electron's
 
 
 
 
 
 

ULTRA-LINEAR
DISTRIBUTED LOAD
HIGH-FIDELITY
AUDIO AMPLIFIERS
 
 
 
 
 

*                *                *



























In the realm of high fidelity audio amplifiers, ULTRA-LINEAR OPERATION, also known as DISTRIBUTED LOAD OPERATION, is a term when applied to single-ended or push-pull vacuum tube audio amplifiers, describes the particular power output stage configuration whereby the Screen Grids (Grid 2) of Tetrodes, Pentodes or Beam Power Tubes are connected to and supplied from a tapping in the primary winding of the output transformer ("SE" single-ended) or each half of the primary winding of the output transformer ("PP" push-pull).

The tap is typically (but not always) at 43% turns or 18.5% impedance when measured from the B+ supply for SE or centre-tap for PP as applicable, instead of from a DC supply either independent to, or common to, the anodes.

An extra tertiary winding may be added to the output transformer to enable the screen grids to be supplied at a selected DC supply voltage different to the plate supply whilst still maintaining the desired design AC signal load impedance ratio.
.
.
The ULTRA-LINEAR AMPLIFIER configuration schematic:

Circuit design modes for ULTRA-LINEAR or DISTRIBUTED LOAD OPERATION are shown in Figures 1 to 4 below:

.
Figure 2 shows a conventional single ended design with the output transformer modified for ultra-linear mode.

Figure 3 shows a conventional push-pull design with the output transformer modified for ultra-linear mode.

Figure 4 shows a tapped choke wired in parallel with a conventoinal output transformer, configured for ultra-linear mode.

Figure 1 shows an extra tertiary winding on the modified output transformer to enable the Screen Grids to be supplied at a selected DC supply voltage different to the Plate supply. This mode is also equally applicable to push-pull amplifiers.

Note: The illustrations show cathode bias designs, however all options are equally applicable to fixed-biased amplifier designs.

Note: Some writers have suggested a resistor/capacitor in series may be substituted for the tapped inductor system above, however there will be obvious linearity issues with the  frequency response and other parameters.

Note: The schematics show Tetrode tubes. In the case of pentodes the Suppressor Grid is usually strapped to the Cathode.
.
.
Strictly speaking, the term "ultra-linear" refers only to a narrow band of screen tap voltage ratios as shown in Fig 6 of the Hafler and Keroes US Patent (more fully explained below). This band is in the range 18-20% plate load impedance.

The sonic properties ("tone") of the ultra-linear output stage configuration - albeit single-ended or push-pull design - are midway between triode and tetrode/pentode/beam power tube . The tone varies with the transformer tapping point - ranging from "bright" (pentode) to "mellow" (triode).

It could be said Ultra-linear mode sounds like a triode - but delivers the power of a tetrode.

Blind listening tests conducted over more than fifty years describe the sound of triodes as "smooth, sweet, mellow, natural" and tetrode/beam power tubes as "clean, bright, sharp, punchy".

Technically speaking, and depending upon tube type, the ultra-linear configuration can deliver between 50% and 100% of power output as for pentode operation of the same tube under the same operating conditions and typically about twice the power output of triode operation of the same tube under the same operating conditions with the same applied DC voltages - but with substantially less harmonic distortion or intermodulation distortion (see comparative performance graph below).

Power output is subject to the inter-active and combined effects of load impedance, tapping point and grid bias. These can be optimised for best performance.

The ultra-linear configuration also offers improved overload characteristics, resulting in more effective power output - ie what the listener actually hears at full power levels.

Output impedance is similar to triodes, allowing minimal or zero negative loop feedback to be used.

The ultra-linear amplifier concept is beautifully described by David Hafler and Herbert Keroes in their 1952 US Patent Application 2710312.

The following extract is included for convenience:

.
 The following illustrations show the conventional Ultra-Linear configuration in three typical arrangements.
.
The first shows the Ultra-Linear output stage with Cathode Bias having a common bypassed Cathode Resistor.
Superior results will be achieved by the use of a separate bypassed Cathode Bias Resistor to each tube.

.
This next illustration shows Fixed Bias Ultra-Linear operation, with each power tube having an individually adjustable bias supply for improved balance in the output stage - essential for large power tubes:

.
Because the Screen-Grids are supplied with DC from taps on the output transformer, the ultra-linear configuration therefore avoids the need for a separate Screen-Grid supply.
.
The third illustration shows the Acrosound Ultra-Linear design, also by Hafler and Keroes, which features a separate and lesser voltage supply to the Screen-Grids for improved performance and cooler operation.

The ultra-linear output configuration is also suitable to single-ended output stages, however only push-pull amplification is discussed in this paper.
.
.
***********************************************************************************************************************
.
.
How It All Began: 1933-1947

F. Langford Smith of "Radiotron Designers Handbook" fame, reported  in the May 1955 edition of Australia's "AWV Radiotronics" Magazine, that the original conceptual design concept had been attributed anecdotally to Australian inventors, R. Lackey and R.R. Chilton of the Australian Radio College in 1933, however the documented evidence has been long lost to the public domain. (If any reader has that information please email it to me)

Consequently, in the absence of proof of the Australian design, the ultra-linear design concept may be attributed by documented evidence to "a British subject" Alan D.Blumlein - see British Patent 496,883  (Application No. 15620/37 - lodged June 5, 1937 and granted December 5, 1938) and US Patent 2218902 (Filed 18 May 1938 and Granted Oct. 22, 1940).

The 1937 Blumlein Patent Application claims:

"in order to obtain a good power output the impedance of the load must be very much less than the characteristic impedance of the valve."

"by choosing a suitable tapping point for the screening grid the characteristic impedance of the anode and screen effectively combined can be given any desired value between the impedance of a pentode or tetrode and the impedance of a triode. A suitable tapping point, in practice, may be found to be such that the voltage swing on the screen is between about a quarter and a half the voltage swing on the anode."

"It is the chief object of the present invention to provide an improved circuit in which the characteristic impedance of valves of the kind referred to can be more nearly matched to their optimum loads and in which the danger of an excessive increase of voltage on removal of the load may be substantially avoided."

Note: In addition to the conventional tapped transformer configuration, this Patent also covered the following configuration:

"A thermionic valve amplifying circuit according to claim 1 or 2, wherein the screening electrode is connected through a resistance to a source of potential and through a by-pass condenser to a tapping point on an impedance, across which the voltage variations occurring at the anode of the valve are set up."

