offset, otherwise their circuits cannot be in equilibrium. But because what comes out is not the same as what goes in, the designer should pause a moment and think about what effect mixing the output with the input might have on performance, stability and capability.

So the message to the audiophile designer is that to effectively design and construct a high-quality audio amplifier using electron tubes, pay close attention to what is happening at each and every point in the circuit and how each individual circuit might interact with every other circuit in the amplifier.

Each individual circuit about each individual stage in the amplifier should be carefelly designed to optimise performance for that stage.

STEP 1:     DETERMINE HOW MUCH POWER (WATTS RMS) IS NEEDED

User needs for final electric power output determine the core requirements for critical components such as tube type, output transformer, power transformer, chassis size, ventilation and gross weight.

These in turn, determine needs for B+ voltage and current, grid bias voltages, rectifier and electrolytic capacitor voltage ratings etc.

Hence, a very apt starting point is to determine power needs.

It is assumed you will be using loudspeakers - which are transducers used to convert alternating current electrical energy into audio or sound pressure energy. Loudspeakers are notoriously inefficient and fragile devices and great care should be taken in their selection, mounting, installation and operation.

Generally speaking, listening to recorded music in an average room at a comfortable level requires only about one watt RMS - which is surprisingly loud with an 85-90 db SPL efficiency loudspeaker - even much more louder with 100 db SPL efficiency units.

Also, to provide for the minimum full dynamic range of recorded music - ie 20 db - a total power rating of 100 Watts RMS will be a reasonable target to enable transients to be reproduced without too much audible distortion above normal listening requirements.

In this age of digital recording, musicians have found it necessary to resort to effects to produce saleable product, hence it is common nowadays for CD's to incorporate extreme levels of low frequencies of all kinds of waveforms, to produce that "thumping" brain-deadening sound so beloved of today's teenie-boppers.

Unfortunately this means that to attain realistic sound reproduction free from audible distortion, a 100 W RMS amplifier and a 100 W RMS loudspeaker system (50 W RMS per channel - stereo) is mandatory.

Of course a lesser level of power may be usefully utilised but transients will be cut-off or truncated, with a corresponding loss of realism.

A more detailed explanation of this requirement is provided by Dave O'Brien, formerly from McIntosh Laboratories, in his description of his "Spectral Fidelity Test for Intermodulation Distortion" (IMD). IMD is what causes the sound to become "fuzzy" at high volumes.

See also an interesting overview at http://www.axiomaudio.com/archives/power.html

Note that recording technologies have always been far ahead of playback technologies. Examples of just how good recording technologies and standards have been in the past are easily heard by listening to recordings made way back in the 1940's and 1950's, or even earlier in some cases. (Some of my CD's include original recordings that go back to 1910). When transcribed to CD format and cleaned up, these recordings contain signal information not previously heard with conventional playback equipment. The information was always there on the original tape, but just not accessible. However thanks to advances in audio recording and transcription technologies, now it is.

Except as otherwise determined from the following information about loudspeaker performance, when used in a typical domestic home situation, an amplifier/loudspeaker combination of less than 100 W RMS will simply produce overload distortion in all of its forms on transients and/or intermodulation distortion on steady state heavy programme material such as sustained pipe organ music.

However this distortion may not be heard in practice because the amplifier may actually simply chop the top off the transient peaks, in which case everything else will sound just fine up to that power output level where the whole system collapses through severe overload. What will be heard is a modified sound, but the modification may not be discernible to all but the most critical listener.

On this basis, a 10 watts RMS per channel TUBE stereo will still do the job well.

In contrast, a 25 watts per channel RMS SOLID STATE amplifier will be needed to produce the same loudness through the same loudspeakers. Why? Who knows, but that's the way it is.

For those who prefer their music through headphones, then the above does not apply. Brain destroying loudness is attainable through headphones at very low power outputs - ie less than one watt.

Pop music tends to be based upon short transients comprising sound forms to which the ear is unfamiliar, so it is easy for the  average listener to tolerate very high levels of distortion - simply because the listener does not know the sound is distorting. Not so with familiar programme material such as classical music, popular programme material from yesteryear, or live recordings of musical instruments and voice well known to the listener.

However, a rational approach to hi-fi is that if you cannot hear the distortion then don't worry about it!!

STEP 2:    SELECT THE LOUDSPEAKER

2.1    SPL Rating

One useful indicator of power needs is the loudspeaker's SPL (sound pressure level) rating. This rating is determined by measuring sound pressure (acoustic) energy under standard defined conditions - usually at 1 metre from the radiating surface with an electrical input power of 1 watt RMS at 400 Hz (cps).

The SPL rating enables a quick reliable comparison to be made between different choices of loudspeaker. For example a speaker having an SPL of  87 db will require TWICE as much electrical input power (RMS Watts) as a similar size loudspeaker having an SPL of 90 db to reproduce the same sound pressure energy level (ie "loudness") in the listening room.

In other words a 10W amplifier/90 db loudspeaker combination will sound just as loud as a 20W amplifier/87 db loudspeaker combination!!! Now there is some food for thought!!

If you can afford electrically efficient loudspeakers like JBL, Altec, Electrovoice, or Eminence, having a sensitivity of 105 db SPL or more, then amplifier power needs will be hugely less - unless you like very loud music!!  Believe it or not, such a loudspeaker will need only 1/63 rd the power of that needed to drive an 87 db efficient loudspeaker. That is, your JBL, Altec, Electrovoice, or Eminence, will be effectively 63 times more powerful with the same amplifier!!

Another way of saying it is that a 105 db loudspeaker driven by a 1 watt amplifier will be just as loud as an 87 db loudspeaker driven by a 63 watt amplifier. Time to throw away those old inefficient drivers!!

For further details of the relationship between power and decibels please refer to the Decibel Chart provided.

A word of warning though - increased loudspeaker efficiency means higher output levels for background hum, noise and hiss. More attention will be needed to the amplifier's design if an acceptable result is sought. Ultimately there will be a trade-off between the competing needs. There is some practical justification then in using a low-cost amplifier to drive a low cost loudspeaker!!

When comparing SPL ratings ensure they have been measured using the same method. Some manufacturers quote SPL ratings at full rated output, which will obviously look better than the SPL produced from 1 watt at 1 metre.

NOTE: Although some specific brands of loudspeakers have been mentioned above as being (relatively) "high efficiency", not all of the models produced by these manufacturers are in that category.

SPL ratings for specific brands and types of loudspeakers vary widely so it is recommended that the manufacturer's catalogue data sheet be studied before forming conclusions.
 

2.2    Power Levels

It is wise to limit power output to a level that the loudspeakers can withstand comfortably, noting that loudspeaker power ratings are generally set at 400 Hz and assume the loudspeaker is installed in a properly designed enclosure that provides adequate cone loading and damping to limit system resonance and prevent cone overshoot.

As a rule of thumb, it is a wise precaution to halve the loudspeaker manufacturer's ratings, particularly if they are published as "peak" watts.

Music power, IHMF and PMP etc. ratings should be appropriately derated to RMS equivalent.

Remember, our SPL rating is measured at 400 cps. The actual frequency response over the full frequency range - particularly in the lower register - needs to be evaluated.

It has been observed that some manufacturers exaggerate the power ratings of their loudspeakers - computer speakers are a current example - eg 1200 W from a 4" loudspeaker - so look for documented verification of quoted or claimed ratings.

Always compare speaker ratings on RMS rated power handling capacity. RMS is an internationally recognised unit of measurement and provides a standard way of rating and comparing.

On the other hand, the commonly used rating of Peak Power has different meanings in different places. It can be anywhere from 1.414 x RMS to 2 x RMS.

PMPO, or Peak Music Power, means the instantaneous power the unit can handle. It refers to the current the unit can handle for a very short interval of time before it fuses or the speaker mechanically self-destructs. PMPO is totally useless and meaningless rating.

The short-time current a conductor can carry during a short-circuit - such that the conductor heats to a predetermined temperature during the duration of the short-circuit, is defined as I squared x T (current squared x time = a constant). The value of the constant is irrelevant - except to determine time.

The International Standard time for determining short-circuit ratings (which require the conductor to remain intact and not suffer visible damage) is ONE second, so for practical purposes the temperature rise of a conductor is determined by the square of the current - which is also the formula for determining power.

