Tubes or Solid State? Or Both?

Tubes

Tubes came first, of course, so there is a certain nostalgia associated with them. Tubes were well understood in terms of electrical characteristics and their ability to amplify signals ever since the invention of the Audion by Lee DeForrest in 1906. By the 1920s radio sets were in almost every home and the need for high-power sound amplification came about with the introduction of "talkie" movies. In order to fill a theater with sound, high-power audio tubes were needed, along with efficient speakers. The most ubiquitous of the early theater amplifiers were made by Western Electric, using their famous 300B triode tubes.

Tubes suffer from some serious drawbacks: there are hot since they have heater filaments, they have a limited lifespan and power handling, they take up a lot of room, they are delicate, require high voltage to operate and they are microphonic. Despite all that, there is a loyal band of tube-head audiophiles who will consider nothing else. Furthermore, the demand for reproduction tubes has soared, with new companies emerging to manufacture them. In fact, Western Electric has been resurrected as a new company making tubes and amplifiers based on original designs. Single new 300B tubes are listed at $699 each! Russian and Chinese companies have been making reproduction vacuum tubes for some time (often of dubious quality), and more recently a small Austrian company called E.A.T. has risen to prominence as a maker of superb hand-built tubes in their Czech factory near Prague.

A new Western Electric 300B

So what do tubes do well in audio circuits? For one, many guitar players favor tube amps for their tonal qualities. This is a bit of a corner case since they include distortion as part of these tone characteristics, so not applicable to audiophile discussions. Aside from that, tubes do have some advantages over solid-state amplification:

For a discussion of tube theory and operation, see this article at engineering.com.

An early Williamson design - the distinguishing feature is the feedback circuit from the speaker output, theoretically reducing transformer-induced distortions and tube non-linearities.

The history of tube amplifier designs is too extensive to relate here in detail, but there were a few topologies of interest. The earliest designs were single-ended, meaning only one active device, transformer-coupled. Later designs used pairs of tubes in a push-pull configuration in an effort to achieve more power. These required center-tapped transformers. Williamson is famously credited with the first of these and his design (from 1944) became a standard. He also is credited with describing global feedback to reduce transformer-induced distortions.

Earlier, Blumlein had come up with a "distributed loading" modification using pentode tubes and an additional tap on the output transformer that was connected to one of the grids. This was later made into a practical design and renamed "ultra-linear", first described by Hafler (the creator of Dynaco and much later solid-state amplifiers bearing his name)  and Keroes in 1951. This resulted in more linear amplification at higher power levels and became a standard design, still in use today. McIntosh and Gow, in the late 1940s, had come up with a way to improve transformer performance by adding additional windings which were connected to the cathodes of the output tubes. This required a much more complicated (and expensive) transformer but is still in use today in McIntosh high-power tube amps, which are regarded by many as the finest available in the high-power category. For a good discussion of some of these old tube designs, see Rod Elliots' website article here.

An "Ultra Linear" tube amp. Similar to the Williamson design, but with transformer taps connected to the pentode screens.

One of the most successful amplifier tube kits for the hobbyist market came from Dynaco, founded in 1955 by David Hafler. In 1959, they released the stereo Dynakit ST-70 (also available assembled) for just under $100. Often called "the poor man's McIntosh", it sold over 350,000 units and is still prized today. The circuit used the same ultra-linear output connections described above and produced 35 watts per channel. Hafler's "secret sauce" was the output transformer. He had previously run a transformer company called Acrosound and had patented a winding technique that resulted in good transient response and wide bandwidth. Modern attempts to duplicate his transformer design have run into difficulty.  At the time, Heathkit only offered monaural tube amps, so once the stereo craze began to take hold, the Dynaco ST-70 was a more convenient option. Dynaco ceased operations around 1980, but clones and improved versions of the ST-70 are still available. Here is one example. See this article for a good history of the ST-70.

Futterman's OTL tube amp - published in the October 1956 edition of the JAES. Note that the speaker output is connected directly to the tubes.

Click for larger version

Another (but not very popular) tube topology is the OTL (Output Transformer-Less) tube design. These are somewhat scarce, but the idea was to get rid of the transformer and drive speakers directly from tubes. This requires tubes to handle much more current than usual, so is stressful and demands high-current devices. Some would claim that this is really the realm of solid-state devices and is a square peg in a round hole. One of the earliest practical designs was the Futterman OTL amplifier, patented in 1953. Futterman hand-built his amps for clients and the design was later commercialized by New York Audio Labs (NYAL). Some complained about stability and reliability problems and the company ceased production around 1987. The Atma-Sphere company created a line of OTL amps starting in 1976 and is still selling them today. These are very high-end devices and are priced accordingly.