Blumlein was employed in the recording industry at that time by EMI (EMI owned the UK and US Patents - Blumlein was the "inventor"). Patent records suggest that during the mid 1930's he was heavily involved in developing recording techniques, including stereophonic sound. It may be assumed his primary object for practical application from this patent was not "high-fidelity" as such, but an attempt to reduce "bass boom" common in domestic receiver pentode output stages (beam power tubes did not yet exist in domestic apparatus) driving loudspeakers in open-backed cabinets, by means of reducing output stage source impedance - now described as "damping factor".

The awful sounding field coil loudspeakers with rigid cones and metal spiders were still in vogue in domestic radio receivers, so the ultra-linear power stage offered improvement to that aspect of the system by delivering better sound quality with more capability to reproduce dynamic peak signals.

To set the stage for the times:

The famous 4 watts audio power output Wireless World "Quality Amplifier" of 1936, designed by W T Cocking,  was based on directly heated PX4 triodes in push-pull mode. The upgraded 12 watt version circuit of 1946 is shown below: C10 and R15 were not included in the original 4 watt  version of 1936 or the 8 watt version of 1937.

.
.
The later more or less identical "Wartime Version" of 1943 also offered an optional pentode connected design (shown below) using the new 6V6 metal can beam power tubes ex USA, however the rated power output of "undistorted" 6 watts from those was less than the upgraded 8 watts PX4 version. On the other hand PX4 valves were in short supply, were expensive and being of a designated class, a government licence was required to purchase them - hardly a favourable environment for the advancement of home hi-fi. If we examine the components, this was a very expensive amplifier to build for its performance class - but competitive at the time.

.
However as the following picture from the December 1937 Wireless World shows, the state of the art at the time was obviously very much "olde-worlde".

The underchassis wiring of the original 1934 model is shown below:

Also in 1936, the British "Leak" company produced a commercial 13 watts P.A.Amplifier using push-pull PX25 triodes. Frequency response was claimed to be 40Hz to 12kHz +/- 1dB - which remains still an excellent performance for PA..

The KT66 beam power tube was introduced commercially in 1937, the same year as Blumlein's patent application and a year after the Wireless World Quality Amplifier of 1936. The KT66 was a British upgraded version of the US metal can type 6L6 beam power tube introduced in 1936. The KT66 was relatively physically and electrically huge - obviously intended for military/industrial applications of the era. It was also expensive

It may be assumed then that when W. T. Cocking prepared his  "Wartime Version of the Quality Amplifier" for the December 1943 edition of Wireless World, he would have considered a modification to replace the PX4 and PX25 triodes with KT66's and/or other pentodes/beam power tubes that were then available.

Other amplifier designs has been published in Wireless World and elsewhere before 1943 however a change to the KT66 would have required a completely new design - ie a brand new Wireless World project - thereby diverting resources from other projects. In any event, being in the middle of WWII,  the KT66 would not have been readily available to home constructors - if at all.

It must be said though that the audio market had already determined triodes were the only way to go for pleasant speech and music reproduction.

This interpretation of history is verified by the January 1946 edition of Wireless World, which offered an upgraded version of the above amplifier using PX25 directly heated triodes for a power output of 12 watts.

The well known post-war Williamson amplifier, an evolutionary more powerful amplifier - and most likely the first real British "hi-fi" amplifier, was still 11 years away. (Williamson was only 14 years of age at the time of Blumlein's patent application).

Williamson's groundbreaking design used triode connected indirectly heated KT66 beam power tubes, intended to replace inefficient and more expensive directly heated triode based amplifiers (the best sound quality at the time). The design was incremental and therefore not patentable.

In his history of the Williamson KT66 amplifier, P. R. Stinson says that Williamson designed the amplifier as a personal project in about 1944/1945 but it was not first published by Wireless World until April and May 1949.

Thus the Blumlein invention, enabling customary triode tone from lower cost pentodes, was a great leap forward. His invention would have offered manufacturers of receivers, particularly EMI, opportunity to offer higher power outputs without loss of triode tonal characteristics (to which the domestic market was conditioned from birth) or increase in costs.

There would have also been opportunities in public address, commercial and cinema applications for improved fidelity.

Given his involvement in recording, it is also likely he may have applied this invention to in-house EMI record cutting head amplifiers - then 78 rpm only.

Blumlein's Patent specified an optimum feedback tapping ratio being between 25 and 50% of plate output voltage to the Screen Grid - ie transformer TURNS ratio - however he did not use either of the terms "Ultra-Linear" or "Distributed Load" in his Patent.

The lack of detailed claims for applications within the patent shows he did not go beyond the conceptual stage with his invention. However regardless of the importance of this invention, Blumlein was a very busy man, whose time and talents were spread across a very wide range of electronic technologies, producing no less than 128 patents in his short career, which explains why this invention received low priority in the overall scheme of things.

In any event, only nine months after this Patent was granted and therefore not long after the invention could be made public, along came WWII, which altered the commercial equation significantly. He is reported to have devoted his talents thereafter to the war effort.

This was his 74th patent since 1928 and during the ensuing 5 years he produced new and novel inventions for a further 54 patents - quite a workload for most mortals.

Unfortunately Alan Blumlein was killed in His Majesty's Service on June 07, 1942, aged 38.

The detailed technologies underlying this important invention were thereafter subsequently lost as a result of Blumlein's death, the war effort and EMI commercial policy.

In contrast to the above documented evidence, the question is - "How many ways is it possible to configure a 4 electrode device in an electronic circuit?"

Of course the answer is that someone, somewhere, at sometime will figure it out.

In this case Blumlein got there first.

For those interested in studying the life and times of this fascinating man see http://www.doramusic.com/index_blumlein.htm
.
.
**************************************************************************************************************************
.
.
The Next-Step: McIntosh

Following the end of WWII and restoration of more normal life, demand for high-fidelity sound equipment increased dramatically. The most technologically advanced manufacturer of hi-fi amplifiers in the world at that time - and most likely ever since - was McIntosh, who on 22 December 1948 lodged Patent Application 2477074 for a "Wide band Amplifier". The patent was essentially for various configurations of cathode coupling in the amplifier output stage featuring a very high quality output transformer.