We can thus say that the power handling capacity of a loudspeaker (or any wire based device such as a transformer) is directly proportional to the electrical current in - up to its maximum capability, when the conductor will fuse (in a loudspeaker this is the voice coil winding)

PMPO ratings simply take an RMS value and divide it into a very short time interval.

If we use that one second nominal time as a basis for determining a PMPO rating, then for a 10 watt RMS rated loudspeaker at 400 Hz (which is the standard rating frequency), we can calculate a PMPO rating for a time interval of one half of an RMS cycle (1/800th of a second for a 400 Hz signal) by the formula:

1 second multiplied by 800 = 800 x 10 watts = 8000 Watts

This is how 10 watts increases to 8000 watts!!

Obviously the rating will vary dramatically with frequency.
 

2.3    Loudspeaker Design and Construction

Many modern loudspeakers have huge magnets and rock hard cones made from heavy plastic, metal or carbon fibre materials.
Unfortunately none of this indicates efficiency and it may well be that a loudspeaker that looks rugged and chunky actually has a low SPL performance resulting in a severe efficiency loss matching the lower price tag.

Often it will be observed that in loudspeakers up to 15 inches in diameter, those having small magnets and paper cones actually outperform those with large magnets and plasticised carbon fibre cones - particularly in the lower bass region. This is because the heavy cone requires more power to drive it. More power input means more electrical current through the voice-coil. More current means larger diameter wire. Larger diameter wire is heavier and produces a lower ampere/turns ratio in the magnetic field. Heavier wire means more mass to shift. More mass to shift requires a stronger magnetic field. A stronger magnetic field requires a larger, stronger magnet. Because there are physical and cost limitations on all of these parameters, the usual approach is to simply use larger wire, which translates into less electrical efficiency in the electromotive energy available to drive the cone.

Another relevant design feature with modern high-excursion cone drivers is that when the cone is driven to its extremity, the voice coil must still be within the magnetic field of the pole piece to electro-magnetically drive and/or control cone movement. It is obvious that the more the cone travels in and out, the longer the voice coil must be, which means there will be fewer turns of wire actually in the magnetic field of the driver at any instant of time. This translates into lower efficiency - unless design techniques are incorporated to maintain the electro-mechanical efficiency of the transducer.

So our modern inefficient 100 W RMS loudspeaker may need 100 W to drive it - but not to actually deliver more sound pressure output than the 20 W loudspeaker of yesteryear.

Cone suspension systems vary markedly but as a rule of thumb, modern rubber surround suspensions used in sub-woofers and car (automobile) audio speakers are not as reliable as the traditional "accordian" suspension. Foam rubber suspensions tend to literally fall apart after a relatively short period of time - particularly if exposed to sunlight.

When selecting a loudspeaker, check to ensure the cone can move reasonably freely - but not "floppy" - and the spider suspension (usually yellow treated fabric around the voice coil at the back of the cone) is of large diameter and capable of free travel.

It will be observed by the very critical listener that paper-coned loudspeaker sound qualities vary throughout the seasons. In summer, when cones dry out, they tend to give brighter, cleaner sound, and sound louder. But in winter, or wet weather, the moisture content in the cone increases, making it heavier and therefore less sensitive to transient peaks, producing a 'duller" sound.

Bass response will tend to improve slightly and resonance reduce from a heavier cone. This feature is incorporated into heavy duty bass speakers, which have heavy cones, heavy voice coil wire and large magnets to maintain efficiency.

Note however that a well designed loudspeaker of modest power rating - eg 25-30 W RMS - does not need a large diameter voice coil or large magnet to be an effective transducer.

For high fidelity reproduction, it is far better to have a light cone move little than a heavy cone move a lot - the further the cone travels the less linear the sound reproduction will be.

It follows that for a given level of power output, a large diameter loudspeaker will be more linear than a small diameter loudspeaker.
 

2.4    Loudspeaker Frequency Response

It has been demonstrated by reputable audio engineers and scientists from numerous tests over the past 50 years or so, that most of the audio POWER produced by a symphony orchestra occurs in the range below 300 Hz.

It can be readily deduced from this that very little of the audio range is reproduced by mid-range (particluarly fully sealed, closed-back mid-range units) or tweeter loudspeakers that are fed through a crossover network filter that prevents frequencies below a predetermined numeric value form being reproduced.

Hence in a typical two or three way speaker system, it is the woofer that is doing most of the work and producing most of the audio power that is actually heard by the listener - (more obvious in two-way than three-way systems).

However in this age of cinema "effects" audio, as provided in cinema sound tracks on DVD (ie either 2, 5.1, 6.1 or 7.1 systems), and the use of physically small speaker systems, it is extremely difficult to accurately reproduce fundamental frequencies below 300 Hz - ie in the form presented by the signal source.

For example, most of the split system "sub-woofer" loudspeakers designed for use in home cinema "effects" audio, are designed for use to about 150 Hz or less. This type of loudspeaker is usually equipped with a rock hard and stiff cone that is incapable of reproducing the nuances of high-definition music. Therefore there is great loss of high-fidelity in much of the fundamental frequencies essential to high-fidelity reproduction.

Whilst this type of reproduction is fine for movies, where the sound-track is an adjunct to the main visual presentation, in the case of purely music reproduction the absence of capability to reproduce significant chunks of fundamental musical signal is cause for concern to the hi-fi audio enthusiast.

The solution lies in using FULL-RANGE loudspeakers - ie having a full-frequency response - when mounted in a suitable enclosure - of at least 40 Hz at the bottom end, to at least 15 kHz at the top end.

This range is basic and is based upon the physical reality that 40 Hz is the nominal bottom note on a double bass violin or bass guitar, and 15 kHz is the upper limit of hearing for much of the population and is also the upper limit for FM radio broadcasting. In fact, many people have difficulty in hearing beyond 8 kHz.

For comparison, the typical low-cost modern "all-in-one" style portable home sound system incorporates three-way speaker systems in each stereo channel. Many of these units are fitted with 6 inch (150 mm) LF drivers so, in terms of frequency response capability, there is no fundamental difference between this class of audio device and a complete floor-standing modular hi-fi sytem.

Thus it is more productive to focus on quality rather than quantity when determining the ideal frequency response for the loudspeaker system.
 

2.5    Frequency Response and Power Response

A frequency response of -0 db at resonance in one unit compared with a frequency response of -20 db at resonance in another (compared to sound output at 400 cps), equates to an electrical power difference of 100 times to attain the same loudness at that resonant frequency. That will be impossible to attain at useful power levels because the loudspeaker will not be able to handle that much power.

Similarly, a frequency response difference of 3 db equates to a need for twice the electrical power to attain the same "loudness" - refer Decibel Chart for power ratios.

So a better (ie 'flatter") bass and treble frequency response curve means cost-free power output. It also means more natural sound - ie higher hi-fi.

An example of this phenomenon is shown in the following typical frequency response graph for a high-quality low-frequency reproducer - ie "Woofer" - SYSTEM - ie Woofer loudspeaker mounted in an enclosure.

IMPORTANT NOTE: These graphs do not show the frequency response of the loudspeaker by itself, and include enclosure resonances and enclosure design impact upon loudspeaker performance - which is what the listener hears in any case.

Note the average SPL of 100 db across most of the usable frequency range, indicating unusually efficient performance over a very flat response curve - by any measure this is a "top of the range" LF Woofer driver unit.

In this case, the loudspeaker is mounted in a medium sized vented enclosure, resulting in two peaks at 33 Hz and 66 Hz about the natural resonance of about 43 Hz. The important feature to note with this enclosure configuration is that even when using a very high-quality transducer offering an average SPL of 100 db, the efficiency at system resonance (40 Hz) drops to 87 db - a reduction of 13 db on the average. Although this may appear to be an alarming performance, it is very typical of vented-enclosure loudspeaker systems. This system would not be suitable for bass guitar applications but may be very satisfactory on hi-fi playback. The slight increase in SPL in the mid-range is of great benefit in speech applications.

It is also noteworthy to comment that the enclosure used for the above test is relatively large by volume. Typical small volume vented enclosures will demonstrate substantially higher peaks above and below the natural resonance - see paper on Loudspeaker Enclosures design.