For a discussion of the problems with tube OTL designs and possible improvements, see John Broskie's blog article at tubecad.com

Some modern tube amps, such as those from Cary Audio, offer flexibility in terms of operating modes. One model offers switchability between ultra-linear and triode modes, for example. Another favorite activity among tube-o-philes is "tube rolling", where different tubes (of equivalent pinout and design) are swapped in and out to test for subtle differences in sound.

For the DIYer, many considerations must be taken in designing and building a tube amp. Some of this is lost art and requires some archeology. For example, you must decide if you want to drive the tube filaments with DC current or AC, Some tubes when their filaments are driven from the AC windings of a power transformer, will inject some 50 or 60-cycle hum into the tube's circuit. Various techniques have been used to reduce this effect, such as a hum balancing pot connected across the filament feeds with the wiper grounded or tied to a DC voltage source. However, some designers have opted to drive the filaments from a DC supply, requiring additional power supply components. Then there is the high-voltage supply. Some designs have supplies ranging from 300 to more than 500 Volts DC, although at fairly low currents. The very best designs have regulated supplies, requiring additional complexity. Recently, some modern designs have utilized high-voltage MOSFETs for regulating tube supply B+ rails. Older designs used shunt regulators such as the OA2 tube, but these were minimally effective. The transformer is one of the most critical parts and must be carefully designed for audio. Various schemes have been devised to overcome transformer limitations (as mentioned in the previous paragraphs), but good transformers are bulky and expensive.

Transistors

So having said all that about tubes, what about transistors? The advantages for portable and low-voltage applications were obvious. Early transistors (first invented in 1949) were based on germanium substrates. Germanium transistors were not very robust and were difficult to use reliably in high-power designs.

A Delco 2N278 Germanium Transistor

Car manufacturers were keen to develop transistorized car radios due to the fragility of tubes and the need to develop high voltages for them, usually using "vibrators" and step-up transformers. In fact, Delco, the electrical subsidiary of GM, developed their own germanium power transistors (designated 2N278) for the power amplification stage in their radios. Their first all-transistor car radio appeared in 1957 and was initially only available in Cadillacs. These radios used transformer coupling in the output stage, which was mounted on a separate chassis.

Schematic of the first all-transistor car radio. Click for a larger version.

Until transistorized equipment became widely available, HiFi largely remained a niche market confined to enthusiasts willing to spend lots of money on tube gear, but then all that began to change.

Early attempts at Hi-Fi audio power amplifiers in the 60s using germanium power transistors were often met with failure, which gave solid-state designs a bad reputation. I'll share a funny story about this: when I was working in a HiFi shop we received an early release of a Marantz solid-state power amplifier for customer demos. By this time, reliability had improved due to the use of silicon transistors. Our head repair guy, who was naturally skeptical and nervous about new "stuff" asked us to set the amplifier up on the bench for testing. Unknown to him, one of the sales guys ran a piece of tubing up through a hole in the bench and out into the adjoining room. When the time came to "fire up" the amp, the sales guy blew a generous puff of cigarette smoke into the tubing which wafted up through the amp. The head repair guy went into a panic and ran for the circuit breakers - only to realize he had been pranked and we all had a good laugh. The amp was fine!

Audio manufacturers had another challenge: they were well familiar with tube designs, but to many designers, transistors were a whole different design game. New circuit topologies had to be invented and many attempts were not great. In the late 60s and 70s Japanese designers were at the forefront of this effort and many good designs began to emerge from Sony, Kenwood, Pioneer, Sherwood, Sansui, and others mostly aimed at the mid-Fi integrated receiver market. The popularity of Japanese gear was made even greater by servicemen returning from overseas, where they were able to purchase equipment at PX bargain prices. Unfortunately, back in those days 1% THD (usually measured at max power) was considered "good enough" for mass-market equipment. Integrated Japanese receivers available at reasonable prices fueled the audio market.