This patent was followed up by the famous incomparable 50 watt McIntosh "Unity Coupled" design utilising beam power tubes.

McIntosh Patent Application 2646467, filed 13 July 1949 - a full three years before the Hafler and Keroes "ultra-linear" application (the concept being still protected by the original Blumlein patent) - covered the McIntosh unity coupled design using very specialised transformers made in-house. Details of this great technological advancement were revealed to the world by McIntosh and Gow in their famous paper published in the December 1949 issue of "Audio Engineering" magazine.

This amplifier had no peers so there was pressure on the hi-fi industry to produce a competitive product.

A line had been drawn in the sand !!

Note: For those who advocate the Williamson amplifier was the world's leading amplifier of the time, it is relevant that the Williamson design, first published in "Wireless World" of April and May 1947 (only eighteen months before McIntosh's patent application filed for examination) - and upgraded in "Wireless World" of August 1949, claimed a power output of about 15 watts whereas the McIntosh design, sold as the model 50W1 and using essentially the same power tubes, produced around 50 watts with considerably less distortion.

However the Williamson was commercially available from numerous manufacturers for a period of about two years before the McIntosh 50W1 became available to the general public. Importantly, the Williamson was still a fine sounding amplifier, was somewhat more affordable to the ordinary consumer, and continued to maintain wide popularity.
.
.
************************************************************************************************************************

Note: The McIntosh design was not described or promoted as an "ultra-linear" amplifier because the conceptual design of its "unity-coupled" output stage is unique.

**************************************************************************************************************************

The much later detailed investigative work of Blumlein's invention by D. Hafler and H. I. Keroes of Acrosound USA (1951) demonstrated practical applications and defined optimum feedback voltage ratios to the Screen Grid for a range of designated tube types in high-fidelity amplifiers.

Acrosound produced a range of amplifiers for PA applications however Acrosound was primarily a manufacturer of transformers.

To our benefit, Hafler and Keroes published a paper on the topic of "distributed load amplifiers" in the November 1951 edition of "Audio Engineering" magazine, detailing much previously unknown technical information on cause and effect relationships within the design concept.

Specifically, they showed the 6L6, 807and KT66 families prefer a Screen-grid load of 18.5% Plate load impedance. (43% turns or 43% Plate Voltage).

Interestingly, they also claim the 6V6 family prefers a Screen-grid load of 5% Plate load impedance. (22.5% turns or 22.5% Plate Voltage).

This concept is fully described in US Patent 2710312, filed May 20 1952 and granted June 7, 1955.
.
.
**************************************************************************************************************************
 

Hafler and Keroes

History shows Hafler and Keroes successfully patented in the USA further development of the general concept that had already been patented by Blumlein in the UK, but had run most of its full term. This Patent was due to expire on 22 October 1955. It is not known if the Patent ran its full term or expired early through non-payment of registration fees - ie if it was allowed to lapse.

The Hafler and Keroes Patent Application lodged May 20, 1952, referenced previous US/British Patent of Blumlien, namely 2,218,902 - Filed 18 May 1938 and Granted Oct. 22, 1940. It was entitled simply "AUDIO TRANSFORMER".

However although Hafler and Keroes referenced in their patent application some of the conceptual technology invented by Blumlein, they confined their patent to the design of the output transformer which, being the key component in the system and being "novel" in its own right, enabled their patent to be successful even though it did not patent the actual ultra-linear technology per-se (thereby not breaching Blumlein's patents).

Since no-one else could manufacture a similar quality transformer they effectively gained control of commercial ultra-linear technology for the duration of their patent.

Their formal claims included a level of impedance loading to the Screen Grids of between 5% and 26% Plate Impedance - a much lower range to that of Blumlien.

One of the Patent Claims states: "6. The combination in accordance with claim 1 wherein the tube types employed are selected from commercial types 6L6, KT-66, 5881, and 807, and wherein said screen grid loading falls in the range 18 per cent to 26 per cent (plate load impedance)".

This ratio was later verified by other respected designers, including GEC/MOV, GE USA and RCA

See also the thoughts of Norman Crohurst.

Of particular interest is the Australian work of F Langford-Smith (of Radiotron Designers Handbook fame) and A. R. Chesterman in 1955, which investigated the Ultra-linear concept in detail for a range of tube types. Their work is a MUST READ and contains many clues as to how to choose the best set of operating conditions for any particular tube type and performance.

See AWV Radiotronics Magazine May and July 1955

This work was summarised in the January 1956 issue of "Wireless World"- Tetrodes with Screen Feedback.

It is worth noting that the original Blumlein US Patent would have remained in force until 1955, so not much would have - or could have - happened regarding practical implementation of this technology in commercial products by other than Blumlien/EMI until after that date.

However, long before Blumlein's patent expired, he was killed in the year of 1942 in an aircraft crash whilst testing a new British radar design. Thus we can only speculate as to what might have been had he survived. It would be reasonably safe to assume though that hi-fi audio would have been allocated a very low priority during the years of WWII during which Britain was struggling to survive.

Further reading on the life and times of this remarkable man is available at http://en.wikipedia.org/wiki/Alan_Blumlein and http://www.doramusic.com/index_blumlein.htm

Since the famous "Williamson" amplifier design appeared some years after that date, it is reasonable to assume the "ultra-linear" concept was either safely locked away, or rejected by the gurus of the time in favour of triode mode.

In fact, Williamson not only rejected the term "ultra-linear" but openly condemned the concept altogether - see Williamson and Walker - Wireless World 1952.

Thus the term "Ultra-Linear" may be rightfully attributed to Hafler and Keroes - an opinion supported by the "Radiotron Designers Handbook 4th Edition. This is the term used in their 1955 US Patent 2710312 (Application date 20 May, 1952) entitled "ULTRA-LINEAR AMPLIFIERS" and is supported in that Patent by technical justification for use of the term.

Importantly, the term "ultra-linear" refers only to a narrow band of screen tap voltage ratios as shown in their graphical explanation of the concept. (Fig 6 in the Patent). This band is in the range 18-20% plate load impedance.

.
.
This graph from the Hafler and Keroes Patent is the only reference I have seen that describes the specific plate/screen-grid ratios that comprise true "ultra-linear".

Plot A shows the tube internal impedance for the various screen taps of 0% (tetrode connection) to 100% (triode connection). The result is expressed as "R" in the table. It can be seen that tetrode connection has a very high tube impedance and triode connection a very low tube impedance.

