Go to home Aspen Amplifiers Go to home
products FAQs discussion buying contact













HiFi Issues
Tubes or Transistors - Science or Art

Hugh R. Dean

I offer a few musings about hi-fi, particularly amplifiers and speakers.  For many of us the hi-fi journey seems effortlessly to span a lifetime - a type of febrile, nervous affliction.  But like so many other absorbing pursuits, it could benefit from a little objective analysis, based on a few practical observations and some enlightened tinkering.  What follows is not an engineering treatise; it is a considered opinion, presented here for interest but without substantive evidence aside from experience.
In 1972, the American Russell O. Hamm wrote a seminal paper comparing tubes with transistors.  The paper is long, but in essence, Hamm, who came from a pedigreed recording background, argued that tubes sound better than transistors as full overload is approached. The reason, he said, is that tubes overload gradually, taking almost 15 decibels of 'grace' before they produce objectionable distortion, whereas transistors are lucky to give more than two decibels of 'grace' before the onset of hard clipping.  While amplifiers using both devices certainly sound terrible at full overload, the nature of music is intrinsically peaky, and the minor overload condition is very common in domestic settings. Thus for music, Hamm argues, tubes will always sound better.

This is probably simplistic and arguable.  In fact, Peter Walker of Quad fame once said that, kept within overload limits, tubes and transistors sound just thesame.  Daniel D'Agostini, the owner of Krell, surely an impressive name in solid-state amplification, recently derided the tube aficionados, saying that hi-fi is about headroom and only solid-state amplifiers can meet this demand.  He builds amplifiers of outrageous power and ruthless, surgical accuracy; the latest reportedly delivers peaks of 24,500 Watts into 1 ohm for short periods.  Notwithstanding, the jury is still out.  However, it should be said that the cost of owning a tube amplifier can be quite high since tubes burn out just like light bulbs.  For many audiophiles this aspect is a deterrent.

Yet there is a large and growing group of music lovers who swear by tubes, and who take pains to point out that in terms of tone, soundstage, stage presence and sheer ease of listening the tube is superior.  After some decades in the solid-state camp, I am now not so sure.  To me it appears to be a matter of taste; and the truth is, I all of us change my tastes according to circumstance.  So whom do I believe, and why?

Many years ago when transistors were new, the design of solid-state amplifiers was dominated by the engineering requirements – and frequently also by a long tradition of tube amplification.  At a time when tubes had reached the pinnacle of their development with the high power beam tetrode, the power pentode and highly refined transformers, the first readily available transistors were constructed of germanium.  They soon proved to be thermally vulnerable.  A great deal of engineering was required to ensure the output stage survived all operating conditions, and in meeting this requirement sonic considerations were necessarily pushed to second rank.  Furthermore, the early devices, even in silicon, exhibited inferior secondary breakdown characteristics, and many pairs of devices were required to ensure reliability - a design constraint still in evidence today.

The active element of a transistor is so small – less than 0.2 gram in even the larger devices – that efficient heat dissipation is crucial to device survival.  This has major repercussions for processing music signals because the electrical characteristics of the semiconductor materials vary considerably with temperature, which largely depends on how hard the device is driven at any instant in time.  In marked contrast, tube elements are large, metallic, and weigh up to fifty grams each.  Furthermore, the parasitic capacitances and leakages of transistors are much more adverse than tubes because of the nature of charge aggregation in a solid-state crystal.  Thus electrical parameters are far more stable in tubes with temperature and voltage effects, and instantaneous effects which might affect the music are vastly less intrusive.  And while transistor amplifiers can be made with vanishingly low distortion by generous use of global negative feedback, it may be that temperature, voltage and associated bias effects, one of the significant differences between these devices and tubes, could influence the way in which they process music.

Over the last twenty years recording technology has advanced to the point where differences in domestic hi-fi amplification can be easily discerned.  While many would hold that modern digital recording is inferior to the best vinyl engineering of yesteryear, it is still true to say that there are some very competently recorded CD’s in today's marketplace which can tax any hi-fi system to its limits. Interestingly, the best recordings are rarely classical, but often jazz and other contemporary popular music.
One of the issues of modern recording techniques has been gradually increasing flexibility - and complexity, particularly since the advent of digital technology.  These developments have enabled the sound engineer to process and mix down multiple tracks days and even years after the recording event.  While this flexibility has conferred greater control over the final recording it has led to a diversity of methodologies - some of them approaching the weird - particularly in sound stage composition.  For this reason, many criticisms of artificial sound staging and improper instrumental and vocal emphasis can usually be attributed to inappropriate sound engineering.  But today's distortion performance and bandwidth are unquestionably superior to those that prevailed twenty years ago.

The differences between tube and transistor sound do not merely concern the devices themselves.  I have found the topology of the amplifier to be very important.  Gordon Rankin of Wavelength Audio, (a successful manufacturer of single-ended tube amplifiers in the US), argues that an amplifier's quality comprises equal parts of componentry, layout, and topology.  If the same components are connected in a different configuration, it is logical to presume the sound will be different.  Thus, all other factors being equal, (and they seldom are!), the push-pull amplifier sounds different to the single-ended (SE) amplifier, and for some reason not fully understood the SE amplifier always features a more prominent and soulful midrange.  I can certainly verify this; I prefer the tube SE or MOSFET SE sound while a number of longstanding audiophile friends  prefer the push-pull, solid-state sound.  