An interesting perspective on what helped created the boom in audio during the early '70s is the following: The '50s were an era of unprecedented economic growth and prosperity in the US and Baby Boomers were somewhat pampered. Their parents wanted to make sure their children did not suffer deprivation as they had during the Great Depression and WWII, so consumer spending on non-essentials (like entertainment) really took off. Boomers grew up in the age of television and were immersed in entertainment (although people only got a few TV stations). When pop music exploded with Elvis, the Beatles, Soul, and the British Invasion, music was something everybody wanted to have access to - and that meant acquiring records and listening equipment. Especially as Boomers started going off to college - every dorm room had stereos. These were the days long before the internet and cell phones, so listening to music filled a void.

Simple push-pull totem pole output stage common in early amps.

Early transistor amps had various topologies, initially transformer-coupled as in the Delco example, but then moved to simpler totem-pole output stages. These were usually NPN transistors (because low-cost power complementary PNP transistors were not widely available) running off a single supply rail with large output DC-blocking electrolytic caps in series with the speakers. One problem with this topology was that it was asymmetrical; the upper transistor functioned as an emitter follower (which is good for low impedance) but the lower transistor was in a common emitter configuration, with the collector (high impedance) connected to the emitter of the upper device. This resulted in non-linear amplification of the positive and negative halves of waveforms.  The cure for this was heavy amounts of negative feedback, which led to other problems. Eventually, these were supplanted by complementary push-pull pairs as better matching and cheaper PNP power transistors became available, allowing direct DC coupling to speakers (and eliminating the coupling capacitor), thus enabling higher damping factors (better low-frequency speaker cone control) and greater power. Also, there is plenty of evidence that large electrolytic capacitors can introduce undesirable audible effects when used in the signal path.

As mentioned, most power transistors available on the market were NPN devices. When complementary power PNP started to come out, an old issue became more apparent: crossover distortion. This occurs when there is a "dead zone" during the signal crossing through zero, which causes a discontinuity in the output waveform. In the rush to push out commercial products, many designers failed to realize that this was the primary cause of "transistor sound" which plagued many of the early products and turned audiophiles (mainly tube aficionados at that point in time) against them. Crossover distortion and clipping, even in AB designs, resulted in harshness and were not as pleasant as tube amps in AB mode. Stability issues were also common, such as oscillation.

Designers resorted to complicated biasing schemes and "quasi-complementary" (mixing NPN and PNP) topologies to reduce the crossover issue, along with heavy negative feedback as with totem-pole designs.

Example Class B Circuit - How Crossover Distortion Happens - Courtesy Wikipedia

One of the earliest Heathkit all-transistor amplifiers - from their 1963 catalog. The kit was $134.95 - about $1300 in 2023 dollars!

For the DIY market, Heathkit began phasing out its tube-based audio kits in 1962 and in 1963 started releasing all-transistor amplifiers and receiver kits. There was considerable marketing hype around these, with phrases like "crystal clear transistor sound". Simultaneously, Allied Radio began to offer some transistor audio gear under their KnightKit brand. Others, such as EICO and Lafayette Electronics followed suit, but Heathkit was the largest supplier of kits; in 1963 they had over 250 kits in their catalog, covering many different hobbyist interests.

As early as 1967, DIYers were treated to articles in Popular Electronics from Dan Myer, the founder of Southwest Technical Products (SWTPCo), describing how to build the "Tiger" solid-state amps and they offered kits. These were early, very simple examples of decent (but not great) transistorized power HiFi amps, so amateurs could get an idea of what solid-state amps should look like. SWTP followed up with a series of ever-more-powerful Tiger amps in the 70s, with increasingly better power and distortion specs.

These amps helped kick off a wave of. solid-state DIY efforts since they were simple and relatively cheap. SWTPCo took advantage of newly released complementary PNP transistors to create simple amplifier circuits, but notice that the 'Lil Tiger ran off a single rail supply and still used an electrolytic blocking capacitor at the output. It did incorporate a "bootstrap" capacitor to help increase the output voltage swing.

Schematic of the SWTPC 'Lil Tiger Amp in the December 1967 Edition of Popular Electronics.

Then there was a race to bring out solid-state amplifiers with ever-increasing power outputs. The technique used was to parallel many output transistors. One early example was Crown, who ca. 1970 marketed a 150 WPC transistor power amp. Unfortunately, it did little to dispel reliability concerns, plus it did not sound very good. One problem was that the Crown designers had chosen to use early integrated circuit op-amps in the gain stages which were clearly not suited to hi-fi, resulting in poor high-frequency slew rates and distortion.