Patchy LF response is definitely noticeable to the musically inclined listener and should be avoided where practicable for hi-fi.

However when mounted in an infinite baffle (ie sealed box) of generous proportions, or in a wall, the response curve will change to something like the following:

Note: This is not the same driver unit as per the above Loudspeaker System 1 but the effect is well demonstrated.

In this case, the low-frequency response falls away rather evenly, with no peaks or troughs in the response curve - AND - the efficiency at 40 Hz (bottom note on a bass guitar or double bass violin) drops away to only 93 db - less than half the reduction in efficiency with a vented enclosure.

Unfortunately the enclosure must be large - ie around 20 cubic feet minimum - to avoid increasing the resonant frequency of the system. For example, an enclosure of  5 to 10 cubic feet volume should increase the system resonance to about 80 to 100 Hz, causing the flat section of the response curve to start to roll-off at about 100 Hz instead of the 60-70 Hz shown.

Provided the amplifier can deliver the required power, such a system can sound quite strong at the bottom end when suitable bass boosting is used - eg tone control

It is also of interest to note the average "flat" section of the frequency response is about 3 db down on the first graph - System 1 - this system thus requiring twice the input driving power for the same loudness.

This system is included to demonstrate the different low-frequency and high-frequency rolloff characteristics to System 2.

Whereas overall SPL is similar, the -3 db point is at 35 Hz in this system, the curve of Loudspeaker System 2 shows a -3 db point of 40 Hz - which is more or less the same in practice.

However at 30 Hz the SPL of System 3 is at 86 db but System 2 is at 90 db - a 4 db difference.

IMHO, for hi-fidelity reproduction, the loudspeaker shown in System 3 is a better investment, simply because the response is flatter over a wider frequency range.

Furthermore, the HF rolloff commencement at 1 kHz is ideal for 3 way systems - ie woofer, mid-range and tweeter drivers with a crossover network.

However for 2 way systems, the characteristics of Loudspeaker System 2 would be better suited because of the higher roll-off frequency, noting Tweeters do not generally commence reproducing below about 2 kHz - mostly higher.
 

2.6    Tweeters

It is customary in two or three way loudspeaker systems to rate the whole system and not the individual components. Hence a tweeter from a 100 W RMS system may actually only have a 20 W RMS rating on its own (within the restricted frequency response of the tweeter), because in typical music the power output is substantially more in the lower register than in the HF range.

It is common to see lower priced drivers rated in "Music Power" in an attempt to match low and high frequency power ratings when processing a "music" signal.

The tweeter and/or midrange unit in fact may have an even lower power rating if it is supplied through a system crossover network, which typically soak up 50% of power input (-3 db insertion loss).

In the case of the three examples above, for an even full-range frequency response, the Tweeter driver would need to have an efficiency matching the Woofer driver unit. In all three cases illustrated, the Tweeter will need to be of very high quality - ie SPL 95-100 db - to match the Woofer performance.
 

2.7    Multiple Loudspeakers

Multiple loudspeakers - preferably of identical design and construction (although satisfactory performance can be obtained with dissimilar but tonally compatible units) - connected in series or parallel as needed - are a very efficient means of adding more power handling capability. It is usually cheaper to add an extra loudspeaker than to double amplifier power. Conversely it is usually cheaper to add an extra low-cost loudspeaker to double power handling capacity than to instal a single high cost higher performing unit. Do the sums!!

There is a cutoff point though - at about 100 W RMS continuous, where the available options for loudspeakers disappears very rapidly. Few manufacturers make loudspeakers capable of handling both the electrical and physical stressors created at this power level. Usually they come in giant sizes - ie 18 and 24 inch drivers, which are beyond the bounds of practicality for most home users. However a true 100 W RMS can easily be handled continuously by four 12 inch drivers, mounted in a single large enclosure, or in multiple enclosures.

Serious consideration must be given to the loudspeaker behaviour at the power level intended to be reproduced. The validity (genuineness) of manufacturer's ratings needs to be carefully researched because it is too late to discover that a manufacturer has exaggerated ratings after the loudspeakers have been purchased and installed.

A little headroom can save an expensive loudspeaker from self-destruction resulting from electrical or physical damage.
 

2.8    Distortion

Loudspeakers are notorious for generating distortion - of many kinds - so the more headroom the better. It follows that multiple speakers will require less cone travel for a given level of acoustic energy (loudness).

Twin-cone loudspeakers incorporating a whizzer cone, tend to change their frequency response with increasing cone travel. At low power outputs the cone and voice coil are very much controlled by the magnetic circuit, but at high cone excursions the cone may overtravel, be less damped, and lose relative drive to the high frequency whizzer cone - resulting in diminished high-frequency response and increased intermodulation distortion. However twin cone loudspeakers offer simplicity of enclosure design, high power output, predictable response and are a preferred low-cost option for public address systems, guitar amplifiers, car audio and home sound systems where full-range reproduction is sought. There is no loss of power into crossover networks.

In twin cone units for hi-fi, it is preferable for the whizzer cone to be of curvilinear, exponential or logarithmic curve shape - ie trumpet like form. Plain simple cones do not provide linearity in frequency response or ideal tonal properties.

Some researchers have suggested that mixing twin-cone loudspeakers with tweeters does not work well, however it has for me so I am converted to accepting mixed cone systems.
 

2.9    Cone Excursion

Remember that the more the cone excursion, the greater the cone over-travel and the greater the back EMF into the amplifier.

Cone travel is a product of its kinetic energy so basically the heavier the cone the harder it is to stop it and reverse its direction - in response to the driving signal power.

If the amplifier is configured with negative loop feedback, the higher back-EMF will cause the feedback system to work harder, which in turn will cause the tonal properties of the sound to change as the amplifier tries to correct the effects of back-EMF from the loudspeaker system.

Another feature of excessive cone travel is substantially increased intermodulation distortion - not so much of a problem for bass guitar or electronic organ, but a definite challenge for hi-fi.

A simple solution to this effect is to use multiple loudspeakers in a totally enclosed or suitably loaded enclosure. The internal air pressure dampens cone travel in both directions, resulting in substantially higher efficiency and reduced back-EMF.

Research has demonstrated that the more cones included in an array of loudspeakers, the greater the low-frequency response from the system.
 

2.10    Crossover Networks

Another design element to consider is that commonly used crossover networks soak up power so it is an essential pre-requisite to determine how much audio power will be lost in the crossover network and add that to total needs.

Commonly seen losses in crossover networks are in the region of 3 db, which is another 50% of power gone down the drain, never to be heard of again!!

Some crossover networks are worse than that.

Crossover networks also suffer roll-off at the crossover frequency - ie roll-off in both upwards and downwards responses, so there tends to be a hole or dip at the crossover frequency.
 

2.11    Enclosures

Horn systems are significantly more acoustically efficient that direct radiator systems - in the order of 30% compared to 5% - ie up to six times more efficient. Accordingly, the amplifier drive requirements for a horn system may be significantly less. However horn systems have different "sound" or tonal qualities to direct radiator systems and that may not be acceptable to all listeners.

All loudspeaker systems require adequately damped cabinets to minimise resonance from the cabinet material and shape. This can occur at any audible frequency - not just in the bass register.

In the 1950's it was the norm for enclosures to be as large as was practicable. Totally enclosed loudspeaker enclosures (infinite baffle) require large volume for low resonance - 10-20 cubic feet volume was considered the minimum essential for resonance free solid bass response down to 30 Hz with a 12" speaker (it still is).

Dedicated audiophiles actually built enclosures having sand-filled walls to attain effective enclosure resonance damping.

Vented enclosures (with or without tunnel or passive radiator) offer substantial reduction in enclosure size but produce two resonance peaks above and below the natural loudspeaker resonant frequency, resulting in a sometimes obnoxious peaky bass response on some kinds of program material. (This is why bass guitar speaker systems tend to use a totally enclosed (infinite baffle) design for best results). But the advent of small apartments for the masses has led to demand for "bookshelf" speakers. Unfortunately bookshelfs do not reproduce sound very well. Get a real loudspeaker and a real enclosure!!