Plot B shows the "undistorted" power output in Watts for the amplifier used in their tests - expressed as "W" in the table.

Plot C shows the level of high-level intermodulation distortion for the amplifier used in their tests - expressed as a "percentage".

Plot D shows the level of low-level intermodulation distortion for the amplifier used in their tests - expressed as a "percentage"

Negative loop feedback was not used in the test amplifier.
.
.
Notwithstanding its title - "Ultra-linear Amplifiers", the Hafler and Keroes US Patent 2710312 describes a design for high-fidelity amplifiers based upon the incorporation of a second, or tertiary, primary output transformer winding feeding the screen grids. That concept represented a great leap forward in amplifier and transformer design. It enables the use of transmitting tubes that invariably have comparatively low maximum DC operating voltages for the screen grids.

Such tubes have more turns of wire for the screen grids than general purpose audio tubes, thereby providing greater and more accurate control over electron flow withinthe tube. This results in more accurate reproduction of high amplitude transient peak signals but with lower distortion. They also offer more accurate alignment of the negative feedback voltage from the plates to counter distortion in the tubes.

In 1951 there were numerous transmitting tubes available to suit this mode of operation.

A great example of this design is the Hafler and Keroes  Acrosound 100 watt ultra-linear amplifier with 6146 tubes and TO-350 transformer.

The 6146 is an outstanding tube for audio however its top cap plate connection can cause pain to careless users and its use in audio amplifiers, like the fabulous 807, has been lost to fashion and user safety concerns. (But if as a DIY'er you want the very best, then the 6146, or its little sister the 2E26, is the way to go. Both have industrial construction, black plates and reliability.)

The famous Williamson triode-connected beam power tube amplifier was later modified by Hafler and Keroes to incorporate the ultra-linear design concept. Details are provided in their paper "Ultra-linear operation of the Williamson Amplifier" by Hafler and Keroes, published in the June 1952 edition of "Audio Engineering" magazine. It seems then that they had the last word, by improving Williamson's design despite his dismissal of the concept. Numerous examples of this particular design are now available on the internet.

Since EMI apparently did not seek to protect their patents it would appear they had little or no interest in it - or maybe they just forgot.

The British did not give up their attempts to maintain technical superiority based upon the Williamson triode design and in 1959 Wireless World magazine published an anonymous scathing and derogatory article entitled "Tetrodes with Screen Feedback - further light on the so-called Ultra-Linear circuit". The essence of the argument is that "ultra-linear" is not ultra-linear - ie what it is claimed to be. This was despite reference in the article confirming the validating work done by the highly respected and more than technically qualified F. Langford Smith and A.R. Chesterman of`AWV Australia. Curiously, the Wireless World article also denounced the claims of Blumlein re feedback.

Meanwhile Acrosound and Dynaco were making bucketloads of money from their very successful products. The international trade war was lost.
.
.
DISTRIBUTED LOAD OPERATION

The Ultra-linear configuration in audio amplifiers is also known as "DISTRIBUTED LOAD" operation.

It appears this term was intoduced by Willamson and Walker of the UK, who objected to the term "ultra-linear".

Of course this term is also technically correct, because the taps on the output transformer primary do in fact present a push-pull reactive impedance load to the Screen Grids, enabling them to contribute to useable power output. This is explained in detail in the Hafler and Keroes Patent and later in the 1955 AWV "Radiotronics" papers via laboratory testing.

However importantly, Mullard - staying with Blumlein's original conceptual model - describe the "distributed load" configuration as a system of negative feedback. Of course in reality it is both.

The term "distributed load" was adopted by Mullard as corporate policy. Their use of this term was published in the May and June 1955 editions of "Wireless World" in their article "Design for a 20 Watt High Quality Audio Amplifier" and later on in their 1959 publication "Mullard Circuits for Audio Amplifiers".

It is interesting to note that this design appeared from Mullard, notwithstanding Williamson's previous condemnation of the concept.

I am unable to verify the specific designer of the 5/20 amplifier but later credits go to Mullard Applications Research Laboratories Engineers Mssrs R. S. Babbs, D. H. W. Busby, P. F. Dalloso, C. Hardcastle, J. C. Latham and W. A. Ferguson. (No mention of Williamson and Walker). It is assumed the commercial pragmatists won the argument.

Mullard continued to use this term through until at least the 1960's when their compendium of audio amplifier and pre-amplifier circuits "Mullard Circuits for Audio Amplifiers" was published as a second edition reprint in 1963.

The following extract from that publication provides an overview of Mullard's design concept.



For an excellent technical introduction to the ultra-linear design concept see http://www.aikenamps.com/UL.pdf

More technical detail, single-ended ultra-linear circuits and non-power tube ultra-linear applications is also available from Glass Audio - "Tube CAD Journal"  Vol. 2 No. 1 - January 2000

For further views and technical details see also http://www.vintageradio.me.uk/amplifier/10watt.htm

Important Note: Performance described as "ultra-linear" is available only under very specific operating conditions - and in the case of popular commonly used audio tubes, usually where the Screen load is 18.5% impedance or 43% turns of the Plate load impedance - measured from the power supply source (B+) terminal.

More detailed information is provided below.

In other cases, the term "distributed load operation" may be more appropriate.

Note however that if my OPTIMISED ELECTRON STREAM © or OPTIMISED ULTRA-LINEAR © TECHNOLOGY is used with silicon diode feed to the Screen Grids, the system becomes one of linear feedback only.
 

Tube Operating Conditions:

The attached graphs, courtesy of GE USA, show the marked difference in operating conditions for pentode, triode and ultra-linear operation of the 6550 Beam Power Tube.

6550 Beam Power Tube Operating Conditions - Pentode Connection


.
.
.
6550 Beam Power Tube Operating Conditions - Triode Connection

.
.
.
6550 Beam Power Tube Operating Conditions - Ultra-Linear/Distributed Load Connection
.

.
Benefits:

The ultra-linear configuration ensures constant stage specific voltage feedback between Plate and Screen grid in the output stage, thus reducing output impedance and distortion, whilst improving linearity and frequency response under variably reactive loudspeaker load conditions.