For the high-end audiophile market, a breakthrough of sorts happened around 1969 when Hitachi developed power "lateral" MOSFETS. These were very linear devices that exhibited tube-like characteristics and sounded better than the bipolar transistor designs of the era. They built some of their own amps with these but also sold devices to other manufacturers, most notably David Hafler - the same guy who was behind Dynaco back in the tube days. Hafler hired Erno Borbely in 1978 to design some excellent and popular cost-effective amps (sold both as kits and completed units) based on the Hitachi MOSFETS.

The Hafler DH200 was a very advanced design - note the use of complementary differential pairs in the input circuitry with constant current sources to increase linearity. The circuitry is fully complementary, no doubt due to Borbely's desire for symmetry in order to achieve low distortion. The output MOSFETs were paralleled to achieve 100 watts per channel.

Hafler DH200 schematic - click for larger version

Around 1973-4 Sony and Yamaha developed their own special transistors for amplifier output stages. These were called "VFETS" or "vertical" MOSFETS and exhibited even better linear transfer characteristics. The name VFET came from the gate structure of devices that resembled V-shaped grooves in the substrate. VFETs of this era were what is called "depletion" mode - meaning a bias is applied to the gate to decrease conduction - and had low drain resistance in the on (conducting) state. Also, they were produced as complementary pairs, meaning both N-channel and P-channel devices, so were suitable for push-pull audio output stages. These found their way into amps produced by Sony, Yamaha, and Sansui, among others. Sadly production of these devices ended around 1980 (mainly due to production costs) but constructors such as Nelson Pass hoarded enough to make them available to DIY hobbyists on a limited basis.

Another variation was the SIT (Static Induction Transistor) MOSFET. There is often confusion between VFETs and SIT transistors. Technically, they are very different. The SIT has a "buried" gate and is similar to a JFET, which usually had a surface gate. These were targeted toward industrial, very high-power applications and were only available in N-channel form. This limits their use in either single-ended designs (requiring an output blocking cap) or bridged topologies. However, they have seen a renewed interest mainly due to the First Watt SIT-3 amp, said to sound very tube-like.

Even though the MOSFET popularity has faded due to the lack of available specialty devices, MOSFET designs served an important function: they forced builders of bipolar transistor designs to improve their products and step up their game. Today, you can still get lateral MOSFETs from Exicon (see my amplifier build here). Another niche MOSFET is the silicon carbide (SIC) variation. These are like SIT devices and are used in high-power high voltage industrial applications, but some DIYers are building amps with them.

Eventually, the high-end solid-state niche market began to emerge from companies such as Luxman, Denon, and others. In the US, we had Conrad-Johnson, MacIntosh, HH Scott, Mark Levenson, and others all getting into solid-state audio.

So transistorized equipment won the lion's share of the market, but tubes have never gone away.

For the input stages of power amps, small-signal FETs were popular early on, due to their low noise and linear characteristics. The famed Toshiba FETs (no longer made) are prized for their linearity and low noise.

Hybrid Designs

What about hybrid amps? Why not have the best of both worlds and build audio gear using both technologies? There have been a few attempts to combine tubes with transistors in amplifiers. These typically use tubes as the input and amplification stages, while transistors are used in the power handling (high current) output stages. Some people claim this results in a better-sounding amp. You can also find some  DACs with tubes as the analog handling part of the chain. Also, some "mostly tube" designs have employed  MOSFETs as voltage regulators

A Tiny Korg Nutube. Why not?

An interesting example of a combination of transistors and tubes is a DIY design for a preamp from Nelson Pass which has FETs on the input and outputs but a Korg NuTube dual triode as the gain stage. The NuTube is actually a tube that looks like a large integrated circuit with a glass cover and you can actually see the filament glowing. The preamp is available as a kit.

There are also commercial examples of headphone amps using a combination of tubes and solid-state devices.

You could argue that a hybrid approach that uses tubes where they are best suited and transistors where they are the best choice is an optimal design choice, but this is rare. If you Google "hybrid tube transistor amplifiers" you'll see a lot of examples, mostly from small manufacturers.

Here is an example of an interesting contemporary hybrid amp - the McIntosh 901, which has 300 watts of tube power and 600 watts of solid-state output, intended for bi-amped systems. Of course, McIntosh does not make tube amps exclusively - they also have products like a 1.2KW Mono all-transistor amplifier!