Finally, in my opinion it is wiser to spend more on the loudspeaker than the amplifier, because the loudspeaker will have greater effect on the final sound. A practical way of evaluating this statement is to check out a typical automobile radio retailer's display setup where a range of loudspeakers can be switched to connect to a common amplifier. The results are profound, and tell us a lot about how loudspeakers "colour" sound.

It also tells us that in the current global trading environment - price and performance are not directly connected.
 

2.12    Tone

Although two and three way systems have always been popular because of their wide frequency response and reduced intermodulation distortion, they do tend to suffer from tonal imbalance between units.

Some systems also suffer from varying efficiencies between units, resulting in a peaky response.

Thus in two and three way systems it is important to select the individual drivers for compatibility of tone.

Another factor is the directionality of mid-range and tweeter units.

To overcome this, which is experienced by noticing that the highest frequencies (including important harmonics) tend to be heard only on the centre-line axis directly in front of the system - a feature not always desirable, when the listener is not able to located directly in front of the speakers.

One method of overcoming this is to use two mid-range drivers, each mounted to point slightly outwards off the front centre axis.

Tweeters should also be two in number, again pointing off axis.

Unfortunately this arrangement may result in an impedance mismatch so arrangements need to be made to overcome this.

The simplest way is to use FOUR each of midrange and tweeter units to attain a matching impedance to the woofer and amplifier.

A simpler solution is to use multiple twin-cone drivers - say 4 x 8" or 4 x 12" per enclosure. Connected in series/parallel they will present the impedance of one unit to the amplifier. eg 4 x 8 ohms speakers in series/parallel connection will present an 8 ohm load to the amplifier.

STEP 3:    DETERMINE WHAT KIND OF "SOUND" YOU WANT.

The "sound" or "tone" of a vacuum tube audio amplifier may be varied at will - or better put, may be customised to suit your personal tastes and hearing capabilities.

However "sound" or "tone" are subjective terms and will vary in conceptual meaning between individuals.

The old adage "one man's trash is another man's treasure" is very apt in this situation, so a person's description, or expression, of a particular sound characteristic or quality will mean something quite different to another person hearing the same sound.

Terms like "bright" tone may be "harsh" to someone else, and "smooth" or "mellow" may be "dull" to another, so avoid (if you can) prescribing tonal qualities as far as you can.

However, the basic styles of output tube configuration do have generic tonal properties that may be fairly recognised as being significantly different to each other.

The choices include:
 

Also there is a choice of interstage coupling design:
 

and yes there are also more exotic options available too!!

Furthermore the tonal qualities will be vastly different for either single-ended or push-pull configurations, each of which offers unique tonal and distortion properties - see article.

Important: The "sound" or "tone" of any amplifier can be dramatically changed by replacing just one component in a circuit. So before replacing your expensive entire amplifier just replace one or two selected components

(the trick is to figure out which ones!!)

The first place to start is the interstage coupling capacitors. Generally speaking, the larger the caps the more "deep or bassy " the tone will be and the smaller the caps the more "bright or trebly" the tone will be. However some commentators suggest excessively large coupling capacitors produce "blocking" distortion.

Other components that can be changed for improvement are the values of plate resistor, cathode resistor, grid #1 resistor (all variable + or - ) and electrolytic filter caps (generally the bigger the better).

Also for a cleaner but softer tone try removing cathode bypass capacitor to all driver stages (not power stage).

Changing any of the above components may cause instability so take care.

My policy is to ensure all Grid #1 resistors are the smallest viable value and never more than 100 k ohms. This reduces hum and noise, increases stability and enables the use of feedback free circuity.

STEP 4.    DETERMINE IF NEGATIVE FEEDBACK IS REQUIRED OR NOT REQUIRED.

Negative feedback offers advantages and disadvantages.

Negative feedback is the commercial solution to reducing hum and noise in audio equipment.

My recommendation is avoid using it - except within the output stage in the form of ultra-linear connection.

NEVER across multiple stages.

Note: Uncompensated (ie no negative feedback) pentodes and beam power tubes usually deliver significant "bass boom" at the resonant frequency of the speaker system. Unless a push-pull triode, or parallel push-pull (ie multiple pairs) tetrode or pentode, or an ultra-linear tetrode or pentode output stage is used, this boom may be intolerable.

Other techniques include using a large heavy construction fully sealed enclosure (ie infinite baffle) having heavy cabinet damping (eg carpet underlay or fibreglass insulation glued to the timber)

None of this is practicable with "bookshelf" sized speakers so either use a triode output stage or get a transistor amp.
 

STEP 5:    DETERMINE THE OUTPUT LOAD IMPEDANCE

This step is quite critical because it will have great effect upon performance and sound properties.

McIntosh and others have determined that the lower the ratio of full primary to full secondary impedances in the output transformer the better the performance.

It is also claimed by some commentators that the more turns in an output transformer the lower the high frequency performance and greater the distortion.

It follows that a significant reduction in impedance ratio can be obtained by the following options:
 

5.1    Parallel operation of output tubes

Every extra pair of tubes reduces the load. The actual transformer load impedance required is calculated by dividing the effective primary load divided by the number of pairs. Thus two pairs (4 tubes) in parallel-push-pull connection will halve the required transformer load impedance and therefore halve the impedance ratio.
Four pairs (8 tubes) will reduce the output load impedance to one quarter that for one pair of tubes.
 

5.2    Series operation of loudspeakers

Every extra loudspeaker increases the effective secondary load impedance by the number of loudspeakers - ie two doubles the load, three triples, four quadruples etc.
 

5.3    Combined Parallel Tube Operation and Series Loudspeakers

Combining both techniques offers substantial advantages. Thus using eight tubes in parallel-push-pull connection under given conditions, provides four times the output power with one quarter the primary load impedance.

When coupled with four series connected loudpeakers the transformer impedance ration reduces to one sixteenth that required for conventional operation with two tubes and one loudspeaker.

eg in a situation requiring 8000 ohms plate to plate primary coupled to an 8 ohm secondary, the turns ration will be the square root of 8000 divided by 8 - ie the square root of 1000 = 31.62:1

If we use eight tubes (four pairs) in parallel push-pull and maintain the same effective load per pair of tubes, the transformer primary load impedance reduces to 8000 divided by 4 = 2000 ohms. If we then connect the four 8 ohm loudspeakers in series we get a load of 32 ohms. The transformer impedance ration will then be 2000 divided by 32 = 62.5 and thus the turns ratio will be the square root of 62.5 = 7.9:1

So the impedance ratio has been reduced from 1000 to 62.5 = 16 times, and the turns ratio has been reduced from 31.62 to 7.9 = 4 times.

In many cases this improved set of operating conditions may be achieved at no extra cost.

Benefits of parallel tube operation include:
 

5.4    Disadvantages of parallel tube operation are few but include:
 

5.5    Connecting Cables

Long connecting leads between the amplifier and loudspeaker may have a DC resistance of up to say one to two Ohms (there and back). In an 8 ohm system this will soak up a significant portion of the amplifier's power output. Hence one can either use welding cable having low DC resistance - or use a speaker system having higher impedance.

It is common in modern SOLID STATE amplifiers for power output to be quoted into a 4 Ohm load. Obviously if one or more Ohms of resistance is present in the connecting cables then not much power is actually going to find its way to the loudspeaker.

In such situations it is true to say that the cable characteristics will be audible - in the sense that the effects of such a cable may be heard. However in a tube amplifier having an output impedance of 16 ohms or more, no deterioration of sound quality will be noticed by the average listener.

For the technically minded, 100 watts RMS into 4 Ohms will require a current of 5 Amperes.
(I2R) 5 Amperes x 5 Amperes x 1 Ohm cable resistance = 25 Watts loss into the connecting leads

On the other hand, 100 Watts RMS into 16 Ohms will require only 2.5 Amperes.
(I2R) 2.5 Amperes x 2.5 Amperes x 1 Ohm cable resistance = 6.25 Watts loss into the connecting leads

So a 100 Watt solid state amplifier feeding a 4 Ohm loudspeaker through a connecting cable having one Ohm resistance will deliver only 75 Watts of electrical energy at the loudspeaker - a reduction of about 1.5 db (hardly audible but significant in reduction in transient response capability and sound reinforcement applications).