Nelson Pass, a prominent US audio designer, believes that the SE amplifier deals with the positive half cycle in a different way to the negative; this mimics differences in the compressive and rarefactive behaviour of air and thus sounds more 'natural'.  I have some difficulties with this statement; by way of a test, what happens if I reverse the speaker leads?  Having swapped compressions for rarefactions, the sound should be very different.  However, when this is done, the SE amplifier remains easily distinguishable from the push-pull.  It is also true that a transformer output stage adds a degree of 'rounding' to the sound; Trace Elliott made a MOSFET guitar amplifier with a transformer output stage and the sound was very similar to an ultra-linear pentode amplifier.

Following on from topology issues is the great feedback debate. Feedback is highly relevant to this discussion; it marks a fundamental difference in the way tubes and transistors are used.  Feedback is much more complicated than first appears, and in few other areas of amplifier design has the debate been more dominated by measured performance.  I have tinkered with the feedback loop of an amplifier and carefully listened to music at varying gains between 5 and 50.  You soon appreciate that feedback is intimately related to sound quality; in general, (and with some caveats), the higher gain, lower feedback configuration sounds better.

Tubes are much more linear than transistors, and can be used in zero feedback designs without difficulty.  In fact, too much feedback in a tube amplifier induces instability because of phase shifts in the output transformer.  But transistor amplifiers require lashes of negative voltage feedback - around 30dB, which represents a reduction in gain of more than thirty times.  While this keeps distortion and output impedance down I believe that in almost every instance a sonic price is paid.  Feedback, whilst commendably improving measured performance, somehow drains the vitality from the music, conferring a flat, uninspiring sound.  The reasons may relate to dislocated phase relationships in the music that are caused by the group delay of the feedback process.  I concluded that a transistor amplifier with less feedback would sound more lifelike, and verified this in my observations.  Measured performance degraded, however - you can't have your cake and eat it too, even in audio!

While amplifier design has clearly been driven until now by measurement philosophies, one can understand this because all electrical phenomena should be measurable, and the approach is founded in rationalist, scientific methodology.  However, using conventional Lin topologies a zero feedback transistor amplifier is almost impossible to construct.  This is because transistors suffer forbidding non-linearities when configured for voltage amplification.  And only in recent years have alternative low-feedback topologies begun to emerge.

The unnerving truth is that almost everything in a hi-fi system, even the hookup wire, makes a difference to the sound, and when this surprising but reasonable observation sinks in, these effects become truly bothersome to the hobbyist and committed designer alike.  I recall once replacing a teflon-insulated, silvered copper earthing wire with a PVC-insulated 'regular' copper wire; the splashiness disappeared and the high frequencies lost their jagged, bright edge.  No other changes were made; a small alteration produced quite astonishing changes in the sound, and none of this was visible on an oscilloscope with any waveform.  I have found similar differences with capacitors - which frankly left me amazed - and now have quite definite prejudices about what to use and even where to use it.  But infuriating as it may seem, different amplifiers respond to different approaches; a successful tweak on a SE amplifier may not work so well on a push-pull amplifier.

A critical difference between engineering and art lies in the final stages of amplifier preparation for the market.  Once an amplifier has been successfully designed and tested, the business of 'voicing' the amp must begin.  Design and testing fulfils the engineering requirements of unconditional stability, reliability, bandwidth, distortion and sensitivity.  But a 'blameless' amplifier may sound quite unappetising; it may be bland, and lifeless.  This applies as much to tube as to transistor amplifiers, and often means the difference between commercial success and failure.  Voicing the amp relates to component choice in particular, with some emphasis on layout, particularly earthing, which can easily destroy an otherwise competent sonic presentation.  

Voicing is not measurable, but bears testimony to the extraordinary discriminating powers of the ear.  Tricks used on one design may not work on another; this also demonstrates quite clearly that distortion and bandwidth measurements, in all their quantifiable glory, are not telling us the whole story.  This is all the more remarkable when I consider that the SE triode amplifier, widely acknowledged as the zenith of sonic realism, boasts very high total distortion of typically 3%, an inferior output impedance characteristic and a somewhat constrained bandwidth.

A promising avenue remains somewhat unexplored in the area of hi-fi amplification; the hybrid amplifier.  The tube is the voltage amplifier par excellence.  The transistor is a wonderful current amplifier, a role in which it happily serves with almost zero feedback.  I remain convinced that the careful combination of the two devices leads to superior amplification. Tube amplifiers are renown for a deep, layered soundstage and subtle, low level detail; transistor amplifiers are admired for their pin-sharp lateral imaging and high-powered transient performance.  It is possible to build such an amplifier using tubes solely for voltage amplification, and transistors for current amplification.  Yet another interesting possibility exists; the integration of a tube into a conventional transistor amplifier, which somehow preserves the powerful advantages of the solid-state amplifier with the tonal subtlety and low level detail of the tube.

I have designed and tested an amplifier to the first topology (the Glass Harmony), and recently completed a solid-state amplifier of 80W per channel using the second topology (the Lifeforce).  The Glass Harmony, a 30W SE amplifier, has matured since the original prototype; and now I believe it well combines the dynamics and stunning resolution of solid-state with the tonal subtlety and warmth of the tube.  The Lifeforce amplifier exhibits low-level detail and a deep sound stage consistent with a tube amplifier and yet preserves the transient response, lateral imaging, efficiency and power of the solid-state amplifier.  It certainly lives up to its name.  There is even evidence of a more prominent, musical mid-range which confers some of the sonic qualities of the SE tube amplifier.

Hugh Dean

© Copyright Hugh R. Dean 1999
All rights reserved