An important advantage is that being single stage-specific, ultra-linear operation avoids the problems of time-delay and phase-shift commonly associated with cascaded stage amplifiers using negative feedback from the loudspeaker for the purposes of reducing distortion and increasing loudspeaker damping.

Ultra-linear output stages provide automatic constant ratio stage-circuit specific feedback, free from the adverse effects of conventional multi-stage feedback systems.

Ultra-linear output stage power output is dependent upon the proportion of Plate load that is applied to the Screen-Grids - typically in the range 75% to 100% that of tetrode/pentode connection at the same plate voltage - but is still twice triode connection for the same tube type under the same operating conditions.

Ultra-linear tone or "sound" approximates that of triodes.

Ultra-linear operation is very forgiving of circuit design and generally provides an acceptable quality sound from average quality output transformers. Note however that the esteemed Norman Crowhurst, in a November 1959 "Audio" Magazine article entitled "Puzzled About Amplifiers", indicates that a high-quality output transformer is essential for ultra-linear operation, to ensure high-fidelity performance over the entire audio frequency range.

To this end, Herbert Keroes of Acrosound developed a special transformer for ultra-linear configuration output stages. The main purpose of this transformer is to enable the use of transmitting tubes, where the Screen-grids must operate at a significantly reduced DC voltage to that on the Plates. Full specifications are provided in US Patent 2791646 (May 7 1957).

A set of high-fidelity amplifier designs was published in the Acrosound Ultra-Linear Transformer Catalogue, published a little later.

The flagship of the range - the TO350 - offers 100W from a pair of 6146 tubes. This is an exceptional amplifier by any standard.
 

Design Concept:

A good example of a typical conventional ultra-linear circuit is shown in the GEC KT88 100W Amplifier.

A more highly developed design is shown in the GEC KT88 400W Amplifier.

The conventional ultra-linear configuration is arranged such that the screen grids have maximum DC voltage applied to them (in practice slightly higher than their corresponding plate/anode voltage because of voltage drop in the primary winding) - thus maximising power output and efficiency, whilst simultaneously receiving an AC negative feedback signal from the output transformer - thus minimising output impedance and distortion.

The magnitude of the AC feedback signal is directly proportional to the percentage turns ration of screen tap to full winding.

The screen tap may be positioned anywhere from 0% (pentode connection) to 100% (triode connection), however performance and tonal qualities change relative to the ratio of the screen tap - see GEC graph below.

In practice, research by the inventors of this method of amplifier configuration suggests that the ideal for most tube types is in the range 40 to 50% turns of each half primary, measured from the primary centre tap.

Further research by Mullard UK for the EL34/6CA7 and EL84/6BQ5, GEC/MOV Valve Co. for the KT88, GE USA for the 6550, and RCA for the 6973 (6CZ5 hi-fi), 7027A (6L6GC), 7591, 7868 (6L6 family), also recommends 43% turns (or 43% of plate signal voltage), or 18.5% impedance.

An example of EL34/6CA7 operation with cathode bias is shown in the following graph - courtesy Amperex Electronic Corporation
.
.

.
.
.
Screen-Grid Operation:

One very important feature of the ultra-linear configuration - different to normal Tetrode/Pentode operation where both Screen-Grids are at nominal AC earth potential - is that when one Screen-Grid in a push-pull pair is driven positive, the opposite Screen-Grid is driven negative by the turns ratio of the output transformer acting about the centre-tap of the transformer, which is at nominal AC earth. This is not a problem because the opposite Plate is also being driven negatively anyway by the action of the push-pull driver/phase splitter.

The opposite applies when the alternating signal reverses polarity.

In Class B operation, the output transformer operates as an auto-transformer, so the opposite Plate and Screen-Grid are still driven to opposite polarity (together in a constant ratio to each other), even though they are not conducting.
 

        *                *                *                *                *                *                *                *                *                *                *

Balanced Amplifier

It should be noted that in the case of a push-pull amplifier, the DC negative bias voltage (even if it is at 0 VDC) applied to the Control Grid #1 is located at the centreline axis of the balanced input signal. This means the inputs are floating and the centre-axis is earthed. That is to say, the input Grids are being driven in a push-pull manner about a common centre-point or axis - eg as seen with a centre-tapped push-pull driver transformer.

Thus in a conventional push-pull Tetrode or Pentode amplifier, any alternations of the balanced input signal to the Control Grids will proportionately increase or decrease current flow in both tubes of the push-pull pair in response to the input alternating waveform shape.

Now in each tube of a push-pull pair, the negative terminal (Cathode) of each tube is AC earthed - even in Cathode bias.

The load on each tube is connected between the Plate and the transformer centre-tap, so the negative terminal of the load - which is at the output transformer centre-tap - is also effectively AC earthed.

But most importantly to ultra-linear operation, in a push-pull amplifier output stage, Screen-Grid behaviour will be similar to that of Grid #1 - ie the Screen-grids will behave as a balanced amplifier - balanced about the centre-tap of the output transformer which, as previously noted, is effectively AC earthed - ie variations in Screen-Grid voltage will produce proportionate variations in electron flow in both output tubes simultaneously.

In conventional Tetrode and Pentode operation, when the DC applied voltage to the Screen-grids is constant, the AC voltage appearing at the balanced Screen-Grid terminals will the product of the electron flow within each tube and will always be a voltage determined by the natural AC voltage gradient applied internally across each tube.

Also in a conventional Tetrode/Pentode output stage, the AC signal voltage appearing at the Screen-Grids is diverted to AC earth at the Screen-Grid terminals via the filter capacitor and is lost as heat. Hence the output voltage appearing at the Screen-Grid terminals is of no consequence to the sound produced by the amplifier - ie it is not reproduced in the output transformer or loudspeaker.

However, in the case of the Screen-Grids in ultra-linear push-pull amplifier configuration this is not so, because the Screen-Grids are located at about 43% of the lineal distance between Cathode and Plate. Thus the AC signal voltage appearing at each Screen-Grid as a result of linear voltage gradient between the Cathode and Plate within each tube will thereby be about 43% of the Cathode to Plate signal voltage.

The Screen-Grids are connected to the load so they will contribute to the sound produced by the amplifier.
 

Class A and Class B Ultra-Linear Operation

In some transmitters, it has been the practice to drive or control the output power tube by means of the Screen-Grid, rather than the Control Grid.

This method offers some benefits to RF situations but is relevant to the explanation of ultra-linear audio amplifiers.