On the other hand, a 100 Watt tube amplifier feeding a 16 Ohm loudspeaker through a connecting cable having one Ohm resistance will produce 93 watts of electrical power at the loudspeaker to be converted to audible energy.

In actual practice the result will be worse for the solid state amplifier because it will actually be presented with a load of 5 Ohms and not 4, thereby reducing power output even further.

Of course to be absolutely fair, the same operating conditions would equally apply to a 4 Ohm tube amplifier or a 16 Ohm solid state amplifier. However 16 Ohms is not a popular preferred output impedance for solid state amplifiers and as explained above, the lower the LOAD impedance of a tube amplifier the more the transformer characteristics will adversely influence operation and performance.

The higher the load impedance the lower will be the losses in connecting leads and the better the output transformer will perform - a win/win solution!!

The simplest way to obtain a high-quality conductor for speakers is to use copper conductor as used in house wiring systems for power circuits. In Australia, this is rated at 32 amps continuous so is good for about 8 kW RMS at 8 ohms.

Copper conductor is made from 100% IACS high-conductivity copper, which is a better conductor than commercially pure silver.

Note 1: Copper water pipe is a relatively poor conductor (80% IACS) because of additives used to improve solderability, however that characteristic is somewhat offset by its increased cross-sectional area compared to small diameter copper wire.

Note 2: Do not use aluminium wire because, on a size for size basis, it has substantially higher resistance than copper (60-80% IACS). It is also difficult to prevent high-resistance joints when terminating aluminium, because the surface of the conductor has a high-resistance coating resultant from oxidisation. if aluminium conductor is used it should be terminated at both ends with professionally engineered crimp lugs, fitted using aluminium jointing compound.

Note 3: For the record, pure silver has the same electrical and heat conductivity as high-conductivity copper. Brass is about 27% conductivity of copper and steel about 7%.

Note 4: In the case of Alternating Current, most of the current is carried in the outer skin of a circular, square or rectangular conductor, regardless of the material. This is why tubular busbars are used for high current applications in electric power switchyards. Unfortunately in the case of copper, the manufacturing process creates a low-conductivity oxide film, that permeates into the surface about 0.1 mm (0.004") deep. Thus in any given copper conductor, the outer 0.2 mm (0.008") of the diameter is a high resistance oxidised material that does not carry current too well (and will get hot). Consequently it is better to use a larger diameter single strand conductor rather than a multi-stranded small diameter conductor. This comment applies to any electrical conductor, including transformers.

STEP 6: SELECT THE TUBE COMPLEMENT

This is also a very crucial step in the design process.

The choice of tubes is very large, however in practice, there are only a few types that have demonstrated over time that they are worthy to be included in a list of preferred tubes for hi-fi amplification.

In my experience most tubes of any given type number have similar characteristics - irrespective of manufacturer - however it is worth mentioning that some current production tubes have different specifications, characteristics and ratings to the original type number specs. In other words, some current manufacturers have abandoned the convention of building a tube that strictly conforms to internationally recognised, "once-only" specifications for any specific type number, permanently determined at the time of original registration.

The standard protocol has always been that an incremental change or variant requires either a suffix or a new type number.

Proceed with caution.
 

6.1 Output Tubes

The style of output stage will have already been made at Step 3 above, and that will determine the first main category of tube - ie triode, tetrode, pentode or beam power tube.

Top cap or plain.

The next step is to select a tube that will deliver the required power output.

Tube handbooks usually provide typical operating characteristics for a range of conditions so the choice must be made to determine actual operating conditions.

Classes of operation, fixed or cathode bias and type of interstage coupling from the driver stage need to be determined.

This set of decisions will determine:
 

6.2 Tube Mounting

Tubes may be mounted:
 

NOTE: In the case of horizontal mounting, tubes must be mounted with the grid wires aligned in the vertical plane - to prevent grid sag when hot and consequent risk of changing grid characteristics - or even creating an inter-electrode short-circuit. This mounting configuration also enhances heat dissipation into the surrounding air, resulting in cooler operation.

WARNING: SOME TUBE TYPES ARE NOT MANUFACTURED WITH CONSISTENT TUBE ELEMENT ORIENTATION IN RELATION TO THEIR BASE PIN CONFIGURATION - THESE TUBES ARE NOT DESIRABLE FOR HORIZONTAL MOUNTING CONFIGURATIONS BECAUSE REPLACEMENT TUBES MAY HAVE A DIFFERENT ORIENTATION THAN THE ORIGINAL USED TO DETERMINE CHASSIS LAYOUT.

In all cases at least 6 mm or 1/4" inch spacing must be provided between tubes to ensure their bulb temperature does not exceed rated limits.

Adequate ventilation is essential. Do not mount tubes near components likely to ignite or to be affected by heat - eg electrolytic capacitors, transformers, cables, plastic components, wooden cabinets etc. - allow at least 2 inches (50 mm) free air clearance. Electrolytic capacitors can dry out then explode.

In the case of rectifiers and power tubes the most common tube internal design configuration is where the Plate assembly is approximately rectangular in section - often with a join in the centre of the wide faces.

The vertical wire posts that support the Plate assembly are nearly always on the hottest faces of the tube - often accompanied by extra cooling fins that extend out from the Plate.

In all cases it is advisable to instal the tube such that the wide face of the Plate - which is the hottest part of the tube - is facing away from adjoining components or other tubes.

The Cathode and Grid wires can sometimes be seen from the outside and these indicate how the tube is constructed, however in the case of Beam Power Tubes the internal wires are usually hidden inside the Plate and Beam electrode assembly.

If a cooling fan (easy nowadays with the availability of low-cost computer fans) is installed, ensure the fan blows towards the narrow face of the Plate assembly so that heat emitted from both sides of the Plate is cooled. Do not instal the fan too close to a rectifier or power tube because the RADIATED heat will extend a significant distance - allow at least 50 mm (2") clearance between a fan and a power tube.
 

6.3 Driver Tubes

Here we have a wide choice, but essentially it is one of triodes or pentodes.

Both styles come in low, medium or high gain varieties.

Triodes are available in single or twin styles (in the same bottle)

Recommended tubes for low-noise, low microphony, ultimate hi-fi are the 12AY7/6072 triode and EF86/6BK8/Z729/M8195 pentode, however very good results will be obtained from more readily available tubes.

In general the 6AV6/12AX7/7025 family provide a "bright" sound, or tone, the 12AT7, or 12AU7 or 12AV7 a "neutral" or "natural" sound, and the 6CG7/6SN7 a "gutsy" or "bassy" sound. All of course, will provide a "flat" frequency response when an amplifier is on test.

Note however that it is well established empirically, that the material and dimensions of the Plate element affect the sound tonal characteristics and many audiophilies have specific preferences in this regard.

The choice of an "industrial" grade or "premium quality" tube does not always provide superior results to a domestic equivalent.

It is wise to consult the tube specs to determine why it has an industrial number, because often the industrial characteristic has no impact on "sound" at all - eg long life. It may be "ruggedised" or have particular selection, inspection or test characteristics directed at specific non-audio applications.

Some of the RCA Hi-Fi circuits use pentodes, such as the 6AU6, 6CB6, 6U8, and 5879. There is also the Euro EF86.

Irrespective of the class of tube - ie receiving, industrial, RF, military or hi-fi - care should be taken in selecting actual tubes - for any given type number there are very wide variations in tube characteristics such as hum, noise, microphony and reliability. The most reliable method is to just plug a tube in and try it.

Sometimes it will be necessary to use shielding cans around driver tubes to eliminate hum and stray RF pickup - however cans tend to increase the risk of microphony (mechanical feedback).

Do not mount driver stage tubes near magnetic components such as transformers or loudspeakers - to prevent magnetic interaction or hum induction.

STEP 7: CLASS OF OPERATION - OUTPUT TUBES

This aspect of design is most important for any amplifier, but particularly for a hi-fidelity audio amplifier.

"For reasons that will eventually become apparent, it is important to recognise that the only difference between Class A, Class AB and Class B is the quiescent bias setting"
- as defined by High-Power Audio Amplifier Construction Manual" - C. Randy Stone - McGraw-Hill 1999.

Control Grid (Grid #1) Bias determines the current flow in the tubes(s) at quiescent (zero signal), maximum signal and all of those values in between.

A modern theoretical analysis on Classes of Operation, Load Requirements and Plate Dissipation by Earles L. McCaul is available at http://www.pentodepress.com/tubes/vacuum-tube-archeology.html
 

However, read-on:

The following standard classes of operation are available - as defined by various authorities.

Class A:

"A Class A amplifier is an amplifier in which the grid bias and alternating grid voltages are such that plate current in a specific tube flows at all times" - as defined by RCA Tube Handbook RC-14 (1940)

Class A operation is the normal condition of operation for a single valve, and indicates that the plate current is not cut off for any portion of the cycle". "The numeral "1" following A indicates that no grid current flows during any part of the cycle, while "2" indicates that grid current flows at least part of the cycle" - as defined by Radiotron Designers Handbook 3rd edition 1940

"A Class A amplifier is an amplifier which operates in such a manner that the Plate output waveform is essentially the same as that of the exciting Grid (#1) voltage.
This is accomplished by operating with a negative Grid bias such that some Plate Current flows at all times, and by applying such an alternating voltage to the Grid that the dynamic operating characteristics are essentially linear.
The Grid must usually not go positive on excitation peaks and the Plate Current must not fall low enough at its minimum to cause distortion due to curvature of the characteristic.
The amount of second harmonic present in the output wave which was not present in the input wave is generally taken as a measure of distortion, the usual limit being 5%." - as defined by Sylvania Technical Manual (1934).
 

Class AB:

"A Class AB amplifier is an amplifier in which the grid bias and alternating grid voltages are such that plate current in a specific tube flows for appreciably more than half but less than the entire electrical cycle" - as defined by RCA Tube Handbook RC-14 (1940)

"Class AB operation indicates overbiased conditions, and is used only in push-pull to balance out the even harmonics"
"The numeral "1" following A or AB indicates that no grid current flows during any part of the cycle, while "2" indicates that grid current flows for at least part of the cycle" - as defined by Radiotron Designers Handbook 3rd edition 1940.

"A Class AB amplifier is one which is over-biased, operating as a Class A system for small signals, and as a Class B amplifier when the signals are large.
The result is that Plate Current flows during appreciably more than half a cycle, yet less than 360 degrees of the cycle."  - as defined by Sylvania Technical Manual (1934).

"Class A is a method of operating a valve so that the Grid remains always negative to the Cathode. The applied signal voltage is small enough to allow the oprating point to remain on the straight portion of the Ia/Va curve, and no Grid-current flows." - as defined by the Radio Society of Great Britain.(1948)
 

Class B:

"A Class B amplifier is an amplifier in which the grid bias is approximately equal to the cut-off value so that the plate current is approximately zero when no exciting grid voltage is applied, and so that plate current in a specific tube flows for approximately one half of each cycle when an alternating grid voltage is applied" - as defined by RCA Tube Handbook RC-14 (1940)

"Class B operation indicates that the valves, which are necessarily in push-pull, are biased almost to the point of plate current cutoff".  "With Class B the numeral "2" is usually omitted, since operation with grid current is the normal condition." - as defined by Radiotron Designers Handbook 3rd edition 1940

"A Class B amplifier is an amplifier which operates in such a manner that the power output is proportional to the square of the Grid (#1) excitation voltage.
This is accomplished by operating with a negative Grid (#1) bias such that the Plate Current is reduced to a relatively low value with no Grid excitation voltage, and by applying excitation such that pulses of Plate Current are produced on the positive half-cycle of the Grid voltage variations.
The Grid may usually go positive on excitation peaks, the harmonics being removed from the output by suitable means." - as defined by Sylvania Technical Manual (1934).

"Class B is a method of operating a valve, by progressively increasing both the steady bias voltage on the Grid, and the applied signal voltage. In Class B, the steady bias is such that, without any applied signal voltage, the Anode current would be reduced to zero. With the signal voltage applied, Grid-current flows and the valve is working beyond the straight portion of the Ia/Va curve. Because of the steady bias, Anode current flows only during the positiove half-cycle of the applied signal voltage." - as defined by The Radio Society of Great Britain.(1948)
 

" As defined previously, Class B pertains to an OPS wherein the output devices are biased to conduct for 180 degrees of the signal cycle. In times past, Class B operation was referred to as push-pull operation (analogous to sourcing-sinking action), but this is a misnomer.
Output devices in a Class B OPS will source current to a load for a half-cycle, but they do not sink current during the opposite half-cycle: they are cut off.
The term push-pull should be confined to Class A type stages.
At least 99 percent of (solid state) audio amplifiers utilise a Class B OPS."

- as defined by High-Power Audio Amplifier Construction Manual" - C. Randy Stone - McGraw-Hill 1999

"A Class B amplifier is an amplifier in which the grid bias is approximately equal to the cutoff value, so that the plate current is approximately zero when no exciting grid voltage is applied and so that plate current in the tube, or in each tube of a push-pull stage, flows for approximately one-half of each cycle when an alternating grid voltage is applied" - as defined by Prof H J Reich in his "Theory and Applications of Vacuum Tubes" - McGraw Hill 1944

Class C:

"A Class C amplifier is an amplifier in which the grid bias is appreciably greater than the cut-off value so that the plate current in each tube is zero when no alternating grid voltage is applied, and so that the plate current flows in a specific tube for appreciably less than one-half of each cycle when an alternating grid voltage is applied" - as defined by RCA Tube Handbook RC-14 (1940)
 

Note 1: "To denote that grid current does not flow during any part of the cycle, the suffix 1 may be added to the letter or letters of the class identification"- as defined by RCA Tube Handbook RC-14 (1940)

Note 2: "To denote that grid current does flow during any part of the cycle, the suffix 2 may be added to the letter or letters of the class identification"- as defined by RCA Tube Handbook RC-14 (1940)
 
 

It has been traditional in hi-fi circles to regard Class A as the preferred mode of operation, however a moment's reflection upon this subject will demonstrate that this has been a false assumption.

If we regard the electron flow in the output tubes as being similar to that of a fluid in a pipe then the term "valve" is relevant to controlling the electron flow.
 

Class A Amplification

A Class A amplifier works like a butterfly valve, as commonly used in a carburettor. This type of valve usually has an axis central to the pipe and when rotated in a flip-flop action, opens or closes, thereby regulating the flow in the pipe. In terms of electron flow in a vacuum tube, this valve may be described as "normally closed", such that when the grid bias is set for zero signal conditions, the plate current will be at or near its maximum permissible dissipation within the rated plate dissipation value for the specific tube type.

When the AC signal alternates positively, the grid bias is increased more positive by the signal and plate current will increase. However the signal power out will also increase so the plate dissipation will decrease.

When the AC signal alternates negatively, the grid bias is decreased more negatively until the plate current decreases to at, or near, its cut-off value.

Hence the limits of operation for Class A are determined by maximum plate dissipation and by the value of negative grid voltage that will cause the plate current to either cut-off, or positive grid voltage that will cause grid-current to flow - whichever occurs first.

It is likely the grid bias will be set at halfway - similar to a butterfly valve - so the axis of the bias will be in between the maximum and minimum limits for grid bias/plate current.

Cathode bias is generally suitable for Class A amplifiers because the plate current does not vary much between minimum and maximum signal conditions - so is often seen in hi-fidelity Class A amplifiers - eg the "Williamson" triode amplifier.

It will be seen also from tube plate characteristic curves that the values of grid bias within the above range are such that most tube types cannot operate linearlyin Class A - ie the positive alternation of signal will not produce the same plate current change as an equal negative alternation of signal.

Furthermore, any transient peak signal having an amplitude several times the average RMS or music signal value will not be reproduced in the output stage load unless it is in the positive alternation of the AC signal - because any signal greater than plate current cutoff cannot be reproduced - ie the negative signal alternation will be truncated.

This is why a single-ended amplifier is fundamentally a waste of time and money for hi-fi.

This is also why a Class A push-pull amplifier is undesirable - unless it has sufficient headroom power to cater for all power output conditions required for high fidelity reproduction at the required listening level.

See paper by Williamson and Walker on the above topics.
 

Class B Amplification

On the other hand, Class B operation works like a "normally closed" valve - such as a gate valve, cock, tap or fawcett. The more the valve is opened, the more fluid flows in the pipe.