Because the Screen-Grid is located much further into the physical Cathode to Plate distance - ie typically nearly centrally between them - it follows that a substantially higher AC signal voltage must be applied to the Screen-Grid if that element is to control the electron flow in the tube.

Operation will be the same as for Grid #1 but at a higher AC voltage.

It follows that if the DC Screen-Grid voltage controls electron flow within the tube and it is varied by means of a superimposed AC voltage, then the Plate Current will vary in response - as for conventional Grid #1 operation.

But what if the Screen-Grid voltage is applied in 180 degrees directly opposite phase to the electron stream within the tube - as is the case for ultra-linear operation?

Obviously the electron stream permitted by the AC signal as applied to Grid #1 is offset by the opposing signal voltage applied to Grid #2 via the tapping on the output transformer - because they are both in the same circuit at the same time.

Consequently behaviour of the tube in response to the controlling signal voltages applied to BOTH Grid #1 and Grid #2 simultaneously, will be different to that of either a Triode or Tetrode.

This is what the above GE graphs demonstrate - ie a change in the operating characteristics of the basic tube, verifying the claim of Hafler and Keroes that they had produced a "virtual" tube, intermediate between a triode and a tetrode.

The situation is however very different between Class A and Class B operation.

In the case of Class A ultra-linear operation, the Screen-Grid of one tube will be driven AC positive but the Screen-Grid of the opposite tube will be driven to an equal, but opposite polarity, voltage by the output transformer balanced output about the AC earthed centre-tap. This will have the effect on the second tube of reducing the effective Screen-Grid control voltage, thereby reducing voltage gain and therefore power output - regardless of the shape of the input waveform to that tube.

But also note that Grid #1 of both tubes is also controlling current in them.

Hence in Class A ultra-linear operation, it is necessary to consider the effects upon Cathode current in BOTH output tubes by  BOTH Control Grids and BOTH Screen-Grids.

However in the case of Class B ultra-linear operation, although the second tube will be similarly driven to reduce its gain, there is no signal in it at that moment in time because its Control Grid #1 is at cutoff bias during the relevant half-cycle of signal input so zero or near zero current is flowing in that tube.

Thus a Class A ultra-linear amplifier will have completely different behaviour to a Class B ultra-linear amplifier.
 

Distributed Load Operation

On the other hand, that portion of the output transformer primary winding between the Screen-Grid tapping and the centre-tap is not subject to cancelling out by out of phase signal in the power circuit.

This portion of the winding - usually 18-19% Plate to Plate load impedance - therefore imposes a load directly onto the Screen-Grids.

Thus a portion of the power output is delivered through this Screen to Screen winding.

However, although the winding and its corresponding load impedance is presented to the Screen-Grids, they are incapable of delivering much power because the Screen-Grids will be aligned in the electron stream such that the small diameter grid wires are not directly in the electron stream - so because of space-charge effects cannot attract electrons.

Furthermore, the Screen-Grids are charged to opposite polarity to the electron stream during half of each signal alternation cycle, which further detracts from their electron collecting capability.

For the record, in their US Patent 2,710,312 Hafler and Keroes state: "It should be noted that power is transferred to the Screen only over part of the signal cycle, i.e. when the absolute value of Plate potential falls below the absolute value of Screen potential. This transfer has the effect of linearizing the Plate characteristics."
 

Power Output

Ultra-linear operation typically (for the popular audio tube types) delivers 100% power output compared with the same tubes in tetrode or pentode connection at the same plate voltage and bias system (GEC) - but sometimes less for other tube types.

For example, Mullard quote the power output for the  EL34 tube as being the same for pentode and ultra-linear connection with 20% turns screen-taps, but for minimum distortion the screen-taps increase to 43% turns, which provides a power reduction of 15% (but distortion is halved).

Note:    Worthy of note is the KT88, which GEC (MO Valve Co./Genalex) claim produces the same power in ultra-linear connection as in Pentode connection at 43% turns. This may be due to the applied Screen-Grid voltage used in ultra-linear operation being twice the recommended value as that for Tetrode operation. Importantly, the electrode structure of the original GEC-MOV KT88 is internally physically identical to the GEC TT21 and TT22 RF transmitting beam power tubes, so the GEC-MOV made KT88 is therefore capable of operating at its rated 600 VDC maximum Screen-Grid DC operating voltage continuously without distress.

The TT21 or TT22, which have a top anode cap and intended for professional broadcast use, may be used in lieu of the KT88 as a direct electrical substitute. Grid to plate capacitance is reduced, which should produce superior audio performance over the KT88.

One of the challenges to the home constructor is that manufacturers' tube manuals and data sheets often quote "typical operation" for fixed bias pentode connection but cathode-bias for ultra-linear connection.
 

Ultra-linear Circuit Characteristics:

The following graph, courtesy of GEC and AWV Radiotronics Magazine (May 1959), shows the comparative characteristics of the KT88 in triode, pentode and ultra-linear connections. These comparative relationships between the different connection configurations should be typical for most (but not all) tetrodes, pentodes and beam power tubes.

Note:

This graph shows that for ultra-linear operation the original GEC/MOV KT88 is unsurpassed!!
 

WARNING: 6550 V KT88


WARNING:


EL34/6CA7 Ultra-linear Circuit Characteristics:

The following graph by Mullard UK shows performance of the EL34/6CA7 valve.

Note the comparative "linear" performance for power out v distortion over a wide range of loads and similar operating conditions.
 

Further information is provided by the following table ex Mullard UK, that shows comparative performance between the Mullard EL34 and EL84 valves in various output stage configurations.


 

Recent Research

I am indebted to Rudolf  Moers, a distinguished Electrical and Electronics Engineer located in the Netherlands, who has made available for us his wonderful recent scientific investigation into the design theory and practice of Ultra-linear audio amplification.

These papers are posted with permission from Linear Audio www.linearaudio.net and their author Rudolf  Moers.

Part 1 - Power Point presentation

Part 2 - Power Point presentation

Parts 1 and 2 of above Power Point presentation - combined in pdf format

Paper - The Ultra-Linear Power Amplifier: An adventure between triode and pentode - pdf format
 

The engineering design methods developed by Mr Moers may be used to determine theoretical plate/screen load ratios for ultra-linear operation of power tubes.
 