In terms of electron flow in a vacuum tube, this valve may be described as "normally closed", such that when the grid bias is set for zero signal conditions, the plate current will be at or near cut-off  for the specific tube type.

Consequently, Class B provides the potential to reproduce any signal - limited only by the amplitude of the signal driving voltage, and the capability of the power supply to maintain voltage and current into the output tubes - up to saturation of the output transformer.

In Class B, the Control-Grid (Grid #1) bias is set such that only the positive alternation of signal input from the phase-splitter/inverter driver (or centre-tapped transformer driver) is reproduced in each output tube.

The negative alternation of the signal input from the phase-splitter/inverter driver (or centre-tapped transformer driver) is not amplified by the output tubes because each tube is biased at or near Plate current cut-off and cannot therefore be driven in the negative direction.

A push-pull driver stage usually features "balanced" alternation, compared to the "unbalanced" alternation of front-end Class A drivers, hence push-pull Class B operation is the only viable option for true high-fidelity reproduction.

Unfortunately Class B does have some extra challenging requirements, such as providing driving power to the output tube grids, a low impedance driver stage,  low DC resistance wide-range driver transformer, high-quality high DC current capability output transformer, fixed Control Grid bias, accurately matched pairs of output tubes, a stable low-impedance regulated power supply, and grid bias set exactly at the plate current cutoff value for both sides of the output stage - at the same time (a challenge for parallel-push pull output stages) - but what the heck if it is for yourself!!

Another challenge is that the grid circuit impedance drops to near zero when the output tube Grid is conducting. Grid-current is caused by the Control-grid voltage being approximately the same as or more positive than the Cathode voltage, causing the Control-grid to behave as a Cathode and become part of the Cathode to Plate circuit electron flow circuit.

A conducting Control Grid presents a low to zero load on the driver stage - a not very comforting thought since this affects driving voltage, frequency response, distortion and absolute total power output. US transformer manufacturer UTC recommends only triodes for this application, explaining that the reflected load from the near zero impedance output tube grids through the driver transformer turns ratio back to the driver stage plates (conventional driver) or cathodes (cathode follower driver) will cause pentodes to deliver dramatically reduced power output together with seriously high distortion.

The voltage/power delivered to the output tube grids will always be limited to the actual driver stage output multiplied by the driver transformer turns ratio.

Better still, is the use of parallel push-pull triode drivers - the lower the Plate or Cathode-follower impedance in the driver stage the better. The non-tube purist may choose to use transistors for this application. Many commercial designs successfully use the transistor driver approach.

On the other hand, Class B is more forgiving than Class A, thus so long as the power supply can maintain reasonably constant plate voltage the output power and transient performance (thus "realism") is more likely to be substantially superior to Class A.

One great benefit from Class B operation is that expensive output tube tube and component life is maximised, because most of the time at home listening levels they will not be working very hard - ie low Cathode current and low heat dissipation.

One major challenge though is to get the Grid bias right. Because the output signal from the phase-splitter is balanced, the central axis of the signal (assuming it is a symmetrical sine-wave) will correspond to the AC neutral point at the centre of the grid-leak resistors, or driver transformer secondary winding centre-tap, to the output tubes.

In theory, the central axis of a symmetrical signal voltage must align electrically with the mid-point of the balanced push-pull input circuit, or else the signal will not be equally amplified by both power tubes.

For true high-fidelity reproduction, this requires the Control-Grid bias to be set such that Plate-Current cutoff occurs exactly at the same time as the input signal is at its mid-point axis - ie equivalent to AC 0 volts - for both sides of the output stage - not an easy task to determine and set up!!

If the bias is set too negative, the first portion of the positive alternation of signal voltage will not cause Plate Current to increase because the tube will still be biased at Plate Current cut-off, so some of the signal will be lost. Because this part of the signal - closest to the mid-point axis - will include a significant portion of the power (area under the curve), the amplified sound will be incomplete - ie some of the signal will be missing.

At first glance this may not seem a significant problem, but in parallel push-pull output stages it is very difficult to have all tubes exactly matched such that they all experience Plate Current cut-off at exactly the same time, so some compromise is necessary. This also applies where a single pair of output tubes is not matched.

As has been previously explained in these pages, it is very difficult when replacing a failed tube to match a new tube with those remaining - because single tubes are not available as "matched pairs" - so an output stage requiring exactly matched tubes will be a very expensive proposition in the long run. Better to incorporate a few adjustments or design for less than optimum tube specs.

Important Note:
Class B power tube output configurations require a low value of Control Grid (Grid #1) resistor - measured between Grid #1 and the Cathode. Too high a value will cause the Grid Bias to go positive - ie less negative - as grid current flows during the input signal cycle. Refer Tube Manuals for details of "Maximum Grid #1 Circuit Resistance."

This value includes any resistances in the Grid Bias circuit, including potentiometers and wire-wound resistors, and is often exceeded in guitar amplifiers in order to gain extra drive voltage into the output stage, but generally results in excessive Cathode Current at full output - particularly in sustained overload mode - and can create instability, with self-oscillation apparent. One trick sometimes used to minimise this effect is to instal a fixed resistor in the Cathode circuit of each power tube - eg 30 to 60 Ohms - which causes positive Cathode Bias to be developed at high output currents, which in turn creates a lesser voltage differential between Grid #1 and Cathode. A bypass capacitor may also be required to each resistor.  This modification will affect the "sound" so for hi-fi proceed with caution.

The driver transformer, if used, should have the best frequency response and lowest distortion characteristics practicable, the lowest practicable primary to secondary turns ratio, together with the lowest practicable DC resistance in the windings.

Another option is the choke/capacitor driver stage coupling system but this has difficulties when trying to obtain high-fidelity performance.

However all is not lost!!
 

Class AB2 Amplification

Class AB2 is a close second to Class B, but avoids the necessity for complex drive circuitry.

Class AB2 also avoids the necessity to set the grid bias exactly at the plate current cutoff value in order to wholly reproduce the amplitude of all signals. In other words, by setting the bias slightly above the cutoff value, full amplitude reproduction of all signals is assured.

Fixed Control Grid bias is essential. If the grid bias is set such that a small current flows at zero signal - eg 10-20 mA per tube - then one can be sure the tube will conduct all signals.

It is generally recommended that a Cathode-follower driver stage be used for Class AB2. This has the advantage of being able to be supplied by a high B+ voltage, such that after voltage drop across the cathode-follower load resistor - eg 100kOhms + normal cathode bias resistor (with recommended grid-leak resistor - ie not direct coupled) to the cathode-follower tube, the maximum permissible plate voltage can be applied across the driver tube - eg 300 VDC Plate to Cathode for a typical triode. This method ensures maximum output voltage from the driver.

Another advantage of a Cathode-follower driver is that it better matches the grid-leak/load presented by the LOW RESISTANCE grid-leak resistors to the output tube grids - eg 100k Ohms maximum per tube - proportionately less for parallel-push-pull.

For adequate low frequency response, very large "fast" coupling capacitors are need - in the order of 10 to 20 uF in Metallised Polypropylene or Polyester film. Oil-filled paper types are not recommended for this application.

A better solution is the tried and true push-pull driver transformer.
 

A final word from C. Randy Stone, from his valuable work "High-Power Audio Amplifier Construction Manual" - McGraw-Hill 1999:

A final word:

Tubes operating in Class A carry full current all of the time. Consequently their Cathodes wear out (by losing the electron emitting surface coating) faster. Loss of emission generally results in progressive loss of high frequency response. In the case of the output power tubes in a good wide-range hi-fi system, this will be about once every one or two years.

So unless you can hear the difference it is more economical to run the tubes in Class AB to preserve their life. In Class AB they should last for many years.

STEP 8. DRIVER STAGE CONFIGURATION

This aspect of amplifier design offers the most creative among us the latitude to do whatever one fancies.

8.1 General Principles

My money goes on installing the phase-splitter as early as possible in the circuit to enable the use of push-pull drivers thereafter - ie to convert an "unbalanced" driver into a "balanced" driver. This approach facilitates more creativity in choosing the methods for amplification of the AC signal driving voltage to the output tubes.

It also enables the use of a voltage amplifier directly driving the output tubes - the result of which is the availability of full output voltage from the selected driver tube (compared with usually reduced output from a phase-splitter) and a significant reduction in the inpout signal voltage required to drive the phase-splitter to the level of voltage needed to drive the amplifier to full output.

Another benefit of a push-pull driver stage is that it can be supplied directly from the B+ supply to the output tubes (because the output tube plates are out of phase with their respective drivers) and thus receive the highest voltage available to its plates.

A useful design rule of thumb is to select a driver stage capable of supplying an AC peak (crest to crest) voltage twice the negative control grid voltage applied to the output tubes. This will cater for pretty much all kinds of signal waveform shapes and reduce distortion from the driver stage by operating it a less than full output. eg if the grid bias is -40 VDC then be sure to instal a driver stage capable of supplying 80 V pp. This applies to each half of the push-pull leg.

My personal preference is a design like the Williamson and GEC 400W, however the first voltage amplifier and phase-splitter, being direct coupled, can present problems relating to inadequate bias to both halves, so some fine tuning may be required for optimum results. The same appliesmto the cathode-follower driver stage. One solution is to delete it and use plate follower, as is the case in the RTVH 100W circuit.

It will be easily seen that using a phase-splitter of a type that can deliver only half or less than half the usual output voltage for that particular tube type is not likely to produce a satisfactory end result.

The exception to the above is pure Class A, where the peak to peak drive signal to the output tubes needs to be only equal to the bias voltage (since the quiescent operating mid-point is about halfway between the grid bias voltage and cathode).

One simple, but effective method, is to use a driver transformer having a grounded centre-tapped secondary at the input (there can be problems of adequately loading driver transformers in Class AB or B output stages), then push-pull voltage amplifying stages throughout the entire amplifier, to provide a "balanced" amplifier throughout. Installing a volume or gain control is a little tricky but not impossible.

The choice of phase-splitter designs is also wide, and it is a case of determining which configuration is the most suitable for the attributes sought.

It is important that the driver stage be capable of delivering adequate signal voltage to the output tubes. This is usually not difficult, however when parallel-push-pull output is used, the Grid # 1 circuit resistance must be proportionately reduced, resulting in a low value of load to the preceding driver stage.

One way to assist this is to use "bootstrapping" techniques to increase driver stage plate voltage, however I just use a high voltage power supply (usually about 400-450 VDC) to deliver maximum rated plate to cathode voltage to the driver stage. Of course the rated plate voltage of the driver tube must not be exceeded.

Note that excessively high voltages to pre-amp stages - (ie more than 250-300 VDC B+) - is not advisable because it produces no real benefit, shortens tube life and can result in excessive background hum and noise levels.

A similar advantage is gained when using a cathode-follower driver stage

8.2 Decoupling

This is another area of audio amplifier design that is often poorly understood by the home constructor, because its design principles are buried in the tube theory textbooks.

Design formula courtesy Radiotron Designers Handbook 3rd edition 1940 Chapter 4

In essence it works like this:

When two circuits operating at the same frequency have an impedance common to both, there is coupling between them, and the phase relationships may be such that the coupling is either regenerative or degenerative. In the former case, instability may result.

If the gain from one stage to the next is more than unity, then positive feedback can occur through back EMF in the B+ supply.

However it is also the case that in a cascade coupled amplifier the output from each stage is usually of opposite AC polarity - ie "out of phase" - to its previous driving stage. Therefore, the textbooks tell us we can have up to two stages feeding from the same point in the B+ supply because the input of the first stage is "out of phase" to the input of the second.

But if we add a third stage, the input of the first stage is synchronised with the output from the third, so positive feedback is likely through the B+ supply.

The usual approach is to instal a cascode coupled decouping resistor between each stage - or two of the three - the resistor having a sufficiently high value to "decouple" the two circuits from each other.

Another method is to use separate decoupling RC network to each stage from a common B+ source, in a similar manner to "star" earthing. But in this case more parts are required. The decoupling resistors need to also have adequate value to achieve the desired result, so the same voltage drop problem remains as with cascade coupling - only .

F. Langford-Smith recommends the decoupling resistor - or choke having adequate equivalent impedance at the amplified frequencies - should have a value of about one fifth the value of plate resistor(s) supplied.

Chokes offer low DC voltage drop and improved regulation in the B+ supply. Resistors cause voltage drop when current flows through the B+ supply (not really an issue with Class A drivers but becomes a problem if the push-pull drivers are driven into Class AB by high signal voltages).

In choke decoupled circuits - like the AWA A515 and some Williamson variants - if the amplifier is intended to reproduce very low frequencies the inductance of the choke needs to be quite high. Note too that when chokes are in series their inductance is added together, so there is some benefit in series connected choke decoupling configurations. To prevent unwanted emf pulses or spikes floating about the B+ supply it is essential for a high level of stability to use chokes having identical physical and electrical characteristics.

One disadvantage with chokes is that if a driven stage B+ voltage falls below its driving stage B+ voltage then some B+ supply current will flow back to the driven stage, thereby reducing the applied plate voltage to the first driver. This phenomena could result in spurious modulation of the signal, heard as distortion.

It is very important to use B+ supply bypass capacitors having a value sufficiently high to ensure there is negligible imedance at the lowest frequency to be amplified AND to provide adequate current to the plates during transient peak signals.

One way to ensure good stability in the supply is to ensure the value of the bypass capacitor - which forms the AC signal return path to the Cathode - is higher in each stage than the value of the bypass capacitor in its driving stage. In this case, although we are dealing with AC signal,  the value of the tube plate resistance and cathode resistor and plate resistor and following grid resistor do not matter because the issue is the impedance between B+ supply and ground.

Another way of explaining this is that the supply end to each stage - usually the junction of the plate resistor and B+ supply - should theoretically be at AC earth potential - ie the negative terminal of the B+ supply. Obviously if there is an impedance inserted between the supply end (nominal AC earth) of the true AC earth (chassis or ground) then there will be a voltage developed across that impedance and the system will not operate as intended. Since in a free system or conductor voltage will appear between positive and negative points, it follows that current will flow beteen those points - in this case through the supply to each driver stage.

The answer is to use adequate values of decoupling resistor or choke AND high values of bypass capacitor.

Note however that high values of capacitor require more current to fill them with charge and take longer to recharge. Also high values of decouping resistor may result in adeqwuate plate voltage to the voltage amplifying driver stages.

The solution is to think it through and proceed with care.
 

STEP 9. CHASSIS AND COMPONENT LAYOUT AND WIRING

Review of past design practice shows us that many designers had no idea of the principles essential to optimising chassis layout.

In a wide range amplifier, the wiring and componentry act as antennae, to pick up stray and induced signals, such as ultra-sonics, RF, hum and noise, from adjacent circuitry.

The basic rules for component layout and wiring are:
 

STEP 10: B+ OPERATING VOLTAGE

One of the most challenging aspects of an amplifier design is to select the most suitable operating voltage for the power tubes.

Most tube handbooks publish a range of optional operating conditions for power tubes and the designer is free to choose from a wide variety of tube types and voltages.

Is it better to run a tube well below its rated maximum limits for long life and cool operation, or to run a tube at or near its maximum ratings to increase power output and efficiency??

Do tubes sound better when pushed?

These are questions only the user/designer can answer, because like "sound" itself, assessment of results is very subjective.

However, because a tube is a voltage amplifier and AC POWER output from the amplifier is calculated as the square of the output voltage divided by the load impedance, it follows that for power tubes, a small increase in voltage will increase power output dramatically.

The limiting factor to applied DC voltage will be the maximum tube ratings.

However a practical solution is to base the B+ voltage on available working voltage ratings of the electrolytic capacitors used in the power supply filter circuit - because these are the most difficult voltage dependent components to acquire.

Rated working voltages for electrolytic caps tend to be available in 50V increments, however there are some popular ratings that are more readily available - particularly in the larger capacitance sizes - such as 350 VDCW, 400 VDCW, 450 VDCW and 500 VDCW.

Not much is available at affordable cost with ratings over 500 VDC.

Hence, for B+ voltages over 450 VDC, it is usually the case that electrolytic capacitors need to be connected in series to withstand the applied voltage. Unfortunately this has the effect of halving the effective capacitance (assuming both caps are identical). Alternatively, consider uning oil-filled paper caps, which are readily available