Application:

In practice, for tube types other than KT88, the real loss of useable output power from the ultra-linear connection is actually significantly less than any power differential measured with resistive loads might suggest (ie a 1.5 db reduction in loudness produced by the loudspeaker), because the ultra-linear connection produces a higher coupling efficiency between the amplifier and loudspeaker than tetrodes or pentodes - ie is more triode like - thus approximating an equivalent "loudness" to pentode connection.

This phenomenom is particularly true of low frequency reproduction, suggesting that ultra-linear connection is superior for double bass violin, bass guitar and general hi-fidelity reproduction down to about 40 Hz - which is the lowest musical frequency normally reproduced in popular music.

The 6L6 family of tubes can be used for ultra-linear connection, but only safely and reliably at reduced voltages not exceeding rated screen-grid voltage. More suitable 6L6 style types include 5881 and 7027A - see manufacturers' data sheets for typical circuit values.

Ultra-linear operation typically delivers about two to three times the power output compared with the same tubes in triode connection at the same plate voltage.

Most importantly, inter-modulation distortion is substantially lower with ultra-linear connection compared with pentode or triode connection for the same tubes.

Note: In the case of hi-fi systems, percussive instruments such as the bass drum, tympany, harpsicord, guitar and the piano present typically short-duration/transient low fundamental frequency signals rich in harmonics. Thus they are less demanding to reproduce than the electronic organ, pipe organ or bass guitar, which produce an essentially long-duration signal, approximate to a sine wave, in the lower register.

Triode connection is still the preferred option for seriously loud organ music because of the lower frequencies to be reproduced, the need for "boom" free bass around the loudspeaker resonant frequency and, most importantly, a need for consistent gain (flat response) throughout the musical scale to ensure all notes are reproduced with equal loudness when they are recorded that way. However triodes deliver substantially higher intermodulation distortion than ultra-linear operation so adequate power headroom is essential to ensure clear mid to upper range reproduction simultaneously with sustained lower register signals.

A major advantage of ultra-linear connection over tetrodes and pentodes is the improvement in tone for recorded music and public address reproduction. The tone is more triode like, being smooth and mellow (but "clearer" than triodes), compared with the harsh tone of tetrodes and muddy tone (read "less definition") of pentodes.
 

Guitar Amplifiers:

Ultra-linear operation is not usually suitable for lead guitar amplifiers because it lacks the "crispness", "harshness" or "bite" in the sound commonly provided by beam power tetrodes and pentodes, however some jazz and country music guitarists may find it preferable where a smooth, natural mellow tone is sought.

However, ultra-linear operation is superior to tetrode/pentode connection for bass guitar applications because low frequency power delivered to the loudspeaker is substantially greater - ie it is "louder" - due to lower output impedance and improved coupling to the loudspeaker.

It also has a "deeper" tone, suggesting improved sub-harmonic performance.
 

Summary:

In summary, ultra-linear connection offers:

Ultra-linear operation also enables lower quality loudspeakers to be used for satisfactory results.

Note that in high power applications - ie more than 100W RMS, transmitting TRIODE tubes such as 805, 809, 810, 811, 812, 833, 845, 8000, 8005, etc may be a more economical and practical solution than trying ultra-linear configuration with tetrodes/pentodes because of simplified wiring, output transformer and power supply requirements.
 

TUBES FOR ULTRA-LINEAR OPERATION

The "Ultra-linear" configuration avoids the conventional conflict between plate and screen voltages by creating a voltage divider network through the output transformer primary to AC earth (transformer centre-tap), ensuring the screen voltage tracks and thus always remains, both below and proportional to the plate signal voltage. By this configuration, the screen is intended to be prevented from exceeding its power dissipation rating.

This statement is subject to the Screen Grid always being operated within its Rated DC Screen Grid voltage and at a DC voltage less than that of the Plate.

Applied plate (B+) voltage for ultra-linear connection should never exceed Grid 2 rated voltage, so standard ultra-linear configuration is only suited to tubes designed for audio applications having a Grid 2 rating approximating either the plate voltage rating (or actual applied plate voltage if less than rated maximum).

Important:    Reference to tube data sheets will show that few output tubes have ratings remotely matching this requirement.

Important:    The use of poor quality output transformers having a high DC resistance in the primary windings may establish a situation whereby the actual DC Plate voltage drop across the transformer primary winding is high, causing the DC Screen-Grid voltage to be higher than the Plate at high signal levels.

Unfortunately, only a small number of tube types are thereby suitable for ultra-linear operation, because in ultra-linear mode the screen grid is operating at or above the plate voltage - a dangerous operating region for any tube.

Only a few tube types were recommended by their manufacturers as being suitable for ultra-linear connection, the most notable being EL34/6CA7, EL84/6BQ5, KT88, 6550, 7027 and 8417 - see manufacturers' data sheets for typical circuit values.

In a typical output tube, the Screen Grid is the ANODE, or positive electrode. It is designed to accelerate electron flow from Cathode to Plate, but is structured in such a way that most electrons pass through it and on to the Plate for collection.

Excessive screen grid voltage attracts excessive electrons to it, thereby resulting in excess grid current, excessive grid power input/dissipation, overheating and melting. The fused screen grid wires may short-circuit B+ to earth, damaging the output transformer and/or power supply components.

Unfortunately, most audio tetrodes, pentodes and beam power tubes are designed such that the Screen Grid may be operated only up to a maximum DC voltage that is well below the Plate voltage - typically 150 to 300 volts.

Plate current is very much controlled by the Screen Grid, thus when the Screen Grid is made ineffective (ie over-active) by application of excessive voltage, the Plate Current is likely to exceed the tube ratings and also melt the plate.

Another way of saying this is that if the Screen Grid voltage is excessive, the capability of Grid #1 (Control Grid) to control electron flow in the tube is diminished - or lost altogether.

Operation of typical power tubes (having a screen grid voltage rating substantially lower than the rated plate voltage) in ultra-linear connection is likely to result in loss of control over electron flow by the screen, resulting in thermal runaway or dynatron action - resulting in self-destruction of the tube. Fire is a constant risk.

For example, tubes designed for RF power service typically have a plate voltage rating many times higher than their corresponding screen grid voltage rating.  This class of tube (eg 807, 2E26, 6146, 4CX series) has the screen grid physically positioned close to the control grid (Grid 1) and tend to self-oscillate or suffer thermal runaway when the screen grid voltage is higher than their rated screen-grid voltage, which is always the case with conventional ultra-linear service, thus rendering them unsuitable for ultra-linear service.

WARNING: When using tubes fitted with a plate top cap (anode cap) in ultra-linear configuration, consider also the risk of self-oscillation and/or parasitic oscillations from the combination of  long leads from the plate top caps and screen grids to the output transformer - particulary significant when using multiple tubes to obtain higher power. Some of this lead length may be avoided by chassis layout however in the case of top cap style tubes the screen connection is always under the chassis, thus ultimately requiring connection in some way from top to bottom of the tube through the output transformer.
 

High Power Ouput:

Where more power is needed it is preferable to use a larger tube that is known to be more suited for ultra-linear circuitry, such as the KT88, KT90 or 813.

Another useful option is to run multiple pairs of tubes in parallel push-pull, such as the arrangement used in the GEC 400W KT88 Amplifier. It goes without saying that normal precautions against instability and parasitic oscillations are essential in layout, lead dress, use of grid stopper resistors wired directly to the pins, and keeping inputs away from outputs. The output transformer must be of high quality with low leakage capacitance and  low leakage inductance between windings. Tubes should be mounted close together to minimise inter-connecting lead length. Grid #1 circuit resistance must be held within the manufacturer's ratings. A low-impedance driver, such as a cathode-follower or transformer is recommended.

For high power applications where high plate voltages are needed, success may be achieved by adding a separate winding for each screen grid to the output transformer, to enable the screens to be AC coupled to the plates thus providing ultra-linear operation, but separated from the DC plate supply, thus enabling the screens to be supplied within their rated voltage from an independent supply.

The ACROSOUND 100W Ultra-Linear Amplifier using 6146 tubes and the TO-350 Transformer is an example of this excellent and innovative design configuration.

One of the benefits of this configuration is that the Screen-Grid supply can be independent to the Plate supply and therefore better regulation can be incorporated into the Screen Grid circuit. Noting the Screen-Grid is the ANODE in a Tetrode or Pentode tube, a regulated supply will deliver improved transient response and a "brilliance" to the reproduction not available with a common B+ power supply as is the case in conventional Ultra-Linear operation.

Bruce DePalma, one of the few Gurus of modern hi-fi amplifier design, presents an interesting and comprehensive commentary on the core design philosophies supporting this approach in his excellent Design Paper - "Analog Audio Power Amplifier Design"

Bruce has developed designs that enable both Ultra-Linear and low Screen-Grid voltage technologies to be successfully integrated in such a way that extremely hi-fi performance results.
 

BIAS

Grid 1 bias for ultra-linear operation is normally higher than that for tetrode/pentode connection, so output stage sensitivity is reduced. Higher output voltage from driver stages is therefore needed.

Grid-stopper resistors to Grid #1 and Grid #2 are still required for ultra-linear operation.

Important: The technique of using a silicon rectifier diode on each Screen Grid in series with the Grid-stopper resistor, as described in my "OPTIMISED ULTRA-LINEAR ©" page, is very helpful in ultra-linear connection. AC signal output voltage from the Screen-Grids is prevented from conducting through to the output transformer, which means that theultra-linearoutput stage operates as a negative feedback system only - ie Plate voltage is fed back to the Screen Grids via the transformer taps but not the other way around. This method ensures all the electron flow goes to the Plates, with all the advantages described previously.

Note:    This configuration has opposite polarity as when zener diodes are installed to reduce screen-grid voltage, so therefore the benefits described are not relevant to the zener diode technique.
 

Important - When using a high B+ voltage:

To ensure Plate Dissipation remains within manufacturer's rating at both zero and maximum signal, it may be essential to use Class B operation - thus introducing further complexity into the circuit design and perhaps offsetting much of the benefit offered from the ultra-linear configuration. It may be more prudent to use a tube having a higher rating.
 

RECOMMENDED TUBES

The following tubes are known to be suitable for conventional ultra-linear operation having nominally equal Plate and Screen-grid DC supply volts:

TUBE TYPE            MAX. ULTRA-LINEAR CONNECTION PLATE TO CATHODE VOLTS

6AQ5A                            285
6BM8                              300
6BQ5/EL84                      300
6BW6                              285
6CA7                               500
6CM5/PL36                      250
6CZ5                                285
6L6GB                             270
6L6GC                             400
6V6GT                             285
5881                                400
6146A, B, W etc               250    (Top cap style)
6550                                450    (Note: GE USA "design maximum" rating. Some brands may not tolerate this voltage)
6973                                300
7027A                              450
7189A                              400
7581                                450
7591                                400    (50% turns recommended by RCA)
7868                                400    (50% turns recommended by RCA)
807/5B-255M                    300   (Top cap style)
8417                                500
EL34                                425
EL84/6BQ5                       300
KT66                                400
8417                                500
KT88                                 600     (Note: GEC UK "design maximum" rating. Some brands may not tolerate this voltage)
KT90                                 600    ("Absolute Maximum" rating)
813                                  1100   (This tube is an excellent option for serious audiophiles but has a top cap and requires a centre-tapped 10 V AC/DC filament supply - 5A per tube. External wiring must be screened to prevent RF induction and parasitic oscillations. Adequate ventilation is essential)
 

In my humble opinion, the most suitable candidates for ultra-linear connection are:

6CM5/PL36    (250 VDC)
6AQ5/6V6GT (275 VDC)
6CZ5/6973      (285 VDC)
5881/6L6GC/7027A/7591/7868    (400 VDC)
KT88/KT90   (600 VDC) (may need to reduce to 450/500 VDC with non GEC manufacture)
813              (1100 VDC)

All are well proven fine quality beam power tubes - each famous in its own right.
 

Other options are the EL34 (425 VDC) or 6CA7 (500 VDC) or EL84/6BQ5 (300 VDC) pentodes, however in my opinion sound from these tubes is not as clean as those above. All of this family of tubes "sounds" similar because of their generally identical electrode construction.

If you have had successful experiences with other tube types that are useful for ultra-linear connection please email your comments.
 

For further information regarding Ultra-linear operation of vacuum tubes see my OPTIMISED ULTRA-LINEAR © OPERATION page
 
 
 

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

BACK TO HOME PAGE
 

This page last modified 24 March 2014

This page is located at http://www.oestex.com/tubes/ul.htm